Comparing Atmospheric Correction Methods For Landsat Oli Data

International Journal of Remote Sensing and Earth Sciences Vol.13 No.
2 December 2016: 105 – 120
COMPARING ATMOSPHERIC CORRECTION METHODS FOR

LANDSAT OLI DATA
Esthi Kurnia Dewi*) and Bambang Trisakti

Remote Sensing Application Center, LAPAN
*)E-mail: esthikd@gmail.com
Received: 29 June 2016; Revised: 27 July 2016; Approved: 30 August 2016
Abstract. Landsat data used for monitoring activities to land cover because it has spatial resolution
and high temporal. To monitor land cover changes in an area, atmospheric correction is needed to be
performed in order to obtain data with precise digital value picturing current condition. This study
compared atmospheric correction methods namely Quick Atmospheric Correction (QUAC), Dark Object
Subtraction (DOS) and Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH). The
correction results then were compared to Surface Reflectance (SR) imagery data obtained from the
United States Geological Survey (USGS) satelite. The three atmospheric correction methods were
applied to Landsat OLI data path/row126/62 for 3 particular dates. Then, sample on vegetation, soil
and bodies of water (waterbody) were retrieved from the image. Atmospheric correction results were
visually observed and compared with SR sample on the absolute value, object spectral patterns, as
well as location and time consistency. Visual observation indicates that there was a contrast change
on images that had been corrected by using FLAASH method compared to SR, which mean that the
atmospheric correction method was quite effective. Analysis on the object spectral pattern, soil,
vegetation and waterbody of images corrected by using FLAASH method showed that it was not good
enough eventhough the reflectant value differed greatly to SR image. This might be caused by certain
variables of aerosol and atmospheric models used in Indonesia. QUAC and DOS made more
appropriate spectral pattern of vegetation and water body than spectral library. In terms of average
value and deviation difference, spectral patterns of soil corrected by using DOS was more compatible
than QUAC.
Keywords: Landsat, atmospheric correction, QUAC, FLAASH, DOS, surface reflectance
1 INTRODUCTION Medan; Tuni et al. (2013) classified land

Over the last decades, Indonesia has closure or land used in North Halmahera;
experienced significant changes on its and Ceccarelli et al. (2013) who examined
land cover due to illegal logging and land land cover in Oristano and Campania,
opening, as well as peat land combustion. Provinces of Rome.
Remote sensing data were widely used for However, Landsat imagery data that
land cover monitoring activities. They were had been recorded by Indonesian
Ginting et al. (2012) who conducted Aeronautics and Space Agency (LAPAN)
research on land cover by utilizing remote could not be used directly to extract land
sensing data, particularly Landsat data to cover information, since the data
monitor in Karo Regency; and Yulius et al. contained noise caused by atmospheric
(2014) who monitored land cover changes effects. Atmospheric noises that might
on Bungus Teluk Kabung Bay. afffect the quality of remote sensing
Furthermore, Tampubolon et al., (2015) imagery were molecules and aerosols
analyzed the changes on critical land of scattering, water vapor absorbtion, carbon
@National Institute of Aeronautics and Space of Indonesia (LAPAN) 105

Esthi Kurnia Dewi and Bambang Trisakti
dioxide, oxygen, ozone, as well as light historical data, and suggest DOS-Iteration
effect and transmittance on the since the method gathered broader
atmosphere (Zhang et al., 2010). spectral range. Lee et al. (2000) who
To eliminate noise which might cause examined Landsat 5 Data with FLAASH,
image distortion, sun correction and Foster and 6S models found that FLAASH
atmospheric correction needed to be done. was the best model. Nguyen et al (2015)
In relation to position of the sun. Sun who reviewed QUAC, FLAASH and 6S
correction was performed by converting models with Landsat 5 Data stated that
digital number (DN) to reflectance value. 6S was the best model which resembled
Then Atmospheric correction was done to on ground condition. Afterall, this
reduce or eliminate atmospheric effect and research aimed to analyze a suitable
to obtain reflectant value on the suface. atmospheric correction model for Landsat
Some atmospheric correction methods OLI data in order to retrieve Indonesian
were, for example, digital number (DN) land cover information which
used directly for transformation reflection corresponded to actual coverage.
by Smith et al (1999), Dark Object To achieve the above result, an
Subtraction (DOS) by (Zhang et al., 2010; analysis was performed on some
Trisakti et al., 2014), Fast Line-of-sight atmospheric correction methods, i.e.
Atmospheric Analysis of Spectral FLAASH, QUAC and DOS for Landsat OLI
Hypercubes (FLAASH), as well as Quick path/row 126/62 in certain area of study
Atmospheric Correction (QUAC), 6S, and compared to USGS’s Surface Reflectance
ATCOR 2-3. data which had corrected. This research
This research utilized atmospheric was aimed to find the most accurate and
correction method which compatible with consistent method in accordance with
data quality, data availability, and actual condition.
research purposes. Some previous studies
on the correction method had been 2. MATERIALS AND METHODOLOGY
performed. Such as Somdatta et al. (2011) 2.1 Data
who implemented both FLAASH and Remote sensing data used in this
QUACC methods on hyperion data. They study were Landsat 8 level LT1 which was
conclude FLAASH was better than QUAC recorded by LAPAN Remote Sensing
in terms of atmospheric correction result. Center, Parepare and Rumpin with 3
Guo et al. (2012) who studied the spot different dates and Surface Reflectance
data with both methods found that data from USGS as the comparative data
FLAASH was more effective to reduce (Table 2-1).
noise than QUAC. Furthermore, Nazeer et
al. (2001) who implemented 6S, FLAASH, Table 2-1: Satelite images used in this research
ATCOR, DOS, and ELM methods on
Landsat 7 Data and compared the results Num Data Explanation Path/Row
with Surface Reflectance (SR) data from ber Acquisition
USGS, found that DOS and 6S generated 1 June 21, 2014 Landsat 8 126/62
the most appropriate and consistent 2 July 26, 2015 Landsat 8 126/62
result. Cui et al. (2001) also conducted a 3 May 9, 2016 Landsat 8 126/62
study on Landsat TM 1, Landsat TM 2, 4 July 26, 2015 USGS 126/62
Landsat TM 3, Landsat TM 4, Landsat TM Surface
5, and Landsat TM 7 with DOS, Reflectance
DOSCOST, DOS-Constant, and DOS-
Iteration. At last, Cui et al choose DOS as For spectral analysis purpose,
the best model to acquire quantitative and sampling process was carried out on
106 International Journal of Remote Sensing and Earth Science Vol. 13 No. 2 December 2016
Comparing Atmospheric Correction Methods for Landsat OLI Data
vegetation (forests), water and soil. The surface temperature and physical surface
data used were acquired in May, June, reflection.
July or during dry season with 1 to 7 FLAASH began with a standard
spectral bands. The spectral bands were equation of spectral glow from single pixel
choosen since they were often used in received by standard Lambertian planar
classifying land cover. The area analyzed (the closest Lambertian planar), which
was Lake Kerinci surroundings, which was accepted by a sensor based on solar
included the lake itself, sea, agricultural spectrum (excluding thermal radiation),
lands, and soil. with the following formula:
Sample observed in this study were
vegetation (forest), body of water and soil, (2-1)
each was selected from three sample
locations (Figure 2-1). Sample taken were
then compared to Surface Reflectance L was the light for single pixel received by
data. Reference value of Landsat OLI spectral sensor; ρ was pixel surface reflectance; ρe
for vegetation, water and settlement were was average surface reflectance for the
shown on Figure 2-2 below. pixel and its surrounding; S was the
accumulation of sunlight reflection and
diffusion by atmospheric particles; La was
the radiance when atmospheric radiation
penetrated sensor. A and B were
coefficient determined by atmospheric and
geometric conditions of the underlying
surface but had nothing to do with
surface reflectance. (Aρ / (1-ρe S))
Figure 2-1: Sampling on water (a), soil (b),
represented radiation energy which
vegetation (c) on Landsat imagery penetrated directly into the sensor from
the target’s surface, which indicated two
cases: reflection occurs when the sun
illuminates target surface; or the
surrounding surface reflected through the
atmosphere and shined in the target
surface target before another reflection.
(Bρe / (1-ρe S)) showed total radiation that
went into sensor from the surface through
the atmosphere. ρ and ρe explained
"proximity effect" of mixed pixel near the
radiation caused by atmospheric
scattering, with the assumption ρ = ρe.
Figure 2-2: Landsat 8 object profile of water, soil However, significant error might occur due
and vegetation Source: Journal
Optics Express, 2014 to fog or strong contrast on the surfaces.
According to Equation 1, surface
2.2 Methods reflection could be calculated pixel by
2.2.1 FLAASH Model pixel. FLAASH used average radian
FLAASH was an atmospheric spatial, ignoring "proximity effect", to get
correction tool to fix visible band, NIR, estimate equation (2) and to predict
SWIR. This model eliminated air and light spatial reflectance. While, Le was the
influences, as well as removed interference average radiant image generated from
due to reflectivity parameter, emissivity, imagery radian and spatial function.
International Journal of Remote Sensing and Earth Science Vol. 13 No. 2 December 2016 107
detected with shadows or cloud free basic

(2-2)
scene. It mean that this correction would
run faster. Below was QUAC Model
Most of the atmospheric correction flowchart (Figure 2-3).
parameters used in this experiment were
metadata of Landsat image, and specific
parameter data as shown in Table 2-2.
After getting the required parameters, the
actual surface reflectance of all imageries
were calculated using Equation 1 and
Equation 2.
2.2.2 QUAC Model

QUAC was an atmospheric correction
method for hyperspectral and
multispectral of visible band, NIR and
SWIR. QUAC initial principles differed
from usual atmospheric correction
method, because this approach was based
on the values of light which penetrated the
scene. QUAC was known as an empirical Figure 2-3: QUAC Model Flowchart
approach based on recording. This Source: ITT Visual Information Solutions (ITT
VIS), “ENVI User’s Guide, Version 4.8”
determined kinds of parameter received
from the atmosphere directly during the
QUAC also utilized the sun elevation angle
recording, without additional information.
and central wavelength. If the sensor did
not have precise radiation or wavelength,
=( + +⋯ + )/ (2-3)
or the intensity of sunlight was unknown,
corrections could still be performed by
The above equation showed QUAC model
using this method within allowed scope of
based on the experience of gathering
accuracy.
average reflectance from diverse content
such as spectrum of each section, n
indicated the amount of spectrum
Table 2-2: Paramater used in FLAASH atmospheric correction
Height of Atmosphere Sun Ground

Data Imagery Time Sensor Visibility
Sensor Model Elevation Elevation
0.78
June, 21 2014 03:23:43 Landsat 8 OLI 705 km Tropical 52.23908383 40 km
km
0.78
July, 26 2015 03:23:36 Landsat 8 OLI 705 km Tropical 53.70803343 40 km
km
0.78
May, 9 2016 03:23:34 Landsat 8 OLI 705 km Tropical 56.82897852 40 km
km
2.2.3 DOS Model dark gray object pixel from grey value of
DOS was an image-based each pixel on the image.
atmospheric correction. Chavez (1996)
stated a basic assumption that few pixels
radiances on covered by cloud image "# = "min + "%& % (2-6)
could be accepted by the satellite due to
atmospheric scattering (path radiance).
Considering the fact that there was very Which:
few target on the Earth surface was Lmin = the light that corresponded to
absolute black, it was assumed that one digital value which was the total
percent of minimum reflektansi was better of all pixel which digital sum
than zero percent. was lower or min DN value.
Landsat 8 data which was LD01% = radiance of dark object which
radiometrically corrected including ToA was assumed to have 0.01
Reflectance dan sunlight correction. ToA reflectance value
Reflectance was corrected by converting
the DN value to reflectance value. Based To calculate Landsat Imagery
on USGS (2014), ToA Reflectance equation
was:
"() = *" ∗ %,() + " (2-7)
'=M + (2-4)
To calculate Object Radiance
Which:
' = ToA Reflectance, without correction
(2-8)
of sunlight angle.
Mρ = REFLECTANCE_MULT_BAND_x, in
which x is Band number So to calculate path radiance of the dark
= REFLECTANCE_ADD_BAND_x, in object
which x is Band number
Qcal=Digital Number (DN) value
(2-9)
Then, image correction was conducted
due to sunlight angle change, to eliminate Assumed that
DN value difference, with the following Tv =1
equation: Tz =1
Edown = 0
(2-5)
= '/(cos(θSZ))= '/(sin(θSE))
Untuk mencari ESUN Landsat 8 OLI
Which: adalah sebagai berikut
! = ToA Reflectance
-./,
SE = sun elevation = (0 ∗ 1 )
SZ = zenith angle of the sun, where SZ = 2 1) 3_* 5)(6( (2-10)
90 ° - SE ∗
237 3 8 3_* 5)(6(
After the sun elevation was

approved, then correction was also Analysis stages of three atmospheric
performed on dark pixel. DOS assumes correction models (FLAASH, QUAC and
horizontal atmosphere on similar images DOS) compared to Surface Reflectance
needed a black target by substracting
data from the USGS were as follow (Figure 2-4).
Figure 2-4: Step To Compare Atmospheric Correction
3 RESULTS AND DISCUSSION

To analyze and compare the results
of FLAASH, QUAC, and DOS atmospheric contrast ratio, which showed that
correction models with SR on Landsat 8 the atmospheric correction was quite
imagery data, it took visual review on the effective. These results were in accordance
models after and before the correction, as with previous research performed by
well the reflectance feature of surface Rahayu et al. (2001), which found brighter
spectral curve. visual effect.
When the results of atmospheric
3.1 Visual Contras Analysis Before and correction were visually compared,
After Correction FLAASH image had the highest contrast
Figure 3-1, 3-2 and 3-3 a, b, c, d are ratio. This fitted to Bong et al (2015)
Landsat imageries with 3 different dates research finding that FLAASH model
before and after FLAASH, QUAC, and DOS visually showed sharp contrast ratio
atmospheric correction with 6, 5, and 2 compared to QUAC model.
RGB bands combinations. Figure 3-4 was Surface Reflectance (SR) of USGS
the Surface Reflectance of the USGS was utilized as comparative model, since it
imagery in combination with 6, 5, and 2 best represented on field condition. In
RGB Bands. Visible changes were seen on term of contrast ratio, image that was
the image, before and after correction. corrected with FLAASH model had the
Before corrected, the image was darker most similarity to SR. However, visual
due to the presence of atmospheric effects judgment only could not decide clearly
that reduced the image’s contrast ratio of whether FLAASH was the best model to
light and dark. After correction, imagery represent the real condition. Thus, it took
data got brighter and clearer. The quality deeper analysis on spectral reflectance
of image was improved due to better
Comparing Atmospheric Correction Methods
M for Landsat OLI Data
curve by testing the quality

lity of images on
each sample, region, and period
per of time.
(a) Before Correction (b) DOS (c) FLAASH (d) QUAC

Figure 3-1: Visual
ual analysis on Landsat 8 image acquired on June
e 21,
2 2014
(a) Before Correction (b) DOS (c) FLAASH (d) QUAC
Figure 3-2: Visual

al Analysis
A on Landsat 8 Image Acquired on July 26
26, 2015
(a) Before Correction (b DOS

(b) (c) FLAASH (d) QUAC
Figure 3-3: Visual
al Analysis on Landsat 8 Image Acquired on May 9, 2016
Figure 3-4: Visual Analy

alysis on Surface Reflectance Image Acquired on Ju
July 26, 2015
International Journal of Remote Sensin

sing and Earth Science Vol. 13 No. 2 December 2016 111
3.2 Analysis Based on Spectral In terms of average difference

Reflectance Curve against SR model, it was concluded that
Further analysis was conducted by the most approaching model was DOS
observing the spectral pattern based on with difference value of 0.009, followed by
sampling test. Spectral results among QUAC with 0.06 and FLAASH with 0.18.
samples with spectral consistency In terms of data diversity, then deviation
between regions and time, was observed. difference of DOS model against SR was
the most similar with 0.0027, followed by
FLAASH with 0.034. QUAC had the
3.2.1 Spectral curve between sampling highest diversity of 0.076 (Figure 3-8).
The spectral pattern of water body For vegetation (forest) sampling,
samples corrected by using FLAASH, spectral pattern of images corrected using
QUAC and DOS models were compared FLAASH, QUAC and DOS models were
with SR image. It was seen that QUAC compared to SR. It showed that DOS had
and DOS images had almost similar almost the same spectral pattern to SR,
pattern with SR image, while FLAASH had followed by QUAC. FLAASH image had the
higher reflectance value (Figure 3-5). highest reflectance value (Figure 3-9).
When the spectral curve of the three When average difference value was
corrected images were compared in terms subtracted from the above models, it
of average difference to SR model, DOS could be concluded that DOS was the
model had the most nearly similar value of most approaching model by 0.0049
0.002, followed by QUAC with 0.004 difference, then QUAC by 0.043, and
difference value, and FLAASH was the FLAASH by 0.128. As data diversity was
least similar with 0.05 difference value. In reviewed, then DOS had the least
terms of data diversity, then image with deviation difference against SR by 0.0035,
the least deviation difference value against followed by QUAC of 0.053. FLAASH had
SR was DOS of 0.001, and followed by the highest diversity of 0.123 (Figure 3-
QUAC of 0.0029. Then FLAASH had the 10).
highest diversity of 0.053 (Figure 3-6). The results of this study was in
Spectral pattern of soil sample images accordance with the research result of
corrected using FLAASH, QUAC and DOS Nazer et al (2014), which stated that DOS
models were compared with SR image. It and 6S had the most appropriate and
showed that DOS image had almost the consistent value with SR compared to
same spectral pattern with SR image, FLAASH and QUAC models.
followed by QUAC and FLAASH with
higher reflectance value (Figure 3-7).
Reflectance (-)
Band (-)
Figure 3-5: Visual Analysis of Surface Reflectance of Water in Image July 16, 2015
Deviation differences to SR (-)

Average differences to SR (-)
Figure 3-6: Average and Deviation Difference of Water against SR
Figure 3-7: Visual Analysis of Surface Reflectance of Soil in Image July 16, 2015
Figure 3-8: Average and Deviation Difference of Soil against SR
Reflectance (-)
Band (-)
Figure 3-9: Visual Analysis of Surface Reflectance of Vegetation in Image on July 26, 2015

Figure 3-10: Average and Deviation Value of Vegetation against SR
3.2.2 Spatial spectral curve similar visible bands, however absolute

Analysis on water sampling band value of NIR and SWIR change.
indicated that QUAC, SR, and DOS had FLAASH had almost the same pattern to
almost similar average value. It mean that SR, but higher absolute value (Figure 3-
all samples were taken from similar 12). When it was reviewed, the spectral
location (Figure 3-11). As it is observed, profile of settlement or soil taken from 3
water spectral profile taken from 3 different locations had almost similar
different locations were almost identical to value with SR (Figure 3-13), which mean
SR, which mean that the samples were that samples were taken correctly and
taken correctly and consistently. FLAASH consistently (Figure 3-14).
image also had similar and consistent Spatial analysis on vegetation
pattern, but there was a difference in (forests) sampling indicated that SR and
absolute value of visible band. DOS had almost the same average rating.
Spatial analysis on settlement or soil It mean that the samples were retrieved
sampling showed that SR and DOS had from similar location. In terms of band
almost the same average rating; which visible QUAC was similar to SR, but
mean that samples were retrieved from absolute value on NIR and SWIR bands
the same location. QUAC and SR had changed. FLAASH had almost the same
pattern, but higher absolute value (Figure Inter (spatial) location observation
3-15). Meanwhile, when spectral profile of here was different to research conducted
vegetation (forest) in 3 different sample by Bongetal (2015) which found that
locations were reviewed, all had almost FLAASH might generate better results
identical value to SR; which mean that than QUAC. This might happen since the
sampling locations were correct and type of aerosol variable and atmospheric
consistent (Figure 3-16). conditions used in Indonesia differed in
previous research areas.
Reflectance (-)
Band (-)
Figure 3-11: Average Surface Reflectance of Water in Three Sample Locations
(a) SR from USGS (b) DOS method
1 1
2 2
Reflectance (-)
Reflectance (-)
3 3
Band (-) Band (-)

Figure 3-12: Water Spectral Profile of Three Sample Locations for SR and DOS
Reflectance (-)
Band (-)
Figure 3-13: Average Surface Reflectance of Soil in Three Sampling Locations
1 1
Reflectance (-)
Reflectance (-)
2
3 3
Band (-) Band (-)

Figure 3-14: Soil Spectral Profile of 3 Sample Locations for SR and DOS
Reflectance (-)
Band (-)
Figure 3-15: Average Surface Reflectance of Vegetation in Three Sampling Locations
1 1
Reflectance (-)
2
Reflectance (-)
2
3 3
Band (-) Band (-)

Figure 3-16: Vegetation Spectral Profile of Three Sample Locations for SR and DOS
3.2.3 Temporal spectral curve almost the same and the spectral pattern
The temporal analysis of images was in accordance with the spectral
corrected using DOS method showed that library, which mean that temporal
absolute value of water, settlement (soil), sampling was consistent and the value
and vegetation (forest) sampling were was correct (Figure 3-17, 3-18, 3-19).
2014
Reflectance (-)
2015
2016
Band (-)
Figure 3-17: Water Spectral Profiles using DOS in Different Years
2014
Reflectance (-)
2015
2016
Band (-)
Figure 3-18: Soil Spectral Profiles using DOS in Different Years
Reflectance (-)
2014
2015
2016
Band (-)
Figure 3-19: Vegetation Spectral Profiles using DOS in Different Years
4 CONCLUSION enough, although it could distinguish

From this analysis, it could be samples due to FLAASH and SR had great
concluded that visual observation difference on the reflectance value. This
indicated visible contrast change on was due to the type aerosol variable and
corrected image. Atmospheric correction atmospheric conditions of Indonesia. Both
on all models effectively reduced noise. QUAC and DOS had more appropriate
Visually compared, FLAASH model water body and vegetation spectral
showed the most similarity with SR image. patterns. However, in terms of soil
However, the best model could not be spectral pattern as compared to the
determined by visual judgment only. It average difference and deviation with SR,
took more evaluation on spectral values. DOS model had more suitable object
Spectral value of object corrected by DOS spectral pattern compared to QUAC.
method was similar to USGS’s SR with
different absolute value of water body ACKNOWLEDGEMENT
0.002, soil 0.009 and vegetation 0.004, yet We thank Remote Sensing
DOS had consistency in terms of location Application Center for support and
and region. Absolute value difference facilities, Remote Sensing Technology and
between DOS and SR was 0.001. QUAC Data Center for data Landsat OLI, and all
also had similar pattern to SR, yet did not reviewers for their valuable inputs.
have location and time consistency.
FLAASH had the highest absolute value, REFERENCES
compared to the other models. FLAASH Caselles V., García MJL, (1989), An alternative
had location and time consistency, but simple approach to estimate atmospheric
when compared to SR it had significant correction in multitemporal studies.
difference on spectral pattern of visible International Journal of Remote Sensing
bands, especially on blue band range. 10(6): 1127-1134. doi: http://dx.doi.org/
Analysis on spectral pattern of 10.1080/01431168908903951.
vegetation, soil and water body showed Ceccarelli T., Smiraglia D., Bajocco S., Rinaldo
FLAASH method was not compatible S., Angelis AD, Salvati L., Perini L., (2013),
Land cover data from Landsat single-date Nazeer M., Nichol JE, Yung Y., (2014), Evaluation
imagery: an approach integrating pixel- of atmospheric correction models and
based and object-based classifiers. Landsat surface reflectance product in an
European Journal of Remote Sensing 46 : urban coastal environment. International
699 – 717. Journal of Remote Sensing 35(16): 6271 –
Chavez PS, (1988), An improved dark-object 6291. doi: 10.1080/01431161.2014. 951742.
subtraction technique for atmospheric Nguyen HC, Jung J., Lee J., Choi S., Hong S.,
scattering correction of multispectral data. Heo J., (2015), Optimal Atmospheric
Remote sensing of Environment 24 (3): Correction for Above-Ground Forest
459-479. doi: http://dx.doi.org/10.1016/ Biomass Estimation with the ETM+ Remote
0034-4257(88)90019-. Sensor. Sensors 15(8): 18865–18886. doi:
Chavez PS, (1996), Image-based atmospheric 10.3390/s150818865.
corrections-revisited and improved. Rahayu, Candra DS, (2014), Koreksi Radiometrik
Photogrammetric engineering and remote Citra Landsat-8 Kanal Multispektral
sensing 62 (9): 1025-1036. Available at: Menggunakan Top of Atmosphere (Toa)
http://www.unc.edu/courses/2008spring/ untuk Mendukung Klasifikasi Penutup
geog/577/001/www/Chavez96-PERS.pdf/ Lahan. Seminar Nasional Penginderaan
(lastaccessed: 24/10/2016). Jauh.
Cui L., Li G., Ren H., He L., Liao H., Ouyang N., Smith GM, Milton EJ, (1999), The use of the
Zhang Y., (2014), Assessment of empirical line method to calibrate remotely
atmospheric correction methods for sensed data to reflectance. International
historical Landsat TM images in the coastal Journal of Remote Sensing 20 (13): 2653-
zone: A case study in Jiangsu, China. 2662,doi: http://dx.doi.org/10.1080/0143
European Journal of Remote Sensing 47: 11699211994.
701-716. doi: 10.5721/EuJRS20144740 Somdatta C., Chakrabarti S., (2011), Pre-
Ginting AY, Latifah S., Rahmawaty, (2012), processing of Hyperspectral Data: A case
Analisis Perubahan Tutupan Lahan study of Henry and Lothian Islands in
Kabupaten Karo (Analysis of Karo Regency Sunderban Region, West Bengal, India.
Land Cover Changes). Peronema Forestry International Journal of Geomatics And
Science Journal 1(1). Geosciences 2(2).
Guo Y., Zeng F., (2012), Atmospheric Correction Tampubolon T., Jeddah Y., (2015), Aplikasi
Comparison of Spot-5 Image Based on Pemanfaatan Citra Satelit Landsat untuk
Model Flaash and Model Quac. Mengidentifikasi Perubahan Lahan Kritis di
International Archives of the Photogrammetry, Kota Medan danSekitarnya. Spektra:
Remote Sensing and Spatial Information Jurnal Fisika dan Aplikasinya 16(2): 15-19.
Sciences, XXXIX-B7: 7-11 Trisakti B., Suwarnana N., Cahyono JS, (2014),
Lee SB, La HP, Eo YD, Pyeon MW, (2015), Pemanfaatan Data Penginderan Jauh
Generation of Simulated Image from untuk Memantau Parameter Status
Atmospheric Corrected Landsat TM Images. Ekosistem Perairan Danau (Studi Kasus:
Journal of the Korean Society of Surveying Danau Rawa Pening). Seminar Nasional
Geodesy Photogrammetry and Cartography Penginderaan Jauh.
33(1): 1-9. Tuni MS, Barus B., Iskandar, (2013), Prediksi
Moran E., (2002), Assessment of atmospheric Perubahan Tutupan Lahan dan
correction methods for Landsat TM data Perencanaan Penggunaan Lahan
applicable to Amazon basin LBA research. Pascatambang Nikel di Kabupaten
ACT Publication No 02-06. Halmahera Timur. Globe 15(2): 146 – 152.
Vincent RK, (1972), An ERTS multispectral Sistem Informasi Geografis. Jurnal Ilmu
scanner experiment for mapping iron dan Teknologi KelautanTropis, 6 (2): 311-
compounds. Proceedings of the Eighth 318.
International Symposium on Remote Zhang X., Yang H., Shuai T., (2010). Comparision
Sensing of Environment, Ann Arbor, of FLAASH versus Empirical Line Approach
Michigan, II: 1239-1247. Available at: for Atmospheric Correction of OMIS-II
http://ntrs.nasa.gov/archive/nasa/casi.nt Imagery. Journal Chinese Academy of
rs.nasa.gov/19730001633.pdf/(last accessed: Sciences, Beijing.
24/09/2014). Zhang Z., He G., Wang X., (2010), A practical
Yulius, Tanto TA, Ramdhan M., Putra A., Salim DOS model-based atmospheric correction
HL, (2014), Perubahan Tutupan Lahan di algorithm. International Journal of Remote
Pesisir Bungus Teluk Kabung, Sumatra Sensing 31(11): 2837-2852, doi: http://
Barat Tahun 2003-2013 Menggunakan dx.doi.org/10.1080/01431160903124682.

Comparing Atmospheric Correction Methods For Landsat Oli Data

Uploaded by

Copyright:

Available Formats

Comparing Atmospheric Correction Methods For Landsat Oli Data

Uploaded by

Document Information

Original Description:

Original Title

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

Comparing Atmospheric Correction Methods For Landsat Oli Data

Uploaded by

Copyright:

Available Formats

International Journal of Remote Sensing and Earth Sciences Vol.13 No.

2 December 2016: 105 – 120

COMPARING ATMOSPHERIC CORRECTION METHODS FOR

Esthi Kurnia Dewi*) and Bambang Trisakti

Received: 29 June 2016; Revised: 27 July 2016; Approved: 30 August 2016

Keywords: Landsat, atmospheric correction, QUAC, FLAASH, DOS, surface reflectance

1 INTRODUCTION Medan; Tuni et al. (2013) classified land

@National Institute of Aeronautics and Space of Indonesia (LAPAN) 105

detected with shadows or cloud free basic

2.2.2 QUAC Model

Table 2-2: Paramater used in FLAASH atmospheric correction

Height of Atmosphere Sun Ground

After the sun elevation was

data from the USGS were as follow (Figure 2-4).

Figure 2-4: Step To Compare Atmospheric Correction

3 RESULTS AND DISCUSSION

curve by testing the quality

(a) Before Correction (b) DOS (c) FLAASH (d) QUAC

(a) Before Correction (b) DOS (c) FLAASH (d) QUAC

Figure 3-2: Visual

(a) Before Correction (b DOS

Figure 3-4: Visual Analy

International Journal of Remote Sensin

3.2 Analysis Based on Spectral In terms of average difference

Deviation differences to SR (-)

Figure 3-6: Average and Deviation Difference of Water against SR

Figure 3-8: Average and Deviation Difference of Soil against SR

Deviation differences to SR (-)

Figure 3-10: Average and Deviation Value of Vegetation against SR

3.2.2 Spatial spectral curve similar visible bands, however absolute

(a) SR from USGS (b) DOS method

Band (-) Band (-)

(a) SR from USGS (b) DOS method

Band (-) Band (-)

(a) SR from USGS (b) DOS method

Band (-) Band (-)

4 CONCLUSION enough, although it could distinguish

You might also like