Understanding Troposcatter Propagation

D~-:F OPY I
v
SECURIT r AIF IitAT NI nr TWI- PArC.
Form Approved
"NTATION PAGE OM8 No. 0704-0188
la. REP
UN, AD-A220 969 lb, RESTRICTIVE MARKINGS
NONE
2a. SEC 3. DISTRIBUTION/AVAILABILITY OF REPORT
APPROVED FOR PUBLIC RELEASE;
2b. DECLASSIFICATION / DOWNGRADING SCHEDULE DISTRIBUTION UNLIMITED.
4. PERFORMING ORGANIZATION REPORT NUMBER(S) 5. MONITORING ORGANIZATION REPORT NUMBER(S)

0 -003
AFIT/CI/CIA-9
6a. NAME OF PERFORMING ORGANIZATION 16b. OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATION
AFIT STUDENT AT Universityj (If applicable) AFIT/CIA
of Colorado I
6c. ADDRESS (City, State, and ZIP Code) 7b. ADDRESS (City, State, and ZIP Code)
Wright-Patterson AFB OH 45433-6583
Ba. NAME OF FUNDING/SPONSORING Bb. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBER
ORGANIZATION (If applicable)
87c. ADDRESS (City, State, and ZIPCode) 10. SOURCE OF FUNDING NUMBERS
PROGRAM PROJECT TASK WORK UNIT
ELEMENT NO. NO. NO. ACCESSION NO.
(UNCLASSIFIED)
11. TITLE (Include Security Classification)
Understanding Troposcatter Propagation
12. PERSONAL AUTHOR(S)

Joseph J. Reynolds
13a. TYPE OF REPORT 13b. TIME COVERED 14. DATE OF REPORT (Year,Month,Day) 15. PAGE COUNT
THESIS/X X TO 1990 245
16. SUPPLEMENTARY NOTATION APPROVEiD FOR PUBLIC RELEASE IAW AFR 190-1
ERNEST A. HAYGOOD, 1st Lt, USAF
Executive Officer, Civilian Institution Programs
17. COSATI CODES 18. SUBJECT TERMS (Continue on reverse if necessary and identify by block number)
FIELD GROUP SUB-GROUP
19. ABSTRACT (Continue on reverse if necessary and identify by block number)
DTIC
ELECTE
D
ST E
S APR2 41990
20. DISTRIBUTION/AVAILABILITY OF ABSTRACT

go z3
21. ABSTRACT SECURITY CLASSIFICATION
o-7o
[3UNCLASSIFIEDUNLIMITED 0 SAME AS RPT. [3 DTIC USERS UTNCLASSIFIED
22a. NAME OF RESPONSIBLE INDIVIDUAL 22b. TELEPHONE (Include Area Code) 22c. OFFICE SYMBOL
ERNEST A. HAYGOOD, ist Lt, USAF (513) 255-2259 AFIT/CI
DD Form 1473, JUN 86 Previous editions are obsolete. SECURITY CLASSIFICATION OF THIS PAGE
UNDERSTANDING TROPOSCATTER PROPAGATION
by
JOSEPH HENDERSON REYNOLDS
B.S., United States Air Force Academy, 1982
A thesis submitted to the Faculty of the Graduate
School of the University of Colorado in partial
fulfillment of the requirements for the degree of
Master of Science
Program in Telecommunications
ACcesiotn F.
1990. 190NTIS . CF.. A,', . .
DfIC 1At-
U' anno. :,;..
Just fictu
By
Distr b.
A ,./
,
Oist
RTCA: Cant Joseph H. Reynolds 14 Mar 1990
SUBJ: Thesis Abstract
TO: AFIT/CIRK (Capt Battlesl
1. IAI AFITR 53-1 (1 Dec 1989), Para 7-7 the following is provided.
Abstract of thesis:
--- Although tioposcattef communications systems, suffer several-shortcomings, this

transmission scheme has found consistent use in several applications. This
thesis discusses troposcatter propagation with emphasis on theory.
characteristics, and prediction tools.
Theoretical understanding of the trfoposcattei, propagation mechanism is rooted

in atmospheri: phenomena; specifically -- refractivity and turbulence. Two
modes of tran.;mission exist: incoherent scatter, if refractivity
irregularitlei exist as turbulent blobs, and quasi-coherent soatter, if
irregularities arrange themselves in layers. Frequency and meteorological
parameters define the dominant mechanism.
One can expect received signai levels (RSL) to exhibit distance and frequency
dependence; short and long-term fading; aperture-to-medium coupling loss; and
diurnal, seasonal, climatic, and meteorological variations. Diversity
techniques are indispensable in thwarting short-term fading. Atmospheric
multipath is known to limit analog system bandwidths yet digital systems are
prone to the related delay spread phenomenon which causes inter-symbol
interference (1SI). Adaptive processing is used to overcome this problem and
can further improve digital performance through implicit diversity.
Most troposcatter prediction methods are rooted in empirical expressions.. The

laborious NBS Tech Note 101 method incorporates climatic variability and
confidence measures. The CCIR methods offer simplicity as do the Yeh, Rider,
Collins, and Bullington techniques. Unfortunately, all the methods suffer
shortcomings with reliance on surface refractivity and incorrect coupling loss
calculations topping the list. The computer based Par method provides
notable theoretical accuracy, yet cannot encompass a range of spectrum slope
values. Accuracy can improve if tools such as the Parl or Radiometeorological
Method-effectively model gradient activity within the common volume. -
3. Key information:
Author: Joseph H. Reynolds

Thesis Title: Understanding Troposcatter Propagation
Rank/Setvice Branch: Capt, USAF
bFOA. 1990
Number of Pages: 245
Degree Awarded: MS Telecommunications
University: University of Colorado
90 04 23 070
4. See attachment I for an abbreviated bibliography; listed in attachment 1
are the sources used most often.
JSEPH H. REYNOLDS, Capt, USAF

Student, AFIT/CIRK
ATTACHMENT 1: Abbreviated Bibliography
SPECIAL NOTE: The entire issue of The Proceedings of the IRE, vol. 43, no. 3
(March 1955) should be referenced.
Anthes, R.A., Cahir, J.J., Fraser A.B., and Panofsky, H.A. The Atmosphere.
3rd ed. Columbus, Ohio: Charles E. Mprrill, 1981.
Boithias, L. Radio Wave Propagation. St. Louis: McGraw-Hill, 1987.
Booker, H.G., and Gordon, W.E. "A Theory of Radio Scattering in the
Troposphere." Proceedings of the IRE, vol. 38, no. 4 (April 1950): 401-
412.
Brauburger, G.H. Jr. (Capt. USAFR). "Meteorological Factors Affecting AN/TRC-

97A Tropospheric Scatter Communications." Technical Report no.
TACCA/XQT-TR-79-1, Langley AFB, Va, 1 August 1979.
CCIR. Volume V. Propagation in Non-Ionized Media. Dubrovnik, 1986.
Crane, R.K. "A Review of Transhorizon Propagation Phenomena." Radio Science,

vol. 16, no. 5 (September-October 1981b): 649-669.
Crawford, A.B., Hogg. D.C., and Kummer W.H. "Studies in Tropospheric

Propagation Beyond the Horizon." The Bell System Technical Journal, vol.
38, no. 5 (September 1959): 1067-1178.
du Castel, F. Tropospheric Radiowave Propagation Beyond the Horizon. New

York: Pergamon, 1966.
Freeman, R.L. Radio System Design for Telecommunications (1-100 GHz). New
York: John Wiley & Sons, 1987.
Frils, H.T., Crawford, A.B., and Hogg, D.C. "A Reflection Theory for
Propagation Beyond the Horizon." The Bell System Technical Journal, vol.
XXXVI, no. 3 (May 1957): 627-644.
Hall, M.P.M. Effects of the Troposphere on Radio Communication. New York:

Peter Peregrinus, 1979.
Military Handbook, Facility Design for Tropospheric Scatter (Transhorizon

Microwave System Design). MIL-HDBK-417, Department of Defense,
Washington, DC, 25 November, 1977. Available through Navy Publications
and Forms Center, ATTN: NPODS, Office of Navy Publications, 5801 Tabor
Avenue, Phildelphia, Pa., 19120-5094.
Monsen, P. "Fading Channel Communications." IEEE Communications Magazine,

vol. 18, no. I (January 1980): 16-25.
Monsen, P., Parl, S., Malaga, A., Tolman, S., and Fetteroll, J. Digitai
Ttoposcattet Performance Model: UserS Manual. Lexingtzn, Massachusetts:
C!G7NATRON, Inc. publication nioi.A288-IS, November, 1983.
Neiburger, M., Edinger, J.G., Bonnet, W.D. Understanding our Atmospheric

Environment. San Francisco: W.H. Freeman, 1973.
Pantei, P.F. Comxunication Systems Design: Line-of-Sight and Tropo-Scatter

Systems. St. Louis: McGraw-Hill. 1972.
Pail, S. A. "New Formulas for Tropospheric Scatter Path Loss," Radio

Science, voi. 14, no. 1 (January-February 1979): 49-57.
Rice, P.L.. Longley. A.G., Norton, K.A., and Barsis, A.P. Transmission Loss
Predictions fir Tropospheric Communication Circuits. Vols. I and II.
Rev. ed. U.S. Department of Commerce, National Bureau of Standards
Tecinlcal Note 10i. Washington: Superintendent of Documents, U.S.
Government Printing Office. 1987.
Roda, G, Tropo)catter Radio Links. Norwood, Maine: Artech House, 1988.
Sarkar. S.K., Dutta. H.N., and Reddy, B.M. "Development of a Prediction

Technique for Tropo-Path Loss from Observed Transhofizon Propagation
Characteristics over India." Third International Conference on Antennas
and Propagation (ICAP 83), Part 2: Propagation, Institution of Electrical
Engineers., London (1983): 229-232.
Memorandum: Estimated Thesis Expenses
Capt Joseph H. Reynolds

AFIT/CIRK
Student, University of Colorado
Telecommunications Program
The following is a conservative estimate of the money I used to complete

my thesis "Understanding Troposcatter Propagation." 1 submit this in an
effort to highlight the inadequate funding received during my program.
I do not expect compensation; perhaps this can serve to justify future
increases in allowances. Any item marked with an asterisk (*) indicates
an estimate; in all cases, estimates are conservative.
Research Expenses
- Library Database Searches: $10.00*
- Travel (USAF Academy Library, US Dept of
Commerce Library): 20.00*
- Copies of Journais/Publications: 70.00*
- Binders for Materials: 28.00
- Telephone Calls 36.00
- Postage 1.00
Prepar ation
- Computer Supplies (Paper, ribbons, floppys): 20.00
- Copies (drafts to committee, figures reduction): 60.00*
- MISC (softbinding, gluestick, staples, clips): 7.00
Defense
- Copies (Slide package): 4.00
- Transparencies (Overhead slides): 18.00
- Refreshments: 4.00
Submittal
- Thesis Bond Paper: 44.00
- Copies
-- On Thesis Bond (3 ea): 28.00
-- On Plain Bond (9 ea): 80.00
-- Binding
--- Hardbound (v Ca): 34.00
--- Softbound (6 ea): 6.00
--- Ringed binder: 3.00
- Postage (4 copies distributed): 15.00*
- Graduate School Fees: 13.00
TOTAL COSTS 498.90
Again, I emphasize this as being quite conservative. Unfortunately I

kept minimum records; regardless, I thought you might be Interested.
estimate that about $135 of the money spent could have been saved if I
did not distribute copies to persons requesting them. However, since
several of these persons provided substantial guidance, I felt
obligated. Notably, a similar breakout of book and supply costs would
yield comparably high expenses.
IOSEPH H. REYNOLDS, Capt, USAF

AFIT/CIRK Student
This thesis for the Master of Science Degree by
Joseph Henderson Reynolds
has been approved for the
Program in
Telecommunications
by
'-Ernest I.Sft
Warren L. Flock
Samul W.Male
Reynolds, Joseph Henderson (M.S., Telecommunications)
Understanding Troposcatter Propagation
Thesis directed by Professor Ernest K. Smith
Early notions that troposcatter transmission
could replace line-of-sight (LOS) communications
links quickly diminished due to high costs, limited
bandwidth, and the onset of satellite communications.
Recently, however, digital processing schemes have
reduced bandwidth concerns; furthermore, troposcatter
systems have found consistent use in less developed
countries and the military. This thesis discusses
troposcatter propagation with emphasis on
communications systems.
Theoretical understanding of the troposcatter
propagation mechanism is rooted in atmospheric
phenomena; specifically -- the index of refraction
and turbulence. It is shown that turbulent eddies
which encompass small-zcale refractivity variations
inhabit the common volume, an area defined by the
intersection of antenna beamwidths. Further, these
turbulent inhomogenities can arrange themselves as
randomly disposed blobs or layers. This engenders
the simultaneous existence of incoherent scatter and
quasi-coherent scatter mechanisms, with the dominant
mode contingent on frequency and meteorological

ivi
conditions. Additionally, diffraction may co-exist

on shorter paths fostering multimode propagation.
One can expect received signal levels (RSL)
to exhibit distance and frequency dependence; marked
short and long-term fading; aperture-to-medium
coupling loss; and diurnal, seasonal, climatic, and

meteorological variations. Diversity techniques are
indispensable in thwarting short-term fading.
Atmospheric multipath is known to limit

analog system bandwidths yet digital systems are
prone to the related delay spread phenomenon which
causes intersymbol interference (ISI). Adaptive

processing is used to overcome this problem and these
processors also introduce implicit diversity, further
improving digital performance.
Most troposcatter prediction methods are

rooted in empirical expressions due to atmospheric
unpredictability. The laborious NBS Tech Note 101

(TN 101) method, the standard for many years,
incorporates climatic variability and confidence
measures. The CCIR offers several simple techniques
as do the methods of Yeh, Rider, Collins, and

Bullington. Unfortunately, all the methods suffer
shortcomings with reliance on surface refractivity
and incorrect coupling loss calculations
V
topping the list. The computer based Parl method
corrects these discrepancies, but cannot include
known, real-time variations in the spectrum slope.
Accuracy can improve if tools such as the Parl or
Radiometeorological Method can effectively model
gradient activity within the common volume.

CONTENTS
PREFACE ....... ...................... xv
CHAPTER
1. INTRODUCTION
1.1 A History of Troposcatter

Propagation ... ............ .
1.2 A General Review of Troposcatter
Systems ...... .............. 5

1.3 Path Geometry for Troposcatter
Systems ..... .............. . 13
1.4 Review of Thesis ... .......... . 15
2. ATMOSPHERIC BASICS
2.1 Introduction .... ............ . 17
2.2 General Description of the
Earth's Atmosphere ...........

... 18
2.2.1 The Composition of the

Atmosphere .... ............ . 21
2.2.2 The Structure of the Atmosphere 24
2.3 The Radio Refractive Index ........ 28
2.3.1 Refractivity ... ........... ... 30
2.3.2 Refractivity Gradient ....... .. 36
2.3.2.1 Effective Earth Radius ....... 36

2.3.2.2 Anomalous Propagation ..... . 38
vii
2.3.3 Small Scale Changes in
Refractivity ... ........... ... 42
2.3.4 Summary of Refraction ....... .. 43
2.4 Turbulence .... ............. .. 44

2.4.1 The Causes of Turbulence ..... ... 45
2.4.2 A Refractivity Turbulence Model . 46
2.4.3 A Conclusion on Turbulence. ... 47
2.5 Atmospheric Effects on
Troposcatter Propagation ........

.. 48
2.5.1 Refraction .... ............ . 51

2.5.2 Absorption .... ............ . 54
2.5.3 Scattering .... ............ . 59
2.5.4 Depolarization ... .......... . 61
3. TROPOSCATTER THEORY
3.1 Introduction .... ............ . 64

3.2 Turbulence Theories .. ........ .. 67
3.3 Reflection Theories .. ........ .. 72
3.4 Mode Theories ... ........... . 78
3.5 Other Theories ... ............ . 81

3.6 Differentiating the Theories. ... 83
3.7 Modern Theory ... ........... . 84
4. CHARACTERISTICS OF TROPOSCATTER
PROPAGATION
4.1 Introduction .... ............ . 92
4.2 Signal Level ... ............ ... 92
4.2.1 Distance Dependence ........ .. 93

viii
4.2.2 Frequency Dependence .........

... 96
4.2.3 Fading ...... .............. .100
4.2.3.1 Short-Term Fading .. ....... .. 102
4.2.3.2 Long-Term Fading ........... .. 104
4.2.3.3 Frequency Selective Fading. . 108
4.2.3.4 Diversity .... ........... .109
4.2.4 Aperture-to-Medium Coupling
Loss ...... ............... .. 116

4.3 Multipath Delay Spread ......... .. 120
4.4 Bandwidth .... ............. .129
4.5 Diurnal, Seasonal, Climatic, and
Meteorological Variables ........ .. 134
4.5.1 Diurnal Variations ........... .. 136

4.5.2 Seasonal Variations .. ....... .. 139
4.5.3 Meteorological Variations . . .. 142
4.5.3.1 Subrefraction ... ......... .144

4.5.3.2 Super-Refraction ........... .. 145
5. TROPOSCATTER PREDICTION TECHNIQUES
5.1 Introduction .... ............ .149
5.2 Prediction Preparation ......... .. 150
5.3 Prediction Techniques .. ....... .. 151
5.3.1 TN 101 ..... ............. .. 154

5.3.2 CCIR Method I.... .......... .160
5.3.3 CCIR Method II ............ 161
5.3.4 The CCIR Chinese Method ........163
ix
5.3.5 The Radiometeorological Method . 164
5.3.6 The Parl Method ... ......... .. 166
5.3.7 The Yeh Method ... ......... 170
5.3.8 The Rider Method .. ........ .. 170
5.3.9 The Collins Method .. ....... .171
5.3.10 The New Bullington Model . . .. 171
5.3.11 Digital Considerations ...... .. 174
5.4 Comparisons of Methods ......... .176
5.5 Conclusions .... ............ 183
BIBLIOGRAPHY ........ ................... 185
APPENDIX
A. MULTIMODE PROPAGATION .... ........... .199
B. THE SYSTEM EQUATION .... ............ 201
C. THE SPECTRUM OF IRREGULARITIES ....... .. 209
D. DEPOLARIZATION ..... .............. 213
E. THE REFLECTION COEFFICIENT .. ........ 216
F. CLIMATIC REGIONS ..... ............ 218

x
TABLES
Table
1. General Characteristics of a Troposcatter
Communications System ... ...........
2. Atmospheric Composition at Sea Level. . . . 23
3. Variation of N with Temperature and
Relative Humidity (1000 mb) ........ . 33
4. Seasonal Variation Time Blocks ......... .. 140
FIGURES
Figure
1. Troposcatter vs LOS ...... ........... 6
2. Terrestrial transmission system domains 10
3. Perspective of various transmission

systems ...... ................. .. 11
4. Transmission system bandwidth comparison. . 11
5. Troposcatter path geometry .........

.... 13
6. Pressure variations with altitude ...... .. 20
7. Atmospheric layers ... ........... ... 25
8. ...
Atmospheric temperature layers ......... 26
9. Tropopause variations with latitude . ... 27
10. Refractivity profiles for model

atmospheres .... ............. ... 34
11. World-wide values for N..

o ......... ... 35
12. Ray bending at various K-factors ........

.. 37
13. Refractive conditions leading to anomalous
propagation events .... ............ . 39
14. Wave paths under various refractivity
values for different earth models . ... 39
15. Spectrum of refractivity irregularities . 46

16. Troposcatter propagation causal framework 49
17. Effective earth radius factor, K, versus

refractivity gradient ... .......... . 52
xii
18. Effective earth radius versus surface

refractivity ..... ............... .. 53
19. Gaseous absorption .... ............. ... 56
20. Combined oxygen and water vapor
atmospheric absorption ... .......... . 57
21. Atmospheric absorption due to rain ........58
22. Atmospheric absorption due to fog, mist
and clouds ...... ............... .. 58
23. Absorption and scattering attenuation due
to rain ........ .............. .. 58
24. Single scatterer radiation pattern ....... 68
25. Scattering by an individual blob ........
.. 69
26. Scattering by a number of blobs ....... .. 69
27. Radiation pattern of a number of
scatterers ...... ................ .. 69
28. du Castel's thin layer or feuillet ..... ... 74
29. Specular and diffuse reflection within a
feuillet ........ .................. .76
30. Radiation pattern for the reflection
mechanism ...... ................ 76
31. Mode theory propagation ... .......... . 79
32. Various mode theory profiles matched with
field data ...... ................ .. 80
33. Fading rates for UHF and SHF links over
the same path ..... .............. 87
34. Signal level variations for: (1) Coherent
scatter, (2) Quasi-Coherent scatter, and
(3) Incoherent scatter ... .......... . 90
35. Troposcatter loss with distance ....... .. 93
36. Scatter loss as a function of scattering

angle ....... .................. . 95
xiii
37. The effect of frequency over similar

paths ....... .................. . 97
38. Short-term effects of frequency ....... .. 99
39. Raleigh distribution showing short-term
troposcatter fading amplitudes ........ .. 103
40. RSL at different times for a 350 km, 300
MHz path ....... ................. .. 106
41. Log-normal density function for U = 8 dB. . 107
42. Standard deviation, a, varying with
angular distance, 0 .... ........... .107
43. Log-normal cumulative distribution
function for various values of a ........108
44. The diversity principle ..... .......... 110
45. Diversity distribution curves .. ....... .. 115
46. Coupling loss ....... ............ ... 117
47. Graphic method for coupling loss
calculation ............ ......... 118
48. Relative RSL comparison for antennas of
different diameters for troposcatter
links ........ ............... ..120
49. Delay scattering regions .... .......... .122
50. Development of the delay power spectrum . . 123
51. Delay spread profiles .... ........... .128
52. Useable bandwidth as a function of
distance and antenna gain ... ........ .. 130
53. Distribution of bandwidths ...........
.. 132
54. Diurnal RSL and surface refractivity, Ns,
variation ...... ................ .. 137
55. Diurnal fading rates ... .......... . . 138
56. Seasonal RSL, temperature, and surface

refractivity variations .... ........ .140
xiv
57. Climate-variation factor as a function of

effective distance and climate type . 156
58. Long term power fading for continental
temperate climate .... .......... .157
59. The standard normal deviate ... ........ .. 159
60. Mean attenuation, overland path, temperate

climate, 1 GHz ...... .............. .. 162
61. Collins method: Basic propagation loss at
1 GHz ....... .................. .172
62. Collins method: Frequency correction
curve ....... .................. .. 172
63. Collins method: Horizon angle loss ......... 173

64. Collins method: Coupling loss nomogram. . . 173
65. Determination of BER for coherent BPSK and

QPSK ........ ................... .176
66. Comparison of TN 101 (NBS) and the Parl
(Turbulent Scatter) methods .. ....... .. 182
67. Comparison of the CCIR (1962 Data),
TN 101 (Longley-Rice), and Bullington
methods ....... ................. .. 182
PREFACE
When, as a Second Lieutenant in the U.S. Air
Force, I arrived at my first duty assignment in a

tactical communications unit, I entered during a
difficult time. Not more than a week prior, the unit

had failed an inspection; the cause -- failure of a
troposcatter link to become operational. This unit

had successfully installed this "tropo shot" on
numerous occasions, but, during the inspection, the
normally gentle Florida climate had turned cold. The
unit, in turn, cited weather-oriented propagation
effects as the culprit for substantially lower
received signal levels (RSL) on this 50 mile, 5 GHz
link. Additionally, it was noted that proximity to
the coast, supplemented by several miles of overwater

transmission, contributed to degraded performance.
The inspection team did not buy this "excuse."
It was interesting to observe subsequent
activity within the unit. As an unseasoned,

unknowledgeable "contributor" I attempted to learn as
much as I could about troposcatter since it provides
the brunt of military, in-theater, tactical

multichannel capability. Material was scarce; many
xvi
people I queried responded that "No one really knows

how it works, but it works, and MIL-HNDBK 417
[Military Handbook, Facility Design for Tropospheric
Scatter, 1977] is your tool in predicting

performance." In the years following my 1982
commission I gained some familiarity with

troposcatter, but the "No one really knows.
aspect remained unacceptable. Now, with the
advantage of being a full-time graduate student under
military funding, I am able to offer a condensed look
at troposcatter, hopefully providing an understanding

to anyone interested -- including Second
1
Lieutenants.
This thesis discusses troposcatter
propagation with emphasis on communications
applications. A brief historical analysis along with
a general description of troposcatter is provided in

Chapter 1. Additionally, the rudiments of path
geometry are given with emphasis on the common volume
and scattering angle. Further discussion requires an

understanding of atmospheric phenomena, especially
IConcluding the story, during a delightfully
mild week, my unit passed its follow-up inspection.
The unit can take some consolation in the fact that
weather can severely degrade propagation; however,
from the inspector's point-of-view the mission was
not accomplished, and a follow-up was necessary.
Interestingly, my following assignment was to the
same inspection team that had earlier failed my unit.
xvii
the refractive index and turbulence. As such, a
quick tutorial covering these essentials is given in

Chapter 2. With the basics established, discussion
in the remaining chapters focuses on propagation
theory, characteristics, and prediction.
By way of summary and to aid the reader in
isolating areas of interest, the rest of this preface
summarizes the essential areas of the remaining
chapters.
Troposcatter systems allow intermediate range
(40-500 mile), point-to-point, multichannel

(commercial grade, 120 channels), high capacity (20
Mbps) communications. Although used for many years,
an understanding of the mechanism causing propagation
has only recently surfaced. Of the many theories
presented in Chapter 3, the simultaneous occurrence
of incoherent scatter and quasi-coherent scatter
modes best explain the existence of the persistent,
weak, beyond the horizon, troposcatter received
fields. In addition, on shorter paths, diffraction

becomes important and will exist in conjunction with
the scatter mechanism in what is termed multimode
propagation. Focusing on the scattering mechanism,
the common volume is known to consist of turbulent
refractive index irregularities which may arrange

xviii
themselves as randomly disposed blobs (incoherent

scatter) or layers (quasi-coherent scatter) dependent
on operating frequency or meteorological conditions.
Chapter 4 offers a lengthy discussion on the
characteristics of troposcatter propagation. A quick
synopsis of this chapter follows (with section
numbers for easy reference):

a. RSL VARIATIONS: (Section 4.2) Crucial in
analog systems, the RSL predictably varies

under the following conditions:
1) Distance dependence (Section 4.2.1) is
difficult to model due to the existence
of several propagation mechanisms,

generally 0.1 dB/km attenuation is used.
2) Frequency dependence (Sections 4.2.2,
3.7) is known to obey an fl attenuation
law up to about 3 GHz with greater
attenuation beyond. Short-term
attenuation variations may exhibit f2 to

f-11 3 dependence.
3) Short-Term Fading (Section 4.2.3.1), the
"trademark" of troposcatter, is caused by

atmospheric multipath and may occur at
rates of a few per minute at VHF, 1 Hz at

UHF, and several Hz at SHF. Amplitudes
xix
closely follow a Raleigh distribution;
losses of 8 dB can occur for 10% of the
time.
4) Long-Term Fading (Section 4.2.3.2) due to
gradient variations and changing
mechanisms show log-normal amplitude
distributions. Losses of 10 dB can occur
for 10% of the time. Fading rates are

discussed under SEASONAL and DIURNAL
VARIATIONS.
5) Frequency Selective Fading (Sections
4.2.3.3, 4.3), not normally associated
with RSL variations, will impact digital
transmission schemes on analog
subchannels if the correlation bandwidth
is exceeded.
6) Diversity (Section 4.2.3.4) is an
economical and essential approach to
reducing the impact of short-term fading
and may incorporate a variety of
techniques. 7 dB gains, 50% of the time
can be realized under quad diversity.
7) Aperture-to-Medium Coupling Loss (Section

4.2.4) occurs for high gain antennas when
narrower beamwidths fail to encompass a

xx
proportionate number of scatterers.
Losses of 6 dB are known to occur.
b. MULTIPATH DELAY SPREAD: (Section 4.3)
Provided a sufficient RSL is present,
performance of digital troposcatter links is
critically dependent on delay spread which
may range in value from 50 to 370 ns. Due to
atmospheric multipath, transmitted pulses
will widen and this creates intersymbol
interference. Clever adaptive processing
techniques exploit delay spread to realize
implicit diversity.
c. BANDWIDTH: (Section 4.4) Bandwidth is limited
by atmospheric multipath which defines the
correlation bandwidth in which uniform fading
exists. A capacity of 4 MHz or 10 Mbps is
typical. Narrow beam antennas will improve
performance; however, coupling loss may
offset this advantage. For digital systems
using adaptive processing, bandwidth is less
important than power (Eb/NO) limitations.
d. DIURNAL, SEASONAL, CLIMATIC AND

METEOROLOGICAL VARIABLES: (Section 4.5)
Conditions may be such that case (a) the
propagation mechanism changes, case (b) the

xxi
gradient changes slightly, or case (c) the

gradient exhibits large-scale changes.
Diurnal and seasonal variations, with
pronounced climatic dependency, or temporary
meteorological conditions (weather) will
foster such conditions. The result is either
enhanced or degraded propagation.

1) Diurnal Variations (Section 4.5.1) can
show daily 10 dB RSL swings in temperate
zones. Fading rates are lowest in the
afternoon; both case (a) and (b)
situations are cited.

2) Seasonal Variations (Section 4.5.2)
provide RSLs as much as 15 dB lower in
the winter months for temperate climates;
however, longer paths exhibit less
seasonal influence. Case (b) situations,
with emphasis on water vapor variability,
are the cause.

3) Meteorological Variations (Section 4.5.3)
are responsible for anomalous propagation

events caused by case (c) conditions.
Weather fronts, temperature inversions,
subsidence, occlusions, and, sometimes,
thunderstorms may signal the occurrence

xxii
of subrefraction, super-refraction, or
ducting. Real-time prediction of RSL
variation is difficult especially in
areas where gradient data is

unobtainable. Degradations may be short-
lived, as with thunderstorms, or may
persist for long (hours, days) periods if

frontal movement is slow.
The final chapter, Chapter 5, briefly
describes the tools available for troposcatter path

predictions. A simple comparison is also offered. A
summary follows:
a. PREDICTING RSLs: (Sections 5.2, 5.3, 5.4)
Assuming proper preparation, the engineer has

a variety of prediction tools at his
disposal; however, in nearly all cases, these
techniques incorporate empirical,

statistically-oriented data which injects
some inaccuracy. Several methods take this
into account through time availability and
service probability confidence measures. The

most commonly used methods, with
descriptions, follow:
1) TN 101 (NBS Tech Note 101, Section 5.3.1)
is a lengthy, well-known document that
xxiii
focuses on long term median path loss

calculations for a variety of climates.
Features include multimode calculations
and confidence measures. The Longley-

Rice method is a computerized version of
this technique.
2) CCIR Methods (Sections 5.3.2, 5.3.3,
5.3.4) include Method I and the Chinese
method (shortened versions of TN 101) and
Method II, a graphical technique
incorporating some climatic variability.
All CCIR models are simple to use.

3) The Radiometeorological Method (Section
5.3.5) is an extremely accurate tool

championed by Boithias and Battesti and
well summarized in CCIR Reports 238-4 and
718-2. Despite great simplicity, it
requires knowledge of the refractivity

gradient within the common volume, a
difficult to obtain parameter.
4) The Parl Method (Section 5.3.6) borrows

from TN 101 yet offers more realistic
predictions through computer flexibility.
Whereas TN 101 is restricted to lower
frequencies (less than 1 GHz), this

xxiv
technique permits manipulation of the

spectrum slope and climatic variables
allowing application to virtually any

configuration. Additionally, more
accurate aperture-to-medium coupling loss
calculations and analysis of digital

modem performance is offered.
Unfortunately, modem analysis focuses on
military systems only. The MITRE method,
used for a specific military system, uses
much of Parl's methodology.
5) Other Methods (Sections 5.3.7, 5.3.8,
5.3.9, 5.3.10) include those of Yeh,
Rider, Collins, and Bullington.
Bullington's model is a new computerized
version of Bullington's initial (1963)
work in propagation theory. The
remaining techniques are very simple with

the Collins method provi ing surprising
accuracy despite its totally graphical
approach.
6) Digital Predictions (Section 5.3.11) will
incorporate any discussed prediction tool
for estimated path loss. In addition,

calculation of delay spread and BER
xxv
performance is needed for complete

digital analysis
b) SELECTING A PREDICTION TOOL: (Sections 5.4,
5.5) Every technique (excluding the
Radiometeorological Method if accurate

gradient data is available) admits to
inaccuracy. Complete comparative analysis is
unavailable; however, some techniques may

appear more attractive due to simplicity.
All CCIR methods, the Yeh, Rider, and Collins
techniques provide this simplicity, yet
computerized versions of TN 101, the Parl,

and Bullington's models make the simplicity
aspect less critical. With respect to

conformity to theory, the Parl method stands
out; TN 101 incorrectly emphasizes surface
refractivity and a singular spectrum slope,
CCIR Method I, the CCIR Chinese method, and
Yeh and Rider's techniques also center on
surface refractivity. The Collins method has

limited application, CCIR Method II works
best in temperate climates, and Bullington's

treatment is untried and based, in part, on
unaccepted mode theory. Finally, all methods
calculating aperture-to-medium coupling loss

xxvi
incorrectly assess this parameter with the
exception of the Parl technique.

CHAPTER 1
INTRODUCTION
1.1 A History of Troposcatter Propagation
The discoveries of three different mechanisms
for beyond the horizon propagation have prompted

engineers to exploit associated capabilities and
theorists to attempt explanations. Earth diffraction
theory (ground wave), formulated in 1910, explained
the suitability of MF and LF for intercontinental
telegraphy systems. In 1927, cross-continent
telephony at HF was accounted for by ionospheric
reflection theory. Finally, since 1950, several
hypotheses for the tropospheric forward scatter or
troposcatter mechanism have endeavored to explain why
transhorizon radio systems in the VHF, UHF, and SHF
domains are possible (Sommerfeld, 1909; Boithias and

Battesti, 1983, p. 657]. This thesis will explore
the last mechanism, troposcatter.

It's probably no surprise that the first
documented observance of unexpectedly high received
signal levels beyond the horizon came from Guglielmo
Marconi. In a series of experiments conducted

2
between 1928 and 1936, this radio pioneer "showed
that microwaves could be reliably propagated to

distances exceeding the optical range by two to three
times" [Isted, 1958, p. 81]. His last paper (1933),
"On the Propagation of Microwaves over Considerable
Distances," summarized the results of one such
experiment aboard the yacht Elettra. Marconi notes,
"[t]he speculations that may arise from such results
concern the entire theory of radio transmission over
distances greater than the optical one" [Carroll
1956, p. 1057]. Earlier (1932), in a paper to the
Royal Institution of Great Britain, he also observed:
In regard to the limited range of propagation of
the microwaves, the last word has not yet been
said. It has already been shown that they can
round a portion of the earth's curvature, to
distances greater than had been expected, and I
cannot help reminding you that at the very moment
when I first succeeded in proving that electric
waves could be sent across the Atlantic Ocean in
1901, distinguished mathematicians were of the
opinion that the distance of communications, by
means of electric waves, would be limited to a
distance of 165 miles [Wiesner, 1957, p. 46].
Unfortunately, several years passed before this

phenomenon was further researched.
During the 1940's problems with high power
UHF television and radar systems resurfaced Marconi's
original observations; however, these unexpected

signals were viewed more as a nuisance than a viable
propagation method. TV frequency allocation plans,
3
based on Van Der Pol and Bremmer's 1937 smooth sphere

predictions, proved inadequate due to interference
caused by beyond the horizon propagation. This

prompted the U.S. Government to "freeze" licensing of
new television stations in 1949. Additionally,

"[t]he advent of radar, which operated at higher
frequencies and higher power, revealed the existence
of reflections well beyond the horizon where the

theory predicted that transmission was impossible"
[Roda, 1988, p. 6]. This "nuisance factor" led
several e..perimenters to further research short wave,
transhorizon propagation.
Carroll [1952b, p. 7] delineates a few of the
early experiments which highlighted short wave,
beyond the horizon propagation research:
a. Katzin's Caribbean experiments in 1945,
performed to investigate ducting over
water for wavelengths of 9 and 3 cm, also

revealed unexpectedly high received
signal levels beyond the horizon.
b. In the Pacific (1946), the Naval Research
Laboratory (NRL) group conducted tests
similar to Katzin's and obtained similar

results. A consequence of this endeavor
was the first use of the term "scatter
4
region." Pekeris used this term to

describe a distance of 30 miles beyond
the horizon in which he concluded that
propagation was due to a scattering

mechanism.
c. During his 1949 experiment in Iowa, Gerks
"reported measurements from a high power
400-mc transmitter at distances of around
100 miles, 50 db and more in excess of
the 4/3-earth theoretical value."

Conclusions again attributed the
phenomenon to scatter.
d. Megaw, the noted English researcher, made
measurements of 10-cm received signals on
a 370 mile link over the North Sea. He

"interpreted the measurements as
scattering by atmospheric turbulence."

e. Finally, Bullington's 1950-51 tests at
500 and 3,700 mc in New Jersey showed
consistent, useable signal levels over

path lengths of several hundred miles.
Bullington went further to predict
available bandwidths of several

megacycles fostering the idea of
5
reliable, short wave, beyond the horizon
communications.
The benefits of this early research was far
reaching. Experiments with troposcatter required

high power transmitters, and, in conjunction with
radar research, development of high power
capabilities (klystrons) proceeded rapidly. Also,
lafge, high gain antenna systems were needed, and

engineers responded. A lot of credit for the initial
development of troposcatter goes to the U.S. Air
Force which, recognizing the benefits of troposcatter
systems, provided the funding for the first active
troposcatter system, Polevault, in 1952 (Gunther,
1966, p. 80]. These efforts also encouraged
theoreticians to suggest a variety of explanations
which dominated propagation rpsearzh in thz carly
1950's. Theory will be discussed later; however, a
general examination of troposcatter will justify the
early enthusiasm for this new propagation mechanism.
1.2 A General Review of Troposcatter Systems

A quick glance at Figure 1 will help explain
the initial excitement over troposcatter propagation.

Simply, a troposcatter path is much longer than
individual line of sight (LOS) radio paths. Roda
[1988, p. 17] shows that this capability has several

6
NcrvcCC, .a Ar:ec
zte: HH-ose, .E , p. 4.
Figure 1. Troposcatter vs LOS
advantages:
a. The ability to "by-pass" inhospitable
terrain to include mountains, jungles,
bodies of water, marshes, swamps, and
even politically sensitive regions.
b. Reduction of the numbers of repeater
sites.
c. Lower overall costs in maintenance,
buildings, access roads, spare parts,
power access, and test equipment due to
fewer sites.
d. Frequency conservation due to fewer
sites.
e. Physical and information security
enhancement. With less sites to defend,

physical security is easier, and with
7
narrow antenna beams, the likelihood of
signal interception is reduced.

Several applications of troposcatter systems
logically follow [ioda, 1988, p. 14; Freeman 1987, p.
139]:
a. Connecting islands with one another and

to the mainland.
b. Military systems, both tactical (mobile
systems in inhospitable terrain) and
fixed (for example, a link into Berlin

from West Germany).
c. Connection to off-shore oil platforms.
d. Commercial networking to sparsely
populated areas where LOS repeater costs
are unjustified.
Other, non-communications oriented
applications of troposcatter propagation exist. For

example, Barrow, et al. [1969] have used Rake
techniques (a wideband signaling method used to
reduce fading and resolve multipath effects) with

troposcatter transmission to probe the scattering
medium. Related applications are reported by Cox
[1969, pp. 905,916] to investigate atmospheric
phenomena such as wind drift and random refractivity

fluctuations. Finally, Crane (1981b, p. 649]
8
describes renewed interest in troposcatter for

interference predictions (the "nuisance factor"
remains); in light of the numerous transmission
systems encompassing our world, all systems will
encounter some scattering in the troposphere thus

interfering with other systems.
Interest in troposcatter communications
peaked in the early 1960's when satellite systems

began to offer improved performance (mostly bandwidth
and range) at less cost. Furthermore, individual
troposcatter sites are more expensive than most
alternatives while interference problems cannot (to a
large degree) be constrained. The cost obstacle can

best be explained by examining the general features
of a troposcatter communications system as presented
in Table 1.
Table 1. General Characteristics of a Troposcatter

Communications System
Very High Path Loss (180 - 260 dE)
Very Sensitive Low Noise Receivers
Severe Fading Effects
VF Channels < 120 (Commercial Grade)
Bit Rate < 10 Mbps
Path Length 50 - 500 Miles
Transmit Power 100 - 50,000 Watts
Diversity At Least Dual
Antenna Size/Gain 6 - 120 ft / 20 - 50 dB
Frequency Range 250 - 6,000 MHz
Adapted from Freeman 1987, pp. 138; Roda, 1988 p. 8

9
This table shows requirements for high power

transmitters, large antennas (doubled if space
diversity is used), and sensitive receivers all of

which force extremely high costs. These costs, along
with the emergence of satellite systems and
insurmountable interference problems, thus engendered

the demise of communications via troposcatter.
Several promising developments have, however,
sparked some renewed interest in troposcatter
communications. Adaptive signal processors on
digital troposcatter systems have limited fading

effects (multipath delay spread) to acceptable
commercial quality. These modems, the "cornerstone"
of digital troposcatter, have improved troposcatter
capabilities to competitive levels; one such modem
was tested on a military system with 15 MHz bandwidth
at rates of 12.6 Mbps in 1980 [Monsen 1980, pp.

16,17]. Additionally, Crane [1988] demonstrated the
feasibility of using higher frequencies (15 GHz) on
troposcatter systems. While one might expect rain

effects to limit capability, it was found that rain
attenuation was insignificant while rain scatter
actually improved performance. At these higher
frequencies, required antenna sizes could be
significantly reduced.
10
Aspects of troposcatter are summarized in
Figures 2, 3, and 4. Figure 2 depicts the domains of
I II
LINE co I DIFFRLcION IR CSCArCR

SIGHT 2CN 2 ONE 20WE OIS4MICE
SoU.ce: Roda, G. Tropseaeter Padij 1.ks.

NO--wed, Maine: Artec1 House, 1558, p. 3.
Figure 2. Terrestrial transmission system domains
the three terrestrial multichannel techniques: LOS,

diffraction, and troposcatter. Interestingly, both
diffraction and troposcatter can occur simultaneously
in multimode propagation. This effect is explained
further in Appendix A; however, while Figure 2 shows

troposcatter loss significantly exceeding that of
free space, useable signals are receivable. Figure 3
places troposcatter in perspective with other
transmission modes (single hop), while Figure 4

provides bandwidth comparisons.
11
a - HF d - Satellite
b - Ionospheric Scatter e - LOS
c - Meteor-burst f - Troposcatter
d
(mi)
1500
1000 a c d
.500
__,
b - -
0
IM lOM looM 10 100 1000 f
(Hz)
Figure 3. Perspective of various transmission
systems
aW
(HZ)
lOOM
tOM
100K
HF nospherlo Meteor Tropoaoatter LOS SatellIte

Sotter Burst
Figure 4. Transmission system bandwidth comparison

12
Satellite transmission offers clear

advantages with respect to range, bandwidth, and area
coverage (hence, not limited to point to point
modes); however, costs can be prohibitive. Meteor-
burst, riding on the crest of sophisticated digital

technology, has experienced renewed interest, but
erratic transmission bursts makes it unsuitable for
voice transmission [Weitzen, 1988, p. 1813]. The
nearly saturated spectrum, limited bandwidth, and
aberrant performance of HF systems disqualify them as
a steady, multichannel performer. Additionally,
ionospheric scatter techniques suffer similar
bandwidth and degradation characteristics. Finally,
while offering high bandwidth, LOS systems
incorporate numerous repeater sites fostering

possibly higher overall costs.
It can be seen, therefore, that troposcatter
communications provides unique capabilities in

certain situations. Military tactical networks still
rely heavily on troposcatter due to its ability to
circumvent inhospitable and/or hostile areas and as a

means to conserve limited radio resources. 1 The oil
lit should be noted that military tactical

troposcatter provides connectivity within the theater
of operation; however, connectivity to "the rest of
the world" is achieved mostly through tactical
multichannel SHF satellite terminals.
13
industry and less developed countries, which don't

need extremely high capacity systems, also find
troposcatter economically attractive. Undoubtedly,
troposcatter will continue to occupy a necessary

niche in transmission system selection, yet it is not
the miracle transmission scheme as initially thought.
1.3 Path Geometry for Troposcatter Systems

An introduction to the path geometry for
troposcatter systems will aid in later discussion.

Figure 5 shows this geometry which can be broken down
into a few key elements. Most critical is 0, angular
SCATTER VOLULII4L
souarce: Rice, P.L., iongley, A.G., Nortcrn.

.Aan.d larsi, A.P. ra'IsmissIJf Lcss PrediioIns
for rcopeAri CC.iL'"U.Icaticl Circuits. yols. I and
II Rev, ;d. tI.S.ztpart? ent of Comme.rce. ?Ewtional
Du ;eAU C Standards Technical Nt~e 101. Washington:
Superintendent cf Dc3011-til U.S. Governme.nt Printing
Office, 196.7, P. 6-8.
Figure 5. Troposcatter path geometry

14
distance, which is simply "the angle between horizon

says in the geat (ircie plane, and is the minimum
diffraction angle or scattering angle unless antenna
beams are elevated" [Rice, P.L., et al., 1967, Vol 1,
p. 6-5]. Relationships for 0 follow:
0 =00 =d/a + et + 0 er, (I-i)
where 0 et and Oer are the transmitter's and

receiver's horizon elevation angles respectively.
Also,
00 = Ds/a +eot + 0or, (1-2)
with 0 or and Oot being "[t]he angular elevation of a

horizon ray at the receiver or transmitter horizon"
[Rice, P.L., et al., 1967, Vol 1, p. 12-10]. Note
the influence of obstacles. A basic rule is that the
smaller the scattering angle, the better the received
signal; therefore, when performing path profiles,
engineers choose sites (if that much flexibility is
given) which avoid path obstacleS. Otherwise,
antenna elevations and, subsequently, scattering
angle values will increase.

The other crucial element, the scatter or
common volume, is related to the scattering angle.
The smaller the scattering angle, the larger the
15
common volume. This region describes the location of

the -c ttering mechanism which ai±ows troposcatter
propagation; logically, then, a large common volume
contains more scatterers. Although larger common

volumes typify enhanced signal levels, degraded
bandwidth performance somewhat limits the benefits of
large volumes. This characteristic, along with
several others are discussed later.
1.4 Review of Thesis

The remainder of this treatment will focus on
troposcatter propagation with emphasis on

communications systems. (Attention will be on pure
troposcatter; however, some discussion of multimode
propagation is appropriate.) Specifically, the
theory behind the propagation mechanism, the

characteristics of troposcattered radiowaves, and the
prediction tools used to assess troposcatter systems
will be analyzed. The purpose is to provide a

consolidated, simple description of radio
communications via the troposcatter mode emphasizing
"how it works" and "what one can expect" when using
these systems.
The characteristics of the atmosphere are
responsible for troposcatter propagation; therefore,
Chapter 2 describes the earth's atmosphere and

16
analyzes its effects on radiowave propagation. Focus

will be on the troposphere with emphasis on the
refractive index and turbulence. Additionally,
natural atmospheric phenomena ("causes") will be
married with propagation phenomena ("effects") to
describe the cause and effect ("causal") nature of
the atmosphere on radiowave propagation.
In Chapter 3 an analysis of theory is
offered. Provided are descriptions of the various

theories developed over the past 40 years to explain
the mechanism(s) behind consistent received signal
levels far beyond the horizon at short waves.
The various characteristics of troposcatter
transmissions are given in Chapter 4. Focus will be
on performance parameters of interest to
communications engineers that are directly influenced
by propagation; equipment will be discussed when
enhancements improve propagation performance.

Chapter 5, the final chapter, will describe
the prediction tools available to determine the
feasibility of planned links to support
communications requirements. A comparison and

criticism of the more common techniques will be
provided.
CHAPTER 2
ATMOSPHERIC BASICS
2.1 Introduction
As radio technology continues to unveil
remarkably sophisticated devices, no innovation has
yet provided complete immunity to the effects of our

atmosphere. Troposcatter systems are particularly
susceptible to this uncontrollable medium, housed
entirely in the most violent of atmospheric layers,
the troposphere. While other external sources can
influence radiowave propagation, "[p]oint-to-point
transhorizon links are usually designed for high-gain
antennas with relatively narrow beams, which are
removed from terrain clutter by at least several
wavelengths" [Military Handbook, 1977, p. 4-109].
The atmosphere then is the critical external
parameter effecting troposcatter propagation;
Appendix B further substantiates this assertion.

The atmosphere can be considered as simply a
transmitting medium for radiowaves, much like a
coaxial cable is the transmitting medium for cable
TV.
18
An electromagnetic wave launched into a

transmitting medium may change in intensity,
phase, direction, or polarization in its passage.
These changes are caused by and are
characteristic of the medium through which the
waves travel; media such as liquid, gases,
solids, and aerosols produce a variety of effects
[Derr, 1972, 9-1].
The above summary further highlights the need to
understand our atmosphere when considering
troposcatter systems. This chapter provides this

basic understanding and further describes the
atmosphere's general effects on radiowave
propagation.
2.2 General Description of the Earth's Atmosphere

The parameters of the earth's atmosphere are
determined through a variety of methods. The

simplest are surface observations made by
meteorologists. "By international agreement these
observations are made simultaneously [synoptically],
using procedures that meet uniform standards of
accuracy" [Neiburger, Edinger, and Bonner, 1973, p.
10]. A more sophisticated method is the radiosonde
which is a balloon launched instrument package
containing pressure, temperature, and humidity

sensors. A radio transmitter is also included to
transmit information and provide tracking information
for wind motion measurements. Radiosonde

measurements are also made synoptically; however, the
19
cost of these packages inhibit worldwide coverage

[Anthes, et al., 1981, p. 38].
Ground and space-based remote sensors top the
list of sophisticated atmospheric research tools, and
can be operated in either passive or active modes.

"[P]assive systems monitor radiation emitted by the
atmosphere in the infrared, microwave, or visible
portions of the spectrum" [Anthes, et al., 1981, p.
42]. These devices can measure, among other things,
temperature layers and gaseous distributions. Active
remote sensors emit electromagnetic energy for which
the
direction of propagation is altered by the
atmosphere. It may be scattered back toward the
source (back scatter) or scattered ahead (forward
scatter). In either case, the characteristics of
the scattered signal provide information about
the part of the atmosphere through which the
signal is passing. This information may include
the intensity of turbulence, the lapse rate, and
the mean wind velocity [Anthes, et al., 1981, p.
40].
Lidar, radar, and sodar are the more common active
remote sensors with lidar using laser beams, radar
operating in the microwave range, and sodar using

acoustic sounding techniques [Anthes, et al., 1981,
pp. 40,41].
Terminology commonly used to describe the
atmosphere are temperature, pressure, and humidity.

While temperature measurements are easily understood,
20
pressure measures the atmospheric force exerted per

unit area on the surface of the earth. Units can be
in inches of mercury (Hg) or millibars (mb) with 1
inch Hg equal to 33.86 mb (Anthes, et al., 1981, p.
29]. Figure 6 shows pressure variations with
altitude. If expressed in miles, these variations
36
32
2S
24
6
20
12
0 200 0 600 Soo 1000

Pfessule (rAb|
Scure, Ltgps, F.K., an.d?&:~muck.!.3. 7ThO
At o ph e: A :ntr due-ion :* a@,CZ2f2y- fd ed.
£Eqloe.d clift. tievJersey: Frentit:.-MA11. 1952. p.
Figure 6. Pressure variations with altitude
follow a general rule up to about 60 miles: 1000 mb

at sea level, 100 mb at 10 miles, 10 mb at 20 miles,
etc (Anthes, et al., 1981, p. 55].

Humidity measures the water vapor content in
air and comes in a variety of forms:
Absolute humidity is stated as the weight of

water vapor in a given volume of air [g/m 3 ].
21
Specific humidity is expressed as the weight of

water vapor per weight of a chosen mass of air
[g/kgl. [R]elative humidity is the ratio of the
air's water vapor content to its water vapor
capacity at a given temperature [%] [Lutgens and
Tarbuck, 1982, pp. 75,76; italics not mine].
Of the three humidity measures, relative humidity is
probably the most familiar, yet is easily
misunderstood.
Relative humidity reaches 100% when the air
is saturated. Saturation occurs at a particular
temperature when water vapor content is the maximum
that a parcel of air can support. If the relative
humidity is less than 100%, the dew point "is the
temperature to which a parcel of air would have to be
cooled in order to reach saturation" (Lutgens and
Tarbuck, 1982, p. 77]. Additionally, if rain occurs
in an unsaturated region, the air cools as energy is
used to evaporate the rain. The wet-bulb temperature
occurs when evaporation ceases and the air is
saturated (100% relative humidity) (Anthes, et al.,
1981, p. 29].
2.2.1 The Composition of the Atmosphere
The culmination of atmospheric measurements
provides an understanding of the structure and
composition of the atmosphere. In general terms, the
atmosphere is a thin, gaseous envelope that surrounds
the earth and is held in place by gravity. Its

22
density decreases with height; 90% of its mass is

contained in approximately the first 20 km of height
while 99.9% resides within the 50 km level [Iribarne
and Cho, 1980, p. 11.

Table 2 shows the composition of clean, dry
air at sea level. Interestingly, up to about 100 ki,

this composition is nearly the same anywhere in the
world [Neiburger, Edinger and Bonner, 1973, p. 24].
Additionally, the region below about 100 km is called
the homosphere since its chemical composition remains

virtually constant, while the heterosphere, above 100
km, exhibits varying composition [Iribarne and Cho,

1980, p.5].
Items marked with cross on Table 2 are
variable constituents introduced by biological
activities and industrial processes. . . . (which
have] deleterious effects even when present in
concentrations as low as one part per million or
less [Neiburger, Edinger and Bonner, 1973, p.
26].
Notably, there is some alarm concerning several of
these variables. Knalil and Rasmussen [1988] report

doubled CO concentrations since pre-industrial times
and warn of increased tropospheric 03 leading to
detrimental climate changes. Volz and Kley [1988]
and Penkett [1988] substantiate the increase of
tropospheric 03. Other potential disturbances such

as depletion of the critical UV blocking ozone layer
23
Table 2. Atmospheric Composition at Sea Level
Content variable
relative to its Molecular
Constituent gas and formula Content, percent by volume normal weight*
Nitrogen (N 2 ) 78.084 - 28.0134

Oxygen (0.) 20.9476 - 31.9988
Argon (Art) 0.934 - 39.948
Carbon dioxide (C0 2 ) 0.0314 44.00995
Neon (Ne) 0.001818 - 20.183
Helium (He) 0.000524 - 4.0026
Krypton (Kr) 0.000114 - 83.80
Xenon (Xc) 0.0000087 - 131.30
Hydrogen (II) 0.00005 ? 2.01594
Methane (C1 4 ) 0.0002 t 16.04303
Nitrous oxide (N 2 0) 0.00005 - 44.0128
Ozone (0.) Summer: t 47.9982
IDo 0.000007
Winter: t 47.9982
0 to 0.000002
Sulfur dioxide (SO2 ) 0 to 0.0001 t 64.0628
Nitrtogen dioxide (NO 2 ) 0 to 0.000002 t 46.0055
Ammonia (NH,) 0 to trace t 17.03061
Carbon monoxide (CO) 0 to trace t 28.01055
Iodine (12) 0 to 0.000001 t 253.8088
rt,.*onitit of' tile gp ijc imjked ,th an airAsk mzsy oslqekro oniicant variation from time to time or rrom pbce to
plute itlallVc IU ticl t wl indt., cJ tmi t1 . caw%.
t ilch ..S. Stlada.d At1t vt%'hcre. t 062
Source: Derr, V.E. Remote Sensing of the

Troposphere. U.S. Department of Commerce, National
Oceanic and Atmospheric Administration, Environmental
Research Laboratories, and the University of
Colorado. Washington: Superintendent of Documents,
U.S. Government Printing Office, August 15, 1972, p.
1-2.
24
as well as acid rain have been noted; furthermore,
"modification of the radiative properties of the

atmosphere" justify consideration from a wide
spectrum of professions [Global Tropospheric

Chemistry, 1984, p. 8].
Derr [1972, pp. 1-2,1-3] notes that not
listed in Table 2, yet meriting mention, are water

vapor, hydrometeor, and aerosol constituents. Water
vapor's absolute humidity rarely exceeds about 25
g/m 3 compared to normal dry air density of 1000 g/m 3 .

However, while small in contribution, water vapor is
the most important and most variable atmospheric
component. Its ability to change states within the
atmosphere (gas to liquid) produces hydrometeors
(hail, snow, and rain). Finally, "[aerosols are

particulate matter suspended or slowly falling in the
gaseous mixture of the atmosphere" [Derr, 1972, p. 1-

3]. Aerosols are differentiated from hydrometeors in
both content and size. Examples of aerosols include
haze, clouds, and smoke all whose droplet radius
sizes are less than 1 micron.
2.2.2 The Structure of the Atmosphere
The atmosphere's structure, Figure 7, is

layered to reflect its physical characteristics. The
homosphere and heterosphere, mentioned earlier,
25
Height Approximate Composition layers and

(kin) temperature Temperature layers electrical charge layers
rK) __________
-~ Magnetosphere
60.000 I (Van Allen
I Radiation Belts)
6000
CL
600 1500 Thermosphere 0
80 180 Mesopause
j E-layers 80-150 krm
Mesosphere T. 0lye CS m
wreJen onlynermittently)
Figre7.atosheri laErsooshr
26
divide chemical composition features. The ionosphere
is layered to show significant levels of electron

content. Finally, temperature layers, more
accurately depicted in Figure 8, relate significant
temperature transitions (transitions occur at layers
ending in -pause). One might expect steadily
I _______re
l;ii Ji i f I I!
I I I
I~ll
f .i
______ I
I 11 j IIt@JUI4
I~~ I I I,
at~00I
11r .M1C 1 I333-U30 -
ML.
Temperature M
9S.rcos Nv itu-g.M. Edimg*qO,J.13., so~r, W.D.

1,rswa"I'g owApAI £flEvP.OfAt. San F,-M'ctec~s 01.1. F,*.1%
Figure 8. Atmospheric temperature layers
decreasing temperatures with height; however, this is

obviously not the case. The stratosphere experiences
temperature increases due to solar absorption by

ozone, while the thermosphere shows increases due to
27
oxygen and nitrogen solar absorption (Smith, E.K.,

1968, p.14; Lutgens and Tarbuck, 1982, p. 17]. While
other atmospheric division schemes exist (Smith,

E.K., 1968 p. 14 mentions division by processes), the
temperature layer scheme is the most common.

Attention is focused on the troposphere
which, as shown in Figure 9, varies in height (about
7 to 17 km) depending on latitude. Knowledge of the
30
27
24
5i21:1
3 j15 T~o9ical Pcpwau-i- I
1 0
\.
-0-1- -50-40-3 - - 2 2
Temperajve JC
Source$ Lutg*118. t.K-. & I.:.

'arbuell~C. rhe
At carhege: An 2,: o4dcCtIe O Met@CreczOO. 214 #.d
1*I.,2..ocd Cliffs. NO. Jersey$ P:entice-fll. 2912.p.
zn
Figure 9. Tropopause variations with latitude
troposphere is extensive since it contains about 80%

of all air mass and originates most weather phenomena
(Iribarne and Cho, 1980, p. 7]. Weather is described
as follows:
The forces that control the physical state of air

are sunlight, surface and air temperatures, water
28
vapor concentration, and cloud cover. Since all

four agents influence one another, the feedback
interactions are complicated. Cause becomes
effect, and vice versa. This never-ending
activity is known as weather [Lynch, 1980, p. 3,
italics not mine].
The term "troposphere," first used by
Treisserenc de Bort in 1908, "literally means the
region where air 'turns over,' a reference to the
appreciable vertical mixing of air in this lowermost
zone" [Lutgens and Tarbuck, 1982, p. 15]. The steady
temperature decrease with height or lapse rate of the
troposphere is significant. Although it can vary
considerably, the troposphere has a normal lapse rate
of 6.5 K/km which, according to Derr [1972, p. 1-9],
indicates large amounts of mixing. This mixing is
both convective and mechanical and ceases at
stratospheric levels. "Thus, the troposphere is a

well-defined thermodynamic environment that does not
interact much with the overlying layer of stable air,
or stratosphere" [Lynch, 1980, p. 3].
2.3 The Radio Refractive Index

Elementary optics describes the process of
refraction as the bending of a light beam as it
travels between different media. Similarly, since

the troposphere generally decreases in density,
temperature, water vapor content, and pressure with

height, radio waves traveling upward, through the
29
troposphere, will be refracted as the medium changes.
Simplistically, under normal tropospheric conditions,

radiowaves will bend downward as "the portion of the
wave in the thinner air starts traveling faster than
the lower portion which is still in dense air"

[Evaluation of FDM and FM Systems, 1983, p. 12-7].
Note that horizontal variations can also occur.

Propagation engineering focuses on the radio
refractive index, n, to quantize refraction. It is
defined as "the ratio of the phase velocity of an
electromagnetic wave in a vacuum [(the speed of

light, c)] to phase velocity [vp] of the wave in the
medium" [Flock, 1979, p. 141]. Mathematic-lly,
n = C/V. (2-1)
Livingston [1970, p. 7] shows vp as equal to i/11

with g, the magnetic permeability, remaining
essentially constant in the atmosphere at 4n/10 7 . C,

the dielectric constant, varies, however, forcing n
to also vary. No single parameter approaches the

radio refractive index in significance concerning
propagation; among others, effective earth radius,
anomalous propagation, and radiowave scattering

calculations rely heavily on the refractive index.
30
The radio refractive index can typically be
measured in two ways. Direct measure is obtained
through the refractometer, an airborne device

developed independently in the early 1950's by
Birnbaum (National Bureau of Standards) and Crain
(University of Texas). This device consists of a
resonating cavity which directly measures the air's

dielectric constant, and, through autocorrelation
analysis, provides refraction measurements [Gordon,
1955, p. 23; Crain, 1955, p. 1405]. Early work using
the refractometer is described by Crain and Gerhardt

[1952] and Birnbaum and Bussey [1955]. Additionally,
McGavin (1962] provides a good narration of some of

the earlier versions of the refractometer. The
second way to measure the radio refractive index is
through surface observations applied to equation (2-
3) as described in the next section.
2.3.1 Refractivity
The refractive index can range from 1.000240
to 1.000400 at sea level, but is commonly converted
to refractivity (N) [Panter, 1972, p. 342]:
N = (n - 1)106. (2-2)
N therefore ranges from 240 to 400 N units at sea
level. Refractivity gains complexity as one

31
considers the variety of random combinations

(temperature, pressure, and water vapor content)
characterizing the atmosphere. Supporting this,

consider
N = 77.6/T(p + 4810e/T) (2-3)
where
p = total pressure in millibars
e = partial pressure of water vapor in
millibars
T = absolute temperature [K] [Smith and
Weintraub, 1953, p. 1035].
A variation of expression (2-3) is
N = 77.6p/T + (3.73 x 10 5 )e/T 2 . (2-4)
While still other equations for N exist, CCIR [1986]

Report 563-3 chooses equation (2-4) as the
recommended expression for refractivity. Hall [1979,

p. 14] adds that this formula
is correct to within 0.5% for atmospheric

pressure between 200 and 1100 mb, air
temperatures between 240 and 310 K, water vapor
pressures less than 30 mb and for radio
frequencies less than 30 GHz.
Expression (2-4) can be thought of as
containing two terms: Ndry = (77.6p/T) and Nwet =

(3.73 x 105 e/T). Bean [1956, p. 32] notes that the
32
dry term varies with atmospheric density and the wet

term varies primarily due to water vapor pressure.
Further, as Table 3 shows,
[a]t very low temperatures Nwet becomes very small

even for saturated air, and so N is almost
independent of relative humidity. As the
temperature rises, there is a slow decrease in
Ndr. but a rapid increase in the saturated value
of Nwet max, At high temperatures Nwet max can
become somewhat larger than Ndry and so N varies
considerably with relative humidity. At high
temperatures and high relative humidity, N is
very sensitive to small changes in temperature
and relative humidity [Hall, 1979, p. 15].
Refractivity variations are therefore attributed

principally to humidity changes; consequently, the
radio refractive index varies more in hot, humid
areas than in colder, dry regions.
Of particular interest is the value of

refractivity in the first km of height where most
troposcatter links propagate. In this case, CCIR
Report 563-3 [1986, p. 108] recommends use of the
average exponential model given by
N(h) = NA x exp(-bah) (2-5)
where
N(h) is refractivity at height (h) in km,
NA = 315 N units,
bA = 0.136 km - I .
33
Table 3. Variation of N with Temperature and

Relative Humidity (1000 mb)
IC WA Alaximuni e.
val,,c uf
0 210 40 60 80 100 NM,
-30 319-3 320-0 320'7 321"4 322'1 322'7 3-4 0-5

-28 316'7 317-S 318"3 319"1 319-9 320"7 40 0'6
-26 314-2 3151 316-1 317,0 317"9 318-9 4-7 08
-24 311-6 3128 313"9 315'0 316'1 317-2 S-S 0"9
-22 309-2 310- 311- 3130 314-3 315'6 6S I1
-20 306-7 308-2 309'7 311-2 312'7 314-2 7-S 1"3
-18 304-3 306"1 307-8 309-6 311-3 3131 8"8 1-5
-16 301-9 3040 3060 388-0 310-1 312-1 10-2 1-8
-14 299"6 3020 304'3 306'7 3090 311-4 11'7 21
-12 297-3 3000 302'7 305-S J08"2 310"9 13"6 2S
-10 295-1 291-2 301-3 304-4 307-5 310-7 "136 2-9
-8 292-8 296-4 300.0 303-6 307-2 310-8 179 3-4
-6 290.6 294.8 298.9 3030 3071 3112 206 3-9
-4 2885S 293-2 297.9 302-6 307-3 3120 23-5 4-6
-2 286-3. 291-7 297-1 302-S 307.8 313-2 26'9 S-3
0 284-2 290-4 296-5 302"6 308.7 3149 30-6 61
2 282-2 2891 296'1 3031 3100 317-0 348 7,0
4 280-1 2880 295.9 3038 311-7 319"6 39-5 8-1
6 278-1 287-1 2960 305-0 313'9 322-9 44-7 9"3
8 276.2 286-3 2964 306-5 316-6 326'7 50.6 10-7
10 274-2 285-6 297-0 3084 319-9 3313 57-1 12-2
12 272-3 2851 298.0 310-9 323'7 336.6 64.3 14.0
14 270-4 2848 299-3 3138 328'2 3.12.7 72.3 25.9
16 268'5 284-7 3010 317-2 333-4 349-7 81.1 18-1
1 266-7 284-9 3030 321"2 339-4 3576 90.9 20-6
20 2648 285-2 305-5 325-9 346-2 3666 101.7 23-4,
22 2631 2858 308-S 3312 354-0 376.7 113-6 26-S
24 2o13 286-6 312.0 337-3 362-7 3880 2267 29.9
26 259-5 287-8 316-0 344-2 3724 400-6 141-1 33-8
28 257-8 289-2 320-6 3531-9 383-3 414-7 13569 38-1
30 256-1 290-9 325-8 360-6 395,4 430'3 174-2 42.8
32 254-4 293"0 331-7 370-3 408-9 447-S 193-1 48-1
34 252-8 295-5 338-3 3810 423-7 466-5 213-7 53-9
36 251-1 298-4 345-6 392-9 440-1 487-4 236.2 60-4
38 249-5 301-7 353-8 406'0 458-1 510-3 260-8 67-S
-10 247-9 3US-4 362-9 420.4 477-9 535-4 287-S 75-4
Source: Hall, M.P.M. Effects of the

Troposphere on Radio Communication. New York: ,Peter
Peregrinus, 1979, p. 16, 17.
34
Expression (2-5) is plotted along with an average
profile at mid-latitude on Figure 10 (A and B plots

respectively).
1" I
O1 I l I O 11
i ' qf~ltl .1411 11111tl
i111,
Source? CC~O. Volume V. Propaqation in Mon-
lonlrod Pdla. G e 1902, 97.
Figure 10. Refractivity profiles for model

atmospheres
Frequently the terms Ns and No are used to

provide refractivity values where N s is the value at
any surface location and N o is the value at sea level
[Panter, 1972, p. 342). In this instance, CCIR

Report 563-3 [1986, p. 109) provides
Ns = N o x exp(-bAhs) . (2-6)
Values for N o vary world-wide and are provided in
Figure 11 for February (top) and August (bottom).

The surface height above that of sea level is used
for h s . As N s is more easily determined (from an

economical standpoint) it is often used in analyzing
refractive effects on troposcatter propagation. For
35
so0W0s 40 120 00 so 60 .0 2060 0t20 -0 60 60 00 -20 40 C0(00
70
we
0 -
3a 30
.0
0 30
01
-0 TT 0
so -
00000o 0 00 00 so 6 0 z20 0 120 60 60 so .00 100 -0 01" 00
Souce CCRoueV0Poaaini
Nonlonize
Mdia Geeva 182,pp.88,99
Fiur 11sol-idoausfo 0 Tp
Ferury botmAgut
36
example, Fitzsimons [1968] reports correlations
between N s and signal levels on an 87 km, 4.7 GHz, 1
kW troposcatter link in Cyprus.
2.3.2 Refractivity gradient
As shown in Figure 10, refractivity models
establish a decrease in refractivity with height.
This decrease is quantified through the refractivity
gradient, AN/Ah, and is "approximated by the
difference between the refractivity values of the
surface and 1 km above the surface" [Military
Handbook, 1977, p. 4-84]. In the troposphere,
parameters normally decrease with height:
pressure p by about 1 mm (1.3 mb) per 11 m,

temperature by about 1 degree C per 200 m,
humidity by about 1 mb per 300 m [David and Voge,
1969, p. 122].
Under these conditions AN/Ah equals about -40 N
units/km. More specific values are obtainable
[Military Handbook, 1977, p. 4-87] :
AN/Ah = -7.32 exp(0.00557N s ) for the U.S., (2-7a)
AN/Ah = -9.30 exp(0.004565N s ) for Germany (2-7b)
AN/Ah = -3.95 exp(0.0072N s ) for the U.K. (2-7c)
2.3.2.1 Effective Earth Radius
The "decrease of N with height bends rays
towards the earth but not sufficiently to overcome
the curvature of the earth" [Hall and Barclay, 1989,

37
152]. The earth's curvature is 1/a o (aO being the
earth's radius, about 6370 km), but refractivity

gives cause to use a modified earth curvature through
the K-factor, K, to produce the effective earth
radius, a:
K = a/a o . (2-8)
Freeman (1987, p.6] further shows that:
K -{1 + (AN/Ah)/157}- I (2-9)
Figure 12 shows the different curvatures for various

K-factors.
K- -1f
I 0N
K. 0 3 d '
' " " '" -
*,un.,s/km
So.,rei freeman, R.L. Radio System Design for

Ta2el olauncations (1-200 GHz). New York: John Wiley
& Sol s, 1987, p. S.
Figure 12. Ray bending at various K-factors
Essentially, the effective radius allows
engineers to assume "an atmosphere of constant

refractive index" [Boithias, 1987, p. 91]. Using
AN/Ah = -40 N units for the first km of height (often

38
termed "standard refraction"), K will be about 1.33.

This leads to the common 4/3 effective earth radius
figure used in special planning charts for

troposcatter path profiles. 1 Realistically, however,
the troposphere is not so predictable and AN/Ah

actually decreases with height. For example, Sarkar,
Dutta, and Reddy [1983] show daily K-factors ranging
from 1.4 to 2.2 on a 240 km, 2 GHz link in India.
Furthermore, extreme variations in the refractivity

gradient (and, hence, the K-factor) can lead to
serious propagation problems, termed, for simplicity,

anomalous propagation.
2..3.2.2 Anomalous Propagation

Three types of anomalous propagation can
degrade troposcattE 'stem performance

significantly. Figures 13 and 14 will aid in this
discuassion (note Figure 14's use of -39 N units/km
for standard refraction). Gradients exceeding -40 N

units produce the first anomalous propagation event,
subrefraction. In this case, paths are less curved;
furthermore, a gradient of zero produces straight
14/3 earth radius charts are but one of

several tools available. 2/3 charts are used under
worst case conditions. Also, flat earth charts are
available but are more difficult to use. Flat earth
models use M, modified refractivity, instead on N
[Hall and Barclay, 1989, p. 153.
39
Normal atmosphere
.- Subrefraction
Super-refraction
Ductincg
No N
Source: Boithias, L. Radio wave Propagation.
St. Louis: McGraw-Hill, 1987, p. 92.
Figure 13. Refractive conditions leading to

anomalous propagation events
00 0
.9
319 -39
.17
157 -157
Real Earth with radius a Representational Earth Flat Earth

with radius 413a
Source: RoithiaC. L. RAdIo move Propagac ion.
St. Louis: MCGCW4LIIH 1987. P. 93.
Figure 14. Wave paths under various refractivity

values for different earth models
40
lines (real earth model) while positive gradients
bend waves upwards [Boithias, 1987, p. 921. This

event will redefine the common volume of a
troposcatter link and can lead to signal fading.
Surface conditions that can invoke subrefraction
typically occur in two forms: Type A - surface
temperatures exceeding 30 0 C and relative humidity
below 40%, and Type B - surface temperatures between

100 and 300 C and relative humidity more than 60%.
Brauburger (1979, pp. 26,27] explains that Type A and
Type B conditions lead to density increases with
height, chiefly in water vapor content, resulting in
subrefraction.
A second anomalous propagation phenomenon,
super-refraction, occurs at gradients less than -40 N
units. Basically, super-refraction results in more
bending, and can, in some instances, obliterate
communications as the bent wave never reaches the
receiver. Super-refractive conditions normally
accompany modified temperature gradients brought on
by the passage of cold fronts, subsidence (vertical,
downward movement of air), or even night time cooling
on the ground [Boithias, 1987, pp. 85,86; Brauburger,
1979, pp. 24,25].

41
Ducting, the final example of anomalous

propagation, is a well documented event and likened
to propagation in a waveguide. Two types of ducts
can occur. In a surface duct, extreme super-

refraction bends the radiowave into the earth and the
earth reflects the bent wave back into the
atmosphere. Conditions permitting (continued,
extreme super-refraction over a reflecting earth),
this process can be repeated over great distances.
Surface ducts are

fairly common over water, especially in the
tropics. For instance it is not unusual to
propagate signals over a distance of 1500 km in
such areas as the Arabian Sea [Staniforth, 1972,
pp. 206,207].
Elevated ducts, the second type, never return
to the earth's surface. They appear under the unique

condition of simultaneous super and subrefractive
conditions at different altitudes. A wave launched
through a subrefractive layer will be bent upward and
later bent downward as it travels to a higher
altitude super-refractive layer. Hall and Barclay
[1989, pp. 155-161] add that refractivity values less
than -157 N units can cause either ducting condition

and further show advection, evaporation, subsidence,
temperature inversions, and weather fronts as
conditions fostering duct creation.
42
2.3.3 Small Scale Changes in Refractivity

To this point, discussion of refractivity has
focused on large-scale variations with height.
Refractivity l1so exhibits localized, fine-scale
variations in both space and time that are crucial in
understanding troposcatter theory.
Therefore equal-index surfaces are not perfect
spheres concentric to the earth [as previous
discussion may suggest], but may take a variety
of forms. Some may even constitute closed
surfaces in which the internal refractive index
may be slightly higher (or lower) than external
values [Boithias and Battesti, 1983, p. 657].
Typically, irregularities have amplitudes of a few N
units above or below the surrounding median; however,

they decrease in both number and amplitude with
height.
Localized temperature and humidity deviations
cause these irregularities. Eklund and Wickerts
[1968] cite humidity as the chief determinant, and

describe refractive index fields as consisting of
sharply bounded air volumes containing excess
humidity. Further,
[ilnteraction between wet and dry air causes
refractive index fluctuations. Through the
turbulence, the strong gradient will be broken
down into a spectrum of fluctuations [Eklund and
Wickerts, 1968, p. 1071].
Section 2.4 describes turbulence and the spectrum
model; however, while it is agreed that this fine-
43
scale structure can cause scattering and

scintillation effects, its precise modeling eludes
theoreticians.
2.3.4 Summary of Refraction

Under normal atmospheric conditions, the
decrease in the radio refractive index causes

electromagnetic waves to experience velocity
increases at higher altitudes in the troposphere.
Consequently, these waves are bent downward with
curvatures defined through the refractivity gradient.
This gradient, as well as the radio refractive index,
is contingent on temperature, pressure, and water

vapor content with the latter displaying the most
influence. Abnormal deviations in the gradient

result in the anomalous propagation events of
subrefraction, super-refraction, and ducting.

Finally, gradient models focus on large-scale
variations refractivity with height; however, small-
scale variations, critically important in

troposcatter, also exist. It bears mentioning that
challenges to the general models used in refractivity

profiles exist. Carroll and Ring [1955] and Crain
and Levy [1987] highlight several objections.

44
2.4 Turbulence
Much of the theory surrounding troposcatter
propagation stems from turbulence theory applicable
to the small-scale refractivity variations mentioned

in Section 2.3.3. This is because "the atmospheric
turbulence behaves like a source of inhomogeneities
which produce scattering" [Tatarski, 1961, p. vii].

Surely turbulence entered propagation theory with
a vengeance in 1949. The troposphere was
pictured as an inhomogeneous medium, that is, one
which could be pictured as instantaneously
containing more or less spherical blobs of air
whose index of refraction differed from that of
the surrounding air by an amount of the order of
one millionth [Carroll, 1952b, p. 81.
Unfortunately, in many respects, atmospheric
turbulence remains "one of the great unsolved areas
of science and engineering" (Derr, 1972, p. 4-1].
Difficulties root themselves in the large numbers of
variables inherent in governing equations which are,
in themselves, quite complex. A sampling of these
variables include geographical dependence and weather
on the large scale, and the small-scale influences of

wind, solar heating, and surface roughness [Derr,
1972, p. 4-1]. This section will cite the causes of

turbulence and relate a model that attempts to
describe small-scale refractivity variations.
45
2.4.1 The Causes of Turbulence

Anthes [1981, pp. 70,122,326] begins that
turbulent eddies or blobs have dimensions of hundreds

of meters to a centimeter, with a time scale ranging
from a few minutes to a second. These eddies are
produced either through thermal or mechanical
mechanisms. Thermal turbulence begins with surface
heating and leads to rising air columns or convection
currents (described by Crain and Gerhardt [1952] as
buoyancy forces). Mechanical turbulence occurs close
to the surface with air flows interrupted by the
stationary ground. This produces horizontal wind
shear. Additional mechanical turbulence is produced
by frictional forces between either thermally or
mechanically created eddies. Both wind shear and
convection currents can rise to higher altitudes and,
here, are larger in scale.
Hopefully, the above description of
turbulence is confusing since turbulence itself is a

random, chaotic event. A typical model of turbulence
therefore shows eddies produced through thermal or
mechanical processes which interact ("mix") to
produced refractivity fluctuations [Crain and
Gerhardt, 1952, p. 51].

46
2.4.2 A Refractivity Turbulence Model

As turbulent eddies mix with one another,
both their size and intensity change. L.F.
Richardson aptly describes this process:

Big whirls have little whirls,
That feed on Ctc Ir velocity;
And little whirls have lesser whirls--
And so onto viscosity [Megaw, 1957, p. 444).
Figure 15, which shows the spectrum of refractivity
irregularities, S(k), plotted as a function of
wavenumber (k), is useful in modeling this process.
SM
- ?.p4t 4
% IL
Source: Wneelon, A.D. "Radio Scattering by

Tropospheric Irregularities." Journal of AtmOspheric
and Terrestrial Physics, vol. 15, nos. 3/4 (October
1959), p. 188.
Figure 15. Spectrum of refractivity irregularities

Blobs created in the input range spawn smaller
structures by the cascading process described
above and eventually the irregularities are
dissipated when their scale size falls below
about 1 mm. These very small structures are
dominated by the influence of viscosity and
diffusion and cannot sustain turbulent activity
(Hall and Barclay, 1999, p. 169].
47
From Figure 15, three distinct areas are
evident. In the input range, large eddies with low

wavenumbers are inserted. They maintain a certain
energy level derived from winds and internal pressure

or temperature gradients. Entering the second
region, the inertial subrange, energy is expended
through creation of smaller eddies and viscous
friction. It is within this region that controversy
about the form of the irregularities' spectrum
resides. Finally, in the dissipation range, the
irregularities are lost through heat conversion
[Picquenard, 1974, pp. 33,34; Ortwein, Hopkins, Pohl,
1961, p. 790].
Gerlach [1984] mentions that eddy sizes of 1

m to 1 km are of most interest, and are in the
inertial subrange. Theories describing

irregularities within this domain include Obukhov's
mixing theory and the mixing-in-gradient theory.

Wheelon (1959], Staras and Wheelon [1959], and du
Castel [1966] provide general descriptions of both.

Additionally, Appendix C further discusses the
spectrum of irregularities.
2.4.3 A Conclusion on Turbulence
Turbulence is easily observed in the
twinkling of stars or the unsettling feeling of a

48
bumpy airplane ride. Unfortunately, it is perhaps

the least predictable of atmospheric parameters
impacting troposcatter propagation. Without it,
however, small-scale refractivity gradients would not
exist and, therefore, neither would troposcatter.
While thought of as a boon in this instance,
turbulence will cause scintillation (described

later); furthermore, the turbulence model, as
outlined above, is the foundation of scattering loss
theory. The turbulence of refractivity centers
around the portion of the spectrum function contained
in the inertial subrange. Equations for this
spectrum (Appendix C) exhibit some speculation as do
the theories describing eddy activities within this
region.
2.5 Atmospheric Effects on Troposcatter Propagation
This concluding section is meant to embrace

previous discussion while introducing some of the
more pertinent atmospheric effects on radiowave
propagation. Figure 16 charts the course this
section will take. Natural phenomena, on the extreme

left includes the observable yet uncontrollable
events characterizing the troposphere. For lack of a

better term, the refractivity event includes both
large and small-scale refractivity changes as
49
c-T-
LLJ Lii
0 0
C-.. < 0-
o 0 o-
00
Ci0 N d. 0
<~0)
rz 01 04
0 Q)0
Q 4F)
CO H00
0 0 44 .
00 K H
Ci) -J4 Mc
LU- LU
<L (
Ch LU; moIr-
ci:
(J) 0
50
described earlier. Crane [1981a] and Bean, Horn, and
Riggs [1962] loosely bundle gases and aerosols under
macroscale (100-1000 km) variations; hydrometeors
within mesoscale (10-100 km) and cloud (1-10 km)
motions; and refractive index fluctuations among
microscale (1 mm to 1 km) changes. Note that dust
and sand constituents, which may occasionally enter
the atmosphere, are not included. CCIR [1986] Report
563-3 mentions that dust and sand propagation
degradations need further research, - nile Report 721-
1 adds that these contributions appear negligible.
In the center of Figure 16 are mechanisms.
These are the effects of natural phenomena on
troposcatter propagation. Finally, on the far right
are system parameters. Appendix B shows external
noise and propagation loss as the only variable
parameters in the system equation; theiefore, all
atmospheric effects on troposcatter propagation are
reflected within either or both parameters. (Noise
due to depolarization is manifested through
crosstalk; however, Figure 16 shows a question mark,
"?", here as no source listed this as a noise
contributor.) IiL essence, then, the effects of the
atmosphere on troposcatter propagation can be viewed
frcm a causal framework: nat,:ral phenomena cause

51
degrading propagation mechanisms which cause changes
in system parameters.
While various approaches describing these
relationships are possible, this section will focus
on the mechanisms. Note should be made that although
Figure 16 neatly categorizes the causal nature of the
atmosphere on radiowave propagation, the entire
process is very complex. Therefore, this treatment
is intended to be more introductory that rigorous.
The multitude of texts devoted to radiowave
propagation confirm this assertion.
2.5.1 Refraction
The ability of the atmosphere to refract or
bend radiowaves was treated earlier in Section 2.3.
Considerations of refractivity therefore include
effective earth radius adjustments and allowances for
anomalous propagation. Figures 17 and 18 can be used
to determine the effective earth radius if either the
gradient of refractivity (AN/Ah) or the surface
refractivity (Ns ) are known respectively. Clavier
[1956, p. 112] notes that effective earth radius K-
factor variations of 4/3 to 0.7 should be expected
due to large-scale refractivity changes.
Anomalous propagation events (subrefraction,
super-refraction, or ducting) can enhance or degrade

52
4IH. t
~ A
7300 k EAR +0 . X 0- 6
300 -~00 N 00 30 40
~ 'L
NDX RAIETW~At. N LNIS ERKIOMTE
Miltary Handbo
Source: Facliy esgnfo
Tropospheric Scatter (Tr~ansoio iroaeSse
~
Design). ~ ~ ~ ~
7~ Dprmeto-Dfne =J -4-DK-
Wahig~n C,2 Nvmbr-17, .4-6
17. Efectie eath aisfcor

Figure~~~~~j4+ ,vru
rercivt4rain
53
11
200
10800
J40
76DO
26 'O 3)0 320 W4 360 1) 4t'
Soure: Mlitay Handbook, Facility Desiarn fcz

-rcpcspheric Fcatter (Tr-ansharizon Micrcwave SyS:C-m
resign). MIL-iiD=K-417, :epartment cf Defense,
'eashington, DC' 25 November, 1977. Available th~rough
Navy Pu~ctcsand .7,rn-s Center, AT-N: N=C'DS,
^ffice of Navy ?P';blications, 5801 Tabor Aven-ue,
Phildelphia, Pa., 19120-50,'4, p. 4-90.
Figure 18. Effective earth radius versus surface

refractivity
54
received signal levels. In rare circumstances, a
surface duct will engender long periods of abnormally

good propagation. More often, however, under extreme
super-refraction (much less than -40 N units),
radiowaves are bent towards the earth surface and
signal blackouts occur as the surface duct "skips"

over the intended receiver. Elevated ducts can also
cause complete signal loss. Roda (1988, p. 147]
concludes that instances of these extreme events
subscribe to no definite rules; therefore, experience
in site specific situations provides the best
prediction tool. He notes an example of ducts over
the North Sea lasting several hours yet occurring

only 0.7% of the year.
More often, however, deviations from standard
refraction (-40 N units) fosters slow signal fades
[Clavier, 1956, p. 112; Roda, 1988, p. 68].

Engineers recognize the inevitability of such
attenuations and include them in the system equation
of Appendix B. This fading is discussed more in
depth in Chapter 4.
2.5.2 Absorption
Electromagnetic energy can be partially

converted into thermal energy through absorption
[Flock, 1979, p. 174]. Signal loss as well as

55
increases in noise result from this process. Figure

16 shows gases, aerosols, and hydrometeors as the
events leading to absorption. For simplicity,
aerosols will be grouped with hydrometeors although
Derr [1972, pp. 1-3 to 1-5] notes physical
dissimilarities between the two events.

Of the gaseous elements listed in Table 2,
only oxygen exhibits significant absorptive
qualities. This is because it possesses "a small
magnetic moment" which, under the influence of a
magnetic field (an iectromagnetic wave), causes it
to "rotate end over end or to oscillate in many other
possible ways" [Livingston, 1970, p. 10; Kerr, 1987,
p. 25]. Furthermore, at certain frequencies, oxygen

will resonate, and absorption peaks to significant
levels. Figure 19 depicts these critical
frequencies, 60 and 119 GHz, and further shows water
vapor absorption which displays similar

characteristics at 22, 183, and 325 GHz. Water vapor
differs from oxygen since it has an electric dipole
moment, yet the absorption process is very similar.
Comparing the attenuation contributions of

oxygen and water vapor, one notes oxygen's dominance
at lower frequencies, about 0.007 dB/km between 2 and
12 GHz, while water vapor remains under 0.005 dB/km

*6~ 56
i% I i I I i
IV,
t W,
Zoi dMda eea 192 p. 141 *.
igr 19 Gaeu bsrto

beo 8iI G___ Beod1 ae ao otiuin
mach thn fo th otprexedtoeo
oxygen ~~
(Egnern Cn irtos 192 p.5]
watr vpo absorptitteuaisfr typca
exicin In an atep t difrniat h w
machnthe, Fork the9mos part

15eeed thse ofh
57
S IO.O00 MC
-J
<.
:7 M: i -.i.. ....... ......
i400 80 00 I00 200 40 160
TOTAL TH DISTANE IN KILOMETERS

Source: Military Handbook, Facility Design for
Tropospheric Scatter (Transhorizonz Microwave System
Design). MIL-HDBK-417, Department of Defense,
Washington, DC, 25 November, 1977. Available through
Navy Publications and Forms Center, ATTN: NPODS,
Office of Navy Publications, 5801 Tabor Avenue,
PhildeRphia, Pa., 19120-5094, p. 4-91.
Figure 20. Combined oxygen and water vapor
atmospheric absorption
effect of small, lossy water droplets at microwave

frequencies is primarily one of absorption."
For wavelengths which are long compared with the

drop size, i.e. in the SHF band, attenuation due
to absorption will be greater than that due to
scatter. Conversely, for wavelengths which are
short in relation to drop size, i.e. in and above
the EHF band, scatter will predominate [Hall,
1979, p. 47].
Figures 21 and 22 show hydrometeor absorption for
earth-space paths traversing the entire troposphere.

Figure 23, more applicable for troposcatter, depicts
58
100 1
2W 100. in
-
100-__2.
bI
10 ZG.n.0 J .0/-
.2 '0 I0
O' 0 ,
0.11
2 0 20 50 100 1 0 10 20 5-D 100

F,.qn . G4 Fw~v Gil
Source: martino, j. Communicationsg Sat*llito
Systems. Englewood Cliffs, NJ:Prentice-Hall,
Inc., p. 115.
Figure 21 (Left). Atmospheric absorption due to
Figure 22 (Right). Atmospheric absorption due to

fog, mist and clouds
to rain
59
cumulative absorption and scatter attenuation in dB
per km. Note in all instances the insignificance of

hydrometeor absorption at troposcatter frequencies.
2.5.3 Scattering
Scattering is essentially the re-radiation of
portions of the electromagnetic energy in directions
deviating from that intended. Obviously, those
signals re-routed from the inended direction
contribute to propa-ation loss. As mentioned,

hydrometeors are a scattering source; additionally,
small-scale refractivity variations will cause

scattering thus making troposcatter propagation
possible. Radar systems profit from this process
since airplanes scatter signals, and, finally,
scattering is a source of interference among

different transmission systems [Lane, 1968, p. 4].
Scattering theory emphasizes particle size
and number. Simplistically, at troposcatter
wavelengths, scatter attenuation due to hydrometeors
increases with drop size (with respect co a certain
wavelength) and the number of drops [Staniforth,
1572, p. 208]. Flock (1979, p. 1751 further shows

that the Rayleigh scattering model applies at
60
troposcatter wavelengths.2 However,

[i]t must be stressed that there is at present no
single model for drop size distributions which is
generally accepted as representing physical
reality - even as a statistical mean over many
rain events (Hall and Barclay, 1989, p. 176].
Crane [198ia, p. 200] contends that the scattering

properties of "the more complex shapes actually
assumed by watez- drops . . . can be approximated only
after long and tedious numerical analysis." However,

he concludes, scattering attenuation can be estimated
through simple models.
As mentioned, hydrometeor effects on
radiowave propagation conventionally include both
scatter and absorption. Again, Figure 23 is

referenced, and the combined attenuation due to
hydrometeors (specifically, rain) is seen as
insignificant at troposcatter frequencies. Hall and
Dowling [1974] offer the same conclusion, but add
that "rain may cause significant changes in the mean

direction of arrival of the scatter signal" in a
troposcatter system. Crane (1988] and Doherty and
Stone [1960] further report enhanced median signal

levels during rain, but Doherty and Stone [1960] add
that fading rates will increase.

2 TheRayleigh scatter model is used when
scatterer size is small compared to wavelength.
Raindrops range from 0.05 to 0.65 cm, at 6 GHz
wavelength is 5.0 cm; thus, Rayleigh scatter applies.
61
The final discussion on scattering highlights
small-scale refractivity changes. These variations

are responsible for the troposcatter propagation
mechanism and also "cause random independent bending
of the antenna beams both horizontally and
vertically" [Roda, 1988, p. 68]. This action results
in rapid fluctuations of received signal levels
termed scintillation [Crane, 1981a, p. 202]. While
small in amplitude, system equation margins must
allow for scintillation induced by scattering.
Boithias [1987, p. 108] expresses scintillation
amplitudes through the structure constant, CN 2 , as
described in Appendix C.
2.5.4 Depolarization
The final mechanism encountered by radiowaves
propagating through the troposphere is depolarization
which may also be termed transpolarization or cross
polarization. This mechanism results in portions of
the propagating energy reappearing orthogonally to
the intended polarization as discussed in Appendix D.
The causes of depolarization, as illustrated in
Figure 16, are aerosols, hydrometeors, and
refractivity (again, aerosols will be lumped with

hydrometeors). It should be noted that this
mechanism arises from a mix of other mechanisms
62
(scatter and refraction) and is difficult to
differentiate; however, its effect is unique,
meriting separate treatment.
Depolarization due to both large and small-
scale refractivity variations is routinely associated
with multipath. Multipath, described in more detail
in Chapter 4, results from a variety of sources such
as scattering, terrain and atmospheric reflection, as
well as refraction. Simplistically, depolarization
due to multipath results from the simultaneous
arrival of the wanted, transmitted wave and an
unwanted, orthogonally polarized wave created under
multipath conditions [Ippolito, 1986, p. 116;
Boithias, 1987, p. 110].
Hydrometeor (rain) depolarization greatly
depends on drop shape. These shapes change from
spherical to oblate spheroids as drop size increases.
Depolarization, then,
is caused by the individual drops falling as

oblate spheroids and producing different
attenuation and phase shift for waves polarised
[sic] parallel or perpendicular to the major axes
of the drops [Hall, 1979, p. 112].
More specifically, if the radiowave's electric field
is not parallel to the axis of the equitorial

plane of the drops, it gives rise, by scatter, to
a wave with orthogonal polarization to that of
the incident wave. The relative level of this
orthogonal wave depends on many factors such as
the angle of incidence of the wave, the canting
63
angle of the drops, the intensity of rainfall,

the distribution of drop diameter, frequency, etc
[Boithias, 1987, p. 138].
Rain depolarization is further discussed in Appendix
D. At troposcatter frequencies, depolarization is
considered unimportant; however, as higher
frequencies are used (see Crane, 1988) it may become
significant [Military Handbook, 1977, p. 4-173].

CHAPTER 3
TROPOSCATTER THEORY
Long experience has . . . taught me not always to

believe in the limitations indicated by purely
theoretical considerations, or even calculations.
These--as we all know--are often based on
insufficient knowledge of all the relevant
factors. I believe, in spite of adverse
forecasts, in trying new lines of research,
however unpromising they may seem at first sight
[Carrol, 1956, p. 1057].
- Guglielmo Marconi, 1932
3.1 Introduction
In the late 1940's, new theories attempting

to explain high signal levels received beyond the
horizon at short wavelengths first appeared. Prior
to these, classical diffraction theory was modified
to include atmospheric refractivity and ducting to
account for this peculiarity. Success was marginal
as experiments showed consistent signal levels many
decibels greater than those predicted by these
modified theories [Bullington, 1953, p. 132; Ortusi,
1955, p. 87]. Investigators realized that nature had
conveniently provided some sort of low altitude

passive reflector, yet were baffled by its nature.
65
Accepted beyond the horizon theory was rooted
in Van Der Pol and Bremmer's 1937 smooth sphere
diffraction predictions which indicated "an
exponential decrease in signal strength at a rapid
rate at and beyond the radio horizon" [Radio
Transmission, 1960, p. 30]. With a predicted decay
of about 1 dB/km at 500 MHz and 2. dB/km at 5 GHz,
long range propagation at and beyond VHF was thought
impossible. Experimentation showed these figures as
somewhat pessimistic, and large-scale refraction was
included which extended the radio horizon through
allowance for an effective earth radius 4/3 greater
than actual (Lane, 1968, p. 2]. However, even this
modification could not adequately explain all
observed data.
After WWII, both radar and television systems
highlighted circumstances where transmission ranges
exceeded classical predictions.
It was also pointed out that the median signal

received at points far beyond the horizon could
not be explained by modifying the effective earth
radius since the data could not be fitted to an
exponential law (Bullington, 1955, p. 1176].
Anomalous propagation events such as super-refraction
and ducting as well as ionospheric influences were
then incorporated. However, these conditions were
sporadic and again could not clarify the persistency

66
of observed beyond the horizon fields [Radio
Transmission, 1960, p. 30].
Sources contend that even today the mechanism

behind troposcatter propagation is not completely
understood [McKay, 1989, p. 220; Roda, 1988, p. 2].
However, what is known can be briefly summarized in

Figure 2.
The only mechanisms for radio propagation beyond
thz horizon which occur permanently for
frequencies greater than 30 MHz are those of
diffraction at the Earth's surface and scatter
from atmospheric irregularities. Attenuation for
diffracted signals increases very rapidly with
distance and with frequency, and eventually the
principal mechanism is that of tropospheric
scatter (CCIR, 1986, p. 367].
In short, troposcatter propagation exists, and is

reliable enough to use for beyond the horizon
communications.
While both diffraction and scatter can
simultaneously exist as multimode, transhorizon
propagation (see Appendix A), this section will
examine pure troposcatter theory only, ignoring
diffraction. Gjessing and Irgens [1964] group
troposcatter theories under "turbulence",
"reflection", or "mode" classifications; this
treatment will do the same. Discussion will avoid

technical analysis, but will provide a physical
understanding of the proposed mechanisms.

67
3.2 Turbulence Theories
Concerning theories on troposcatter

propagation, no single citation is referenced more
than Booker and Gordon's 1950 IRE article, "A Theory

of Radio Scattering in the Troposphere." However,
their suggestion that constant, beyond the horizon
fields are due to scattering from small turbulent
blobs was initially forwarded by Pekeris [1947]

[Johnson, 1958, p. 166]. Additionally, at about the
same time as Booker and Gordon, Megaw [1950]
independently published a similar theory. A summary
of turbulence theory proponents is offered by du
Castel [1966, p. 12].

Simplistically, turbulence theory purports
that within the common volume, portions of

transmitted energy are scattered by small-scale
refractive variations "much as fog or moisture seems
to scatter a searchlight on a dark night" [Freeman,
1975, p. 265]. Tatarski summarizes the problem:
A plane monochromatic electromagnetic wave is
incident on a volume V of a turbulent medium;
because of turbulent mixing within the volume V,
there appear irregular refractive index
fluctuations, which scatter the incident
electromagnetic wave. It is required to find the
mean density of the energy scattered in a given
direction [Tararski 1961, p. 59].
Booker and Gordon pictured the atmosphere as
consi~ting ot a col cctio uf

upherical blobs of
68
diameter 1 (the scale of turbulence) with each blob

having a refractive index slightly different from
that of the mean. "Incident waves cause polarization

of the elementary volumes, converting them to
electric dipoles, which in turn radiate energy"
[Picquenard, 1974, p. 32]. Figure 24 depicts a

scatterer's radiation pattern which decreases very
rapidly from the transmission beam's axis. 1
IND ICATIVYE
RADILATION ctdanl
PATTERN
Source: Roda, G. Troposcatter Radio Lirks.

Norwood, Maine: Artech House, 1988, p. 63.
Figure 24. Single scatterer radiation pattern
Individual blob scattering is plotted as

scattered power (P) versus scattering angle (0) on
Figure 25. When all blobs within the common volume
are considered, the plot becomes smeared as in Figure
26 and the radiation pattern resembles Figure 27.
iThis, along with the fact that more
scatterers are located in the lower portion of the
tropospherc, explains th= need for a small scatter
angle.
69
Source: Gordon, W.E. "A Simple Picture of

Tropospheric Radio Scattering." IRE Transactions on
Communications Systems, vol. CS-4, no.1 (March 1956),
p. 98.
Figure 25 (Top). Scattering power by an

individual blob
Figure 26 (Bottom). Scattering power by a number of

blobs
ransmission
Source: du Castel, F. Tropospheric Radiowave

Propagation Beyond the Horizon. New York: Pergamon,
1966, p. 48.
Figure 27. Radiation pattern of a number of

scatterers
70
Although most power is radiated in the forward
direction (0=0), the receiver is located at (0>0);

therefore, Booker and Gordon assert that the
mathematical expression of Figure 26, the scattering
coefficient or the scattering cros3 section (0), is

critical.
The scattering coefficient answers Tatarksi's
question by giving the scattered power per unit solid

angle, per unit incident power density, per unit
volume [Picquenard, 1974, p. 32]. When integrated
over the scattering volume, it provides the power

scattered to the receiver [Gordon, 1955, p. 98]. A
number of references provide different expressions
for a; however, these will not be provided due to the
complexity involved. 2 Regardless, the crucial

parameter in expressions for a relates to turbulence
modeling. Wheelon [1959, p. 193] gives two equations
for a using either the spectrum of irregularities,

S(k), as described in Appendix C, or the spacial
correlation function C(r).
Booker and Gordon used an exponential C(r)

model [Picquenard, 1974, p. 31]:
C(r) = exp(-r/1) (3-1)
2 See
Booker and Gordon [1950]; Gordon [1955];
Wheelon [1959]; and du Castel [1966] for example.
71
The scale of turbulence 1 is considered to be the

distance at which the cross correlation between
the two sets of simultaneous refractive index
fluctuation falls to i/e of the value obtained at
zero separation [Crain and Gerhardt, 1952, p.
50].
Thus the correlation function falls between unity
(r=O) and zero (r=-). However, this choice was
somewhat speculative as the ability to measure small-

scale atmospheric refractivity variations was, at
that time, marginal [Chisholm, 1956, p. 8;
Picquenard, 1974, p. 33]. Later, after the
development of the refractometer, other models were
developed with the Bessel model, described by Staras
and Wheelon (1959, p. 81], providing better agreement
with experimental data.
Another adjustment to Booker and Gordon's
work is due to their disregard for the time
variations of refractive index irregularities.

Emphasis in turbulence theory shifted to the spectrum
function, S(k), which is especially suited to depict

these temporal fluctuations [Staras and Wheelon,
1959, p. 80]. This function, described in Appendix
C, "is the Fourier decomposition of the space

correlation function" [Wheelon, 1959, p. 186].
Ortwein, Hopkins, and Pohl (1961] and Staras and

Wheelon [1959] summarize both spacial correlation and
spectrum methods in modeling turbulence.

72
While the above description of turbulence
theory does not provide rigorous treatment, a few

important conclusions are evident. Primarily,
turbulence theory attributes persistent, beyond the

horizon fields to scattering from spherical blobs
whose refractive indexes slightly differ from the
mean. Received fields are calculated through

integration of the scattering coefficient over the
critical common volume. Crucial in this theory is
the modeling of turbulence which has undergone
numerous revisions and exhibits several variations
(see Appendix C).
3.3 Reflection Theories
The idea that beyond the horizon fields at
short wavelengths are caused by reflections from

tropospheric layers was first suggested by Norton
[1948]. Later, Feinstein [1951, 1952a, 1952b],
launched a similar investigation but his ideas were
adapted to subsequent mode theories (discussed

later). Serious endeavors include the work of Friis,
Crawford, and Hogg [1957] and French researchers led

by du Castel [1966, p. 12].
In general, proponents assert that layers
within the troposphere are formed by relatively sharp
gradients in the refractive index. These layers are

73
numerous in number, display limited yet varying

dimensions, and present random position and
orientation.
The number and size of the reflecting layers, as

well as the magnitude of the discontinuities in
the gradient of dielectric constant which form
them, influence the received power [Friis,
Crawford, and Hogg, 1957, p. 628].
An understanding of the character of a layer aids

received power calculations.
When Friis, Crawford and Hogg first
postulated their theory, refractometer data
substantiated the existence of steep variat ' -s in
the refractive gradient. However, the actuai shape
of the layers formed by these variations remained
speculative. In short, then, calculations were

performed for three general cases: large, small, and
intermediate size layers with respect to Fresnel zone
dimensions. 3 Later, du Castel [1966] offered a more
thorough, and, quite elegant, description of these
layers.
Figure 28 displays du Castel's layer or
feuillet. This thin, stable layer forms when laminar

air flow "within a layer produces a change in the
3 Fresnel
zones may be viewed as three-
dimensional ellipsoids encompassing radiated energy,
and dimensionally dependent on wavelength. Boithias
[1987], Livingston [1970], and Kerr [1987] further
describe this basic concept.
74
z A
0 0
0 0-------
O 0 0 L

1966, p. 154.
Figure 28. du Castel's thin layer or feuillet
mean refractive index across its thickness" (du

Castel, 1966, p. 26]. As shown on the left of Figure
28, the refractivity gradient contains a
discontinuity; therefore, within the layer, the
refractive index differs slightly from that above and
below the layer. The surface has primary

irregularities, 1 and h, both on the order of tens of
meters, and secondary irregularities, 11 and hl,
several kilometers and a few hundreds of meters

respectively. Activity within an irregularity may be
likened to that of a turbulent blob described
earlier; however, although turbulent irregularities

75
are crucial in liyer formation, in this instance the

existence of a layer is more critical than individual
blobs. A simplified layer model, consisting of
numerous irregularities, is of size L and thickness H
and is shown at the bottom of Figure 28 (du Castel,

1966, pp. 154,155].
A feuillet's reflection mechanism is easily
understood through application of elementary optics.
While the main flow of transmitted energy propagates
(refracts) through the layer, a portion is reflected,
hence, partial reflection occurs as predicted in
Snell's law for mediums of differing dielectric

constants. Furthermore, depending on the nature of
the reflecting layer, either specular or diffuse
reflection occurs [Barton, 1962, p. 335]. 4 Figure 29
shows a number of elements, a (not to be confused
with the scattering coefficient described uider
turbulence theory), within a surface, S, with

transmitted energy incident at a grazing angle a.
The primary reflections at the surface elements a
within a solid angle co, add together in phase to
give specular reflection in a smaller solid angle
W', or add randomly to give diffuse reflection in
a solid angle (o [du Castel, 1966, p. 1591.
4 Boithias
[1987, p. 49] simplistically
differentiates between specular and diffuse
reflection.
76

Propagation Beyond the Horizon. New York: Pergarncn,
1966, p. 159.
Figure 29. Specular and diffuse reflection within a

feuilet
Tronsmission Reception

1966, p. 48.
Figure 30. Radiation pattern for the reflection

mechanism
77
Figure 30 gives a general look at troposcatter
propagation via specular and/or diffuse partial
reflection.
Calculations for the received field rely

exclusively on evaluation of the reflection
coefficient, p. Appendix E discusses p in greater
detail. However, while determination of reflected
power is a basic exercise, the complexities
introduced by "the spatial distribution of layers, in
size and intensity distributions, and in
distributions of curvature for nonflat layers" make

computations more statistically oriented [Cox, 1969,
p. 907]. Thus, a simple calculation (see Friis,

Crawford and Hogg [1957]), is again made difficult
due to the unpredictability of the atmosphere.
Recapitulating, when the normally smooth
refractivity profile exhibits sharp discontinuities,
layers consisting of turbulent irregularities are
likely to form. These layers act as partial mirrors,

reflecting a portion of the transmitted power.
Normally basic calculations for received power are
complicated by a reflection coefficient that must

consider the complexities of the variable shapes and
motions of atmospheric reflecting layers (see
Appendix E).
78
3.4 Mode Theories

When Bremmer [1949] mentioned that 4/3 earth
curvature theory ignored the reflected component,
partial coherent internal reflection or mode theory
took root. 5 Feinstein [1951, 1952a, 1952b], Carroll
[1952a, 1952b, 1952c], and Carroll and Ring [1955]

were enthusiastic (most notably, Carroll) of this
idea which challenged classical notions.
Initially, one will find some similarities

between mode and reflection theories. Both
incorporate Snell's law of reflection and refraction;
however, mode theory introduces no "new" concepts,

such as turbulent irregularities. Instead, it
focuses on large-scale refractivity noting that not
only a refracted (see Section 2.3) but a series of
reflected waves must be produced due to the
refractivity gradient [Feinstein, 1951, p. 1292].
This in turn causes various modes of propagation
similar to surface ducts, the distinction being that

only the reflected, not the refracted, wave is
considered. Figure 31 simplistically depicts one ray
for several modes; however, it should be noted that
each mode will exhibit a family of such rays, and
these rays will experience reflection (external), and
5 For
the earth's atmosphere, an upgoing wave
reflects an internal wave, and a downcoming wave
reflects externally [Carroll, 1952b, p. 9, 10].
79
Source: Katzin, M. "Tropospheric Propagation

Beyond the Horizon." Transactions of the IRE,
Professicnal Group on Antennas and Propagation, ?GAP-
3 (August 1952a), p. 116.
Figure 31. Mode theory propagation
therefore attenuation, as they propagate downward
[Katzin, 1952a, pp. 112,113].

The object of the internal reflection hypothesis
is to relate the average received field to the
rate of decrease in atmospheric density, and
hence to eliminate some of the arbitrary
parameters required by the scattering concepts
[Bullington, 1963, p. 2850].
Mode theory proponents were particularly
critical of then accepted refractive index profiles

which showed "decreases linearly with height
indefinitely . . . [causing] the index of refraction
to become less than 1 or even negative at

sufficiently large heights" [Carroll, 1952c, p. 93].
Initially, then, a bilinear model was introduced
which showed linear refractivity decreases to 30,000
feet and, above this sharp break in gradient, the

index was assumed to be unity. Determining the
received field involved an infinite summation of mode
80
contributions from successive layer reflections using

Fresnel's summation rules. This eventually gave a
reflected field "equivalent to half the reflection

from the half-wave layer of maximum contribution"
(Carroll, 1952c, p. 84]. Although proponents admit
nonrigorous calculations, close theoretical agreement
with experimentation was provided by Colwell and
Anderson [19521.
Attempts to provide a better fit with field
data (Megaw's points) included use of trilinear and
inverse square profiles as depicted in Figure 32.
.ANTENNA
:1IG)ITS ?tFy..ITFt
G
1n95.
) 0 36p3 10.
3le~ Winter aupntl
It
fideld .;A daft
6O0 forms
source: Chisholm, . 1. II P"rogreaS Of

T'ropospheric Propagalti~n Research Related to
Conatinications Beyond the ilorizon." in s rransactions
on Communications Systems, vol. CS-4, 1-0-1 (March
1956), p. 10.
Figure 32. Various mode theory profiles matched with

field data
81
However, Chisholm (1956, p. 81 and du Castel [1966,

p. 132] point out that the mode of maximum
contribution occurs at the gradient discontinuities

of the bilinear and trilinear models and that such
discontinuities do not exist in nature. Further, the
inverse square and a subsequent parabolic profile
yield poor results [Picquenard 1974, pp. 52,531.
These "hit or mic>" endeavors contradict
Bullington's assertion that mode theory offers less
arbitrary parameters, and provides the first

criticism of mode theory. Other objections exist.
Turbulence, which is know to persist, is not
incorporated. Carroll and Ring [1955] admit to this
shortcoming and use it as a possible explanation for
high troposcatter fading rates. Saxton [1956] is
critical of mode theory's nonrigorous approach, and
Katzin [1952a, 1952b] shows the internally reflected
component to be of negligible magnitude. Finally, du
Castel (1966, p. 132] notes an absence of frequency

dependence which contradicts experimentation.
3.5 Other Theories

While turbulence, reflection, and mode
theories represent the brunt of ideas explaining the

troposcatter mechanism, other perceptions have been
published. Most notable is that of Bullington [1963]
82
which incorporates the "best" of all three theories

while eliminating inherent arbitrary factors.6 In
essence, Bullington's approach assumes an exponential
refractivity profile on which turbulent fluctuations
are superimposed. The primary mechanism is that of

partial internal reflection with the statistical
fluctuations explaining both fading and frequency
dependence. Turbulent irregularities are analyzed

through long-term empirical data, and the reflection
coefficient is related to the variance of this
turbulent activity. Christopher and Debroux [1987]
have found Bullington's predictions both suitable and
simple. However, due probably to the use of mode
theory, sources do not mention further work in this
direction.
Attempts were also made to explain beyond the
horizon fields through terrain irregularities. In
fact, Bullington [1947, 1953] pointed out that
surface roughness could, in some instances, produce
enhanced fields, yet troposcatter paths over smooth
paths such as water were known to exist [Jowett,
1958, p. 92]. Thus, pure diffraction theory, as
6 Bullington
considers scatter and reflection
theories as basically the same; however, his use of
the reflection coefficient favors reflection theory.
This hesitancy to differentiate the two is explained
in Section 3.6.
83
championed by Ortusi [19551, never amounted to much.

A similar fate befell de Belatini [1959] who
envisioned turbulent toroidal shapes, a few
kilometers in diameter, which acted as weak lenses.
These atmospheric lenses were said to cause vertical
divergence of transmitted waves which allowed beyond
the horizon propagation.
3.6 Differentiating the Theories
Judging from presented criticisms and the

quantity of supportive sources, two theories,
turbulence and reflection, best seem to explain the
troposcatter propagation mechanism. At this time
it's productive to further distinguish the two,
since, as mentioned earlier, there is some difficulty
differentiating them. Scattering may be thought of
in two ways. Normally, one thinks of scattering as a
random spread of energy from some source through
space. If, however, some coherence is evident, then
the terms reflection, forward scatter, and focusing

may be used [Rice, et al., 1967, vol. 2, p. IV-l;
Bullington, 1959].
Turbulence theory therefore refers to the
random spread of transmitted energy from a single
irregularity or blob. Although Booker and Gordon
originally assumed isotropic scattering, turbulence

84
theory backers admit anistropy. The summation of

power scattered by the numerous blobs within the
common volume are purported to add incoherently.
Conversely, reflection theory is essentially
scattering from a layer consisting of numerous
turbulent irregularities. Due to specular and
diffuse components, power in this case is said to add
in a quasi-coherent fashion. Therefore, while both
theories recognize "scattering" as the propagation
mechanism, the scattering source, individual blobs or
collective layers, is in question. Taken one step
further, the question does not center on turbulence
since turbulence is known to exist; rather, the

question is whether turbulent irregularities exist as
single, randomly disposed entities or arrange
themselves into layers. Finally, following Waterman
[1970] and the above discussion, turbulence theory
will henceforth be termed incoherent scattering while
reflection theory is called quasi-coherent
scattering.
3.7 Modern Theory

Although early work presented two dominant
theories, a few cautious individuals like Saxton
[1956], M. A. Johnson [Discussion on Tropospheric

Propagation-I, 1958], and even Carroll [1952b]
85
acknowledged the possibility of several concurrent

mechanisms. Indeed, both incoherent and quasi-
coherent scattering will occur simultaneously with
dominance related to operating frequency. However,
this now accepted conclusion took many years to prove
and arrived mostly through assorted experiments and
theoretical "adjustments" designed to strengthen
proposed theories.
After development of the refractometer,
incoherent scatter promoters saw the need to alter
Booker and Gordon's original proposal since the scale
of turbulence was larger than initially thought.

Gordon [1955] and Booker and deBettencourt [1955]
incorporated this change in modified theories;
furthermore, the latter provided experimental results
verifying their ideas. Later, however, de Belatini
[1959], Bolgiano [1958], and Rice, et al. [1967,
vol.2] challenged incoherent scatter based on
uncorrelated theoretical versus experimentally

derived wavelength dependencies (discussed in Chapter
4).
Evidence supporting quasi-coherent scatter
includes a sizable effort from Crawford, Hogg, and
Kummer [1959]. Of particular note are the wavelength

dependence results [p. 1092] which, for intermediate
86
size layers, closely match theory to data, while
small layers, likened to single blobs, show poor
correlation. Additionally, with respect to fading

rates (Figure 33), these researchers show rather
smooth signal levels at UHF -- a difficult result for
incoherent scatter to explain, but causing no problem

in quasi-coherence. Yet, as Appendix E demonstrates,
quasi-coherent scatter (a reflection theory) is

insignificant beyond UHF since reflection
coefficients reduce substantially. Further, the
uncertainties of layer size, orientation, and motion

inject considerable guesswork in calculations.
Of scatter concepts as a whole, then,

Bullington asserts that they are
a statistical framework that can be adjusted by

arbitrary parameters to fit almost any consistent
experimental data and, in fact, the values of the
parameters have been changed significantly when
the need to accommodate unexpected experimental
results has arisen [Bullington, 1963, p. 2849].
Additionally, in a superb review of troposcatter
field tests, Cox [1969] provides several conclusions:
1. The most striking feature observed in all the
experiments is the wide range of variation in the
measured parameters for different time periods,
regardless of what the parameters are.
2. [SIimilar measurements were interpreted in

different ways.
3. (D~uring some time periods the atmosphere
appears to be in a state of nearly isotropic,
homogeneous, nonchanging (stationary) turbulence,
87
-t ac V6 GOMC
'° " I ________At
-120 I.
0 5 I0 IS 20 25 30 35 '0
TIME IN 31ECONDS
(a) RAPID FADING
-70
-lOG
" -, i...0...... I I i
-, W,
, ,PJ,,.I,.j 4110,.,
1...
"11 1 1 'lP1fl I I I 'I 1 ' VA, I

1200 5 10 is 20 25 30 35 40
TI m IN SECONOS
(b) SLOW FADING
O -70
' I
,'-I %, , ' I46
,~ , I C VG
TIME IN SECONDS
(C) VERY SLOW FADING
Source: Crawford, A.B., Hogg. D.C., and Kummer

W.H. "Studies in Tropospheric Propagation Beyond the
Horizon." The Bell System Technical Journal, vol.
38, no. 5 (September 1959), p. 1119.
Figure 33. Fading rates for UHF and SHF links over
the same path
88
• . but during other time periods the

turbulence may be anisotropic and/or
inhomogeneous and/or changing with time. During
still other time periods (often at night) the
atmosphere may tend toward a stable state with
the major irregularities being caused by
stratified layers having different
characteristics [Cox, 1969, p. 921].
In short, then, two shortcomings are noted. First,
experimenters are "guilty" of partiality to their

theory, and, second, the source of scattering is of
such unpredictability that complete theoretical
agreement is currently impossible. Ignoring the
first accusation, a closer look at the scattering
source is desirable.
Certainly the atmosphere is unpredictable.
In fact, in his recent (1988) book, Troposcatter
Radio Links, Roda is satisfied in concluding that two

types of discontinuities coexist within a
sufficiently large common volume: turbulent motion
comprising blobs, vortices, or eddies; and laminar

motion composed of feuillets or small reflecting
surfaces [p. 63]. Physical evidence supporting

small-scale turbulent motion is found in voluminous
refractometer profiles conducted over many years.
Furthermore, Hooke and Hardy [1975] have detected
thin, wavelike disturbances produced through gravity

waves, and Crane [1981] adds that newer radar systems
enable direct observation of such layers. Finally,

89
indirect evidence of dual discontinuities occurred as

far back as 1961 in Ortwein, Hopkins and Pohl's
multi-frequency measurements. Undoubtedly, then,
both incoherent scatter and quasi-coherent scatter
explain short wave, beyond the horizon propagation.

What remains is determination of the dominant
propagation mechanism. Waterman [19701 uses
generalized meteorological conditions to
differentiate the two. Sarkar, Dutta and Reddy
[1983] experimentally prove that the spectral slope

(m, see Appendix C), refractivity gradient, or
wavelength can be used as discriminators. Most

sources use frequency (wavelength) as the
differentiator; above a certain operating frequency
incoherent scatter will dominate. Cizne [1981, p.
656] shows the transition wavelength as
X = 2Lovsin(0/2) (3-2)
where Lov, the outer scale of turbulence,

dimensionally describes the largest eddies entering
the inertial subrange of Figure 15. Again, then,

uncertainty enters the picture as Lov will vary
depending on atmospheric conditions. Eklund and
Wickerts' [1968] tests place the transition between 1
90
and 3 GHz, while Monsen, et al. [1983] use a 2 GHz
cutoff.
By way of summary, Figure 34 shows signal

level variations versus frequency and distance for
(1) coherent scatter, (2) diffuse scatter, and (3)

incoherent scatter. (Coherent and diffuse scatter
add to give quasi-coherent scattering.)
db ____ ____ ___ 0____

.10 ___ -10__
0 "0
-4C 0t.
L
-
- .05 ____ ______
100 j00 1000 3000 Mc/s 100 Z00 300 500 1Ickm
Frequency Distance

1966, p. 171.
Figure 34. Signal level variations for: (1) Coherent

scatter, (2) Diffuse scatter, and
(3) Incoherent scatter
Three conclusions may be drawn. First, the
transition frequency resides between 1 and 3 GHz

contingent on acmospheric conditions. Second, both
mechanisms have similar ranges. Finally, as will be
explained in Chapter 5, predictions would benefit

91
from using incoherent scatter received power
calculations as it provides a worst case approach.
Thus, uncertainty induced by the atmosphere
continues although great progress has transpired.
For some, years of theoretical work has been
fruitful; however, many sources state the need for
theoreticians to cease pursuit of a universal
approach, instead focusing on practical measures.
Some may argue that the methods offered in Chapter 5
supply such vehicles; however, until atmospheric
phenomena is completely determinable, it is likely
that some effort will persist.

CHAPTER 4
CHARACTERISTICS OF TROPOSCATTER PROPAGATION
4.1 Introduction
Although a complete understanding of
troposcatter remains elusive, this transmission

technique continues to enjoy world-wide use. Chapter
1 surveyed these applications; an excellent summary
of early troposcatter systems is provided by Gunther
[1966]. This chapter will summarize pertinent

characteristics of troposcatter transmission systems,
emphasizing those parameters directly influenced by

propagation. Specifically, it will highlight
attributes of interest to communications engineers,

providing typical values while describing expected
performance. Equipment will be discussed only when
enhancements, designed to reduce the effects of

degraded propagation, are possible.
4.2 Signal Level
Critical in evaluating the feasibility of a

transmission system to support communications
requirements is the received signal level (RSL).

Appendix B and Chapter 5 analyze RSL calculation.
93
This section looks at the distance and frequency
dependence, fading characteristics, and aperture-to-

medium coupling loss of troposcatter received
signals. Since the RSL is expressed as a statistical

median, this section will not emphasize the
"temporary" effects of diurnal, seascnal and
climatological influences. These events are covered

elsewhere in this treatment; however, their influence
cannot be understated.
4.2.1 Distance Dependence

Figure 35, derived experimentally, more
accurately depicts attenuation for the scatter loss

portion of Figure 2.1 It can be construed,
db db
UO - f-220 Mcs 150 f-3300 Mcs
h-lCOc m h-2OOm
.i 200 20C .
250 250
0 250 N0 750 km 1000 0 250 500 750 1000 k.

Dis'orc!
Source: du Castel, F. ,:opospheric RadIweV

propagatIon Ssyond tie Horizon. New York: ?ergamCn,
1966, p. 64.
Figure 35. Troposcatter loss with distance
IThe term "scatter loss" describes the loss

exceeding that of a free space, line-of-sight link;
it is more commonly used as attenuation relative to
free space.
94
therefore, that troposcatter links exhibit distance

dependence with losses ranging from 150 to 250 dB.
This dependence is difficult to express in simple
mathematical terms since different losses per unit
distance occur for different link ranges [Boithias

and Battesti, 1983, p. 658].2 For example, in 1960
the Joint Technical Advisory Committee (JTAC)

reported a 20 dB/100 statute miles loss for links of
100 to 300 miles, and 11 to 12 dB/100 miles in the
subsequent 200 to 300 miles ["Radio Transmission,"

1960, p. 35]. Most sources generalize to an inverse
power relationship such as the d -4 law (d, distance)
giying about 0.1 dB/km [David and Voge, 1969, p.
148].
Other estimates are available. For links
covering 150 to 300 km, d- 5 ,5 may suffice for

antennas at zero degree elevation; however, an
additional loss following h- 2 , where h is the height
of the common volume, is included for antenna beams
above the horizon [Picquenard, 1974, p. 41].

Surprisingly, Sarkar, Dutta and Reddy [1983] found
distance exponents ranging from -7 to -13 for a 240
km, 2 Ghz link in India (median value not given).
2 Explainable
by multimode and/or simultaneous
incoherent and quasi-coherent scatter propagation.
95
This unexpected variability is attributed to changing

refractivity gradients.
Another method which uses angular distance
(8, the scatter angle) to predict loss was introduced

by Norton, Rice and Vogler (1955]; however, in some
instances (most notably, the influence of terrain)
calculations are unsatisfactory [du Castel, 1966, p.
67]. Regardless, Parl [1979] relates ranges of 8-2
to 9-6, and Flock [1979] provides Figure 36 showing
scatter loss as a function of angular distance for
900 MHz links. A correction factor, 10 log (f2 /900),

110
10 -
90 -
S80-
70 -
0
so-
C 40-
30 f
20O
10 -
0 I I I I 1I 1 1 I! 1 1 1 1I 1 I I ! 1 ! Il l l !
1 2 3 4 5 10 20 30 50 70 100
Scattering angle. 9 (mrad)

Source: Flock, W.L. £1eectromagnetics and the
Environment: Remote Sensing and Telecommunications.
Englewood Cliffs, New Jersey: Prentice-Hall, 1979, p.
190.
Figure 36. Scatter loss as a function of scattering

angle
provides additional loss in dB for any frequency, f2 ,
other than 900 MHz. Finally, note should be made

96
that for short links where smooth earth diffraction

can dominate (multimode), an exponential law may be
appropriate.
The mentioned high loss of troposcatter links
requires exceptionally precise antenna siting to
avoid attenuating obstructions while keeping
elevation angles at an absolute minimum. In
addition, compared to typical line-of-sight systems,

troposcatter links necessarily use 30 to 50 dB more
transmission power, receivers with nearly 12 dB more
sensitivity, and much larger antennas with aperture
diameters of up to 120 feet [Azurza, 1975, pp.
18,20].
4.2.2 Frequency Dependence
Troposcatter links normally operate in the

250 to 6,000 MHz range (see Table 1). This domain is
limited by antenna size on the low end and absorption
and depolarization on the upper, although Crane

[1988] recently showed the feasibility of higher
operating frequencies [Hall and Barclay, 1989, p.

207]. Concerning the effect of frequency on signal
levels, it was shown above (through the correction
factor for angular distance dependence) and on Figure

34 that loss increases (RSL decreases) as frequency
increases. However, as Figure 37 discloses, this

97
30cblrret spre (C)
120 km
0 200/s
MI
- -- 72mI
2 kmk L 20 Mi
0i 410
"-- I__ "I s
C _
S'urce: du Castel, F. rcpoapheric Aadiavav*

r:cpagation 9eyond th. MorIlzon. )4ew York: Per;aw,3A
19E6. P. 6,9.
Figure 37. The effect of frequency over similar

paths
increase is subtle at lower frequencies (below 2.2
GHz in this instance). CCIR Report 238-5 (1986, pp.

377, 381] notes that troposcatter path attenuation
relative to free space generally obeys an fl law up

to 3 GHz. 3 Expressions for higher frequencies were
not found.
3 Similarly,
the ratio of power received
(including scatter loss) to the power one should
receive after free space loss only, Pr/Pfs, is
proportional to f 1 . One should be aware that the
exponent's sign will change when expressing this
ratio as signal level versus attenuation; discussion
uses attenuation relative to free space only.
98
Short-term variations of this general law are

known to exist, however, and Report 238-5 notes such
instances. As an example, two links in France
operating at 460 and 2220 MHz, observed f2 , 1% of the

time and fl, 99% of the time for the longer (325 km)
link. The shorter link (165 km) had a dependence of
fl.5, 1% of the time and fO.50 , 99% of the time. In
addition, Eklund and Wickerts [1968, p. 1068, their

Figure 3] revealed short-term frequency dependences
ranging from f2 to f-1/3 for a 259 km path at 1 and 3
GHz. Figure 38 clarifies short-term frequency

dependences by showing distributions for the
attenuation ratio of two frequencies traversing the
same path; agreement with Eklund and Wickerts is
4
notable.
As mentioned in Chapter 3, conflicts between
theoretical and experimentally derived frequency

dependence values (specifically, the exponent "n" of
fn) were an early source of controversy among
theoreticians. Depending on the turbulence model
used for incoherent scatter theories, attenuation
4 Again,
confusion may develop. In Figure 38
the attenuation ratio used corresponds to differences
in RECEIVED POWER between opposing frequencies.
Cited values by Eklund and Wickerts show ATTENUATION
WITH RESPECT TO FREE SPACE; hence, the exponents of X
on the far right of Figure 38, in Eklund and
Wickerts' terms, would change sign.
99
db
-100 km
-5 A'A'-
270
.4501000 4..
-cis
1 10 50 90 99%
Percentage of cases where the level exceeded the ordinate
1966, p. 70.
Figure 38. Short-term effects of frequency. The

received power ratios 200/2000 Mc/s and
450/4000 Mc/s with their subsequent
attenuation coefficients are shown. In
terms applicable to discussion, a
fr gyency ependence range of about
f- 3 to f5 is depicted.
relative to free space was predicted to obey either

f 113 or fl laws [du Castel, 1966, p. 147; Rice, et
al., 1967, p. IV-7]. Likewise, quasi-coherent

scatter theory predicts exponent (n) values of either
2 or 4, although in the particular case of diffuse
reflection from intermediate size layers a value of 1
is obtainable [Eklund and Wickerts, 1968, p. 1067;
Crawford, Hogg, and Kummer, 1959, p. 1092].

Further, when short-term variations, as discussed
100
above, were noted, evidence mounted against a single,
dominant propagation mechanism.
Once both theories were incorporated and
theoreticians recognized the variability of
turbulence, these short-term fluctuations were
understandable. Consequently, the results of Sarkar,
Dutta, and Reddy's [19831 experiments on a 240 km
link in India are explainable. For a single
frequency (2 GHz), they report exponent values
ranging from -2.1 to -0.9 under mostly turbulent
conditions, and 0.4 to 2.5 during layered situations
(median values were not provided). Regardless, as
one can deduce from Figure 37 (attenuation relative
to free space), the long-term effect of frequency on
signal level will be generally small below 3 GHz, but
becomes important at higher frequencies.
4.2.3 Fading
Probably the most distinguishing parameter of
troposcatter propagation is the extreme fading
observed in the RSL. As discussed earlier,
scattering within the common volume directs portions
of energy toward the receiver; furthermore, the
nature of these scatterers is quite variable. As a
result, received signals, which are the sum of these
numerous reradiations, exhibit variations directly

101
related to the activity within the common volume. As

will be discussed, the RSL is then characterized by
short and long-term fluctuations about a median
value, and, in some instances, frequency selectivity
is observed. Fading is a critical concern for analog
communications systems; however, as discussed later
(Section 4.3), digital systems are troubled more by a
different parameter.
Due to the unpredictability of the
atmospheric phenomena responsible for fading,
expressions for these variations are rooted in

statistics. Common techniques depict cumulative
amplitude distributions over hourly periods while
stating the median value as that level equaled or
exceeded 50% of the time ["Radio Transmission," 1960,
p. 34]. Many RSL prediction techniques, detailed in
Chapter 5, heavily rely on these statistical
distributions. Additionally, while short and long-
term amplitude variations exhibit Rayleigh and log-
normal distributions respectively, the actual
instantaneous signal level is difficult to express
mathematically but may be presented graphically as
described in Panter [1972, pp. 365-369].
Generally speaking, short-term fading rates

will increase with frequency (see, for example,
102
Figure 33); however, this can be reduced by using
antennas with narrow beamwidths. Further, as later
highlighted in Section 4.5, long-term fading will
show diurnal, seasonal, and climatological dependency
[du Castel, 1966, p. 86; Crawford, Hogg, and Kummer,
1959, p. 1118]. Since fading can, at times, severely
impact transmission quality, techniques to improve
performance are essential. Among these, aiversity
reception offers the greatest utility.
4.2.3.1 Short-Term Fading
Persistent short-term fades, much like
scintillation mentioned in Section 2.5.3, are the
"trademark" of troposcatter propagation. Fading

rates of a few per minute at VHF, 1 Hz at UHF, and
several (up to 20) Hz at SHF are not uncommon.
Studies of fading rates use statistical formats by
which the mean fading frequency has been shown to
vary "between 0.044 (exceeded for 99% of the time)
and 4.3 (exceeded for 1% of the time) times its
median value" [Hall, 1979, p. 142]. Additionally,
CCIR Report 238-5 shows that short-term fading rate
distributions vary as the hourly median RSL itself
changes (a long-term fade). In short, when the
median RSL falls above or below its median, fading

103
either decreases or increases respectively (CCIR,

1986, p. 384].
The amplitudes of short-term troposcatter
fades will closely conform to a Raleigh distribution.
This follows logically when one examines the cause of
rapid troposcatter fades. The numerous scatterers
within the common volume each provide individual
contributions which, upon arrival at the receiver,

add constructively or destructively to the RSL. In
this way, atmospheric multipath emerges, and due to
the randomness of the individual phase contributions,
a Raleigh distribution, as provided in Figure 39, is
established (Livingston, 1970, p. 121]. 5 Typically,
- t I.ll
i
I H+ , I111*
C csOal S i)0- 1 0O '

.0 Wu~607 s o " f" 14
Sczret FreetSLo, .a ign te,

SCr' 191 P,
Figure 39. Raleigh distribution showing short-term

troposcatter fading amplitudes
5 0ften,the term phase interference fading is
used instead of short-term or Raleigh fading; this
explains why [Peterson, et al., 1966, p. 51.
104
one can expect short-term fading amplitudes to exceed

the median by 5 dB, 10% of the time, while 90% of the
time the fade will not fall below 8 dB of the median;
30 dB fades are shown to occur less than 0.1% of the
time [Saxton, 1956, p. 588]
4.2.3.2 Long-Term Fading

Short-term fades are superimposed on long-
term or power fades which feature cyclic changes on
the order of hours, days, and even seasons. The
primary cause is variations in the refractivity
gradient which will show high climatological
dependence, while operating frequency provides little

influence (Hall, 1979, pp. 139,11; Livingston, 1970,
p. 120; Bullington, 1955, p. 1179]. As the gradient
c' -nges, bending will essentially redefine the common
volume; hence, the number of scattering sources will
change. In addition, gradient fluctuations will

influence the intensity of individual scatterers
(related by the structure constant, CN 2 , in Appendix
C), and both situations will cause RSL deviations.

An additional cause of power fading is that
propagation mechanisms (multimode, incoherent
scatter, and quasi-coherent scatter) can change
105
[Rice, et al., 1967, vol. 1, p. 10-1].6 Figure 40

provides evidence of mechanism changes on a 350 km,
300 MHz path.
Long-term fading is usually expressed as an
hourly median which is derived from a "distribution
of the hourly medians over a period of at least one
full year" [Panter, 1972, p. 353]. When plotted on a
decibel scale, the hourly median loss (that is, the
value exceeding the long-term median) will display a
log-normal density function [Saxton, 1956, p. 588].
Figure 41 shows this function for a standard
deviation, Y, of 8 dB. Hall [1979, p. 140] notes

that depending on path length and climate, a typically
varies between 4 and 8 dB. This variation is
depicted with respect to angular distance, 0, in
Figure 42. Figure 43 provides the cumulative
distribution for several values of a. Interpreting,
for a standard deviation of 8 dB, the hourly median
value will exceed the long term median by 10 dB 10%
of the time, while 90% of the time it will not drop

below 10 dB of the same long term median.
6 Explainingmechanism changes, gradient

variations will cause bending that may enhance the
diffraction mechanism (multimode). Atmospheric
conditions may foster extensive stratification
(quasi-coherent scatter) or extreme turbulence
(incoherent scatter) [Watterman, 1970].
106
Is I
-90~
-110
is
Source: du Castel, F. Troposph~eric Radiowave

1966, pp. 82, 83.
Figure 40. RSL at different times for a 350 kmn, 300
MHz path. (a) Specular dominant, quasi-
Coherent scatter; (b) and (c) Diffuse
dominant, quasi-coherent scatter; and (d)
Incoherent scatter.
107
003
oo, - -- - i / \
002 -.---
00; OOeb / :t:
0 0 .0 N 0Q q 0D N0
I T T I I I + + + + + + + +
-3(7 -2o- -10 +2. *30,
Scurce: Panter, P.F. Communication Systems

Design: Line-of-Sight and Tropo-Scatter Systems. St.
Louis: McGraw-Hill, 1972, p. 355.
Figure 41. Log-normal density function for F = 8 dB
I) 2 , ,, 50- 60 . -
L Mc raw-Hill, 2 p 3
a I ,
Fiur 42"ion < ,a , rying ith
Standard deviat,'I,
Loui:
M~ra-Hit, I72,p. 358.
Figre
2. tanarddeviation, a3, varying with
angular distance, B
108
Sgg~
g tg 0 !Rca
22
S10
4I IQ I I
II i I 1 1 11l l. V E
1111IL.....J.....A
I
I
TV YFi
,A',
iii_ I_
-
I 1
i I N,
Ii
-HjIl~r.JZ 1I- I 0
0M 2O.~2 a R " O 0
Pelct'd of Time i , otO.:e 's e-ceeded
Sou.ee: 1ante-, P.7. Coa.Municarion Systems

Design: LI.,e-of-Sighe and Tropo-Sact*er Systers. St.
Louis: Mcrsw-MIll, 19?2, p. 357.
Figure 43. Log-normal cumulative distribution

function for various values of a
4.2.3.3 Frequency Selective Fading

For troposcatter FDM-FM schemes (which still
comprise a sizable portion of in-use systems),

performance is also related to frequency selective
fading characteristics. Essentially, the transfer
function of a troposcatter channel, showing amplitude
with respect to frequency, reveals random variations
within the transmitted, RF, bandwidth. Although
close correlation among adjacent frequencies is

typical (the so-called "correlation bandwidth"),
independent fading occurs for frequencies
sufficiently separated.
109
Due to this independent fading, certain

subchannels may fade and intermodulation noise on the
baseband signal results. As a result, while not
cited as critical in analog techniques, digital data
transmitted on individual subchannels via low rate
modems will experience an error rate that cannot be

reduced. Bello [1969a, 1969b] and Bello and Ehrman
[1969] should be referenced for detailed information;

additionally, Section 4.3 further discusses frequency
selective fading.
4.2.3.4 Diversity
While an obvious method to combat the effects
of fading would be to build in large power margins, a
more economical and less detrimental (more power
increases chances for interference) solution is
diversity reception. This well documented technique

provides substantial improvement with respect to
short-term and, to a lesser degree, frequency
selective fading. While several parameters are
available, the diversity principle remains the same:
separate signals over virtually the same path will

exhibit different short-term variations although
their mean levels are nearly identical [Boithias,
1987, p. 208]. The combination of the best of these

distinct signals results in improved overall
110
I I
iI
_ __ L I i 1RH$ 2"__ __ LEVEL ii
IIt I
FADE
Source: Roda, G. Troposcatter Radio Links.

Norwood, Maine: Artech House, 1988, p. 84.
Figure 44. The diversity principle
performance as Figure 44 illustrates.

Clearly, the true value of diversity lies not in
the median gain (of a few decibels), but in the
fact that it flattens out the no-diversity
Raleigh distribution. With diversity, short-term
signal fluctuations are greatly reduced, leaving
only long-term variations to be contended with
[Panter, 1972, pp. 351,352].
Diversity reception is an indispensable tool for
troposcatter communications systems.

Diversity techniques may be physically
visible, termed explicit diversity, or inherent in
the channel itself, termed implicit diversity

(Monsen, 1980, p. 18]. Analysis will focus on
explicit techniques; Section 4.3 discusses implicit
diversity. Single, explicit techniques (often called
dual diversity) consist of the following:
1ii
a. When two or more antennas are sufficiently

separated, space diversity is established and
advantage is taken of the different signals
produced by slightly different common volumes
["Evaluation of FDM and FM Systems," 1983,
pp. 12-11, 12-12] .
b. Frequency diversity uses .different

frequencies which cause scatterers to
reradiate at differing phases ["Evaluation of
FDM and FM Systems," 1983, pp. 12-11,12-12].
c. Polarization diversity, which uses separate
polarizations (horizontal and vertical),
provides limited improvement unless used in
conjunction with space diversity [Monsen, et
al., 1983].
d. Angular diversity incorporates separate

feedhorns on a single antenna; however, "the
upper beam from the antenna generally
receives a slightly lower mean level"
[Boithias, 1988, p. 209].
e. Finally, time diversity may be of benefit if

real-time transmission is not essential
[Hall, 1979, p. 147].

The use of two techniques simultaneously is termed
quad diversity.
112
While each technique offers both benefits and
shortcomings, space and frequency diversity are most
often used. Both techniques add certain equipment
(receivers, diplexers, combiners, waveguide, etc) to
the transmission system; however, space diversity
costs are generally higher due to the additional
antenna(s). Conversely, frequency diversity may be
difficult as it increases spectrum use. Regardless
of whether space or frequency diversity is used,
adequate parameter separation is essential. Formulas
to derive this separation stem from correlation
analysis. 7 Correlation coefficients describing the
degree of independent signal fades with respect to
one another will range from 1 (fully correlated) to 0
(no correlation). While sources disagree over
expected values, acceptable diversity occurs at
values as high as 0.6 [Engineering Considerations,
1972, p. 57; Roda, 1988, p. 851.
From the correlation coefficient, a scale
length, 1 in meters, is derived. Horizontal
separation distances for troposcatter space diversity
antennas are then given by
Ah = 0.36{D 2 + 4(1)2}1/2 (4-1)

7A
superb example is found in Monsen, et, al.
(1983] for a quad diversity (dual space, dual
polarization) tactical military system.
113
where D is the antenna diameter in meters.
Typically, 1, for horizontal space diversity, will be

about 20 m. For frequency diversity,
Af = (1.44f/Od) (D2 + 12)1/2 (4-2)
is used with f, being the operating frequency in MHz;

0, the scattering angle, in mrad; d, path length, in
km; and 1 approximately 15 m in this case. Caution
is advised in using these expressions below 1 GHz
[CCIR, 1986, pp. 381,382]. As a general rule, a
horizontal space separation of 100k or a frequency
separation of 2% or more suffices [Roda, 1988, p.
87].
The diversity improved RSL is realized
through a combiner which can be of several
configurations ["Evaluation of FDM and FM Systems,"
1983; "Tropospheric Scatter," 1979].
a. Simplest is the selector or switching
combiner which selects then switches to the

receiver having the best signal. Therefore,
the selector method is not a combiner per se,
and since it does not combine receiver
outputs, it is the poorest performer of

combining techniques.
114
b. An equal qain combiner adds contributing

signals to provide a single output. Provided
individual signals are all within about 6 dB
of one another, equal gain combiners

outperform selector combiners. However,
under deep fades (about 30% of the time)
equal gain combiners are less effective
unless the poorer performing path's receiver
is squelched.
c. Finally, the optimum ratio, ratio squared,
variable gain, or maximal ratio combiner also
adds individual signals but only after

conditioning circuitry manipulates individual
receiver inputs. The manipulation is in
proportion to the noise contributions of
separate signals. Ultimately, a portion of

the weaker signal(s) is added to the
strongest signal giving the best performance
of all combining techniques.
Figure 45 compares the performance of these
methods for (left) dual diversity and (right) quad
diversity. Generally, dual diversity will yield 4 dB

improvement 50% of the time and 17 dB, 0.1% of the
time. Quad diversity improves performance to about 7
and 27 dB respectively under the same time

115
Is 402, T 44 U
re I Q)4t
"T~: 2 1L
I /A
7 _1
--Tl--K I-I-= r -; t~co

.o
U)
U0 1.0
IN---;TTT-
T7T
Ii II U) 0
- ~.L0
13m4wH2l M oo 73 3.4 10081 of

116
percentages [Panter, 1972, pp. 18,19]. Studies in

diversity and combining techniques enjoy voluminous
treatment; excellent initial sources include Roda
[1988] and Panter [1972].
4.2.4 Aperture-to-Medium Coupling Loss
Completing the components that impact the RSL

is aperture-to-medium coupling loss (coupling loss).
This term, first coined by Booker and deBettencourt
[1955], describes 'he apparent degradation of
theoretical gain for large aperture, troposcatter
antennas. Generally, this degradation occurs only
for antennas with gains exceeding 30 dB, and shows
little distance dependence between 150 to 500 km
[Boithias and Battesti, 1983, p. 662]. Discovery of

this oddity, unique to troposcatter, was
comparatively recent since it becomes appreciable
only at apertures exceeding a hundred wavelengths
which are, currently, physically realizable at
microwave frequencies or higher [Crawford, Hogg, and
Kummer, 1959, p. 1080].

A widely accepted explanation of coupling
loss is provided by Booker and deBettencourt [1955]
with subsequent improvements by Boithias and Battesti
(1983]. Figure 46 helps this discussion; however, a

more realistic view incorporates a common volume full
117
Source: Gordon, W.E. "A Simple Picture of

Trcpospheric Radio Scattering." ITR Transactions on
Communications Systems, vol. CS-4, no.1 (March 1956),
p. 99.
Figure 46. Coupling loss illustration
of scatters with the majority in the lower portion of
the volume. Nevertheless, one would expect
proportional gain at the receiver, R, as antenna
beamwidths decrease from PL to PC- However, as PC
decreases to PS, the antenna gain increase is
countered by a decrease in scatterers. In essence,
less scatterers are illuminated and therefore less
scatterers participate in reradiation. Other
explanations that focus on wavefront distortions due
to antenna proximity to terrain and anisotropic
turbulence have been discussed by Freeman [1987),
Picquenard [1974], and Staras [1957]. Regardless,
the work of Boithias and Battesti [1983], recognized
in CCIR Report 238-5, provides predictions shown
experimentally as quite adequate; however, when
incorporated in overall link predictions, Parl [1979]

118
reveals substantial shortcomings in these empirically

derived methods (see Section 5.3.6).
Sources use several techniques with which to
calculate coupling loss. Two will be presented here.
Boithias and Battesti [1983] conclude that Figure 47

(empirically derived) can be used for 150 to 500 km
links. If, for example, transmit and receive
I. . . . _. . . ..
a. A/
I ____ ______N
Surn o: nttn &gains imfiec ip-. 0 G, (dB)
eirce.: 3o~thlaS, 1. Aedlo W&a. Pro#pAaicn.

St. bout$? M1eraw-Mll, 1 S?, p. 210.
Figure 47. Graphic method for coupling loss

calculation
antennas have a cumulative free space gain of 80 dB

(40 dB for each antenna), the effective gain (Geff)
is about 74 dB representing a 6 dB coupling loss.

Another method provides
119
Ag = 0.07 exp{0.055(Gt + Gr)} (4-3)
with transmit and receive antenna gains, Gt and Gr

respectively, given in dB. For the same example
mentioned above, coupling loss, Ag, computes to 5.7
dB. Expression 4-3 is limited to individual antenna

gains of 55 dB, and both antennas should exhibit
similar gain. Notably, coupling loss decreases above
path lengths of 500 km, and vanishes above 1000 km
[Roda, 1988, pp. 71, 72].
An interesting comparison of practical
importance results from discussion thus far. It is
well known that there is a basic transmission loss
between isotropic antennas in free space (see

Appendix B); further, antennas exhibit a free space
gain depending on aperture size. Finally,
troposcatter coupling loss combines with these

events, all of which show direct frequency
relationships, to produce Figure 48. In this manner,

optimum antenna selection for a chosen operating
frequency is possible by weighing relative RSL
benefits with the costs incurred by larger apertures

(Boithias, 1987, pp. 212,213; Hall, 1979, pp.
138,1391.
120
+10 '--
1
--
IOM
-0.1 0.2 0.5 1 2 5 10
Frequency (GHz)
Source: Boithias, L. Radio Wave Propagation.
St. Louis: McGraw-Hill, 1987, p. 212.
Figure 48. Relative RSL comparison for antennas of

different diameters for troposcatter
links. Comparative reference set at 0
dB for two, 10 m antennas at 1 GHz.
4.3 Multipath Delay Spread

Multipath, which has been mentioned briefly
several times, is the cause of many detrimental

effects. These include fast fading or scintillation,
bandwidth reduction (as will be discussed in Section
4.4), and delay spread. Multipath describes the
slightly delayed arrival of portions of the
transmitted wave. Therefore, individual rays of
varying amplitude and phase arrive simultaneously and
contribute in a positive manner, adding to the vector

121
sum of all arriving rays, or detrimentally, depending

on phase relationships. They can be produced by
partial reflection from refractivity gradient

discontinuities, specular or diffuse ground
reflections, and, always present in troposcatter,
scattering from a number of randomly disposed
inhomogeneities within the common volume (atmospheric
multipath) [Freeman, 1987, p. 18]. In short,
the power scattered at different positions in

space arrive at the receiving antenna at
different times, due to the various path lengths
involved. This can be detected as the broadening
of a transmitting pulse [Gjessing and McCormick,
1974, p. 1330].
This broadening, termed delay spread, can
catastrophically impact high speed digital

transmission.
Delay spread, sometimes called multipath-
spread, differential-delay, or time-smear, measures

"the duration of the channel impulse response, or .
the maximum delay difference between the first and
last significant path" [Baghdady, 1970, p. 3]. It
can be affected by frequency, antenna diameter, path

length, and scattering angle [Monsen, et al., 1983,
p. 6-11]. Representative values range from 50 to 370

ns although 1 us widths have been experimentally
shown to occur [Lemmon, 1989, p. 30-6]. Figure 49

depicts typical delay scattering regions, note how
122
04,
0 00
Ai l r ;
r4) I On
1"4 2 0
0 c
C
>-4
I-cJ --
123
different portions of the common volume (shaded

region) contribute different path delays. Notably,
spread values exhibit seasonal variations due mainly
to changes in the structure constant, CN2 (see
Appendix C) [Monsen, et al., 1983, p. 6-11].
Measurement techniques for delay spread are discussed
by Lemmon [1989].
The formation of delay spread in a
troposcatter link is simplistically depicted in
Figure 50. A short transmitted pulse (impulse)
?)
A~krpeO rrcm: Xoda. Z. frocpaie,r Radlo 1.1ks.

Nc.-ead. g~A:.ch )tc*4s.194 p. IA.; 4o~ gr A.,
T'@yte@ht.P.J. * Ler:. G., LSVit1%St*fl.
C.C. and
WSeLA~j9"2ir. P.. *A.M/~--C tr.tia' LiA Snf.fecr-4
d.fCrd , '48., 11 Jan :S95. 5.

$
Figure 50. Development of the delay power spectrum

124
scattered by the common volume will be received as
gn(T) (the power impulse response) with a width much

greater than that transmitted; amplitude will also be
distorted. A subsequent pulse, transmitted shortly

after the first, will exhibit a similar shape since
high speed transmission rates greatly exceed short-
term fading rates. However, additional received

pulses will eventually change shape and the average,
Q(T), of all impulse responses establishes the delay

power spectrum.
The spectrum width is commonly expressed as
twice the rms spread, or the 20 spread, of the delay

power spectrum. Typical values were given
previously; theoretical values incorporate formidable
expressions rooted in the work of Bello [1969a] and
well summarized by Freeman [1987] (see Section
5.3.10).8 Most importantly, however, the impact of
delay spread is essentially the widening of received
pulses. Thus, components which arrive late will

spill over into adjacent time slots creating
intersymbol interference (ISI), and, ultimately,
degrading bit error rate (BER) [Freeman, 1987, p.
216]. It is significant to note that the delay power
8 The
interested reader would do well to
reference IEEE Transactions on Communications
Technology, vol. COM-17, no. 2 (April 1969) in which
much of Bello's work is conveniently packaged.
125
spectrum (a time domain representation) relates to
the correlation bandwidth (a frequency domain

expression) through Fourier transformation [Bello,
1969a, p. 130]. While the correlation bandwidth
limits analog capabilities (see Sections 4.2.3.3 and
4.4), the delay power spectrum can be exploited by
digital schemes in a unique way.

In Section 4.2.3.3 it was shown that FDM-FM
systems which have RF bandwidths that exceed the

medium's correlation bandwidth would experience
frequency selective fades on certain subchannels.
This can now be explained as due to multipath since

multipath establishes the delay power spectrum which
is known to be the Fourier transform of expressions
for the correlation bandwidth. In this way,

multipath is thought of as detrimental to
troposcatter communications systems. However, if one

were to take advantage of multipath in which
different portions of the frequency band fade

independently, a form of diversity, implicit
frequency diversity, emerges. Similarly, since
identical information will arrive over a spread in
time, implicit time diversity is also possible

[Monsen, 1980, p. 18; Roda, 1988, p. 101]. Hence,
although multipath can induce severe problems in both
126
analog and digital transmission, it can, through

rather clever techniques, be used to actually improve
digital transmission.
A variety of techniques are available through
which implicit diversity as well as ISI reduction is
realized. Roda [1988, pp. 100-103] lists several.
Additionally, Monsen [1980] provides a general
description of probably the most widely used

troposcatter technique, adaptive processing. The
basic principle behind adaptive processing is that
successive pulses differ very little from one another
due to the relatively slow, short-term fading rates
(slow with respect to transmitted data rates) of the
medium. Incoming pulses are compared with previous
ones through adaptive equalizers, correlation

filters, or maximum likelihood detectors and pulse
shapes are corrected as the processor "adapts" to
fade induce distortions. The net effect is to

minimize ISI while adding diversity improvement
(implicit frequency and time). Roda [1988, p. 185]
mentions that adaptive processors are available that
can support up to 12 Mbps, exceeding typical
correlation bandwidths; furthermore, although high in
cost, the improvement is well worth the expense.
127
A preeminent example of the benefits gained
through adaptive processing is given by the Raytheon

distortion adaptive receiver (DAR) used on the
transportable military AN/TRC-170 troposcatter
terminal. Two time-gated information-bearing
waveforms are transmitted on separate frequencies
(subcarriers). Through time interleaving, the

transmitted frequency is of nearly constant amplitude
allowing 3 dB improvement over traditional, single-
gated schemes. Distorted received pulse shapes are
reformatted through the adaptive process described
above and ISI can be circumvented up to a certain

delay spread value, contingent on data rates.
Finally, implicit diversity improvement is provided
as multipath delay spread worsens (again, up to
certain delay spread values). On Figure 51 three
delay spread profiles are provided. The AN/TRC-170,
operating between 4.4 and 5 GHz, 2 kW, quad explicit
diversity, 1.024 Mbps, with 9.5 ft antennas will
support total transmission losses of 237.4 dB for
profile (A), 240 dB for (B), and 241.1 dB for (C)
(Freeman, 1987, p. 218]. Discussion of the
intricacies of the Raytheon DAR modem is provided by
Monsen, et al. [1983].
128
A
0-
30
oe~a) ,n Psec
Source: Freeman, R.L. Padio System Design for

Telecomunications (1-200 GMz). New York: John Wiley
& Sons, 1987, p. 218.
Figure 51. Delay spread profiles. (A) 65 ns,

(B) 190 ns, (C) 380 ns
Current trends in communications systems

probably dictate greater emphasis on digital
performance. However, for digital troposcatter,

sources are limited, consisting mainly of military
sponsored research. Additionally, the in-place
commercial communications infrastructure consists

mainly of line-of-sight networks rapidly being
augmented by fiber technology. For this reason,

research material in troposcatter will probably never
provide the multitude of documents typifying early
work. Furthermore, with respect to troposcatter
propagation (the thrust of this thesis), dated
material maintains significance, regardless of the

129
transmission scheme. Simply, no matter how
sophisticated terminal peripherals may be, systems
will always require RSLs above threshold to operate.
Regardless, the impact of digital equipment
peripherals, such as adaptive processors, has
rejuvenated troposcatter systems through exploitation
of otherwise detrimental propagation effects. The

subsequent discussion on bandwidth will further
amplify this assertion.
4.4 Bandwidth
Another parameter of interest is the capacity
or bandwidth of troposcatter systems. Generally
speaking, the transmittable bandwidth is related to
climatology and distance [CCIR, 1986, p. 385].

Additionally, some original work by Booker and
deBettencourt (1955] identified improved bandwidth
performance with higher gain antennas. Figure 52

provides a broad look at bandwidth availability as a
function of distance and free space antenna gain.
Within several hundred kilometers, bandwidths of 3 to
4 MHz are typical; however, variability is noted.
Although limited in capacity, analog troposcatter
systems can provide multichannel operations, up to
120 commercial grade voice subchannels. Further, a
demonstration of the ability of troposcatter to

130
a Ro 0 60 goo
1979,6
p. 45
Figure 52. Useable bandwidth as a function of

distance and antenna gain
support television transmission was provided as early

as 1953 by Tidd [1955). Adaptive processing has, on
the other hand, diminished the limitations of
bandwidth on digital troposcatter schemes allowing
data rates exceeding 10 Mbps.
Explaining the comparatively small capacity
of an analog troposcatter channel, one must again

look at the propagation mechanism. Atmospheric
multipath, produced by the random disposition of
common volume scatterers, causes time variations in
portions of the received signal. This, as mentioned,

131
introduces frequency selective fading in the
frequency domain and delay spread in the time domain.
Within the correlation bandwidth, frequency selective
fading is not a concern since "all the frequency
components of a signal would maintain substantially
fixed relative amplitude and phase relationships and
fluctuate practically in step" [Baghdady, 1970, p.

31. This correlation bandwidth is an attribute of
the propagation mechanism, and, as such, it is
limited by the dimensions of the common volume.
If the transmitted bandwidth exceeds the
correlation bandwidth, intermodulation noise on the
analog baseband signal increases, decreasing system
performance [Panter, 1972, p. 488]. Likewise, delay

spread increases, and unless adaptive processing is
used, ISI increases on digital systems. Either the
correlation bandwidth or delay spread can therefore
be used to determine bandwidth. Additionally, due to
the volatile nature of the scatterers within the
common volume, it should be evident that bandwidth
will exhibit variability. This gives rise to the
notion of an instantaneous bandwidth, and, therefore,
a distribution of bandwidths as depicted in Figure 53
(du Castel, 1966, p. 72].

132
0
0 2
8_ _ 5
2300 Mc15
300,km 10
12
1 10 50 .90 99%

Propagaticn Beyond the Horizon. New York: Pergamon,
1966, p. 71.
Figure 53. Distribution of bandwidths. (2.3 GHz,

300 km, 20 m Antennas)
Calculating bandwidth is a simple matter as

Hall [1979, p. 144] and Picquenard [1974, p. 43]
show. If At represents the difference in time
between the arrival of the longest and shortest

reradiated rays, the bandwidth is 1/At. One can use
the 2a figure of the delay power spectrum as At and
will get a worst case capacity of 2.7 MHz for 370 ns
and a best case bandwidth of 20 MHz for 50 ns.
Modern test equipment readily provides 2a values;
however, earlier techniques, such as that used by
Crawford, Hogg, and Kummer [1959], relied on

frequency-sweeps that provided frequency domain
representations. Additionally, Gordon [1955]
furnishes a way to theoretically calculate bandwidth
through trigonometry.
133
Mentioned distance and antenna gain
dependence also merit treatment. Distance dependence

is easily explained: as path length increases, common
volume dimensions increase, and, therefore, so does
At (bandwidth decreases). For antenna gain, large

beamwidth (low gain) antennas illuminate a large
volume with numerous scatterers. As beamwidths
decrease, less scatters participate and At therefore
decreases (bandwidth increases). However, as Section
4.2.4 discusses, coupling loss develops for antennas
with 30 dB or more gain; therefore, gain is
sacrificed for bandwidth and the engineer must
balance the two for an optimum mix [Booker and

DeBettencourt, 1955, p. 289; Boithias, 1987, p. 213].
Concluding, it is interesting to note that
bandwidth, once isolated as the major impediment to
troposcatter applications, has diminished in
significance for digital systems due to adaptive
processing (see Section 4.3). Through adaptive
equalization and implicit diversity, competitive data
rates are available. The primary limiter now is
power: energy-per-bit to noise power density ratios
(Eb/No), critical in BER performance (see Section

5.3.11), is directly influenced by transmitter power
134
and this is physically limited by klystron

capabilities [Monsen, 1980].
4.5 Diurnal, Seasonal, Climatic, and Meteorological

Variables
This final section focuses on the conditions
or variables under which propagation takes place and

the impact of those variables on several of the
previously mentioned parameters. It may help to
define the variables to be discussed:
a. Diurnal variations describe daily changes in
link parameters.
b. Seasonal variations describe seasonal (e.g.
summer to winter) changes in link parameters.

c. Climatic variations describe the regional
differences in parameter values. These are

provided in terms of geographical zones or
latitude. Appendix F describes the climatic
zones considered under troposcatter

propagation.
d. Meteorological variations describe weather's
influence on link parameters.
Close relationships characterize several of these
variations; for this reason, climatic variations are
not discussed separately but are included with

135
diurnal and seasonal variations. Discussion will

clarify this treatment.
Regardless of the variable considered, the
cause of parameter variation is rooted in atmospheric

phenomena. In general, three cases are cited:
a. Propagation mechanisms may change. Depending
on conditions, scatterers within the common

volume may resemble small, random, turbulent
eddies that produce incoherent scatter
thereby increasing short-term fades while
9
possibly lowering the median RSL.
Conversely, under stratification, scatterers
may arrange themselves into layers that
promote quasi-coherent scatter which provides

higher median RSLs and less scintillation.
b. The entire path may be influenced by

refractivity gradient changes which will,
through wave bending, alter the position of

the common volume. Since there are more
scatterers at lower heights the median RSL

may improve or degrade. Additionally, on
shorter paths, multimode propagation
9 From
Figures 33 and 40 it is evident that
incoherent scatter (dominant above about 2 GHz) fades
more rapidly. Similarly, from Figure 34 and 40 it is
evident that incoherent scatter provides lower RSLs
below about 2 GHz.
136
(Appendix A) may occur or diffraction may

dominate. Finally, the intensity of the
scatterers within the common volume, as
measured by the structure constant (CN2 , see
Appendix C), will change as the gradient
changes and RSL is effected.

c. Substantial variations in the refractivity
gradient may introduce anomalous propagation;

while enhancements are possible, normally,
RSL will degrade and short-term fading will
increase.
Discussion will show how variables can instigate
these events during propagation. For consistency

(and for ease of analysis), RSL deviations bear the
brunt of parameter analysis. Literature more

commonly uses long-term fading, transmission loss, or
attenuation from a median value in similar studies.
4.b.1 Diurnal Variations

Daily variations in both RSL and short-term
fading rates typify troposcatter links. Generally

speaking, the RSL will be lowest in the afternoon and
highest in the early morning. For desert climates,
hourly median differences of 20 dB between these
extremes are common. Temperate zones can show a 10

dB swing; however, equatorial regions exhibit nearly
137
insignificant changes [Bothias and Battesti, 1983, p.
658; CCIR, 1986, p. 3771.

An example of diurnal RSL variation is
provided in Figure 54. This link (2 GHz, 240 km in
- 1390
CLUAIJULATIVE
DISTRIBUTION
1
.60
." REFRACTIVIY) 37
o DIURNALPA TTERN
OF SIGNAL 30
00 04 08 12 Is 20
IST( LT)
0 I 10 20 50 80 90 99
PERCENTAGE PROBABILITY
Source: Sarkar, S.K., Dutta, H.N., and Reddy,
B.M. "Development of a Prediction Technique for
Tr:po-Path Loss from Observed Transhorizon
Propagation Characteristics over India." 7hird
international Conference on Antennas and Propagation
(IC..P 83), Part 2: Propagation, institution of
E-ectrical Engineers, London (1983), p. 231.
Figure 54. Diurnal RSL and surface refractivity, Ns,

variation. (2 GHz, 240 km link in India)
India) shows swings of 30 dB. Fading rates are also

known to vary diurnally, and this is illustrated in
Figure 55. While RSLs are lower in the afternoon,

fading rates are higher as expected for incoherent
scatter dominance (see case fa) above) [Sarkar,
Dutta, and Reddy, 1983].

138
26
24
22 I
20 1
:6
614
1S2
101
4
00 I .I I I I
03 04 06 12 16 20 CIO
TIME. :ST
Source: Sarkar. S.K., D,.tta, M .., and F*edy,
3... "-*velcpmeeF cf a Fredicticn Techrqnje tcr
Tropc-1ath Lcs frcm Observed Transhcr!zon
Frcapaticn Characterstics cver India." Third
2nteruatirnaI Conferente cn Antennas and Fr:pagaticn
(ICP 83), ;!rt 2: ProcFa9ticn, InS ttin of
Etectrical Engineers, L.7dn (1983), p. 232.
Figure 55. Diurnal fading rates. (2 GHz, 240 km

link in India)
The causes of diurnal RSL and short-term

fading deviations are explained mostly through case
(a), with some contributions by case (b). Probably

of greatest significance, as mentioned in Section
2.4.1, daily solar heating increases turbulent
activity with convection currents bringing turbulent

eddies to higher levels; therefore, incoherent
scatter occurs. Nighttime movement is obviously

reduced and this would increase the likelihood of
quasi-coherent scatter. Case (b) is also likely to

occur on a diurnal basis. It was demonstrated in
Section 2.3.2.1 that daily K-factor (therefore,

139
refractivity gradient) variations are common.

Therefore, the combination of cases (a) and (b),
brought on by routine atmospheric phenomena (some of

which are cited here), contribute to daily RSL and
fading variations.
4.5.2 Seasonal Variations

As with diurnal variations, seasonal
variations in RSL vary with climate. In temperate
zones, RSLs are lower by as much as 15 dB for 150 to
250 km paths in the winter than in the summer.
(Notably, diurnal RSL variations are also lower in
the winter [CCIR, 1986, p. 377].) The opposite is
true for desert regions where RSLs are lowest during
the hottest season. Tropical zones exhibit minimum
RSLs during the rainy season, whereas monsoon
climates provide lowest RSLs during the rainy and dry

seasons. In all cases, the influence of seasonal
variation diminishes with path distance [Boithias and
Battesti, 1983, p. 658]. For purposes of analysis,
seasonal variations are divided among eight time

blocks as provided in Table 4; time block 2 is most
often used as it provides worst case figures for
temperate climates [Freeman, 1975, p. 267].
140
Table 4. Seasonal Variation Time Blocks

Number Month Hours
1 Nov-Apr 0600-1300
2 Nov-Apr 1300-1800
3 Nov-Apr 1800-2400
4 May-Oct 0600-1300
5 May-Oct 1300-1800
6 May-Oct 1800-2400
7 May-Oct 0000-0600
8 Nov-Apr 0000-0600
Source: Freeman, R.L. Telecommunication Transmission
Handbook. New York: John Wiley & Sons, 1975, p. 267.
An example of seasonal RSL variation for a

4.1 GHz, 171 mile link in the northeastern portion of
the U.S. is given in Figure 56. Also depicted is a
Q.O L /JANTENNA TN
>_.
0-10
Lit . i.- . -
> 320
IIZiK! -J I
EPERATURE
-
-
''~
120w
JAN FES MAR APR, MAY JUN JUL. AUG SEP OCT NOV DEC. JAN
1956 1957
Source: Crawford, A.S., Mog. D.C., and Kumm er

W.H. "Studies in Tropospheric Propagation Beyond th~e
Morizon." The Bell SysteU Techntical Journal, vol.
38, no. 5 (September 1959). p. 1CSS.
Figure 56. Seasonal RSL, temperature, and surface
refractivity variations. (4.1 GHz, 171
mile link in northeastern U.S.)
141
close correlation between RSL and surface

refractivity, Ns. This analysis may be misleading
since changes in the refractivity gradient along the
entire path and within the common volume, as in case
(b), explain seasonal variations10

To clarify, one may analyze any region where
certain months display higher humidity and
temperature. Humidity (the measure of water vapor
content) and temperature variations will effect
refractivity as noted in expression (2-3). Water
vapor variations are by far the largest contributor;
therefore, during periods of high humidity

refractivity will increase. Evidently, this increase
(or a subsequent seasonal decrease) is not
proportionally distributed with height and,
therefore, the refractivity gradient, AN/Ah, differs

from its median (-40 N units/km for the first
kilometer of height). Super-refraction will occur
(displacing the common volume downward where more
scatterers are found), and scatterer intensity as
10 From Figures 54 and 55, one might conclude

that surface refractivity is a reasonable indicator
of RSL deviations. In numerous situations, it is
[Bean and Dutton, 1966]. However, one must remember
that troposcatter propagation is influenced by
gradient changes along the path and within the common
volume; therefore, the refractivity gradient and
structure constant are the best measurable indicators
of RSL deviations.
142
well as population will increase. The cumulative

result of super-refraction and scatterer variations
is improved RSLs during more humid periods. CCIR
Report 563-3 [1986] provides evidence of seasonal
gradient variation in Figures 5 through 8 of that
report. In short, seasonal variations within the
atmosphere alter the refractivity-gradient generating
a case (b) situation.

Finally, the distance dependence (less
seasonal influence with longer path distance) is

explained through the displacement of the common
volume to greater heights as path distance increases
[Boithias, 1987, p. 208]. Although gradient changes

may reduce this height, less scatterers are present
than for a comparable shorter link. In effect, a

change in RSL will occur for both links (in this
case, an increase) but will be proportionally less

for the longer of the two.
4.5.3 Meteorological Variations

Research provided little material addressing
the effects of weather on troposcatter performance.

Most efforts concern the diurnal and seasonal
deviations mentioned above. This is predictable;

theoreticians and engineers are more concerned with
143
long-term performance. However, users express
concern only when the system fails to support

requirements. Assuming proper equipment maintenance,
the engineer must be aware of the weather conditions
sponsoring degradations, and, somehow, he must
explain this to the user. This section will attempt
to help the engineer in this unenviable position.
Discussion thus far has described typical
diurnal and seasonal climatic variations. The

introduction of temporary weather effects may,
however, radically alter traditional performance.
While there are a variety of weather phenomena,
analysis boils down to refractivity; specifically,

anomalous propagation. Section 2.3.2.2 discussed
anomalous propagation, a brief summary follows:

a. Subrefraction - When the refractivity
gradient exceeds the norm (greater than -40 N
units/km) wave bending is decreased.

b. Super-refraction - When the refractivity
gradient is less than the norm (less than -40
N units/km) wave bending is increased.
c. Ducting - As gradient values less than -157 N

units/km are encountered, surface or elevated
ducting occurs.
144
Any one of these events will signal the onset of case
(c) degradations (or possibly enhancements, but these
will be ignored). The key is to determine the
weather event(s) that produces these anomalies. Such
events were also briefly discussed in Section
2.3.2.2. This treatment will include that

11
information with some amplification.
4.5.3.1 Subrefraction
Under Type A conditions (surface temperatures
exceeding 3 0 0 C and relative humidity below 40%),
convection currents will raise layers of high
humidity and, therefore, water vapor concentrates at
increased heights. This most ofter. occurs in the
late afternoon. Type B conditions (surface
temperature between 100 and 300 C and humidity
exceeding 60%) will occur mostly in the early morning
"in constant trade wind and sea breeze areas where
differential heating Jr land and sea results in
advection of air which is warmer and more humid than
the normal surface layer" [Brauburger, 1979, p. 27].
liThe conditions producing super-refraction

can, depending on severity, invoke surface ducts.
Additionally, the simultapeous occurrence of sub- and
super-refractive conditions may create elevated
ducts. In both instances, propagation can deviate
substantially and degrade troposcatter performance.
Section 2.3.2.2 should be referenced for further
information on ducting.
145
It's apparent that Type B conditions also occur when
weather fronts are present. The net effect of Type A
or B conditions is an increase in density, primarily
water vapor content, with height producing an
increase in the refractivity gradient. This action

will redefine the common volume to higher levels
where scatterer population decreases [Brauburger,
1979].
4.5.3.2 Super-Refraction
Figure 8 indicates that within the
troposphere temperature normally decreases with
height. If, however, "cooler air either passes over
or develops over a warm body of water or over warm
moist soil" a temperature inversion will occur
[Brauburger, 1979, p. 24]. Hence, temperature rises
with altitude producing large density decreases with
height. This can occur at night as the water or
ground cools faster than the air, or can occur
anytime a cold front moves into an area. Another

weather phenomenon, subsidence, develops when the dry
a r mass of a high pressure system slowly settles
from high altitude. "This air mass is warmed by
adiabatic compression, over-laying and trapping a
cooler air mass saturated by surface moisture"
[Brauburger, 1979, p. 25]. In both instances,

146
density decreases with height much more than normal.

This will decrease the refractivity gradient, more
bending will occur, and troposcatter propagation can
be severely degraded. Clues toward recognizing
super-refractive conditions include cold front
activity and low altitude layering of smoke, haze, or

mist [Brauburger, 1979].
From a more general perspective, it should be
apparent that weather fronts can provide poor
propagation conditions. Albrecht [1967] reports RSL
decreases by as much as 24 dB with the passage of
cold fronts (900 MHz, 278 km link). Occlusions,
formed as a cold front overtakes a warm front,
provide similar effects, Additionally, thunderstorms
within the common volume produce results much like a
cold front. Since rain attenuation at troposcatter

frequencies is, for the most part, negligible,
thunderstorm degradation is better described by case
(b): as the storm moves through the common volume,
gradient changes will alter the intensity of the
scatterers. On the other hand, compared to cold

fronts, warm fronts have a less pronounced influence.
Speculatively, if a link is located on a slow moving

or stationary front, prolonged degradation is likely.
147
Summarizing, a variety of weather effects
influence troposcatter propagation; however, the key

determinant is refractivity gradient changes. Thus,
determining meteorological influences on troposcatter

propagation is difficult. Accurate real-time
analysis is hampered by the need to accumulate and

correlate RSL and gradient data. Additionally, using
visual or generalized indicators such as haze layers,
fronts, costal breezes, and subsidence is, at best,

guesswork. Finally, while surface refractivity
measurements are easily obtained and known to
correlate with RSL deviations, noted limitations (du

Castel [1966, p. 80] and Hall [1979, p. 180])
introduce speculation that prohibits complete
reliability (see also footnote 10, this chapter).
The accumulation of more empirical studies, termed
radiometeorology, is neatly summarized in CCIR Report
563-3; however, the focus is rather general, and not
specific to troposcatter propagation.
The author has, on several occasions,

experienced the frustrations of weather oriented
degradations on tactical military systems. Links
that normally performed well were degraded for

periods of several days, and, at least from a
military perspective, this is unacceptable. In this
148
situation, analysis is worthless: operations were
conducted regardless of the predicament faced by
communications personnel; furthermore, in austere
locations, meteorological data giving gradient
behavior is unobtainable. There is a need for the
development of systems whose attributes mirror
troposcatter (comparatively long range, high quality,
multichannel, and cost effective). Although the

AN/TRC-170 mentioned earlier offers improved
capability, it cannot avoid the effects of extreme
super-refraction or ducting. Until development of
such a system, the user would be well advised to
"wait until the weather changes."

CHAPTER 5
TROPOSCATTER PREDICTION TECHNIQUES
5.1 Introduction
Armed with the knowledge provided in
preceding chapters, one might suspect that predicting
troposcatter link performance is difficult. Due to

atmospheric complexity, troposcatter propagation
offers a variety of deviations through multipath,
diurnal, seasonal, climatic, and meteorological
conditions. On top of this, it is known that the
mechanism for propagation can include a mix of
diffraction, incoherent scatter, and quasi-coherent
scatter depending on such variables as path distance,

operating frequency, and atmospherics. h bwever,
certain situations dictate the use of troposcatter.
Thus, the engineer must have a tool for analyzing the
feasibility of the proposed link to support

requirements.
Fortunately, there is a simple and quite
reliable tool available. Unfortunately, the single
variable needed to use this tool, the refractivity
gradient within the common volume, is difficult to

150
obtain. (This technique, the Radiometeorological

method, will be discussed later.) Therefore, an
assortment of methods have been proposed, most of
which rely on empirical, statistically oriented
relationships established through a rather massive

accumulation of data [Hall, 1979, p. 135].
The shortcomings of these experimentally
derived techniques are readily apparent: no single

method can encompass the multitude of variables
impacting propagation in all areas of the world.

Recognizing this problem, several techniques include
uncertainty parameters which project the confidence
which one can place in the prediction. Additionally,
with the advantage of small, powerful computers one
can provide these predictions rather rapidly, but,

naturally, the product is only as good as the method
employed and the data input. In short, the

prediction must be viewed as an estimate, not
absolute in accuracy since the result is rooted in
statistical analysis [Military Handbook, 1977, p. 4-
321].
5.2 Prediction Preparation

Assuming the selected technique provides
suitable accuracy, one must be adequately prepared to

begin the prediction. This first step, the
151
accumulation of data, is the most time consuming and
critical step in troposcatter propagation prediction.
An initial desktop analysis supplemented by an actual
site survey provides inputs to include great circle
calculations, path profiles, and, most importantly,
antenna elevation angles. In addition, such

variables as antenna height, sources of interference,
location of the common volume, and typical
meteorological conditions supplement the list of
necessary data. In this treatment, it is assumed
that this necessary Lirst step is completed. An
excellent document describing data accumulation is
Military Handbook, Facility Design for Tropospheric
Scatter [19771.
5.3 Prediction Techniques
The "grandaddy" of troposcatter prediction
tools is Transmission Loss Predictions for
Tropospheric Communication Circuits (Rice, P.L., et
al. (1967]) commonly referred to as NBS Tech Note 101
(TN 101). This document, first published in 1965,
accumulated data from many years and consolidated
related studies (most notably Barsis, Norton and Rice
(1961, 1962] and Gordon [1955]). Additionally,

152
Scatter [1977] uses the same technique provided in TN

101.
CCIR Report 238-5 offers three techniques.
Method I is a simplified version of TN 101. Method
II reflects the work of Boithias and Battesti [1983].
Finally, the Chinese method appears in Annex II of
the CCIR report. Notably, the earlier mentioned

Radiometeorological method appears in Report 238-4
but not in the more recent, 238-5 version (it was
moved to Report 718-2).
A newer technique, based on the ideas of Parl
[1979] and Monsen, et al. [1983], still use portions
of TN 101. However, whereas TN 101 is rather limited

in application (explained in Section 5.4), this more
contemporary approach incorporates variability

through the spectrum function. Additionally,
aperture-to-medium coupling loss is calculated in
more precise terms. Finally, calculations emphasize

the incoherent scatter mechanism as this provides
worst case analysis. The brunt of this effort was
funded through a military contract and focuses on

military systems; however, the basic approach is
applicable for all troposcatter systems. Although
most work was done through contract with SIGNATRON,
Inc., this tool is termed the Parl method in

153
deference to Dr. Steen A. Parl's [Parl, 1979] earlier
efforts; note also that Parl co-authored the 1983

SIGNATRON document [Monsen, et al., 1983].
Additionally, the MITRE Corporation specialized the

Parl method for the AN/TRC-170 military tactical
LOS/troposcatter terminal. Hence the MITRE method is
limited mostly by frequency (4.4 to 5 GHz) and
equipment (the DAR modem).

A collection of other methods are available.
These include the Yeh, Rider, Collins, Bullington,
MIT, RCA, Page, ITT, and Longley-Rice models. The
first four will be discussed. The remainder (except

Longley-Rice which is a computerized version of TN
101) are briefly considered by du Castel [1966, pp.
209-213].
Before selecting a method, one must be aware

that some techniques may exclude several important
factors. These include atmospheric absorption,
intermodulation noise, delay spread, aperture-to-
medium coupling loss, diversity improvement, short-
term fading, and diurnal and seasonal effects. As an

example, the brief description of the system equation
in Appendix B provides the essential ingredients of
TN 101. After some scrutiny, the reader will observe
the inclusion of coupling loss and absorption;

154
however, other factors such as diversity improvement

and delay spread (if digital) are absent and can be
overlooked. This is not a criticism of TN 101 since
this technique is one of the most thorough offered.
It is imperative, however, that the engineer be aware
of the applicability of the selected method, making

adjustments as needed. As a final word of caution,
other techniques, not listed here, exist and perform
well in specific situations; in short, beware of site
specific tools.
The remainder of this section provides a
brief look at most of the prediction methods listed

above. It will highlight the concepts incorporated
in several of these techniques and, for others, it
will provide the actual calculation. This approach
is taken for brevity and simplicity: TN 101 and the
Parl methods are complex and rather intimidating
whereas the others are easily understood.
5.3.1 TN 101
Using data accumulated in the 1950's and

early 1960's, the authors of TN 101 developed
empirical models "for the estimation of the
cumulative distribution of annual hourly median
transmission loss as a function of frequency and path
geometry" [Crane, 1988, Appendix A, p. 3]. Critical

155
in this approach is the description of hourly, long-
term median RSL which, as shown through long-term

fading in Section 4.2.3.2, displays a log-normal
power density function. A standard deviation of
nearly 6 dB is noted, and this deviation gives cause
for the development of confidence measures
(Christopher and Debroux, 1987, p. 43.3.4). Before

the calculation of confidence measures, however, TN
101 considers the reference basic transmission loss
(Lbsr) given in Appendix B, expression (B-6).
Although simplistically presented, calculations for
Lbsr entail much greater complexity. Regardless, for
the pure troposcatter propagation mode (TN 101 also
analyzes multimode and diffraction paths), the
calculation is then extended to accommodate the
desired time availability and service probability.1
For simplicity, it is assumed that Lbsr is
calculated as described in Appendix B. From there
the following is applied:
Ln(O.5,50) = Lbsr - Vn(O.5,50) (5-1)
iWhen this is done, the system equation of

Appendix B (equation {B-l}), which is an expression
of the power required for a specified carrier-to-
thermal noise ratio at any instant in time, is
modified as described in Military Handbook [1977, pp.
4-174 to 4-177].
156
where Ln(0.5,50) describes the basic transmission

loss with a service probability of Q = 0.5 and a time
availability of q = 50%. A time availability of 50%

signifies that the path loss will meet or exceed L n
half of the time. A service probability of 0.5
indicates that half of a sample of links under

identical conditions will meet the specified time
availability. The term Vn(0.5,50) is a climate-

variation factor. It is provided in Figure 57 for
climate zones (subscript n) defined in Appendix F
2
(Freeman, 1987, pp. 146-178].
it I I i I
M 0 (2)and (6) 1 ...
.a J -T
(7a)
. _ __ , _ , _
> -10 i
0 00 200 300 400 500 600 700 800 900
Effective Distonce, da , in km
Tropospheric Scatter (Transhorizon Microwave system
Design). MIL-HDBK-417, Department of zefense,
Phildelphia, Pa., 19120-5094, p. 4-154.
Figure 57. Climate-variation factor as a function of

effective distance and climate type.
(Code numbers provided in Appendix F)
2 Figure
57 uses different nomenclature for
the climate-variation factor. This difference stems
from terminology, and is not consequential in this
analysis; therefore, consider V (0.5,50) = V(0.5)
[Military Handbook, 1977 p. 4-191; Freeman, 1987, p.
178]. The effective distance, de, used in Figure 57
is simple to calculate; see Freeman (1987, p. 173].
157
With Ln(O.5,50) one now knows the basic

troposcatter transmission loss exceeded 50% of the
time for 50% of all links under similar conditions.
This grade of service is unacceptable for most
applications; therefore, time availability and
service probability are scaled-up to desired levels.
A simple approach to extend time availability is
realized through curves similar to Figure 58. An
-'EUNT UICAh o
to T1 VC9 I I
l' 1-7
FK+i
....... F CUUCES
S..~w,ACOICTEO ALUESOf0' t
I I ~ ~ II4.0~a2 73V(0.10
oo) V(09'~ 19.1 .Z-
11
1CC
I~ I_
131WWL NO,
FigureMVI
58.AC Long ter powr adngfo cntneta
temperate climate (Region 6),

Frequency > 1 GHz
additional loss, Yn(q,de), is found through the curve

corresponding to the desired time availability (q, on
the far right) and is included in expression (5-1).

158
A family of such curves for different climates and
frequency is available in TN 101; however, as more
data accumulates, changes are made [Military

Handbook, 1977, p. 4-153]. Furthermore, these curves
are depicted for general frequency ranges; therefore,

a more refined approach, as summarized by Freeman
[1987, p. 178], uses a frequency correction factor
that can improve accuracy by more than 2 dB.

The service probability, included since there
are additional variations for which the model cannot
account, is enhanced through the following
semiempirical formulas:
ac2 (q) = 12.73 + 0.12Yn 2 (q,de) (5-2)

and
2 (q ) 2 )
(Irc(q) (Sc + sr 1/2 (5-3)
The first step is to solve expression (5-2), using

curves similar to Figure 58 for Yn(q,de), giving
ac 2 (q), the path-to-path variance between predicted
and actual median transmission loss. Next, equation
(5-3) is used to find yrc(q), the error between
predicted and actual median transmission loss for q
percentage of time. sr 2 is 4 dB 2 for new links and 2

dB 2 for established ones.
159
Finally, arc (q) is multipied by the standard

normal deviate, Zmo(Q), given in Figu--e 59. one
A4 IL
. -
- ? r..:-
-3 T: :U.
-41
0 0001 0 001 0.01 0020.050.1 0.2 0.3C.40,5 0.7 0.8 0.9 0.95 0.99 0.998 0.9999
C0005 06 038 0.999
Service Protibily. 0
Source: Freeman, R.L. Padio System Design fcr

Telecc.,mnicaticns (1-ICO G.~z). New York: John Wiley
& Sons, .987, p. 155.
Figure 59. The standard normal deviate
simply selects the desired service probabilit, Q,

finds the product Zmo0 .Q)arc(q), and gets a value in
dB that must be included in expression (5-1) along
with the scaled-up time availability factor Yn(q,de)-
The result is a prediction yielding the desired time
availability, q, and service probability, Q.
Normally, Q = 0.95 and q = 99.99% for high qLIlity
transmission; obviously the transmis-sion loss f-r

Ln(0.95,99.99) will ._.ncrease from the original
Ln(0.5,50) (Freeman, 1987, pp. 193-197].

160
Despite this admittedly shallow treatment of

TN 101, one should now understand the fundamental
concept used in this prediction tool. That is, if

enough data is accumulated on a variety of
troposcatter links, empirically derived models will
afford adequate prediction. Further, recognizing
inherent shortcomings, estimates should include time
availability and service probability to bolster the

confidence of the prediction. Affirming this
summary, a family of curves, similar to Figure 58, as
well as a climate-variation factor (Figure 57)
characterize TN 101's underlying approach.
5.3.2 CCIR Method I

A "short course" in TN 101 is offered by the
first CCIR prediction technique, CCIR Method I.
Although no match for TN 101's laborious
calculations, resulting predictions are close enough

to make it perhaps the most popular of all tools. If
compared with expression (B-6) in Appendix B, one
will note many similarities in this method's
fundamental formula:
L(50) = 30 log f - 20 log d + F(qd) - Gt
- Gr + Lc - V(de) (dB). (5-4)

161
For convenience, components are again delineated:

f operating frequency in MHz,
d path length (km),

F(qd) the attenuation function (see Appendix
B, Figure B3),
Gt, Gr transmit and receive antenna gains (dB),

Lc aperture-to-medium coupling loss (see
Section 4.2.4, expression {4-3)), and

V(de) the climate-variation factor (see
Section 5.3.1, Figure 57).
After solving for L(50), CCIR Method I uses a similar
yet simpler time availability and service probability

technique than TN 101 to establish prediction
confidence.
5.3.3 CCIR Method II

At about the same time as the first
publication of TN 101, Boithias and Battesti
completed work on an easy-to-use technique based on
worst month statistics. This graphical method
requires knowledge of "the equivalent path distance.

. ., the time percentage required, the type of
climate, and the frequency" (Hall, 1979, p. 142].

CCIR Report 238-5 incorporates this tool as CCIR
Method II.
162
Using Figure 60, the basic transmission loss
(l{q}) for the least favorable month at a selected

time availability is derived. There are three other
_0____'K- I ~tII t
i190 =
.,2Br ' I -
i 10
Figure 60. Mean attenuation, overland path,

temperate climate, 1 Gliz
such curves, one each for humid tropical, equitorial,
and desert climates. These curves are for a 1 GHz

path with zero degree elevation angles. For other
frequencies (less than 5 G~z), a factor of 301ogf
(GHz) is added. In addition, the term 201og(d/dq) is
included where
163
dq = d + 8.5(01 + 02). (5-5)
dq is the equivalent path distance (km), and, again,
d is the path distance (km). 01 and 02 are the

elevation angles of the transmit and receive antennas
respectively. Summarizing, for calculation of the
actual transmission loss
L(q) = l(q) + 30 log f + 20 log (d/dq) - Geff (5-6)
where Figure 47 provides the effective gain (Geff)

which is the sum of both antenna gains and the
coupling loss [Roda, 1988, pp. 125-129; CCIR, 1986,
pp. 372-376].
5.3.4 The CCIR Chinese Method
Researchers in China have suggested use of

the following empirical formula (in dB):
L(50%) = 30 log f + 30 log 0 + 10 log d - Geff
+ 124.6 - 0.08(N s - 320) + N(H). (5-6)
Here, f is the operating frequency in MHz, 0 is the
scattering angle in radians, d is the path distance

in km, Geff is the effective gain from Figure 47, and
N s is the annual mean surface refractivity value.
Additionally,
N(H) = 20 log(5 + 0.3H) + 0.65H (5-7a)

164
and
H = 8dld 2 /d (5-7b)
with dl and d2 being the distances from each antenna
to the crosspoint of the radio horizon rays (at the

lowest portion of the common volume). Additionally,
this method calculates the variability of the hourly

median losses. In this instance, these are also
shown to be log-normally distributed but standard
deviations, a, are given through terminal location,
radio horizon path distance (ds), and the difference
between mean summer and winter surface refractivity
(AN s ) [Roda, 1988, pp. 132,133; CCIR, 1986, pp.
386,387] (in dB):
a = 7.0 + 0.09AN s exp(-0.003d s ) (flat areas) (5-8a)

a = 4.6 + 0.06AN s exp(-0.003d s ) (mountains) (5-8b)
a = 8.8 + 0.11AN s exp(-0.003d s ) (over the sea) (5-8c)
5.3.5 The Radiometeorological Method

The work of Battesti and Boithias, along with
a third cohort (P. Misme), was again recognized by

the CCIR in report 238-4 for efforts in troposcatter
prediction. Recognizing that refractivity gradient
changes will not only displace the position of the

common volume but also alter the intensity of
individual scatterers, their efforts produced an
165
exceptionally accurate technique termed the

Radiometeorological method [CCIR, 1982, p. 137;
Boithias and Battesti, 1983, pp. 659, 660]. While it
has been stressed that the best indicator of
troposcatter performance loss is the gradient, this
is a parameter of substantial variability and is,

currently, impossible to model for every conceivaile
path. Regardless, if one knew the yearly gradient
variations within the proposed link's common volume,
he would apply
L(dB) = 30 log f + 30 log d + 102 + 1.5G c (5-9)
using f, the operating frequency in MHz; d, the path

length, in km; and Gc, the common volume's gradient,
in N units per km. Obviously, L(dB), the basic
transmission loss, is only as accurate as the
3
sounding data used for Gc.
Sarma [1988] shows the criticality of this
sounding data in the design of a 300 km, 2 GHz link
with a common volume situated above Kanpur, India.
3 CCIRReport 238-4 recommends referencing the

original work of Battesti, Boithias, and Misme
("Calcul des Affaiblissements en Propagation
Transhorizon A Partir des Param~tres
Radiom~t~orologiques" {"Calculations of Attenuations
in Trans-horizon Propagation from Radiometeorological
Parameters") in volume 23, number 5-6, pages 129-140
of Ann. des T616comm.) for discussion of gradient
soundings.
166
Using a plane mounted, state-of-the-art

refractometer, Sarma obtained accurate refractivity
gradient data for winter (worst case) conditions.
Then, employing the Radiometeorological method, he
compared results obtained through both the airborne
configuration and available radiosonde measurements.
The result was a discrepancy of nearly 35 dB; the
costs of several days of data collection (use of a

Cessna 182 H Skylane) are probably going to be less
than the savings in transmitter configuration. Thus
the importance of sounding data is demonstrated, and
the advantages of the Radiometeorological method are
well represented.
5.3.6 The Parl Method

It has now been well established that
prediction accuracy largely depends on knowledge of
the refractivity gradient within the common volume.

However, many techniques, such as TN 101, do not
stress this parameter due to its variability and
simply assume a specific value. This, according to
Parl [1979], limits prediction reliability. However,
if one were to focus on activity within the common
volume, specifically, the spectrum function (see
Appendix C), accuracy would improve. This is the
feature of the Parl method since the structure

167
constant (CN2 ) is known to relate directly to changes

in the refractive index, and, further, since values
of the spectrum slope (m) indicate whether incoherent
or quasi-coherent scatter dominates [Eklund and
4
Wickerts, 1968; Sarker, Dutta, and Reddy, 1983].
Noting variability in the spectrum function,
Monsen, et al. [1983] overhauled portions of TN 101
to include values of the spectrum slope and structure
constant that allow accurate predictions for
virtually all types of troposcatter systems. Parl
[1979] mentions that TN 101 uses a spectrum slope of
m = 5 which corresponds to quasi-coherent modes, he

then adds that at m = 11/3, incoherent scatter
dominates. Incorporating this mechanism would yield
more accurate predictions, since many systems use
higher frequencies and incoherent scatter dominates
above about 1 GHz; or more conservative analysis,

since, even at lower frequencies, turbulent
atmospheric conditions may temporarily foster the
notably poorer performing incoherent scatter
mechanism. Then, using m = 11/3, Monsen, et al.
4 The MITRE method incorporates the Parl

method for specific application to the military
tactical AN/TRC-170 terminal. Documents explaining
the MITRE method include Dube and Hotzel [1985],
Coirin and Kessler [1987], and Neuburger, et al.
[1985]; however, these publications are limited in
distribution.
1%8
[1983] developed an expression for CN 2 (see

expression {C-2a} of Appendix C) that allows
calculation of path loss in a manner similar to TN
101 (similar method, different results). Notably,
however, the Parl method allows use of any spectrum
slope value since it is available on a rather
flexible computer program called TROPO.
Another concept introduced in the Parl method
highlights aperture-to-medium coupling loss.
Basically, Parl [1979] asserts that coupling loss has
been incorporated in other techniques (TN 101 for

example) as an "afterthought" when, in reality,
volume integrations used to derive path loss
calculations must also account for narrow

veamwidths. 5 This is because most initial
calculations feature integrations for wide beamwidth
antennas which illuminate more scatterers, then, if
narrow beams are used, only antenna gain and
empirically derived coupling loss formulas are

manipulated without regard to the initial
integration. The error of this approach is readily
seen; furthermore, in the majority of situations,
5 Thisthesis has purposely avoided integral

formulas; however, the basic methodology of all
techniques begins with an integration of the
scatterers within the common volume.
169
this aspect is a more significant differentiator than
the variability of the spectrum slope.
Along with the advantage of being computer
based, the Parl method also includes:
a. Calculations for a variety of climates to
include those delineated in TN 101 or any
user defined zones.
b. Computations for horizon angles, antenna
elevation angles, antenna heights, diversity
separations, coherence bandwidth, and yearly
path loss and RSL distributions.
c. The ability to analyze multimode paths.
d. Performance predictions for specific modems
(DAR and the MD-918, used on military
systems) as well as delay spread
calculations.
e. Time availability and service probability
analysic.
Again, however, it bears mentioning that the accuracy
of this method is only as good as the data input.
Furthermore, the uncertainty of the values used for
the spectrum slope, as highlighted in Appendix C,
once more introduces prediction skepticism since the
Parl method performs only singular calculations for
specific spectrum slope values. Echoing this

170
concern, Crane (1981a, 1981b] projects that only when

accurate CN 2 profiles are available will prediction
accuracies substantially improved.
5.3.7 The Yeh Method

In 1960, Dr. Luang P. Yeh offered a simple
prediction tool (in dB):
Lmp = 30 log f + 20 log d + 108 + Lc
- 0.2(N s - 310) + 57. (5-10)
The parameter breakout for this calculation of the

median path loss, Lmp, follows: f, operating
frequency, in MHz; d, path length in miles; 8,
scattering angle, in degrees; Lc, coupling loss; and
NS, the surface refractivity [Panter, 1972, p. 393;
Freeman, 1987, p. 268].
5.3.8 The Rider Method

C. C. Rider's 1962 solution for the median
path loss is as follows:
Lmp = 30 log f + 30 log d + 50 log 8 + Lc

- 0.43(N s -330) + 49 (5-11)
where f, d, 0, and N s are the same as in Yeh's

method. Rider did not initially include coupling
171
loss, L.; however, it is included here for
comparative purposes [Panter, 1972, p. 394].
5.3.9 The Collins Method
Another prediction tool is the U.S. Army's
quick and easy Collins method as developed by the
Collins Radio Company. This totally graphical
technique exhibits surprizing accuracy through the
use of four graphs (Figures 61 through 64). By
adding the values obtained from all four graphs,
Collins offers field expediency typifying tactical,
on-the-move military communications [Panter, 1972, p.
394].
5.3.10 The New Bullington Model
The concept of using a combination of scatter
and mode theories in predicting troposcatter
propagation was offered by Bullington in 1963. This
was discussed in Section 3.5 where it was pointed out
that not much was done to further this idea. Two
employees of the MITRE Corporation, Christopher and
Debroux [1987], have reexamined Bullington's ideas,
adding foreground reflection and diversity
considerations to the hypothesis that predictions can
be based on an exponentially varying refractivity
profile. The result is a computer driven prediction

172
170
291 2X _5 '00
19
ISO
220 2 X 0 50 60 70 80 90
250, - - Ds-orce, sts.ute

mies
S260 y _____ -
I
2701
A I I____
290 I _ L.......
,_ I II
100 50 200 250 300 400 500 600 700 SO0 900
D:s'once, sic e niles
Source: Panter, ?.F. Communication Systems

Figure 61. Collins method: Basic propagation loss at

1 GHz
15-
5o 1i T< I I
10-
'00 200 330000 :00 2,000 330 5000 '0.:00
Source: Panter, P.F. Communication Systems

Design: Line-off-Sight and Tropo-Scatter Systems. St.
Figure 62. Collins method: Frequency correction

curve
173
Sj" of .%oe
an "~;:es
Source: Panter, P.F. Communication Systems
Figure 63. Collins method: Horizon angle loss
Aleere'Ce'e'u
50- s,
losi~ .5
-:6-
ce0.4-5
_C9 06 2
,.-20
90 1 08-
:1 2-
!.4 3
;5 4'
S204 12]
Souce Patr:.. C m un c t o ytm

Desin:
ineof-SghtandTrop-Sctte Sysems St
Lois 5McrwHil 292,p037
6425l n
Figure1% e h d ou l n o s n m g a
174
tool, simple to use, yet rivaling the accuracy of

more complex methods.
5.3.11 Digital Considerations

Completing this compendium of troposcatter
prediction tools are digital transmission
considerations. In many respects, digital
troposcatter transmission systems use the same
prediction tools as analog systems. Transmission
loss (including coupling loss), path profiles, and
antenna siting calculations can use any one of the
described tools; however, delay spread and BER

analysis, not included above, warrant consideration.
A method to calculate delay spread, as
mentioned in Section 4.3, is offered by Freeman
[1987, p. 205]:
) )
A = 5.21d(Ot + ir (4ao + 40o + Ot + 2r 10 - 8 . (5-12)
Here, d is path length in km, ao and 0o are the
transmit and receive main beam angles corrected for
above horizon elevation (MOO and Poo of Figure 5) in

radians, and Qt and~r are the transmit and receive
antenna 3 dB beamwidths in radians. From this
calculation one should ascertain the effects of delay
spread as discussed in Section 4.3; adaptive
processing considerations are essential.
175
To calculate the expected BER use the

following:
Eb/No = RSLdBW - 10 log(BR) - (204 dBW + NFdb) (5-13)
This expression for the energy-per-bit to noise power

density ratio should be somewhat familiar: BR is the
bit rate in bps, NFdB is the noise figure of the
receiver or low noise amplifier (LNA), and RSL is
determinable through
RSLdBW = EIRPdB W - TL + Gr - LLr. (5-14)
For this equation, the effective isotropically

radiated power (EIRP) includes the sum of transmitter
power, antenna gain, and miscellaneous device losses

(all in dB); TL is the total transmission loss as
determined by a prediction tool; Gr is the receive
antenna gain; and LLr is the line loss to the LNA
[Freeman, 1987, p. 214; Flock, 1987).

All that now remains is to find the BER using
the calculated Eb/No and Figure 65 (note the effect
of different modulation schemes). One may see the
importance of accurate prediction in the TL term of
expression (5-14); other factors remaining constant,
the accuracy of the predicted BER is contingent on
the accuracy of the prediction technique.

176
D,FF RENTiALLY CODED
CODED AND
RAN~DOMIZED0
T7iiORETICAL
-
10 $
I0"'
"
'a' I I g I
3 4 5 6 7 8 9 10 11 2 13 14 15
Source: Freeman, R.L. Radio System Design for

Telecommunications (1-100 GHz). New York: John Wiley
& Sons, 1987, p. 215.
Figure 65. Determination of BER for coherent BPSK

and QPSK
Furthermore, if delay spread is sufficiently

neutralized, the BER is the single most important
parameter in assessing digital system performance.
5.4 Comparison of Methods

Based on information presented thus far, it
would seem prudent to choose a simple tool such as
the Collins or Yeh method. However, simplicity

smatters of shortcuts, shortcuts translate into a few
dB, and these few dB can be critical in link

prediction. A rigorous comparison of all techniques
177
is not available; furthermore, such an undertaking

would be quite laborious and this, compounded with
lack of commercial interest, means such comparisons
are not likely. Some analysis exists, however, and
these will be presented. The intent is to provide
the engineer with enough information from which to
choose an applicable method. A good way to begin is

to analyze the advantages and limitations of proposed
techniques.
Discussion will center on TN 101 as it forms
the nucleus of many techniques including Military
Handbook, Facility Design for Tropospheric Scatter
[1977], CCIR Method I, the Parl, and the MITRE

techniques. The primary advantage of TN 101 is its
historical methodology. Years of empirical data are
accumulated to provide prediction curves yielding
accurate long term median path attenuation. This,
however, can be a disadvantage since most data is
gathered in temperate climates. Sarkar, Dutta, and

Reddy [1983] report large discrepancies between TN
101 predictions and actual data for a link in the
Indian sub-continent. Fitzsimons [1968] offers
similar criticism with data gathered from an 87 km
link in Cyprus.
178
Crane [1988] also notes that much of the data
used in TN 101 is for lower (below 1 GHz) links.

These frequency limitations were highlighted earlier:
since TN 101 uses a spectrum slope of m = 5 and since
this corresponds to quasi-coherent scatter which is
dominant only at lower frequencies, TN 101 should be

cautiously applied above 1 GHz [Parl, 1979]. This,
as mentioned, is where the Parl and MITRE methods
differ since the more conservative m value of 11/3 is

used (or any other value for the Parl method).
Hence, the frequency limits of TN 101 are offset if
one uses the similar Parl or MITRE methods. In
addition, except for the Parl and MITRE models, all
techniques incorporating aperture-to-medium coupling
loss incorrectly embody this parameter and this can
be exceptionally significant when narrow beamwidths
are used. Unfortunately, subsequent calculations for
modem performance in these methods are restricted to
military systems; extending their use on a more
general level would require alteration.
Simplicity is always a consideration and TN
101 and its derivatives (CCIR Method I excepted) do
not offer this luxury. Notably, however, the

Longley-Rice, Parl, and MITRE methods are available
through computer applications. Further, there is no

179
reason why any of the models, with the exception of

the Collins method, cannot be adapted for computer
use. If, however, simplicity is crucial and computer
programs are unavailable, any of the CCIR methods or
the Yeh, Rider, or Collins techniques may be

considered. These will not, however, offer
comprehensive (or any) confidence measures typifying
TN 101 and climate differentiation may not be
incorporated.
Another limitation of many techniques, to

include TN 101, is that they do not quantitatively
embrace short-term Rayleigh fading characteristics.
The new Bullington technique does; furthermore, this
method offers simple, compact results through
computer applications [Christopher and Debroux,

1987]. The newness of this technique may, however,
generate skepticism; additionally, it is based, in
part, on unaccepted mode reflection theories.
A final criticism of most methods is the
dominating influence of surface refractivity, Ns,
which has been shown fallacious in several instances.
Methods that make this "mistake" include, to varying
degrees, all TN 101 derivatives as well as the

methods of Yeh and Rider. Neither the Collins, CCIR
Method II, or the Radiometeorological method include

180
Ns . However, despite its simplicity, the Collins
method exhibits limited application for long (greater
than 400 mi) links. Furthermore, the
Radiometeorological method is known to be quite
accurate and simple only if refractivity gradient
data within the common volume is available. Finally,
CCIR Method II provides good "worst case" analysis,
at least for temperate climates [Boithias and
Battesti, 1983, p. 660; CCIR, 1986, p. 3761.
A 1968 comparison of several methods produced
some surprizing results [Larsen 19681. Data from

fifteen operational links in various regions were
compared with predictions for those links using the
techniques of TN 101, CCIR Method I, Yeh, Rider, and
Collins. Notably, only two of the links had
operating frequencies above 3 GHz and two more above
2 GHz. Despite several manipulations of results to
provide equivalent solutions, Larsen states:
[t]he conclusions which may be drawn at this

stage indicate that the N.B.S. [TN 101) and CCIR
are almost equal in performance, although CCIR is
much more convenient to use. The CCIR method
provides important estimates of fading ranges,
and a combination of CCIR and Yeh gives good
estimates of median path loss. The Collins
method suffers from having a restricted range of
parameters, but gives very good results when used
inside this range. All the calculation methods
show remarkable failure to account satisfactorily
for climate variation, and indicate the need for
further investigation of this point [Larsen,
1968, p. 114].
181
Larsen's calculations indicated reasonable

correlation for all methods except Rider's; it was
considerably inferior. The Collins method

outperformed all, with Yeh's being slightly better
than TN 101 and CCIR Method I. However, Larsen
cautions that his link sample could have
significantly influenced results.
Fitzsimons [1968] may also be referenced for
a comparison between actual data on an 87 km, 4.7 GHz
link in Cyprus and the predictions of TN 101,
Collins, Yeh, and an earlier CCIR model. Results

were all on the pessimistic side: Yeh by 7 dB,
Collins by 9 dB, and TN 101 by 11 dB.

Among the newer methods, Parl [1979] and
Christopher and Debroux [1987] o-fer some
comparisons. In Figure 66 a comparison of TN 101
(NBS) and the Parl (Turbulent Scatter) methods are
given. The difference can be attributed to coupling
loss calculations and the conservative approach of
using incoherent scatter in all cases instead of
quasi-coherent scatter as in TN 101's case. Figure

67 compares the scatter losses of the new Bullington
method with an earlier (1962) CCIR method and TN 101
(Longley-Rice). As a reference, the "Chisholm" curve

represents actual data for several links at 400 MHz.
182
2.2,
Source: Par2., S.A. "ew Formulas for

Tropospheric Scatter Path Loss." Radio Science, vol.
14, no. 1 (January-February 1979), p. 53.
Figure 66. Comparison of TN 101 (NBS) and the Par
(Turbulent Scatter) methods
MILES
100
- I
200 300 400 500
I 600
I 700
TI 800
T_ 900
-60 I I
-J
u N, %, BULLINGGTON8 PRiDICTION
-100-
LONGLEYRICE-
-120
-140
0 200 400 600 800 1,000 1,200 1,400 1,600
PATH LENGTH IN KILOMETERS
Source: Christopher, P.F., and Debroux, P.S.

"New Applications of Bullington's Tropospheric
Reflection Model.." MILCOM 87, IEEE Military
Communications Confference, vol. 3 (Oct 1987), p.
43.3.6.
Figure 67. Comparison of the CCIR (1962 Data), TN

101 (Longley-Rice), and Bullington
methods. "Chisholm" represents actual
data for several links at 400 MHz.
183
5.5 Conclusions
Conclusions are less than obvious. None of
the techniques offer complete, "any situation"
accuracy except the Radiometeorological method. Yet

this statement is made with a notable absence of
complete comparison. The poor performance of TN 101

in Larsen's analysis leads one to suspect similar
performance in the Military Handbook [1977], CCIR

Method I, Parl, and the MITRE methods. 6 However,
given Larsen's limited analysis, heavily favoring
lower frequencies, it is erroneous to include the
Parl and MITRE methods in this allegation since these
contemporary approaches fundamentally differ from TN
101. Further, the limited samples used by Larsen and
Fitzsimons nullify any conclusive arguments.
One should note that with respect to
theoretical considerations, the Parl method stands
out. Through computer flexibility, this method
allows variable inputs, provides theoretically sound
coupling loss calculations, and does not emphasize
surface refractivity or singular spectrum slope
values as does TN 101. CCIR Method I, the CCIR
6 One
must assume that Larsen correctly
applied the techniques being compared for each of the
15 links analyzed.
184
Chinese method, and Yeh and Rider's techniques center
on surface refractivity. The Collins method has

limited application, CCIR Method II works best in
temperate climates, and Bullington's treatment is
untried and based, in part, on unaccepted mode
theory.
To conclude, recognizing that the majority of
prediction techniques are simply estimates based on
empirical analysis, the best procedure would be to
choose a variety of methods. Final analysis would

include projections from several techniques, and a
best case/worst case forecast would establish
prediction confidence. This recommendation is based

on the limitations presented; modifications to
techniques such as the Radiometeorological or Parl
methods supplemented with development of accurate
methods to profile common volume gradient variations
could allow significant improvement.

BIBLIOGRAPHY
Albrecht, H.J. "Correlating Tropospheric Scatter

Data and Cyclonic Parameters." Proceedings of
the IEEE, vol. 55, no. 10 (October 1967): 1768-
1769.
Anthes, R.A., Cahir, J.J., Fraser A.B., and Panofsky,

H.A. The Atmosphere. 3rd ed. Columbus, Ohio:
Charles E. Merrill, 1981.
Azurza, J.J. "Tropospheric Scatter Communications in

the Oil Industry." MS thesis, Univ. of Colorado,
1975.
Baghdady, E.J. "Performance Characteristics of

Multipath Channels." The Electromagnetic Science
Series of Short Courses Lecture Notes 5.2:
Tropospheric Radio Propagation. Sponsored by the
U.S. Department of Commerce, Environment Science
Services Administration, Institute for
Telecommunication Science and the University of
Colorado, Electrical Engineering Department, 27
July - 7 August, 1970.
Barrow, B.B., Abraham L.G., Cowan, W.M. Jr., and

Gallant, R.M. "Indirect Atmospheric Measurements
Utilizing Rake Tropospheric Scatter Techniques-
Part I: The Rake.Tropospheric Scatter Technique."
Proceedings of the IEEE, vol. 57, no. 4 (April
1969): 537-551.
Barsis, A.P., Norton, K.A., and Rice, P.L.

Performance Predictions for Single Tropospheric
Communication Links and for Several Links in
Tandum. U.S. National Bureau of Standards,
Technical Note, no. 102 (Aug 1961).
Barsis, A.P., Norton, K.A., and Rice, P.L.

"Predicting the Performance of Tropospheric
Communication Links, Singly and in Tandum." IRE
Transactions on Communications Systems, vol. CS-
10, no. 1 (March 1962): 2-22.
186
Barton, J.H. "A Selective Survey of Soviet Bloc

Scatter Development. "IRE Transactions on
Antennas and Propagation, vol. AP-10, no. 3 (May
1962): 335-339.
Bean, B.R. "Some Meteorological Effects on Scattered
Radio Waves." IRE Transactions on Communications
Systems, vol. CS-4, no. 1 (March 1956): 32-38.
Bean, B.R., and Dutton, E.J. Radio Meteorology.
National Bureau of Standards, U.S. Department of
Commerce, Monograph 92, 1966.
Bean, B.R., Horn, J.D., and Riggs, L.P. Synoptic
Radio Meteorology. National U.S. Department of
Commerce, Bureau of Standards Technical Note 98.
Washington: Superintendent of Documents, U.S.
Government Printing Office, 1962.
Bello, P.A. "A Troposcatter Channel Model." IEEE

Transactions on Communications Technology, vol.
COM-17, no. 2 (April 1969a): 130-137.
Bello, P.A. "Selection of Multichannel Digital Data
systems for Troposcatter Channels." IEEE
COM-17, no. 2 (April 1969b) : 138-161.
Bello, P.A., and Ehrman L. "Error Rates in Diversity
FDM-FM Digital Troposcatter Transmission." IEEE
COM-17, no. 2 (April 1969): 182-191.
Birnbaum, G., and Bussey, H.E. "Amplitude, Scale,
and Spectrum Refractive Index Inhomogenities in
the First 125 Meters of the Atmosphere."
Proceedings of the IRE, vol. 42, no. 10 (October
1955): 1412+.
Boithias, L. Radio Wave Propagation. St. Louis:

McGraw-Hill, 1987.
Boithias, L., and Battesti, J. "Propagation Due to
Tropospheric Inhomogeneities." IEE Proceedings,
vol. 130, part F, no. 7 (December 1983): 657-664.
Bolgiano, R., Jr. "The Role of Turbulent Mixing in
Scatter Propagation." IRE Transactions on
Antennas and Propagation, vol. AP-6, no. 2 (April
1958): 161-168.
187
Booker, H.G., and deBettencourt, J.T. "Theory of

Radio Transmission by Tropospheric Scattering
Using Very Narrow Beams." Proceedings of the
IRE, vol. 43, no. 3 (March 1955): 281-290.
Booker, H.G., and Gordon, W.E. "A Theory of Radio
Scattering in the Troposphere." Proceedings of
the IRE, vol. 38, no. 4 (April 1950): 401-412.
Brauburger, G.H. Jr. (Capt, USAFR). "Meteorological
Factors Affecting AN/TRC-97A Tropospheric Scatter
Communications." Technical Report no. TACCA/XQT-
TR-79-1, Langley AFB, Va, 1 August 1979.
Bremmer, H. "The Propagation of Electromagnetic Waves
Through a Stratified Medium and its WKB
Approximation for Oblique Ircidence." Physica,
vol. 15 (August 1949): 593-608.
Bullington, K. "Radio Propagation at Frequencies
Above 30 Mc/s." Proceedings of the IRE, vol. 35
(1947) : 1122+.
Bullington, K. "Radio Transmission Beyond the
Horizon in the 40- to 4,000-MC Band."
Proceedings of the IRE, vol. 41, no. 1 (January
1953) : 132-135.
Bullington, K. "Characteristics of Beyond-the-
Horizon Radio Transmission." Proceedings of the
IRE, vol. 43, no. 10 (October 1955): 1175-1180.
Bullington, K. "Status of Tropospheric Extended
Range Transmission." IRE Transactions on
Antennas and Propagation, vol. AP-7, no. 4
(October 1959): 439-440.
Bullington, K. "Reflections from an Exponential
Atmosphere." The Bell System Technical Journal,
vol. XLII, no. 6 (November 1963): 2849-2867.
Carroll, T.J. "Internal Reflections in the

Troposphere and Propagation Beyond the Horizon."
Transactions of the IRE, Professional Group on
Antennas and Propagation, PGAP-2 (March 1952a):
9-27.
188
Carroll, T.J. "Normal Tropospheric Propagation Deep

into the Earth's Shadow: The Present Status of
Suggested Explainations." Transactions of the
IRE, Professional Group on Antennas and
Propagation, PGAP-3 (August 1952b): 6-11.
Carroll, T.J. "Tropospheric Propagation Well Beyond

the Horizon." Transactions of the IRE,
Professional Group on Antennas and Propagation,
PGAP-3 (August 1952c): 84-100.
Carroll, T.J. "Marconi's Last Paper, 'On the
Propagation of Microwaves over Considerable
Distances'." Proceedings of the IRE, vol. 44,
no. 8 (August 1956): 1056-1057.
Carroll, T.J., and Ring, R.M. "Propagation of Short

Radio Waves in a Normally Stratified
Troposphere." Proceedings of the IRE, vol. 43,
no. 10 (October 1955): 1384-1390.
CCIR. Volume V. Propagation in Non-Ionized Media.

Geneva, 1982.
CCIR. Volume V. Propagation in Non-Ionized Media.
Dubrovnik, 1986.
Chisholm, J. H. "Progress of Tropospheric

Propagation Research Related to Communications
Beyond the Horizon." IRE Transactions on
Communications Systems, vol. CS-4, no.1 (March
1956): 6-16.
Christopher, P.F., and Debroux, P.S. "New
Applications of Bullington's Tropospheric
Reflection Model." MILCOM 87, IEEE Military
Communications Conference, vol. 3 (Oct 1987):
43.3.1-43.3.7.
Clavier, A.G. "Microwave Communication Beyond the
Horizon." Electrical Communication, vol. 33, no.
2 (June 1956): 108-116.
Coirin, M.R., and Kessler, J.A. "User's Guide for

the Personal Computer Version of the AN/TRC-170
Initial Link Engineering Manual." The MITRE
Corporation Working Paper WP-25855, Supp. III,
Fort Huachuca, Az, May 1987.
189
Colwell, J.F., and Anderson, L.J. "Concerning the

Radio Field Due to Internal Reflection in the
Stratified Troposphere." Transactions of the
IRE, Professional Group on Antennas and
Propagation, PGAP-3 (August 1952): 117-125.
Cox, D.C. "Review of Techniques in Transhorizon

Propagation Research and their Relationship to
Parameters of the Troposphere." Radio Science,
vol. 4, no. 10 (October 1969): 905-925.
Craig, K.H., and Levy, M.F. "The Modeling of

Transhorizon Anomalous Propagation Conditions."
Fifth International Conference on Antennas and
Propagation (ICAP 87) Part 2: Propagation,
Institution of Electrical Engineers, London
(1987): 347-350.
Crain, C.M. "Survey of Airborne Microwave
Refractometer Measurements." Proceedings of the
IRE, vol. 43, no. 10 (October 1955): 1405-1411.
Crain, C.M., and Gerhardt, J.R. "Measurements of the
Parameters Involved in the Theory of Radio
Scattering in the Troposphere." Proceedings of
the IRE, vol. 40, no. 1 (January 1952): 50-54.
Crane, R.K. "Fundamental Limitations Caused by RF
Propagation." Proceedings of the IEEE, vol. 69,
no. 2 (February 1981a) : 196-209.
Crane, R.K. "A Review of Transhorizon Propagation

Phenomena." Radio Science, vol. 16, no. 5
(September-October 1981b) : 649-669.
Crane, R.K. Tropospheric Scatter Propagation at 5

and 15 GHz: Preliminary Results for a 161 km
Path. Interim Report on Research Conducted Under
the Post-Doctoral Program, Rome Air Development
Center, Hanscom AFB, Maine. Hanover, New
Hampshire: Dartmouth College, 1988.
Crawford, A.B., Hogg. D.C., and Kummer W.H. "Studies
in Tropospheric Propagation Beyond the Horizon."
The Bell System Technical Journal, vol. 38, no. 5
(September 1959) : 1067-1178.
David, P., and Voge, J. Propagation of Waves. New
York: Pergamon, 1969.
190
de Belatini, P.C.M. "Inadequacy of Scatter

Mechanisms in Tropospheric Radio Propagation."
Nature, vol. 184, no. 4698 (November 14, 1959):
1558-1559.
Derr, V.E. Remote Sensing of the Troposphere. U.S.
Department of Commerce, National Oceanic and
Atmospheric Administration, Environmental
Research Laboratories, and the University of
Colorado. Washington: Superintendent of
Documents, U.S. Government Printing Office,
August 15, 1972.
"Discussion on 'Tropospheric Propagation Beyond the

Horizon--I'." The Proceedings of the Institution
of Electrical Engineers, Pt. B, Supp. 8, vol. 105
(1958) : 122-126.
Doherty, L.H., and Stone, S.A. "Forward Scatter from
Rain." IRE Transactions on Antennas and
Propagation, vol. AP-8, no. 4 (July 1960): 414-
418.
Dougherty, H.T. "The Refractive Index Gradient and
Surface Refractivity." The Electromagnetic
Science Series of Short Courses Lecture Notes
2.1: Tropospheric Radio Propagation. Sponsored
by the U.S. Department of Commerce, Environment
Science Services Administration, Institute for
Telecommunication Science and the University of
July -7 August, 1970.
du Castel, F. Tropospheric Radiowave Propagation
Beyond the Horizon. New York: Pergamon, 1966.
du Castel, F., ed. Progress in Radio Science 1960-
1963, Volume II, Radio and Troposphere. New
York: Elsevier, 1965.
Dube, R.J., and Hotzel, P.M. "TRI-TAC Equipment
Interface Plan for AN/TRC-170." The MITRE
Corporation MTR-85A00004, McLean, Virginia, May
1985.
Eklund, F., and Wickerts, S. "Wavelength Dependence
of Microwave Propagation Far Beyond the Radio
Horizon." Radio Science, vol. 3 (New Series),
no. 11 (November 1968): 1066-1074.
191
Engineering Considerations for Microwave

Communications Systems. 2nd ed. San Carlos,
California: GTE Lenkurt, 1972.
Evaluation of FDM and FM Systems. Air Force
Communications Command (AFCC) textbook for AFCC
30000-1, Wright-Patterson AFB, Oh., January 1983.
Fehlhaber, L. "Results of the Theory of Tropospheric

Scattering, Derived from Propagation
Measurements." Conference on Tropospheric Wave
Propagation, 30 September - 2 October, IEE,1968:
37-42.
Feinstein, J. "Tropospheric Propagation Beyond the
Horizon." Journal of Applied Physics, vol. 22,
no. 10 (October 1951): 1292-1293.
Feinstein, J. "The Role of Partial Reflections in
Tropospheric Propagation Beyond the Horizon."
Transactions of the IRE, Professional Group on
Antennas and Propagation, PAGP-2 (March 1952a):
2-8.
Feinstein, J. "Partial Reflections in Tropospheric
Propagation." Transactions of the IRE,
Professional Group on Antennas and
Propagation,PGAP-3 (August 1952b) : 101-111.
Fitzsimmons, T.K. "Observations of Tropospheric-
Scatter Path Loss at C Band Frequencies and
Meteorological Conditions in Cyprus."
Proceedings of the IEE, vol. 115, no. 7 (July 7):
247-251.
Flock, W.L. Electromagnetics and the Environment:
Remote Sensing and Telecommunications. Englewood
Cliffs, New Jersey: Prentice-Hall, 1979.
Flock, W.L. Propagation Effects on Satellite Systems
at Frequencies Below 10 GHz. Second Ed., NASA
Reference Publication 1108(03), 1987.
Freeman, R.L. Telecommunication Transmission
Handbook. New York: John Wiley & Sons, 1975.
Freeman, R.L. Radio System Design for
Telecommunications (1-100 GHz). New York: John
Wiley & Sons, 1987.
192
Friis, H.T., Crawford, A.B., and Hogg, D.C. "A

Reflection Theory for Propagation Beyond the
Horizon." The Bell System Technical Journal,
vol. XXXVI, no. 3 (May 1957): 627-644.
Gerlach, A.M., ed. "Objective Analysis and

Prediction Techniques - 1984." Air Force
Geophysics Lab Report No. AFGL-TR-84-0328,
Hanscom AFB, Ma., (30 Nov 1984).
Gjessing, D.T., and Irgens, F. "On the Scattering of
Electromagnetic Waves by a Moving Tropospheric
Layer Having Sinusoidal Boundaries." IEEE
Transactions on Antennas and Propagation, vol.
AP-12, no. 1 (January 1964): 51-64.
Gjessing, D.T., and McCormick, K.S. "On the

Prediction of the Characteristic Parameters of
Long-Distance Tropospheric Communication Links."
IEEE Transactions On Communications, vol. COM-
22, no. 9 (September 1974): 1325-1331.
Global Tropospheric Chemistry: A Plan for Action.
National Research Council, Global Tropospheric
Chemistry Panel Board of Atmospheric Sciences and
Climate Commission on Physical Sciences,
Mathematics, and Resources. Washington: National
Academy Press, 1984.
Gordon, W.E. "Radio Scattering in the Troposphere."
Proceedings of the IRE, vol. 43, no. 1 (January
1955): 23-28.
Gordon, W.E. "A Simple Picture of Tropospheric Radio
Scattering." IRE Transactions on Communications
Systems, vol. CS-4, no.1 (March 1956): 97-101.
Graham, J.W. "Troposcatter Transmission." Course
notes from Session 7 of the MITRE Institue
Course: Introduction to Digital Communication
Systems, 20 October 1987.
Gunther, F.A. "Tropospheric Scatter Communications
(Past, Present, and Future)." IEEE Spectrum,
vol. 3, no. 9 (September 1966): 81-100.
Hall, M.P.M. Effects of the Troposphere on Radio
Communication. New York: Peter Peregrinus, 1979.
Hall, M.P.M., and Barclay, L.W., eds. Radiowave
Propagation. London: Peter Peregrinus, 1989.
193
Hall, M.P.M., and Dowling, G.R. "Effect of Rain on

4GHz Troposcatter Radio Paths." Electronics
Letters, vol. 10, no. 11 (May 30, 1974): 210-212.
Hooke, W.H., and Hardy, K.R. "Further Study of the
Atmospheric Gravity Waves over the Eastern
Seaboard on 18 March 1969." Journal of Applied
Meteorology, vol. 14, no. 1 (February 1975): 31-
38.
Ippolito, L.J. Jr. Radiowave Propagation in

Satellite Communications. New York: Van Nostrand
Reinhold Co., 1986.
Iribarne, J.V., and Cho, H.-R. Atmospheric Physics.

Boston: D. Reidel, 1980.
Isted, G.A. "Guglielmo Marconi and Communication

Beyond the Horizon: A Short Historical Note."
The Proceedings of the Institution of Electrical
Engineers, Pt. B, Supp. 8, vol. 105 (1958): 79-
83.
Johnson, M.A. "A Review of Tropospheric Scatter

Propagation Tieory and its Application to
Experiment." The Proceedings of the Institution
of Electrical Engineers, Pt. B, Supp. 8, vol. 105
(1958) : 165-176.
Jowett, J.K.S. "The Measurement and Prediction of
VHF Tropospheric Field Strengths at Distances
Beyond the Horizon." The Proceedings of the
Institution of Electrical Engineers, Pt. B, Supp.
8, vol. 105 (1958): 79-83.
Katzin, M. "Tropospheric Propagation Beyond the

Horizon." Transactions of the IRE, Professional
Group on Antennas and Propagation, PGAP-3 (August
1952a) : 112-116.
Katzin, M. "The Insignificance of Continuous

Internal Reflections in Tropospheric Propagation
Beyond the Horizon" Transactions of the IRE,
Professional Group on Antennas and Propagation,
PGAP-4 (December 1952b): 14-18.
Kerr, D.E. Propagation of Short Radio Waves. Rev.

ed. London: Peter Peregrinus, 1987.
194
Khalil, M.A.K., and Rasmussen, R.A. "Carbon Monoxide

in the Earth's Atmosphere: Indications of a
Global Increase." Nature, vol. 332 (March 17,
1988): 242-244.
Lane, J.A. "Review of Tropospheric Propagation."
The Electromagnetic Science Series of Short
Courses Lecture Notes VI-1 and VI-2. Sponsored
by the U.S. Department of Commerce, Environment
Science Services Administration and the
University of Colorado, Electrical Engineering
Department, 17 June - 3 July, 1968.
Larsen, R. "A Comparison of Some Troposcatter

Prediction Methods." Conference on Tropospheric
Wave Propagation, 30 September - 2 October, IEE,
1968: 110-117
Lemmon, J.J. "Propagation and Performance

Measurements Over a Digital Troposcatter
Communications Link." A paper reprinted from
Conference Proceedings No. 419, Advisory Group
for Aerospace Research and Development (AGARD),
(1989).
Livingston, D.C. The Physics of Microwave
Propagation. Englewood Cliffs, New Jersey:
Prentice-Hall, 1970.
Lutgens, F.K., and Tarbuck, E.J. The Atmosphere: An

Introduction to Meteorology. 2nd ed. Englewood
Cliffs, New Jersey: Prentice-Hall, 1982.
Martin, J. Communications Satellite Systems.
Englewood Cliffs, N.J.: Prentice-Hall, Inc.,
1978.
McGavin, R.E. A Survey of the Techniques for
Measuring the Radio Refraction Index. National
U.S. Department of Commerce, Bureau of Standards
Technical Note 99. Washington: Superintendent of
Documents, U.S. Government Printing Office, 1962.
McKay, L.R. What Every Engineer Should Know about
Electronic Communications Systems. New York:
Marcel Dekker, 1989.
Megaw, E.C.S. "Scattering of Electromagnetic Waves

by Atmospheric Turbulence." Nature, vol. 166,
no. 4235 (December 30, 1950): 1100-1104.
195
Megaw, E.C.S. "Fundamental Radio Scatter Propagation

Theory." The Proceedings of the Institution of
Electrical Engineers, Part C, vol. 104 (1957):
441-455.
Scatter (Transhorizon Microwave System Design).
MIL-HDBK-417, Department of Defense, Washington,
DC, 25 November, 1977. Available through Navy
Publications and Forms Center, ATTN: NPODS,
Phildelphia, Pa., 19120-5094.
Monsen, P. "Fading Channel Communications." IEEE
Communications Magazine, vol. 18, no. 1 (January
1980): 16-25.
Monsen, P., Parl, S., Malaga, A., Tolman, S., and
Fetteroll, J. Digital Troposcatter Performance
Model: Users Manual. Lexington, Massachusetts:
SIGNATRON, Inc. publication no. A288-15,
November, 1983.
Neiburger, M., Edinger, J.G., Bonner, W.D.
Understanding our Atmospheric Environment. San
Francisco: W.H. Freeman, 1973.
Neuburger, A., Freyheit, P.J., Leroux, G.,
Livingston, D.C., and McLaughlin, K.P. "AN/TRC-
170 Initial Link Engineering Manual." The MITRE
Corporation Working Paper WP-25855, Bedford, Ma.,
16 Jan 1985.
Norton, K.A., Rice, P.L., and Vogler, L.E. "The Use
of Angular Distance Transmission Loss and Fading
Range for Propagation through a Turbulent
Atmosphere over Irregular Terrain." Proceedings
of the IRE, vol. 43, no. 10 (October 1955): 1488-
1526.
Norton, K.A. Advances in Electronics, Academic
Press, New York, vol. 1 (1948): 381-423.
Ortusi, J. "The Various Theories on the Propagation
of Ultra-Short Waves Beyond the Horizon." IRE
AP-3, no. 2 (April 1955) : 86-91.
196
Ortwein, N.R., Hopkins, R.U.F., and Pohl, J.E.

"Properties of Tropospheric Scattered Fields."
Proceedings of the IRE, vol. 49, no. 4 (April
1961): 788-802.
Panter, P.F. Communication Systems Design: Line-of-

Sight and Tropo-Scatter Systems. St. Louis:
McGraw-Hill, 1972.
Parl, S. A. "New Formulas for Tropospheric Scatter

Path Loss." Radio Science, vol. 14, no. 1
(January-February 1979): 49-57.
Pekeris, C.L. "Wave Theoretical Interpretation of
Propagation of 10 cm and 3 cm Waves in Low-Level
Ocean Ducts." Proceedings of the IRE, vol. 35
(1947): 453+.
Penkett, S.A. "Increased Tropospheric Ozone.
"Nature, vol. 332 (March 17, 1988): 204-205.
Peterson, C.F., Farrow, J.J., Capps, F.M., and
Samson, C.A. Propagation at 5GHz in Europe - A
Comparison of Predictions and Measurements for
Various Types of Paths. U.S. Department of
Commerce, Environmental Science Services
Administration, Institutes for Invironmental
Research, ESSA Technical Report IER 2 ITSA 2.
Boulder, Co: Institue for Telecommunication
Sciences and Aeronomy, July 1966.
Picquenard, A. Radio Wave Propagation. New York:
Macmillan, 1974.
"Radio Transmission By Ionospheric and Tropospheric
Scatter." A report of the Joint Technical
Advisory Committee (JTAC) in Proceedings of the
IRE, vol. 48, no. 1 (January 1960): 4-44.
Rice, P.L., Longley, A.G., Norton, K.A., and Barsis,

A.P. Transmission Loss Predictions for
Tropospheric Communication Circuits. Vols. I and
II. Rev. ed. U.S. Department of Commerce,
National Bureau of Standards Technical Note 101.
Washington: Superintendent of Documents, U.S.
Government Printing Office, 1967.
Roda, G. Troposcatter Radio Links. Norwood, Maine:

Artech House, 1988.
197
Sarkar, S.K., Dutta, H.N., and Reddy, B.M.

"Development of a Prediction Technique for Tropo-
Path Loss from Observed Transhorizon Propagation
Characteristics over India." Third International
Conference on Antennas and Propagation (ICAP 83),
Part 2: Propagation, Institution of Electrical
Engineers, London (1983): 229-232.
Sarma, S.B.S.S. "Troposcatter Link Design Using Air-

borne Microwave Refractometer Observations over
Kanpur." Indian Journal of Radio & Space
Physics, vol. 17 (June 1988): 103-107.
Saxton, J.A. "Tropospheric Scatter Propagation."

Wireless World, vol. 62, no. 12 (December 1956):
587-590.
Silverman, R.A. "Fading of Radio Waves Scattered by
Dielectric Turbulence." Journal of Applied
Physics, vol. 28, no. 4 (April 1957): 506-511.
Smith, E.K. "Introduction to the Physics of Media:

The Atmosphere, The Ionosphere." The
Electromagnetic Science Series of Short Courses
Lecture Notes II-1 and 11-4. Sponsored by the
U.S. Department of Commerce, Environment Science
Services Administration and the University of
June - 3 July, 1968.
Smith, E.*K., and Weintraub, S. "The Constants in the

Equation for Atmospheric Retractive Index at
Radio Frequencies." Proceedings of the IRE, vol.
41, no. 8 (August 1953): 1035-1037.
Sommerfeld, A.N. "The Propagation of Waves in

Wireless Telegraphy." Ann. Phys., series 4, vol.
28 (1909): 665+.
Staniforth, J.A. Microwave Transmission. New York:
John Wiley & Sons, 1972.
Staras, H. "Aperture-to-Medium Coupling Loss." IRE
AP-5, no. 2 (April 1957): 228-232.
Staras, H., and Wheelon, A.D. "Theoretical Research
on Tropospheric Scatter Propagation in the United
States, 1954-1957." IRE Transactions on Antennas
and Propagation, vol. AP-7, no. 1 (January 1959):
80-87.
198
Tatarski, V.I. Wave Propagation in a Turbulent

Medium. New York: McGraw-Hill, 1961.
Tidd, W.H. "Demonstration of Bandwidth Capabilities
of Beyond-Horizon Tropospheric Radio
Propagation." Proceedings of the IRE, vol. 43
no. 10 (October 1955): 1297-1299.
"Tropospheric Scatter, Satellite, and VLF/LF
Communications." USAF Technical Training School
Study Guide and Workbook KEO 3000-119, Keesler
AFB, Ms., June 1979.
VanZandt, T.E., Green, J.L., Gage, K.S., and Clark,
W.L.. "Vertical Profiles of Refractivity
Turbulence Structure Constant: Comparison of
Observations by the Sunset Radar with a new
Theoretical Model." Radio Science, vol. 13, no.
5 (September-October 1978): 819-829.
Volz, A., and Kley, D. "Evaluation of the Montsouris

Series of Ozone Measurements made in the
Nineteenth Century." Nature, vol. 332 (March 17,
1988): 240-242.
Waterman, A.T. Jr. "Advanced Research in Propagation
Techniques Quarterly Report: Transhorizon
Propagation Techniques." U.S. Army sponsored
Research and Development Technical Report ECOM-
0138-2, October 1970.
Weitzen, J.A., and Ralston, W.T. "Meteor Scatter: An
Overview." IEEE Transactions on Antennas and
Propagation, vol. 36, no. 12 (December 1988):
1813-1819.
Wheelon, A.D. "Radio Scattering by Tropospheric
Irregularities." Journal of Atmospheric and
Terrestrial Physics, vol. 15, nos. 3/4 (October
1959) : 185-205.
Wiesner, Jerome B. "New Methods of Radio
Transmission." Scientific American, vol. 196,
no. 1 (January 1957): 46-41.
APPENDIX A
MULTIMODE PROPAGATION
Two modes for terrestrial, beyond the horizon

propagation are diffraction and troposcatter. While
most troposcatter paths are engineered to avoid
diffracting obstacles (unless diffraction is
desired), situations may dictate otherwise. Roda
[1988, p. 142] notes that at small scatter angles

(less than about one degree) diffraction can, for a
certain time percentage, dominate.

For short periods (minutes) the mechanism with
the lower loss is dominant if the two losses
differ by 6 dB or more. For longer periods
(days, months), in view of the variability of the
signals, the dominant mechanism is definitely
that with the lower loss if the median path
losses differ by more than 20 dB [Roda, 1988, p.
142].
While troposcatter signals exhibit deep fades,
diffracted signals remain comparatively stable

therefore limiting the effects of fading on some
links.
Freeman [1987, pp. 145,146] discusses two
ways to differentiate dominant propagation modes.
a. Assuming low (ground level) antenna

height, the transition distance (d) at
which distances exceeding d is
predominantly troposcatter (less than d

200
implies diffraction) is given in
kilometers by
d = 65(100/f)1 / 3 , (A-l)
where f is the operating frequency (MHz).
b. If the scatter angle exceeds 20 mrad

(about 1.1 degrees), diffraction effects
cease to be important.
Using (A-1) for small scatter angles, Table A.1 is
developea:
Table A.1. Mode Transition Distances

Frequency (MHz) Distance (km) Distance (mi)
500 38 24
1000 30 19
3000 21 13
5000 18 11
Other situations may allow diffraction
contributions to be disregarded. If more than three

diffracting obstacles are in the path, troposcatter
is the dominant mode. Also, at microwave frequencies
over a somewhat smooth earth, diffraction losses are
large enough to allow consideration of troposcatter
only [Monsen, et al., 1983, p. 2-55].

APPENDIX B
THE SYSTEM EQUATION
The troposcatter system equation, as provided
by Military Handbook, Facility Design for

Tropospheric Scatter [1977] (which borrows from Rice,
P.L., et al., [1967]), is used for path predictions
and can also serve to illustrate the impact of the
earth's atmosphere on troposcatter propagation. This
equation,
Pt = R + B + L, - Gp - 204 + F + Lbcr (dBW), (B-1)
specifies the required transmitter power, Pt, to

provide a specific carrier-to-noise ratio, R, at any
instant in time. Assume, for this example, that R
and the following parameters are fixed:
B bandwidth (dB);
L1 the sum of waveguide, duplexer, and other
system element losses (dB); and

Gp effective antenna gain (dB), this
parameter will also include degradations

introduced by aperture-to-medium coupling
loss described in Chapter 4.
The constant -204, termed Johnson noise power per
hertz, is found through (-10 log kT) with k being
Boltzman's constant and T as the ambient reference

202
temperature, usually 288.5 K. The remaining

parameters of (B-i), F and Lbcr, are of particular
interest since these include the uncontrollable
effects of the atmosphere.

F, the effective noise factor in dB, is found
through
F = 10 log f (dB) (B-2)
and
f= fam - 1 + icl 1 fr (B-3)
where
fam is the external noise factor,

ic is antenna circuit loss,
11 is transmission line loss, and
fr is the receiver noise.
All constituents of (B-3) are expressed in power
units. Critical in this analysis is fam which is

depicted on Figure B! in dB as Fam. While a variety
of noise sources factor into Fam, note the
contributions of sky noise due to oxygen and water
vapor. These atmospheric related contributors appear
negligible at troposcatter frequencies. However, the
sun, although not considered an atmospheric
component, merits mention since its effect can be
substantial (25 dB at 2 GHz) when it passes through

203
30
ESAAE
.JSINESS
ANREAS0 FO O 0 o
MAUO THIEIMREHOCOMM: REMVDH TDA

REEIIN - -EM SUN)E AQtiE
o ELEVTIO
ESTIMFEQUENC
BUINS AREA
Tropopheri ScatSe (Transhorion!A M)roa Syste
Desgn) MI-DJ-1 eatet0fDfne
~ NoebrT97.
Washington, ~ ~~ AaObl S NOC,
hog OM
M Lo
T PulCato
ofGffic
Navy 80 Avenue ENAN,%TE
Pa.R 19120-I094, p.RO
ChENeTERa 4-205.RNARO
BAM ExtrnlEoieNouce
FigureEAM
204
antenna beams. Therefore, although atmospheric noise
contributions are insignificant for troposcatter,

certain circumstances (direct solar noise) preclude
complete disregard of all external noise
contributors.
While external noise sources are, for the
most part, nominal, the atmosphere's effect on the
calculated reference basic transmission loss, Lbcr,

is more severe. For any instant in time, this loss
is expressed as
Lbcr = Lbf + Aa + Acr (dB) (B-4)
where
Lbf is basic free space loss in dB (Figure
B2),
Aa is oxygen and water vapor absorption in

dB (Figure 20), and
Acr is attenuation below that of free space
due to atmospheric and terrain features
in dB.
Acr is a rather versatile figure contingent on the
mode of propagation. Ignoring pure diffraction and
multimode possibilities, focus is on pure
205
IOOOO-
5000-
20500-
1000-
-00-
10 20 30 50 100 200 300 500 1000

PATH DISTANCE, d, IN km

Tropospheric Scatter (Transhorizon Microwave System
Design). MIL-HDBK-417, Department of Defense,
Office of Navy Publications, 5801 T bor Avenue,
Phildelphia, Pa., 19120-5094, p. 4-L7.
Figure B2: Basic Free Space Transmission Loss
206
troposcatter and Acr is then developed. 1

Tc conformn with ciLed work, Acr is renamea as As
(scatter loss) for a pure troposcatter path, and is
the attenuation relative to free space in dB:
A s = Lbsr - Lbf (dB). (B-5)
Note that the basic free space loss, Lbf, will drop
from (B-4) when used with (B-5). For typical
troposcatter links (exceptions as noted in Military
Handbook, 1977, p. 4-142) the reference basic
transmission loss, Lbsr, is
Lbsr = 30 log f - 20 log d + H o + F(Od) (dB), (B-6)
with f as frequency in MHz and d as path distance in

km. H o , the frequency gain function is significant
only for low effective antenna heights and will be
ignored for this analysis. (Military Handbook, 1977,
pp. 4-144, 4-145 describes H o for these rare
instances.)
The attenuation function, F(ed) is plotted as
a function of the product of the scatter angle, 0 in

iBy ignoring diffraction and multimode a
substantial portion of the path prediction tools
provided by Military Handbook, Facility Design for
Tropospheric Scatter (19771 and Rice, P.L. et al.
(1967] is neglected. This is more in line with the
thrust of this thesis (pure troposcatter); however,
the engineer cannot neglect other mechanisms unless
profile data dictate otherwise.
207
radians, and d (Figure B3). The scatter angle was
described in Section 1.3. Note che influence of Ns,

surface refractivity, a parameter described in
Section 2.3.1 and dependent on atmospheric

conditions.
As an example of atmospheric effects on the
calculated reference basic transmission loss, Lbcr,
consider a 200 km link, operating at 4.4 GHz, with a
scatter angle of 0.039 radians. Clear sky
atmospheric absorption, Aa, would cause about 1.5 dB
loss (see Figure 20) with rain adding no more than 2
dB (see Figure 23). However, F(Od) would range from
about 155 to 169 dB, depending on Ns, which in turn
forces variations of 3 to 17 dB for As. Overall,
Lbcr might vary from 4.5 dB (best case) to 20.5 dB
(worst case) for this example.
Concluding, while several parameters in (B-l)

are manageable, two parameters, F and Lbcr, exhibit
uncontrollable and, at times, damaging traits

contingent on atmospheric conditions. Fortunately,
both F and Lbcr are predictable and can be nearly

circumvented through engineered margins. It remains,
however, that atmospheric phenomena emerge as a
crucial consideration in troposcatter link design.

208
.
...
..
..
....
13 N. 10
I."I
10
§ 0 000 3 0 0 1 0
Source:~~Req
Mliay adooFc l Desnd
sa r
Scate
TrpspNi 10asorznMirwveSs
Desgn0 2I-DK4,Dprteto0Dfn
Washington,~~~~
25Nvmer.97.Aalbe DC.. hog~ ..
~ ~ an7om7etrAT:NO
Navy ~ ~ ~Pulcain
~ ~ 1
Ofic .av of Pulctos 580 Tao Avenue,
7hleiha Pa,1i0304 .4 13.
FiueB150d eru h Pout~

APPENDIX C
THE SPECTRUM OF IRREGULARITIES
Referring to Figure 15, the equation for the
spectrum function, S(k), often shown as cD(k), is of

the general form
S(k) = bCN 2 k - m (C-1)
within the inertial subrange. Discerning individual

parameter values is difficult as references provide
several variations [CCIR Report 563-3, 1986; Crane,

1988; Derr, 1972; Eklund and Wickerts, 1968; Flock,
1979; Hall, 1979; and Parl, 1979]; however, sources
are in general agreement with the following:
b - a constant, typically 0.033;
k - wavenumber, 2n/k;
m - the slope of the spectrum;
CN 2 - The structure constant.
CN 2 indicates the variability of amplitude

fluctuations of the refractive index. Its value,
valid below 2 km, ranges from 10- 1 5 to i0- 10 (m- 2 /3 )
(Boithias, 1987, p. 108; Hall, 1979, p. 44]. Several

formulas are available [Monsen, et al., 1983, p. 2-
19; Gerlach, 1984, p. 49]:

210
CN 2 = 8 x 10-14 h -1 /3 exp(-h/3200) (C-2a)
for m = 11/3, and, more generally,
{N(X I ) - N(X2)}2 = CN 2 r2 / 3 . (C-2b)
In expression (C-2a), h is the height above the
earth's surface to a scatterer in meters, and in

expression (C-2b) "N is the fluctuating refractive
index at two points X1 and X 2 separated by distance
r" [Gerlach, 1984, p. 49]. VanZandt, et al. [1978]
explore the structure function from a practical
perspective.
The spectrum slope, m, is a figure of
considerable uncertainty, values of which have

engendered much of the controversy in predicting
troposcatter paths. An often used figure of 11/3

originates by assuming an isotropic homogeneous
atmosphere applied to the turbulence theories of
Kolmogorov and Obukhov [Fehlhaber, 1968, p. 38; Lane,
1968, p. 10].1 However, Rice, et al. [1967, p. IV-7]
note shortcomings in the use of the 11/3 spectrum,
although Parl [1979] reaffirms otherwise. Most
lImmediately apparent is the error in

assuming isotropic scatterers; however, ease of
calculation forced this initial assumption and
consequences were felt negligible. Note that Figure
27 more realistically shows anisotropic scattering.
211
realistically, Gjessing and McCormick [1974] show m

as quite variable:
m = 40.6 + 708v 2 (C-3)
where
v2 = g/T(dT/dz + g/cp). (C-4)
Here v2 is the Vdis~ld-Brunt frequency used to

designate atmospheric stability, g indicates gravity,
T shows temperature, dT/dz is the temperature lapse
rate, and cp is specific heat at constant pressure.
On several links in Norway and Denmark, Gjessing and

McCormick's findings relate values ranging from about
5/3 to 20/3 with a mean near 11/3. Similar findings
are reported in Bolgiano [1958], du Castel [ed.,

1965], Cox [1969], and Sarkar, Dutta, and Reddy
[1983].
Referring again to Figure 15, a specific
value defined by
k = (4n/X)sin(0/2) (C-5)
is used (0 is the scattering angle). It is at this

value (wavenumber) that the spectrum must be
evaluated for troposcatter paths since it gives the

"intensity of the vertical spectrum" at the operating
212
wavelength [Lane, 1968, p. 9; Ortwein, Hopkins, and
Pohl, 1961, p. 789].

Finally, a summary of the features of the
spectrum is provided:
1. Variability of the form with time (hour to
hour and season) especially at low heights of
the common volume.
2. Variability of the slope of the spectrum with
wave-number range within the inertial
subrange.
3. Increased constancy with height [du Castel,
ed., 1965, p. 73].
Thus, variability is the key term describing
atmospheric turbulence.
APPENDIX D
DEPOLARIZATION
The depolarization mechanism, as depicted in
Figure Dl, results in received, copolarized (Ell and

E2 2 ) and crosspolarized waves (E2 1 and E1 2 )- "The
Ot~AA0VAIN - I
I I
TIANSMITIO RECEIVD
I VICT- TO
Iii.
I
!U - COPO.LARIZD WAVES
1,1 931 - CROSS.POLAAIZEO WAVES
Source: rppolito, L.J. jr. Radiowave

Propagation in Satellite Communications. riew York:
Van Nostrand Reinhold Co., 1986, p. 94.
Figure Dl: The Depolarization Mechanism
ratio of the power of the wanted or copolarized wave

to the power of the unwanted or crosspolarized wave"
is termed cross polarization discrimination (XPD)
[Flock, 1987, p. 4-42]:
XPD = 20 log (Eli/E1 2 ) (dB) (D-l)
This equation is applicable to the other fields (E2 2
and E2 1 ) in a similar manner. Additionally, higher

XPD values (such as 25 dB) are desirable.
Figure D2 shows depolarization due to rain.
Drop canting results from wind forces and "normally
.... .. . .. ............
. .....
. .......
... .. ........
........... ----- ... .
--------...
. ... ... .............. .
214
- 0 O0O
% 000
0 I
- 0 -
, 0 CICEIVID
0 I
WAVES
CANT(O
OILAIT SrIFOIDAL
Source: Ippolito, L.J. Jr. Radiowave

New York:
Propagation in Satellite Communications.
p. 96.
Van Nostrand Reinhold Co., 1986,
Figure D2: Depolarization Due to Rain
exhibit a distribution of canting angles" (Flock,

1987, p. 4-43]. Basically, the received waves
consist of vertical and horizontal components (co-
and crosspolarized waves) which sum to give ER1 and
ER11. Rain depolarization produces both attenuation
and phase differences which depend on droplet size

and shape. Furthermore, at decreasing frequencies
(below 10 GHz) depolarization remains a factor as the

phase difference remains high at low frequencies
(Flock, 1987, p. 9-33].
Figure D3 shows XPD changes on earth-space
paths with respect to copolarized attenuation due to

rain. Notably, troposcatter links, which normally
215
20 / - .
CA:vLLA
1N?.
35 A~XI /30
-~ C141
0 0 0 30 ' 10 20 30 40 10 20 30 40
CO.POL&A.10 ArTEN. CO.POLAA.Z!2 ArrEN. CO.POLMRZED ATlEN.

INCco INcil IN :S
Source: Ippolito. L.J. Jr. Radlowave

Propagation in Satellite Communications. New York:
Van Nostrand Reinhold Co., 1986, p. 99.
Figure D3: Depolarization Effects Due to Rain
operate below 5 GHz, will experience negligible XPD

degradation.
APPENDIX E
THE REFLECTION COEFFICIENT
Determining the reflection coefficient, p,

for tropospheric layers offers unique complexities
due to variability of the atmosphere. Rather than
presenting a smooth, flat surface, atmospheric layers
exhibit dimensional irregularity and constant motion
which compound normally simple calculations.
It can be shown that for an abrupt change in

the atmosphere's refractive index, An, Fresnel's
equation for reflection can be reduced to
P0 = An/2a 2 (E-l)
for 2An o a 2 [Picquenard, 1974, p. 491. Here, a is

the grazing angle. More often, however, layers can
be formed by less abrupt changes in refractivity and
po will be reduced by a form function,
Fp = sin (47raAh/X) /4OaAh/X, (E-2)
providedAn o 7r 3 Ah/X and layer thickness < )/(4a)

[Hall, 1979, p. 39]. This gives:
p = PoFp (E-3)
which is normally much less than unity [Hall and
Barclay, 1989, p. 168].

217
Exceptionally important for purposes of
differentiating troposcatter propagation mechanisms:

"[b]ecause the magnitude of F is inversely
proportional to AhA, layer reflection is not of much
consequence above UHF" [Hall, 1979, p. 40, emphasis

mine]. Finally, Hall [1979, p. 40] shows that for
troposcatter paths where there are a number of
reflecting surfaces, N:
p = {2N/dX f p0 2 F 2
2
2 22dV1 / 2 (E-4)
where 12 is the shorter length of the Fresnel

ellipsoid illuminating the reflecting layer and V is
the common volume.

APPENDIX F
CLIMATIC REGIONS
Troposcatter link performance varies

considerably in different parts of the world. For
this reason, some prediction techniques have divided
the planet into various regions depending on
climatology. From there, analysis is made for the
specific region in question. Analysis in NBS Tech
Note 101 [Rice, et al, 1967] provided the first
consolidated effort to categorize according to

climate. Since then, changes have been made as
empirical studies demanded modification. Both
Scatter (Transhorizon Microwave System Design) [1977]
and CCIR Report 238-5 [CCIR, 1986] provide agreement
on a current division scheme. The following bullet

summary is extracted directly form the CCIR report,
all divisions are by latitude in both hemispheres
unless indicated otherwise.
1. EQUITORIAL:
- Between 100 N and 100 S.
- Slightly varying high temperature.
- Monotonous heavy rain.
- Permanent humidity.
2. CONTINENTAL SUB-TROPICAL:
- Between 10 and 20".
- Dry winter.
- Rainy summer.
219
3. MARITIME SUB-TROPICAL:
- Between 101 and 20).
- Lowland areas near the sea.
- Influenced by the monsoon.
4. DESERT:
- Two land areas roughly between 200 and 300.
- Yearly, semi-arid conditions.
- Extreme diurnal and seasonal temperature
variations.
5. MEDITERRANEAN:
- On desert zone fringes, near the sea, between
300 and 400.
- Fairly high temperature.
- Marked influence by the sea.
- Nearly no rain in summer months.
6. CONTINENTAL TEMPERATE:
- Between 30u and 600.
- Pronounced diurnal and seasonal variations.
- Western portion:
-- Strongly influenced by sea.
-- More moderate temperature variations.
-- Year round rain.
- Eastern portion:
-- More temperature variation.
-- Less rain in winter.
- Can include Japan.
7a. MARITIME TEMPERATE, OVERLAND:

- Between 301J and 60w where prevailing winds are
unobstructed by mountains bringing moist
maritime air inland.
- Typical regions:
-- U.K.
-- North American west coast.
-- European west coast.
-- African northwest coast.
7b. MARITIME TEMPERATE, OVERSEA:

- Similar to 7a, focus on costal and oversea
areas.
- Propagation path oversea although terminals
inland.
8. POLAR:
- Between 600 to poles.
- Low temperature.
- Low humidity.

Understanding Troposcatter Propagation

Uploaded by

Copyright:

Available Formats

Understanding Troposcatter Propagation

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Understanding Troposcatter Propagation

Uploaded by

Copyright:

Available Formats

D~-:F OPY I

4. PERFORMING ORGANIZATION REPORT NUMBER(S) 5. MONITORING ORGANIZATION REPORT NUMBER(S)

Wright-Patterson AFB OH 45433-6583

12. PERSONAL AUTHOR(S)

19. ABSTRACT (Continue on reverse if necessary and identify by block number)

20. DISTRIBUTION/AVAILABILITY OF ABSTRACT

JOSEPH HENDERSON REYNOLDS

B.S., United States Air Force Academy, 1982

A thesis submitted to the Faculty of the Graduate

School of the University of Colorado in partial

fulfillment of the requirements for the degree of

SUBJ: Thesis Abstract

TO: AFIT/CIRK (Capt Battlesl

--- Although tioposcattef communications systems, suffer several-shortcomings, this

Theoretical understanding of the trfoposcattei, propagation mechanism is rooted

Most troposcatter prediction methods are rooted in empirical expressions.. The

Author: Joseph H. Reynolds

JSEPH H. REYNOLDS, Capt, USAF

Boithias, L. Radio Wave Propagation. St. Louis: McGraw-Hill, 1987.

Brauburger, G.H. Jr. (Capt. USAFR). "Meteorological Factors Affecting AN/TRC-

CCIR. Volume V. Propagation in Non-Ionized Media. Dubrovnik, 1986.

Crane, R.K. "A Review of Transhorizon Propagation Phenomena." Radio Science,

Crawford, A.B., Hogg. D.C., and Kummer W.H. "Studies in Tropospheric

du Castel, F. Tropospheric Radiowave Propagation Beyond the Horizon. New

Hall, M.P.M. Effects of the Troposphere on Radio Communication. New York:

Military Handbook, Facility Design for Tropospheric Scatter (Transhorizon

Monsen, P. "Fading Channel Communications." IEEE Communications Magazine,

Neiburger, M., Edinger, J.G., Bonnet, W.D. Understanding our Atmospheric

Pantei, P.F. Comxunication Systems Design: Line-of-Sight and Tropo-Scatter

Pail, S. A. "New Formulas for Tropospheric Scatter Path Loss," Radio

Roda, G, Tropo)catter Radio Links. Norwood, Maine: Artech House, 1988.

Sarkar. S.K., Dutta. H.N., and Reddy, B.M. "Development of a Prediction

Capt Joseph H. Reynolds

The following is a conservative estimate of the money I used to complete

TOTAL COSTS 498.90

Again, I emphasize this as being quite conservative. Unfortunately I

IOSEPH H. REYNOLDS, Capt, USAF

Joseph Henderson Reynolds

has been approved for the

Understanding Troposcatter Propagation

Thesis directed by Professor Ernest K. Smith

Early notions that troposcatter transmission

could replace line-of-sight (LOS) communications

links quickly diminished due to high costs, limited

bandwidth, and the onset of satellite communications.

Recently, however, digital processing schemes have

reduced bandwidth concerns; furthermore, troposcatter

systems have found consistent use in less developed

countries and the military. This thesis discusses

troposcatter propagation with emphasis on

Theoretical understanding of the troposcatter

propagation mechanism is rooted in atmospheric

phenomena; specifically -- the index of refraction

and turbulence. It is shown that turbulent eddies

which encompass small-zcale refractivity variations

inhabit the common volume, an area defined by the

intersection of antenna beamwidths. Further, these

turbulent inhomogenities can arrange themselves as

randomly disposed blobs or layers. This engenders