ISH Journal of Hydraulic Engineering

ISSN: 0971-5010 (Print) 2164-3040 (Online) Journal homepage: http://www.tandfonline.com/loi/tish20

Real-time OGC compliant online data monitoring

and acquisition network for management of
hydro-meteorological hazards

Subhrasankha Dey, Nilanjana Bhattacharya, Saswat Chakrabarti &

Dhrubajyoti Sen

To cite this article: Subhrasankha Dey, Nilanjana Bhattacharya, Saswat Chakrabarti &
Dhrubajyoti Sen (2016): Real-time OGC compliant online data monitoring and acquisition
network for management of hydro-meteorological hazards, ISH Journal of Hydraulic
Engineering

To link to this article: http://dx.doi.org/10.1080/09715010.2016.1264896

Published online: 16 Dec 2016.

Submit your article to this journal

Article views: 3

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tish20

Download by: [Gazi Universitesi] Date: 18 December 2016, At: 09:34

ISH Journal of HydraulIc EngInEErIng, 2016
http://dx.doi.org/10.1080/09715010.2016.1264896

Real-time OGC compliant online data monitoring and acquisition network for
management of hydro-meteorological hazards
Subhrasankha Deya , Nilanjana Bhattacharyab, Saswat Chakrabartia and Dhrubajyoti Senc
a
g S Sanyal School of Telecommunications, Indian Institute of Technology, Kharagpur, India; bcivil, Structures and offshores unit, Shell Technology
centre, Bangalore, India; cdepartment of civil Engineering, Indian Institute of Technology, Kharagpur, India

ABSTRACT ARTICLE HISTORY

Natural calamities are undesirable events which often cause significant damages to the society and received 5 July 2016
environment. One of the prime drivers for such a situation is rainfall which, when in excess, is responsible accepted 22 november 2016
for initiating hydro-meteorological hazards like flash flood, landslides. In this context, a suitably
KEYWORDS
designed real-time, telemetred and open-source standard rainfall and water level data acquisition Hydro-meteorological;
network for flood-prone areas may be indispensable in managing the likely hydro-meteorological multiple data-logging;
hazards. This paper demonstrates the design of such a web-based online low-cost high sampling OGC remote data; web-based;
compliant data acquisition and signal transmission system with multiple remote data logging points ogc compliant
to capture information of rainfall distribution over a pilot scale catchment. The network also monitors
the water level at the exit end of the basin to facilitate the estimation of catchment run off. A study is
also made to compare the distributed rainfall observations of tipping bucket rain gauges with that
of an ordinary rain gauge. Finally, the possible applications of the observed data in managing hydro-
meteorological hazards are indicated.

1. Introduction a flash flood-prone semi-arid watershed in Southern Spain.

The work has been implemented to collect water level data in
Robust technology-based systems capable of mitigating the
Rambladel Albuj′on watershed. The authors have developed
adversities of hydro-meteorological catastrophes are important
a low-power and long-range communication device for this
in saving lives and preventing social and economic disruptions.
purpose. Remote communication has been established using
A short burst of intense rain may result in a flash flood or a
User Datagram Protocol (UDP)/Internet Protocol (IP) which is
landslide in a mountainous region or water logging in low-lying
low-cost and low-complex than Transmission Control Protocol
urbanized regions like a coastal city. Thus, any information in
(TCP)/Internet Protocol (IP). Shariff et al. (2015) demonstrated
real-time of rain and water level of a potential hazard zone is
the implementation of a web-based wireless monitoring system
helpful in providing early warnings of an impending disaster.
for grid connected photo-voltaic converter along with user-
Recent developments of internet-based online communication
friendly web application for easy access of monitored data over
technology facilitates such attempts for acquiring rainfall water
internet. The methodology has been administered with reliable
level and other relevant meteorological data from remote places
and robust technological applications (e.g. implementation of
to within easy reach of the potential users. The Open Geospatial
Zig-Bee-based technology). The system is limited to photo-vol-
Consortium (OGC), a public domain data exchange platform
taic converter with no provision for failure detection and fault
further helps in archiving and sharing such spatially distrib-
tolerance. Pallugna et al. (2013) developed a low-cost remote
uted information in a standardized and uniform format so that
rain and stream data acquisition system to collect stream flow
potential beneficiaries of the data may gain access to it with
and rain water information for building a flood warning sys-
least hindrance. For example, a real-time online OGC compli-
tem. The stream and rain data acquisition system presented
ant meteorological data acquisition network can be of help to a
in the paper has been implemented using Gizduino micro-
third party like the state or district administration which may
controllers. GSM modules (Telit GE 846 QUAD) are used for
like to implement the data in running a flood warning model
remote data transfer through SMS. As the proposed scheme
or a landslide prediction model for providing prior advisories
tends to be costly over a long period of time, sampling rate has
to the people of affected region.
been reduced significantly. However, a low sampling rate data
A recent survey of existing hydro-meteorological data
acquisition system may not serve as a reliable flood-warning
acquisition and transmission systems has been carried out by
system due to the uncertainty of hydro-meteorological dis-
Sorg and Kunkel (2015) in implementation of meteorological
asters. Recently, urban conglomerates and communities have
data monitoring and data acquisition system and OGC com-
also started tapping the power of information and communi-
pliant data collection network. Marin-Perez et al. (2012) have
cation technology (ICT) to monitor environmental parameters
described the design and development of a real-time meas-
including rain, run off, air pollution, etc. Notable amongst these
urement system for flood monitoring and its deployment in

CONTACT Subhrasankha dey subhrasankha.dey@iitkgp.ac.in

© 2016 Indian Society for Hydraulics
2 S. DEY ET AL.

are the urban observatories of the UK (Urban observatory information over a pilot scale micro watershed within the cam-
2006) or the German network of terrestrial environmental pus of IIT Kharagpur. The collected data from several rain-
observatories (TERENO 2011). fall sensors and one water level sensor have been made open
Important characteristics of a real-time data monitoring source using OGC compliance such that the acquired data may
network can be divided into sub-functions, like data monitor- be consumed by any organization according to their format
ing method, data transfer mechanism, sampling intervals, data for carrying out possible useful analysis like flood forecasting
logging mechanism, program development software, data anal- and issuing relevant warning. Numerous other hydrological
ysis and interpretation for numerous applications (Shariff et al. applications are also possible by analysing the spatial-temporal
2015). Of these parameters, sampling plays an important role variability of rainfall and water level data if the above objective
in rain as well as other metrological parameter measurements. is met.
Tipping bucket rain gauges (TB rain gauges) are often used
for ground-based rainfall measurements (Habib et al. 2001).
In a tipping bucket, rain gauge, rain is collected into a pair of 2. Pilot study area, measured variables and
collecting buckets balanced on a spindle that tips and drains measurement techniques
each time when one of the buckets gets full. Number of tips A part of the campus of IIT Kharagpur is used as a pilot test
occurring during a pre-specified period of time (e.g. number bed for the data acquisition process. The first in the chain of
of tips per minute) may be counted or the occurrence of each the Indian Institutes of Technology, IIT Kharagpur is located
tip is recorded with a certain time resolution (e.g. the micro- in the District of West Medinipur, in the State of West Bengal,
controller connected to the rain gauge checks every second about 125 km south-west Kolkata (erstwhile Calcutta). Figure
whether there is a tip or not). The accuracy may be increased 1 shows a map of the campus area of the institute with the
in the latter case by minimizing the sampling interval. locations of the instruments marked with different symbols,
Both wired and wireless data transfer mechanisms have as indicated below.
been previously tested (Shariff et al. 2015). Wired data trans-
fer mechanism promises reliability, security, high-speed data (1) Tipping bucket rain gauges for online measurement
transfer, inexpensive installations and maintenance with of rainfall: Red marker.
superior performance. However, the required infrastructure (2) Tipping bucket raingauges for offline measurement
sometimes needs to run cables in challenging environments. of rainfall: Green marker.
On the other hand, wireless communications promise greater (3) Rectangular weir with pressure type digital water
mobility with a neater working environment. GSM (Global depth sensor for online measurement of discharge:
System for Mobile Communications), GPRS (General Packet Blue up arrow.
Radio Service), satellite communications, Wi-fi, Bluetooth
Figure 1 also indicates the location of one ordinary rain
and Zig-Bee-based devices are some of the popular technol-
gauge (location indicated with a brown marker) which records
ogies in wireless communication. Data transmissions over
the daily rainfall. This rain gauge, maintained by the depart-
satellite communications has been reported to be slow and
ment of physics, has several years of archived daily rainfall data.
costly (Shariff et al. 2015). Other economic considerations have
The tipping bucket rain gauges for both online and offline data
shown that GSM/GPRS systems have high operating costs.
collection conform to the following specifications:
Bluetooth, though simpler and cost-effective but covers only
short-range communications. Wi-fi is costlier than Bluetooth • Diameter of collector: 20.32 cm.
and Zig-Bee along with higher power consumption rate. Zig- • Tipping bucket size: 297.18 cm.
Bee mesh network supports capacity of up to 65,536 wireless • Least count: 0.254 mm.
nodes compared to 7 and 32 nodes for Bluetooth and Wi-fi,
respectively (Sung and Hsu 2011). The rectangular weir for flow measurements comprises a
The device that stores data is one important component of sharp-edged weir crest and sides made of 5 mm thick alumin-
a data acquisition system. Collected data from sensors are gen- ium plates fixed to the walls of a minor drain, just upstream of
erally logged locally (Hloupic et al. 2011). Remote data logging a culvert near the northern boundary of the campus. Schematic
method has been introduced to increase the ease of accessibil- of the weir in two views are shown in Figures 2(a) and (b).
ity of remote data. Data may be transmitted using RF (Radio The location ensures that the flow taking place over the weir
Frequency) link and/or wired connections to a server computer is always ‘free-flow’ without any submergence under any flow
situated in a remote location after being stored in the available condition.
local memory available (Benghanem 2009; Marin-Perez et al. The rectangular weir has a pressure tapping point as shown
2012; Rosiek and Batlles 2008; Shariff et al. 2015). in Figure 3, which is connected to a piezo-resistive pressure
Although as shown above, several researchers have dis- sensor of the following specifications:
cussed in detail about the development of meteorological • Sensor Output: milli-volts.
data monitoring and acquisition systems, not much is reported • Pressure Max: 6894.76 N/m2.
about the standards followed to make the acquired data acces- • Pressure Measurement Type: Differential.
sible to the public. Since availability of the data in OGC format • Operating Voltage: 12 V.
is easily accessible to all, it is desirable that hydro-meteorolog- • Supply Current: 2 mA.
ical data are not only acquired without error, but also made
accessible to the public through OGC compliancy (OGC Although for direct calculation of discharge, the approach
2016). The objective of this paper is to present a model archi- water head (H) would have been more convenient, but due
tectural design of a low-cost real-time web-based online data to the absence of any suitable arrangement to fix the sensor
acquisition network with multiple remote data logging points at such a location, the pressure head at the weir plate (h)
for capturing spatially distributed rainfall and drain water level was measured. By deducting the depth from the tapping to
 ISH JOURNAL OF HYDRAULIC ENGINEERING 3

Figure 1. locations of the spatially distributed sensing nodes in IIT Kharagpur campus (each grid cell represents 200 m square).

Figure 2. details of rectangular weir (a) Stream-wise view; (b) longitudinal section.

brink of a sharp-crested weir is approximately related to

the approach water head (H) by the following relation (refer
also Figure 2),

h∗ V2
= 0.85 − 0.83
2/ 2/ (1)
H + V 2g 2g(H + V 2g )

For the present application, the approach velocity is found

to be a maximum of around 2 m/s with an approach 2/
head
of around 0.5 m. Hence, the velocity headV 2g , which
evaluates to around 0.2 m, appears to be rather significant
with respect to the water depth and cannot be ignored in the
commutations. Here, Vis the average velocity in the channel.
Thus, two discharge equations are used: (a) one at location
on the upstream of the weir, Equation (2) and (b) the other at
the location of the weir, Equation (3), to relate and solve the
unknown variables for a measured value of the time-varying
Figure 3. rectangular weir dimensions with respect to ground level, water depth variable.
at crest (h) and water pressure head (H) of the overfalling nappe.
Q = B1 HV (2)
the weir crest (d), one obtains the effective depth of water � 3
as h – d (say, h*) above the weir crest at brink. According 2 √ � H
Q = K1 K3 B 2g 0.611 + C1 H ∕2
to Ranga Raju (1993), the depth of flowing water at the 3 W (3)
4 S. DEY ET AL.

In equation (2) and (3), the following values of constants are cable to upload the same on the flash (non-volatile) memory
used, according to the dimensions of the channel and weir: of the microcontroller.
where Q: Discharge passing through the weir V: Average veloc- For reading and transmitting online rain gauge data, the
ity in the channel, K1: To be read from Figure 8.10 of Ranga microcontroller is coded to count the number of tips of the
Raju (1993), but an average value may be taken as 1.05, K3: An oscillating bucket of the rain gauge within a time span of 5
average value suggested by Ranga Raju (1993) is 0.95, B: Width min. Of course, this time duration is flexible and may be pro-
of weir opening = 0.910 m, B1: Width of channel at the location grammed to smaller or larger values, as desired. The micro-
of weir = 1.030 m, g = 9.81 m/s2, acceleration due to gravity, controller is equipped with a real-time clock which, at the end
C1 = 0.07, as read from Figure 8.12 of Ranga Raju (1993) for of 5 min, signals the end of data collection for that interval
B
B1
= 0.9, W: Height of weir crest above upstream-bed level of and the number of tips recorded is multiplied by 0.254 mm
channel = 0.05 m. (the least count of the tipping bucket rain gauge) and saved to
The three unknowns Q, H and V, are to be solved from the an SD card along with a real-time clock stamp. The controller
three equations (1), (2) and (3), and needs to be evaluated for simultaneously transmits the data through an Ethernet con-
every value of water depth (h * ) at brink. Thus, a known pressure nection to a central server taking advantage of the institute’s
head (h*) is converted to an equivalent discharge, which is then Local Area Network (LAN).
acquired using the system described in the following section. For reading and transmitting the pressure head data from
the sharp-crested rectangular weir (the discharge is later com-
puted at the server end during analysis), the same Arduino
3. Data acquisition, transmission and local storage
Uno microcontroller board has been employed for reading the
Two types of data acquisition methods have been employed data at intervals 15 s, which may be changed according to the
in the present study: (a) offline, and (b) online. The former, requirement. The data thus acquired is not only written locally
that is the offline system, is used for some of the rain gauges to an SD card along with time-stamp but is also transmitted
(shown in Figure 1) for storing rainfall data at intervals 15 min to the central server using the institutes LAN. However, the
locally in an SD card. This is done with the help of PIC micro- location of the nearest LAN port being somewhat far away,
controllers. The online data acquisition system followed for the data are first transmitted wirelessly through a X-Bee-Pro
some of the rain gauges as well as for discharge measurement (DIGI 2016) radio frequency (RF) transmitter to another sim-
using a sharp-crested weir uses an ATmega328p microcon- ilar unit near a LAN port through using a second Arduino
troller-based Arduino Uno board(Arduino 2016) for data Uno board. The X-Bee pro RF modules operate within the ISM
acquisition, transmission and storage. The microcontroller is 2.4 GHz frequency band. Figure 4 describes the architecture
programmable and may conveniently be coded using an inte- of the data acquisition transmission and storage units for the
grated development environment (IDE) loaded on to a PC and systems described above. It also demonstrates the formatting
transferring the executable version of the code through a USB that is required for the acquired data according to the open

Figure 4. architecture of the data acquisition and signal transmission system developed for the present project.
 ISH JOURNAL OF HYDRAULIC ENGINEERING 5

geospatial consortium (OGC) recommendations, a feature that Hyper-text Transfer Protocol (HTTP) requests generated by
is described in the next section. web clients in a web browser automatically reaches the micro-
controller attached to the web server. The requested real-time
data are displayed in the web browser of the requesting web
4. Data storage in a central server and OGC
client. The data collected as variables in HTML script are web
compliance of data
scrapped with time stamp and positional details like latitude
As described above, the local microcontrollers attached to each and longitude of the observation point and saved in .txt or .csv
online sensor (rain gauge or water pressure sensor at the weir) file format for future reference.
feeds the acquired data along with a time stamp to a Hyper- For the implementation of the above feature, a suitable
Text Mark-up Language (HTML) script and passes on to a sampling interval of 5 ms has been implemented at the client
network-controller. This data are sent through an Ethernet side for data acquisition. The remote data collection task has
(LAN) access point which facilitates a wired data communi- been implemented using the coding platform Python 3.4. Web
cation using the TCP/IP protocol. At the receiving end, the scrapping has been done using the Beautiful Soup v4.2 mod-
data from all the sensors available in the network are collected ule (Beautiful Soup 2016) integrated with Python 3.4 (Python
at a server computer which saves them in its own memory. A 2016) Apache HTTP server v2.2 has been employed for remote
Web User Interface (Web UI) has been developed to moni- data access and online monitoring.
tor the deployed sensor node individually. The collected raw A screenshot of the web user interface and the data acquired
data are then connected to a Geographic Mark-up Language for a typical sensor node as viewed from a monitor of a web
(GML) file which complies with the requirements of the browser as shown in Figure 5. It may be noted that the sensor
Open Geospatial Consortium (OGC 2016).Before generating location for which the data collection details are shown on the
the GML, the converter consults a registry service called the right of the figure, also has some additional data acquisition
Catalogue Service for the Web (CSW), which is responsible for units like humidity, temperature and wind speed apart from
validating the GML file. The data are then transferred to a Web rainfall. By deploying the web application in any computer
Feature Services (WFS) portal, where a transaction service connected to the institute LAN, it is possible to save data locally
function is called upon to upload the data into the correspond- at any number by logging location.
ing database. This enables the data to be easily accessible for
any organization to use and consume according to their own
6. Utilization of rainfall data for estimating regional
format. The aforesaid procedures are summarized in Figure 4.
precipitation characteristics
Further details of the OGC compliance of data may be found
in OGC (2016). In the preceding sections, the techniques for data acquisition
from different rainfall measurement devices and a flow meas-
uring weir was described. Several applications of these data
5. Multiple data logging
exist, like rainfall-run off analysis, derivation of intensity–
The previous section describes a form of local logging of the duration–frequency (IDF) curves, depth–area–duration–
data transmitted by the online rainfall and water pressure sen- curves, etc. In this section, a general analysis of the rainfall
sors on to a central server. However, it may be often required be data is demonstrated for the IIT Kharagpur campus. Further,
to log the data at some other remote PC connected to the LAN. an analysis for estimating the accuracy of the tipping bucket
This kind of a multiple logging may be used when different rain gauge is presented using sample rainfall records obtained
analyses are made using the same or part of the data by differ- from the sensors. It is also of interest to see the type and quality
ent agencies having different objectives of utilizing the data. of data acquired from the sensing locations. This is shown in
In the present case, such a feature has been implemented Figure 6 for three typical rainfall sensing nodes in three dif-
by writing an HTML script which contains the real-time data ferent locations and the water level sensing node at the sharp-
from the sensors and one which is available in the web servers. crested weir.

Figure 5. a screenshot of the developed the web server showing sampling locations and sample data from one typical location.
6 S. DEY ET AL.

7. Rainfall patterns and characterization in study those of Kharagpur based on the 10-year daily rainfall record
area (2004–2013).
Next, the rainfall characteristics of the study area of IIT
A 10-year record of daily rainfall spanning from 2004 to 2013
Kharagpur is demonstrated using tipping bucket (TB) rain
was used in the current study to understand the pattern of
gauge records set to capture rainfall at 10-min interval. Such
rainfall in Kharagpur. This data, obtained from the ordinary
sampling at high frequency is often helpful in carrying out
rain gauge (ORG) maintained by the Physics Department of
probability analysis of maximum rainfall values required
the Institute, assists in analysing the overall rainfall character-
for the design of hydraulic structures, rainwater harvesting,
istics of the IIT Kharagpur campus study area. The distributed
flood management and other similar applications based on
point rainfall data obtained from the tipping bucket rain gauge
extreme value occurrences in rainfall events. For these hydro-
(TB) is used to obtain finer rainfall patterns, and is discussed
logical applications, distributions such as Gumbel (from the
in subsequent sections. Location of the ORG and the TB-s is
Generalized Extreme Value, GEV, family) and Log-Pearson
indicated in Figure 1. Figure 7 depicts the time series of yearly
are seen to be most effective especially for predicting rain-
rainfall during the study period. Annual rainfall is seen to vary
fall volumes for periods longer than the recorded duration
between 1000 mm and 2300 mm over the years.
(Olofintoye et al. 2009; Singh et al. 2012; Vivekanandan and
The monsoon period in India, which generally coincides
Roy 2013). The annual maximum daily rainfall values from
with the months of June to September, shows an average of 76%
the years 2004 to 2013 for the given study area were evaluated
of the annual rainfall at Kharagpur compared to the national
average of 74%. The analysis is based upon the 10 years data
Table 1. average monthly rainfall comparison.
between 2004 and 2013, it was estimated that the average mon-
soon rainfall for Kharagpur happens to be 1238 mm with a June July August September
coefficient of variation of 0.34. In Table 1, the national mon- Kharagpur (mm) 249 301 259 305
soon rainfall data (Attri and Tyagi 2010) is compared with national (mm) 163 286 251 168

14 4.5
4
Rainfall in mm

Rainfall in mm

12
3.5
10
3
8 2.5
6 2
4 1.5
1
2 0.5
0 0
16:27:35
4:47:37 PM
5:07:38 PM
5:27:40 PM
5:47:56 PM
6:08:13 PM
6:28:14 PM
18:48:16
19:08:17
19:28:19

16:27:35
4:47:37 PM
5:07:38 PM
5:27:40 PM
5:47:56 PM
6:08:13 PM
6:28:14 PM
18:48:16
19:08:17
19:28:19

(a) (b)
Pressure Head in mm

7 700
6 600
Rainfall in mm

5 500
4 400
3 300
2 200
1 100
0 0
16:27:35
4:47:37 PM
5:07:38 PM
5:27:40 PM
5:47:56 PM
6:08:13 PM
6:28:14 PM
18:48:16
19:08:17
19:28:19

16:27:35
4:47:37 PM
5:07:38 PM
5:27:40 PM
5:47:56 PM
6:08:13 PM
6:28:14 PM
18:48:16
19:08:17
19:28:19

(c) (d)

Figure 6. (a), (b) and (c) Sample data obtained by rainfall sensors (values given in mm measured every 5 min) at three locations as on 15 July 2015; (d) corresponding
pressure head (h* in mm) at weir.

Figure 7. yearly rainfall time series.

 ISH JOURNAL OF HYDRAULIC ENGINEERING 7

with five different probability distributions viz. Normal, Log- 8. Error analysis of tipping bucket rainfall records
Normal, Gumbel (EV1), Log–Pearson III and Log–Gamma
The methodology described by Habib et al. (2001) is used to
Distributions, to obtain the best-fit model. In this study, a com-
find the accuracy of tipping bucket (TB) rain gauges using
parison of the different distributions is made by evaluating the
Equation (5) and Equation (6) to calculate absolute error and
goodness of fit of the given data, by following three tests: (a) the
Equation (6) to calculate relative error, respectively, using a
Kolmogorov Smirnov test, (b) the Anderson Darling test and
reference rainfall data for the location.
(c) the Chi-squared test explained by Anderson and Darling
(1952), Sharma and Singh (2010) and Vivekanandan and Roy ETBi = RTBi − RORG (5)
(2013). Following Olofintoye et al. (2009), a ranking procedure
is used where the distributions are ranked from 1 to 5 for each
test according to the goodness of fit. The total rank is calculated RTBi − RORG
by the summation of ranks for individual goodness of fit tests RETBi = (6)
and summarized in Table 2. RORG
Thus, from Table 2, Log-Gamma Distribution is found
In the present case, a reference field measure is available in the
to be best fitted to the given data-set since it has the highest
form of the daily rainfall recorded by the Symon’s ordinary
cumulative rank in the table. Next, the annual maximum daily
rain gauge (ORG), recorded by the Physics Department of the
rainfall for return periods from 2 to 25 years is computed from
institute. This value, considered as a standard, is compared with
the inverse cumulative probability function of the Log-gamma
distribution.
Let the data-set R represent the annual peak rainfall values
from 2003 to 2013. If F(r) represents the cumulative probabil- Table 3. Maximum daily peak rainfall expected for different return periods.
ity function for all r ∊ R, then the annual peak daily rainfall Return period
rT for return period T (in years) is given by Equation (4). The (Years) 2 3 4 5 10 15 20 25
estimated rainfall values for different return periods, calculated Estimated rain- 169 217 251 278 364 417 457 490
according to this formula are presented in Table 3. Figure 8 fall (mm)
plots the relationship between rT versus T and can be used to
predict estimated annual maximum daily rainfall for interme-
diate return periods. 500
Maximum daily rainfall (mm)

450 R² = 0.9991
1 400
F(r > rT ) = 1 − (4)
T 350
300
250
200
Table 2. Probability distributions ranked for goodness of fit. 150
100
Kolmogorov Anderson
Distribution 50
Smirnov Darling Chi-squared Total
0
log-gamma 1 1 2 4 0 5 10 15 20 25
log-Pearson 2 2 1 5
log-normal 3 3 3 9 Retun period (year)
gumbel 4 4 4 12
normal 5 5 n/a n/a Figure 8. Maximum daily rainfall versus return period.

Figure 9. Error spread for TB1, TB2, TB3 and TB4.

8 S. DEY ET AL.

the rainfall recorded by four of the TB rain gauges deployed |

around the study area of IIT Kharagpur campus. 𝜎RETBi |RTBi
e= | (10)
Denoting RTBi as the cumulative rainfall for 24 h recorded √
m
by the ith TB and RORG as the 24 h rainfall recorded by the
ORG, the error ETBi is defined as difference between the sums The plot of RETBi ||RTBi for the four TB gauges measured in per-
of the two and expressed as Equation (5). The error statistics centage figures are shown in Figure 10. Clearly, TB4 shows
are calculated based on the relative error RETBi as given in least average relative error for all rainfall intensities, although
Equation (6). corresponding values for TB3 are also close.
As depicted graphically in Figure 9, TB4 has the least range From the rainfall volume data collected in the monsoon
of error while TB2 shows maximum spread of error calculated of 2013 from the tipping bucket gauges installed in the IIT
from Equation (5). Kharagpur campus, it is observed that the average daily rain-
For certain simulated rainfall intensity, the relative error fall intensity on a typical rainy day is about 25 mm/day. Since
defined by Equation (6) achieves its maximum overestimat- from Figure 10 it is seen that data of TB4 is most reliable for
ing limit when the reference rain gauge is not recording any intensities in this range and higher, the recorded data from this
rain, i.e. RORG = 0. This situation may be avoided if the rainfall TB rain gauge is used to develop IDF curves to characterize the
intensity range is divided into bins so that the mean and stand- rainfall intensity pattern in IIT Kharagpur. The Generalized
ard deviation of RETBi can be expressed in a conditional sense, Extreme Value Distribution is best suited to seasonal trends in
| rainfall and includes the Weibull and Gumbel type distribu-
RETBi ||RTBi (Equation 7) and 𝜎RETBi |RTBi (Equation 8), where RTBi
| tions as recommended by Nhat et al. (2006) and Sharma and
is the rainfall intensity of the ith bin. Expressing the error in
Singh (2010). The maximum intensity of varying durations
a relative sense and in a conditional way is more convenient
during a typical monsoon at different return periods is meas-
from the practical point of view.
ured here and the Gumbel distribution is used to fit the data
1∑ to a probability distribution which is the standard distribution
RETBi ||RTBi = (RETB )j ||(RTBi ) (7)
n j for this purpose cited from Lumbroso et al. (2011) and Nhat et
al. (2006). It is observed that there is high degree of correlation
in time between the rainfall recorded at the gauges identified as
√ TB1, TB2, TB3 and TB4. Thus, spatial variation of IDF curves
| 1 ∑
𝜎RETBi |RTBi = [(RETB )j ||(RTBi ) − RETBi ||RTBi ]2 (8) is ruled out in the IIT Kharagpur campus.
| n j
Based upon the 10-min interval rainfall data recorded by
TB4 for different storms, the maximum intensity recorded for
where, n = Sample size of rainfall record from each TBi
rainfall events of durations from 10 min to 24 h are found and
i ∈ {1;2;3;4}; j ∈ {1;2; :::;n} (9) fitted statistically to find the corresponding return periods (in
days) and summarized in Figure 11.
If there are m data points in a particular bin, the likely error
in the computed mean is calculated by Equation (10) from
statistical theory (Hutchinson 1993).The bin size was varied 9. Application of real-time hydro-meteorological
from 10 to 30 mm/day at different magnitudes of average rain- data for disaster management
fall intensities such that the error for each range is minimized,
The rainfall and discharge data acquisition in real time and
where e is defined as,
making the archived data OGC compliant, as described in the

Figure 10. RETBi ||RTBiestimates vs daily rainfall intensity (mm/day).

 ISH JOURNAL OF HYDRAULIC ENGINEERING 9

Figure 11. Maximum daily intensity–duration–frequency curves at Kharagpur.

previous sections, helps any external agency to access such under research grant [NRDMS/11/1978/2012/1], Dt.13-03-2013 (Sensor
data easily for further analysis. For example, a similar system Network and Web-Enablement for Geospatial Technology Based to
Development and Disaster Management: A Pilot Study).
employed in the mountainous regions of a country, like the
state of Uttarakhand or Sikkim of India in the Himalayas, may
help the state administration to picture the real-time scenario ORCID
of impending disasters like floods or landslides. Of the possible Subhrasankha Dey http://orcid.org/0000-0001-5494-1111
hydro-meteorological hazards, the prediction of flash floods
is one of the important objectives of different missions. The
References
primary data that is required to drive the flood forecasting algo-
rithms is rainfall. Many algorithms exist, like the physics-based Anderson, T., and Darling, D. (1952). “Asymptotic theory of certain
model of Koussis et al. (2003) or the data-driven approach of ‘goodness of fit’ criteria based on stochastic processes.” Ann. Math.
Stat., 23, 193–212.
Duncan et al. (2011) to convert rainfall data into flood flows. Arduino. (2016). https://www.arduino.cc/.
The data obtained through the study described in this paper is Attri, S. D., & Tyagi, A. (2010). “Climate Profile of India”, Met Monograph
intended to be used for setting up and calibrating such models No. Environment Meteorology-01/2010, India Meteorological
as a part of an ongoing project at IIT Kharagpur. Department, Ministry of Earth Sciences, Government of India. (http://
www.indiaenvironmentportal.org.in/files/climate_profile.pdf).
Beautiful Soup. (2016). https://www.crummy.com/software/Beautiful
10. Conclusions Soup/.
Benghanem, M. (2009). “Measurement of meteorological data based on
This paper presents a low-cost implementation of wireless wireless data acquisition system monitoring.” Appl. Energy, 86(12),
monitoring system for various meteorological parameters. 2651–2660.
Further, system with multiple data logging points and making DIGI. (2016). https://www.digi.com/support/productdetail?pid=3430/.
Duncan, A., Chen, A.S., Keedwell, E., Djordjevic, S., and Savic, D.A.
it OGC compliant is shown to be useful to the community (2011). “Urban flood prediction in real-time from weather radar and
at-large, especially for applications to hydro-meteorological rainfall data using artificial neural networks.” IAHS Red Book series
disaster management. An application of the acquired rainfall no. 351, Presented at the Weather Radar and Hydrology International
data in space and time within a pilot test bed located in IIT Symposium, International Association of Hydrological Sciences, Exeter.
Kharagpur is also demonstrated. Similar inexpensive data Habib, E., Krajewski, W.F., and Kruger, A. (2001). “Sampling errors of
tipping-bucket rain gauge measurements.” J. Hydrol. Eng., 6(2), 159–166.
acquisition, transmission, storage and distribution system Hloupic, G., Stavrakas, I., Moutzouris, K., Alexandridis, A., and Triantis,
through OGC compliancy may help developing countries to D. (2011). “WSN open source development platform: application to
set-up and operate hydro-meteorological decision support sys- green learning.” Procedia Engineering, 25, 1049–1052.
tem conveniently at low-cost. Hutchinson, T. (1993). Essentials of Statistical Methods in Forty One Pages,
Rumsby Scientific Publishing, Sydney.
Koussis, A., Lagouvardos, K., Mazi, K., Kotroni, V., Sitzmann, D.,
Acknowledgements Lang, J., Zaiss, H., Buzzi, A., and Malguzzi, P. (2003). “Flood forecasts
for Urban basin with integrated hydro-meteorological model.”
The authors thank Mr. Debabrata Bej for his invaluable contribution in J. Hydrol. Eng., 8(1), 1–11. doi: http://dx.doi.org/10.1061/(ASCE)1084-
the installation of digital tipping bucket rain gauges and assimilation of 0699(2003)8:1(1).
rainfall data collected at IIT Kharagpur. Lumbroso, M., Boyce, S., Bast, H., and Walmsley, N. (2011). “The
challenges of developing rainfall intensity-duration-frequency curves
and national flood hazard maps for the Caribbean.” J. Flood Risk
Disclosure statement Manage., 4(1), 42–52.
No potential conflict of interest was reported by the authors. Marin-Perez, R., García-Pintado, J., and Gómez, A.S. (2012). “A real-
time measurement system for long-life flood monitoring and warning
applications.” Sensors, 12, 4213–4236.
Funding Nhat, L.M., Tachikawa, Y., and Takara, K. (2006). “Establishment of
intensity-duration-frequency curves for precipitation in the monsoon
This project is financially supported by the Department of Science and area of Vietnam.” Annuals of Disaster Prevention Research Institute,
Technology, Ministry of Science & Technology, Government of India Kyoto University, 49, 93–103.
10 S. DEY ET AL.

OGC. (2016). “Making location count”. http://www.opengeospatial.org/. Sharma, M.A., and Singh, J.B. (2010). “Use of probability distribution in
Olofintoye, O., Sule, B., and Salami, A. (2009). “Best-fit probability rainfall analysis.” New York Sci. J., 3, 40–49.
distribution model for peak daily rainfall of selected cities in Nigeria.” Singh, B., Rajpurohit, D., Vasishth, A., and Singh, J. (2012). “Probability
New York Sci. J., 2, 1–12. analysis for estimation of annual one day maximum rainfall of
Pallugna, R.C., Cultura, A.B., Gozon, C.M., and Estoperez, N.R. (2013). Jhalarapatan Area of Rajasthan, India.” Plant Archives, 12, 1093–
“Low-cost remote rain and stream data acquisition system for mapping 1100.
of potential micro-hydro sites.” Energy Power Eng., 5, 56–62. Sorg, J., and Kunkel, R. (2015). “Conception and implementation of an
Python (2016). https://www.python.org/. OGC-compliant sensor observation service for a standardized access
Ranga Raju, K.G. (1993). “Flow Through Open Channels”, Tata McGraw- to raster data.” ISPRS Int. J. Geo-Inform., 4, 1076–1096.
Hill Publishing Company Limited, New Delhi, Channel controls and Sung, W.-T., and Hsu, Y.-C. (2011). “Designing an industrial real-time
transitions, 227–233. measurement and monitoring system based on embedded system and
Rosiek, S., and Batlles, F.J. (2008). “A microcontroller-based data- ZigBee.” Expert Syst. Appl., 38(4), 4522–4529.
acquisition system for meteorological station monitoring.” Energy TERENO. (2011). http://teodoor.icg.kfa-juelich.de/overview-de.
Conv. Manage., 49, 3746–3754. Urban observatory. (2006). http://www.urbanobservatory.ac.uk/future.
Shariff, F., Rahim, Nasrudin Abd, and Hew, Wooi Ping. (2015). “Zigbee- Vivekanandan, N., and Roy, S. (2013). “Assessment of probable maximum
based data acquisition system for online monitoring of grid-connected precipitation using Gumbel distribution and Hershfield method.”
photovoltaic system.” Expert Syst. Appl., 42, 1730–1742. Bonfring Int. J. Data Mining, 3(1), 1–5.