Machine Learning-Based Flood Prediction of Assam

Deovolenty Myrchiang, Moziihrii Ado. Khwairakpam Amitab

D(•pc,nment of luformmion Tc.•clmology
North-Easfem Hill University
Shillong . India
2023 International Conference on Computational Intelligence, Networks and Security (ICCINS) | 979-8-3503-1379-6/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICCINS58907.2023.10450018

deovole.ntymyrchiang03 @gmai I.com

A l,s trtltl- Flood pr<'dktion is cruci,d in dis,-.ster managemeot The. vul ne.rability of Assam to fre.que.n t floods has motivate.d
,rnd mitigation, partkularly in flood-prone regions like Ass,\m. us to in"estigate advance.d ML models. leveraging historical
This paper presents a study on machin e learning-b,-.sed flood rainfal I data and geosp.ttial infom1;.ttjon to develop accurate.
prediction using A~am·s historkal raioran :rnd geospatial dat.-...
Among the five ornchine-le,-..rning algorithms a~essed, the Rao- and re liable Hood prediction models. ML algorithms have.
dom Forest algorillun was lh t best-1X'rt'orming model for flood gained JX>pularity in flood prediction d ue to the ir capability
prediction. The proposed syst em can predict the risk or floods to anaJyze complicated info rmation and uncover patterns and
in the oext fr,•e days by using t'orte.,-.st n\iof,\11 dah\ obt:iioed linkages 181. The prediction capability can be furthe r enhanced
from the ()penWe,Hher API, total n\inf::\11 data , and 1>re1u·ucessed by adding geographical infom1;.ttjon such as digital e levation
geographical data. A US<'r-friendly web ap1)lic:Hioo was ,dso
created to enlrnnce its usabilHy ,rnd practkalily. The proposed models. slope. aspect. and curvature. The ML models can
S)'Stem c,rn accurately predict floods, which will hell) disaster c.apture. the spatial fe atures that influence. flood occurre.nce.
preparedness in Assam a nd other flood-prone regions worldwide. The combination o f ML algorithms and geospatial data allows
flood risk e.valuation. leading to enhanced flood prediction
Index Terms- Flood prediction, machine learning, geospatial and control me.asure.s. Ac.curate flood prediction can assist
data, rainfall dah\, A~am, Random Forest
authorities and sta keholders in adopting prompt mitigating
plans. resource. allocation. and infrastructure development to
I. I NT RODUCTION
limit Rood d isaster [3 1-
Floods are repetitive. natural calamitjes that cause extensive.
destruction to infrastructure. agriculture. and human exis- II. L ITERAT URE R EV IEW
r
te nce I 1. Flood prediction is cn1c.ia.l in disaster management.
p,uticul,.u-ly in Hood -prone. regions such as Assam. Timely There are few existing studies on flood prediction for
and accurate flood prediction c an significant.ly contribute to Assam . Some of the. notable ones are discussed in this se.c tion.
effective decision-making. mitigation measures. and proactive.
disaster preparedness (2 ). Traditional flood predictjon methods Gupta and Dixit !"9] address the crucial job o f estimat·
often re.Jy on historical data and empirical mode.ls. ,vhich ing flood risk in the Assam re.g ion. Multi-C,i te ria Dedsion
may have. limita tions in accuracy and re.l iability [J I. In recent Analysis-Analytical Hierarchy Process (MCDA- AH P) was
years. combining Machine Leam ing (ML) techniques and used to prioriti1..e flood-prone regions and de.velop Geographic
geospatial data has shown promising results in enhancing Information System (GIS)-bas-ed flood risk maps. To evaluate.
flood prediction capabilities (41. The main objecti,•e o f this the mode.I. they have. used digital ele.v;.ttjon. hydrologicaJ. his·
study is to develop a machine learning-based flood prediction torical Rood . Land Use/Land Cover (LULC). soil type maps.
model for Assam using rai nfoH and geographical data. Ass,u11 rainfal I data. e tc. These. datasets serve. as the basis for studying
is located in northeaste.m India and encompasses an area of flood risk features such as geogr.tphy. proximity to bodies o f
78.438 square kilometers and has a latitudinal extension o f 240 water. land use. p.ttte.m s. and pre.c ipitation patterns. The authors
N to 280 N and a longitudinal extent o f 89045/E to 96000/E. prioritized sites de.pending on the. seve,i ty of Hood risk and
(5). The. distinct ge.ographical characteristics o f Assam . such developed spatially explicit flood risk maps at the re.g ional and
as its proximity to the Bay of Bengal and the. Himalayan administrative levels. In addition. the. study seeks to provide.
foothills. impact the. state•s annuaJ rainfoJI patte.rns [2"1. Wan11 valuable sugge.stions fo r flood risk manageme.nt and mitigation
and moist air from the Bay of BengaJ collides with cooler approaches to help decision-makers and stakel10lde.rs.
air masses from the. Himalayas during the. monsoon season. Manoharan et al. I .I 01 explore the application of Random
resulting in significant precipit;.ttjon throughout the region. Forest (RF) classifie r in mapping the. Hailakand i. Assam flood
This geographical convergence gives a natural pathway for during 20 18 using Sentinel· .I Synthetic Aperture Radar (SAR).
monsoonal moisture to mig r.tte into Assam. resulting in heavy The research ers initially processe.d the SAR data to con·
and protracte.d rain (51 . The. Brahmaputra valley is a natural struct acc urate Digital Elevation Models (DEMs) and create.
b.tsi n. gathe.ring water from se.ve.r al rive.rs and enhancing Hood a baseline wate.r Je.ve.l model. They combined pre-flood and
pote.ntial 16). Assam ·s soil type. and land use patterns aJso post-flood imagery to identify flooded areas using bac kscatter
considerably impact rainfoH runoff and absorption 1"7]. coefficients acquired from SAR data. More than 200 pixels.

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:31:52 UTC from IEEE Xplore. Restrictions apply.
including ,vale r bodies and olher sutfaces. were considered in absolute e1Tor. and root mean squ,u·e. e rror. The study result
training the RF model. showed that ANFIS-ACO has belier accuracy and reliability
11-.e. researc h arljcle [ I I] focuses on mapping and assessing than the ANFIS model.
floods using mulljte.mJX>ral SAR images in lhe Brahmapu- Based on the v;.uious studies d iscussed above. limited ,vork
tra river in Assam. India. The sludy utilizes Se.n linel- lA has employed ML te.c hniques for Hood inundation mapping.
SAR data to develop a methodology fo r flood mapping. 11-.e. Furthermore. only a fe.w studies exist on flood prediction in
researchers integrate SAR and G IS methods to accurately Assam. Integrating machi11e. learning approaches with rain-
delineate flooded and non-flooded areas. They use image. fall and ge.ographical data ca n increase the. accuracy and
processing te.c hnique.s like densily slicing to generate Hood effectiveness o f Hood risk assessme.nts. Advanced research
inundation maps. The. flood maps were veri fied using visual of historical rainfall patterns. land cover changes. and other
inte.rpretation and ground lruth data. TI1e article highlights ll-.e. critical geographical v,ufables can be. conducted utilizing ma-
importance. o f integrating remote sensing data with hydraulic chine learning methods. This interface would en able analytical
models for accurate flood exte.n t mapping and discusses vari - automation. adaptive mode.ling to capture. dynamic flood risk
ous me.lhods for speckle filte.ring of SAR images. Overall. ll-.e. fac tors. and dala-driven insights to be derive.d fo r more thor-
article provides insights inlo flood mapping using SAR data ough flood risk assessments in Assam.
and e mphasize.s the. significanc.e. of geoinfom1ation fo r Hood
Il l. M ACH INE L EARNING AND GEOS PATtAL DATA
assessmen t and manage.menl.
Machine le,u ning techniques are capable o f le,u·ning fro m
TI1is study 1121 utilized Sentinel- I SAR data to map and
histoiical flood and Geospatial data [3]. ML models can
monitor flood inundation in Kaziranga National Park (KNP).
learn patterns from the. data and use them to pre.d iet flood
Assam. SAR dala was more e.ffective than optic.al data in
identifying high moisture. and inundated areas. The dataset acc urately (81 . Examples of ge.o spatia.l data include geographic.
used in this study includes sixteen dual-polarized (VV and coordinate.s. elevation. slope. aspe.c.t. and land cover. Ge.o sp a-
tial data give. information on ll-.e. physic.al properties of the.
VH) SA R images acquired fro m June lo September 20 17
and Land sat-8 OLI data acquired in August 20 17 . The study landscape and its surroundings (151. It is ctitical in flood
prediction models lo capture flood-pro11e. locations' spalial
identified two Hood waves in July and August 20 17 . with
the se.c ond wave causing significant flooding in the p,uk variabilily. Digital elevation models. for example. can aid in
the ide.n tification o f low-lying areas and rive.r courses prone.
The k-mean d ustering algorithm is applied to each SAR
color composite image. for classification. SAR data for Hood to flooding (16-1. Combining mac hine learning algorithms with
monitoling can provide valuable information for decision- geospatial data improves the accuracy and locali zation of flood
making and disaster response. TI1e study reveale.d thal 6% prediction mode.ls.
of the area was covered by pe.m1,rnenl water bodies. 33% A. Mc,chine Leaming Techniques
by se.asonal flood. and 61 % by land categories within KNP. Following are the machine. learning techniques explored in
The authors presenled the. usefulness o f SAR data fo r Hood this work fo r predic.ting floods in Assam.
monitoling and mapping in ll-.e. sludy area. I) logistic Regression (LR): LR is a binary classification
Using a mulli -criteria analysis. the aulhor [ 13] disc.usses model. It calculales the probability of an event belonging to
a case sludy on flood risk assessmenl in the Kopili Ri,•e r a class (flood o r no flood) and ulili1..es a thre.shold to create.
Basin. Assam. The study utilizes satellite. socio-economic. binary predic.tions. The mode.I is trained on hisloric.al data to
and LULC data to identify flood hazard zones and assess discover the best coefficients (weights) and is optimized using
vulnerabilily in the re.gion. A total o f 181 spatio multi- the sigmoid fu nction. Once. trajned. it can pre.d iet floods for
te mporn.J satellite datasets comprising IRS Resourcesal- 1/2 new data by evaluating the. probability of occurrence. and using
AWiFS and microwave SAR data of both IRS Serie.s RISAT-1 the threshold [ 17].
and Canada-based RADARSAT- 1/2 series were used fo r the. 2) Supporl Vec10r Machine (SVi\1): SVM is a supervised
study. 11-.e. results show thal a significant area of cropland and ML for flood prediction based on rainfall and geographical
several villages fall into high and moderate-high-risk zones. data. ll cakulates the optimal decision boundary to split
The study also assesses social vull-.e.rability and infrastructure. diffe re.n l classes. maximi1...ing the maf(Jin between them. SVM
vulnerabilily. particularly roads. 11-.e. authors claimed the. ap- c.an handle non-linear data using the kernel method [ 18). Afler
proach can be applied in other river basins to mitigale Hood training. it predicts flood occ.u tTences by cl assifying fresh data
risks. Additionally. the study emphasizes the importance of points based on their position relative to the decision border.
reducing vulnerabilily and imple.menling land use regulations 3) k-Nearest Neighbo,:r (kNN): kNN is a simple and ef-
for effective. flood risk reduc tion measures. fective supervise.d ML algorithm for Hood prediction based
Agnihotri e t al. I 141 carried oul a case study to predict Hood on rainfall and geographical data. kNN is non-parametric and
at the Matijuri gauge. station of the B,u·ak River basin. Ass,u11 straightforw,u·d to const11.1cl but computationall y expe.nsive for
using Adaptive Neuro-Fu1..zy lnfere.nce Syste m (ANFIS) and large datase.ts. It discovers ll-.e. •k• nearest neighbors lo a new
Ant Colony Optimi1.ation (ACO) algorithm . The performance. data point and e mploys a voting mechanism: it assigns the data
of ANFIS and the hybrid implementation o f ANFIS-ACO were. point lo the most prevale.nt class within its 1-.e.arest neighbor
compared using ll-.e. coefficienl of determination (R2). mean set I 191 .

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:31:52 UTC from IEEE Xplore. Restrictions apply.
 4) Decision Tree (DT): DT is an inte.rpretable technique. A. Daw Collection and Preprocessing
for flood prediction based on rainfall and geographic.al data. It Historic.al flood data was collected from Assam State. Dis-
builds a tree-like structure by separating data into subgroups aster Management Authority (AS DMA) [25) and Indian flood
based on featu re.s. Each node represents a characteristic. while. inventory [26]. These. re.c ords include vital details regarding
leaves indicate the. projecte.d d ass (e.g .. flood incidents).
pre.vious Hood events. such as their geographic d istribution
Decision trees can handle differe nt data types but may overfit and periods. laying the groundwork for furtJ-.e.r research and
due to noise (201. The. overl-itting can be reduced by pruning. flood risk assessment. \Ve use Google. Earth Engine. (GEE)
limiting tree de.pth. and e nsemble. me.thods like RF. [27) platfom1 to access Climate Hazards Group lnfraRed
5) Random F01~s1 (RF): RF is an ensemble ML technique.
Precipitation with Station (CHIRPS) dataset [28 1to explore the.
made. up o f decision trees. Random Forest is use.d fo r cfassi-
relationship between rainfall patterns and Hood disasters. This
fication and regre.ssion. It builds numerous decision trees by satellite-based dataset provides high-resolution precipitation
sampling the. data and selecting random features for each tree.
estimates. e nabling .t thorough inve.stigation of how rainfoJI
A majority vote makes the. final decision for classification or
affects flooding. 111e next step im•olves getting geographical
average. for regression. Random Forest e.n hances accuracy and
data from the USGS. specific.ally the DEM data. It includes
generaJization by merging dive.rse decision tree.s (2 1].
valuable detai Is about the topography and terrain charac-
B. Geosp(lJia/ dflla teristics o f the area. including elevation. slope. aspect. and
TI,e. United Stales Ge.o logical Survey (USGS) 1221 provides curvature. Identifying flood-pro ne ,u e.as and appreciating flood
geospatial data about the region·s topography and terrain. threats requires understanding the area•s physicaJ features.
The DEM. from the USGS provides data on elevation. slope. The. acquired flood data from ASDMA and the. Indian Flood
aspect. and c urvature. which are crucial in understanding Inventory dataset were. used to labe.l tJ-.e. location data for
flood-prone areas. These metrics are essentiaJ te1Tain and point sampling ope.rations. A total of 984 points. as shown in
land su1face featu re.s. assisting in understanding a region's Figure 1. represent recorded flood events. while 1020 points.
topography and hydrological behaviour. c.ategori 1..ed as non-flood points. were generated randomly. as
I) Elevation: Elevation is the. height o f the Earth•s surface. shown in Figure 2.
above o r below the sea level. Elevation data is critical for Hood
prediction models since it affects wate.r accumulation and flow.
Higher altitude areas are. less prone to flooding. while. lower
e.levations are more vulnerable to inundation during floods
[231.
2) Slope: 111e steepness or inclination of the. terrain is
represented by a slope. It is computed as the. vertical change.
in elevation divided by the hori1..ontaJ d istance between two
places. During heavy rains. steeper slopes c.an acceler..\le water
flow and increase the danger of surface nmoff. pote.n tially
leading to floods. Gentler slopes. on the. other hand. allow
wate.r to flow more. gently. lessening the danger of floods 1"24).
]) Aspec1: 111e direction a slope faces is called aspect. It
is the azimuth angle measure.d clockwise from the north to
the sharpest slope direction. The distribution of solar radiation Fig. I. Flood Localion
and the accumulation o f snow or water on the slope. is
influenced by aspe.c.t. North-facing slopes. for example. receive.
less sunshine and may keep snow or water longer. influenci ng
wate.r availability and Hood patterns 1"23).
4) Curw1ture: The variation in slope along a sud·ace is
represented by c urvature. It aids in determining ,vhether the.
te1Tain is concave o r convex. Positive. curvature. indicates a
convex surface. (such as a hilltop). whereas negative. curvature.
indicates a concave surface (such .ts valley). Curvature directs
wate.r flow. influences ,vater accumulation or drainage in a
location. and is helpful in flood mode.l ing (241 .
IV. M ETHODOLOGY
11-.e. steps involved in Hood prediction fo r Assam include.
gathering data. preprocessing. developing the l\{L classifiers.
training the classifie.rs. incorporating c.o ordinate.s provided by
the users. and rainfall forecast data. Fig. 2. No~Flood Location

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:31:52 UTC from IEEE Xplore. Restrictions apply.
 TI,e. DEM data is processed with ArcGJS [29]. ArcG JS is
also use.d to ge.n erate additional ge.ographic layers such as +
elevation (Figure 3). slope. (Figure 4). aspect (Figure 6). and
c.urvature. (Figure 5). TI1e final datase.t consists o f the past five.
days• rajnfall data for flood classification and the next five.
days• rainfall data for flood prediction. c umulative. five days•
rainfall data. longitude. latitude. elevation. slope. aspect. and

.---
curvature.

.....
-----·
.......
+ __
...._,........

--
---·-.
,..._,..,
......
.......
Fig. 6. Aspccl Profile

8 . Classifimtio11 of Flood

Fig. 3. Elcvalion Profile

+
.-.....
............. -----------~
,_ ..,.,..,._,._,.. . ,.._
..... ...
0,,,,....- ....
• i un
- t-.UM

.,,. ..,,......
- Bll!>.it3f

. ,,,,..no. Fig. 7. Flood ClllSSifkation ~fodel

• n - •m
• •m•••n
. ...... ,.n> The. steps involved in the cl assification of flood are shown
in Figure 7. Five ML classification models are. explored in this
Fig. 4. Slope Profile study: LR. SVM. KNN. DT. and RF. The dataset is divided
into training and testing sets in a 70:30 ratio. The. mode.ls
are trajned to determine. whether or not a specific location
is prone to Hooding based on rainfall and geospatial data.
The model's perfom1an-ce and generalization capability are.
e.valuated on the testjng subset using d iffe rent per1·om1an-c.e.
indicators. including accuracy. pre-c.ision. recaJL and Fl-score.
The ped·ormance algorithms are. compared to detem1ine which
algotithm exhibits the. best overall ped·ormance fo r the flood
classification as sho\vn in Table. I. The best-per1·om1ing model
RF is picke.d and utilized in flood predictjon.
C. Floo(l Prulic1io11
\Ve have adopted a systematic data-drive.n approac h to
forecasting probable Hoods in a particular place. as shown
in Figure 8. The core of the Hood prediction me.thodoJogy in-
Fig. 5. Cuivnture prolilc volve.s applying a mac hine learnjng model trained on historical
flood data and pertjnent features. The daily rainfall forecast

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:31:52 UTC from IEEE Xplore. Restrictions apply.
 implementing timely interve.n tions to mitigate. lhe. impact o f
.,,.,, .....w.-. . ,.. r-.........,,..1
1,,,..;.,o1,, ,..,..,,...., flooding.

Flood Prediction

Fig. 8. Flood Prediction ~fodcl 11.am

...........
info rmation for lhe following five. days is obtained by using
OpenWeather API [301. The proposed system utilizes latitude.
and Jongitude coordinates lo ac.cess location-sped fie. forecast
data. ensuring prec.ise flood prediction. TI1e collected forecast
data undergoes pre.processing to exlracl the. daily rainfalJ
Fig. 9. We~ A pp From Page
values. which are then aggregate.d to calculate lhe. cumulative.
rainfall for the. sele.c.led location O\'er the 5-day period. l11is
The. proposed flood prediction syslem can antidpate. floods
c.umulati\'e rainfall data is of par,u11ount importance. in evalu-
in Assam for the. nexl five. days. To make the. predic.tions
ating Hood danger accurately. By c.onside,i ng the cumulative.
accessible and user-f,iendly. we employe.d the. Flask web
rainfall O\'er the 5-day d uration. lhe flood prediction models
frame.work (3 1-1 to create a use.r-frie.ndly i nted·ace. As shown in
can ac.c.o unl for polential saluration o f the ground and lhe.
Figure 9 . users c.an input latilude and longitude coordinates to
impact o f sustained rainfall o n flood oc.currences. re.suiting in
acquire. flood predic.lions for the. specific location. The sysle m
more robust and reliable flood risk assessme.n ts. The. geospatial
uses rainfall prediction data from the OpenWeather APL total
data about the. study area. such as slope. elevation. aspect.
rainfal I data. preproc.e.ssed geographic a.I data. and the. Random
and c urvature. are also use.d in addition to rainfall data in
Forest model. the.n combines these. data to forecasl the. chanc.e.
flood prediction. The fore.c.asl rainfall and geospatiaJ data ,u·e.
of Hoods based on the user input TI1e Hood pre.d iction and
then combine.d and properly aligned for acc urate analysis and
rainfall fo recasl for the. next fi"e days in their selecte.d region is
prediction. l11is comprehensive. data integration ,Uds in under-
displayed to the users as shown in figure 10. l11is user-friendly
standing the pote.n tial flood risk in lhe. given location. 11-.e.
strategy promotes flood pre paration and decision-making for
prediction result is displayed as a binary result. cate.gorizing authori ties. disaster managemen t teams. and communilies.
the area as flood-prone or not. This melhodology o ffe rs a
practical and economical method o f predicting floods by com-
bining diffe renl dala sources and ulHiz.ing machine learning Flood Prediction
techniques. there.by enhancing catastropl-.e. prepare.d ness and
response efforts.
\I. R ES ULT AND A NALYSIS

For ll-.e. flood classification. ,ve analyzed and compared five.

ML aJgorithms: LR. SVM. KNN. DT. and RF. The assessment
was based on lhe cl assification ped'o rmance crite.ria: accuracy.
precision. recall. and F l -score. The. aim is to pick the best
model among these algori thms fo r flood pre.diction in ll-.e.
future.. P11!dit liur1:
'rh• -11•1 , ...... no flood t.. thi• lo.,.lion,
Model Accuracy Precision Recttll Fl •Scorc
LR 0.940! 0 .9'2 0.96 0.94
SVM 0.9377 ().9 1 0.97 0.94
K NN 0.9451 o.<n 0.97 0.95
DT 0.9426 ().93 0.95 0.94 Fig. 10. Resuh
RF 0.9626 0.94 0.9:s 0.96
TABLE I
MODEL P e RFORM AXCI!
\1 1. CONCL USION AN D FUTURE WORK
This sludy presents a study on flood forecasting using
Base.d on ll-.e. table I the d assifkalion pe.rformance melrics. machi11e. learning. hislorical data. geographical data. and rain-
the RF model achieved the highest ac.c.uracy. precision. recall. fall forecast data. Five popular machi11e. learning classifiers
and F I -score among all five algo1i lhms. So . we ha\'e selected were expJore.d: Logistic Re.gression. Support Vector Machines.
RF for flood predic.tion. The proposed flood prediction system K-Nearesl Neighbours. Decision Tre.e. and Random Forest
can aid authorities in making well-infom1ed decisions and The best-perfom1ing Random Forest model was selected for

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:31:52 UTC from IEEE Xplore. Restrictions apply.
use. in o ur propose.d flood pred ict ion model. Histo,ica.l Hood ll 51 E. Nohani. ).,I. Moharrami. S. Sharofi. K. Khosmvi. B. Pradhan. B. Pham.
data was collected from the ASDMA and the Indian Flood S. Lee. and A. ).klesse. ..La~lide suscep1ibili1y mapping lL,;.ing dilfer•
em gis•ba.,;.cd bivariate models.... \.\bier. vol. 11. 07 20 19.
Inve.ntory. The ge.o sp atial data was downloaded from USGS: Ll6 1 M.-J. Kroak and F. Omtt"ling. Cortogrophy: ,·i.u1t1/i;,tui<m of ,:eo.(J)OlitJ/
we have used ele.vation. slope. aspect. and c urvature. 11-.e. dt1tl1. C RC Press. 2020.
daHy rainfall forecast was obtained by using Open\\leathe r Ll 71 M. I. H. Jnti. P. B. Sanloso et al.. "Prcdk-1ion or llood nrt"fls lL,;.ing
the logis tic rc-gre~<;..,;.ion method (case study of the pro,1 inccs b..1ntcn. dki
APL A user-friendly inte rface. was also develope.d using t.1-.e. jakana. nnd west java): · in )(Jl,nwl of P/ty.(ics: Omference Serie.c. vol.
Flask web frame.work. The. \Veb application allows end users 1367. no. I. IOP Publis hing. 20 19. p. 012087.
to input latitude and longitude data to get Hood predictions l l81 M.-J. C hang. H.-K. C hang, Y.-C. c 1....n. G.-F. Lin. P.•A. c 1....n. J.-S.
L1i. nnd Y.-C. Tan. ..A support vector machine forcca.,;.ting model for
for the specifie.d Jocation in Assam. Our study can boost typhoon fl ood inundation m.1pping aOO early llood warning system.,;: ·
flood forecast e fficiency. allo\ving proac tive. de.c ision-making \.\bier. vol. 10. no. 12. p. 1734. 20 18.
for flood management. Ll9 1 P. C unningham aOO S. J. Dclnny. "k-ncarcs.1 neighbour cla.,;.sifiers-a
nnoria1: · ACM comp111in,: .~111wrys (CSUR). vol. 54. oo. 6. pp. 1- 25.
Despite. the positjve findings o f this study. certain limitations 2021.
c.an be. add resse.d in future. research. The predicted performance. [201 B. Ch..<irbuty nnd A Abdulnzccz. "'C.'lllSSilkation ba.,;.ed on decision
of the model can be. improved further by includ ing more. tree nlgorithm for m.1chinc lt"flming: · Joun1t1I of A1>J)Ue,I Sde11ce mui
Tec/i,w/o,:y Jj·emis. ,•ol. 2. no. 0 1. pp. 20-28. 202 1.
geospatial and e.nvironmenta.l data. such as river characteris- [211 S. J. Rigaui. ..Random fot('St: · Jountlll Qf huura,ice Me,iici,ie. vol. 47.
tics. including flow rate.s. ,vate r Je.vels. discharge. ra tes. soil oo. I. pp. 31-39. 2017.
moisture data. etc. A flood susceptibility map depicting t.1-.e. [221 USGS • U.S. Geological. (2022. 12) Earthcxplorer. IOnlinc l. Availabk :
USGS-U.S.Gcological
likelihood and severity of flooding in a g iven place would ,Ud [231 P. V. Bolstad and T. Stowe. ..An c\'aluation of dcm acc-umcy: clevntion.
in proactive. flood control and risk assessment. slo1:,c. and a.,;.pec1: · Phmo,:ramme1ric £11,:ineering & Remme Sen.(illg.
, ,0 1. 60. no. 11. pp. 1327- 1332. 1994.
[241 \V. J. Pierson Jr nnd R. A. Stacy. ..The elc\'alion. slo1:,c. aOO c-rnvaturc
REF ER ENCES spcc1rn of n wind roughened sea surface:· NASA. Tech. Rep.. 1973.
[251 Go\'ernmcm of Ass.am. (2023. 06) A,;.sam s.tate disa.,;.ter m.1n.1gcment
11 1 ScknccDircct. (2023. 06) Flood 1>rediction. Web- au1h:>ri1y. Website. (Online (. Availabk : h11p://sdn1.,ssam.nic.inlproj«1_
site. (Online). Availnblc: hu1>s://W"\\1w.scicnccdirec-t.com/topics/ flood.lnml
eanh-and· plnnctary• scicnccs/llood• 1>rediction [261 M. Sah:lrin. A Jain. R. R. Bais hya. S. Hnobam. 0 . Srccjith. D. P;ii. aOO
121 A,;.sam Flood. (2023. 06) A hislorical understanding of tL.;sam·s A. Rnlieein~ab. "lndin llood im1entOl)1 : the crt"fltion of a mulli-sounx
lloods. Website. IOnlineJ. Avnilabk: httpc.:1/wW"\\1.epw.inlengngc/aniclC'/ mtion.11 gcosp,.1tial databa.,;.e to fac-ililate compreh:-nsi\'e llood research:·
his torical. unders.tarxling-nss..,ms• floocb. Nomral flt1:.ard.(. vol. 108... no. I. pp. 619-633. 2021.
131 B. Mahcsh. ..Mnchinc learning algorithm.-.-a review:· lmem01io110/ )01,r• [271 0 . Mulanga aOO L. Kumar. ..Google eanh <cngi™" applications: · p. 59 1.
11al ofScie,ice m1d Reuarch (/.fSR).//menie1J. vol. 9. no. I. pp. 381- 386. 2019.
2020. [281 C. Funk. I~ Peterson. M. Land.,;.fefd. D. Pedtcros. J. Verdin. S. Shukla.
141 U.S . Gcolog.ical Survey. (2023. 06) How are floocb. G. HtL-.ak. J. Rowland L. Hnffison. A. Hoell ernl.. ·--nx- elim:itc h..<i:,.ards.
predicted? (Onli,.... I. !Online (. Avnilabk: h1t1>.<;. ://www.us~.gov/focp/ infrared pnxi1>it.1tion with stations-a new environntt"nlal record for
how-ore• lloo~-prcdk-ted monitoring t'Jttrcmes:· Scie-111[/ic , i'1UJ. vol. 2. oo. I. pp. 1- 2 1. 2015.
151 K. Gogoi and K. Rao. ··Analysis of rainfall tre~ O\'er assam, nonh east [291 Bri. ..An".Cils : · https://www.esri.com'en-us/arcgis/about•arcgis/
india:· c11 ej(,1mwl.org. IOnlincl. A\'ailablc: htt1>.<;. ://www.c-wejoum.<il.org/
1
overview. 2023.
vol I 7 no2/p,.1n..<ilysi.,;.-of-roinfal1-tre~-0\'er-nssmn• nonh•ea.<;.l• irxtinp [301 OpcnWeatherMap. ..Opcnwt"flthcnnap api: · lmpc.:1/opcnwt"flth:-rmap.org/
161 C. Borpujari and A. Bora. ·•Gis-ba.,;.cd digital terrain analysis of assam• api. ()6 2023.
meghafaya foothill.,;. in kanuup district. assam: · NM11re £11\·ironmem a.ml [3 11 M. Grinberg. Fla.tk "eb ,ierelopmem: dt!l•eloping weh t1J)plica1iom wil/1
R)/1111io11 Tec/mQ/ogy. vol. 20. no. 4. pp. 1765-177 1. 2021. py1llon. •• O'Reilly Media. Inc: ·. 20 18.
171 D. Kumar. ..Moniloring and nsscssmcm or lnnd U§.C' and L1nd cover
changes (I 977 • 20!0) in kamrup dis.tric-t or a.<;..,;.am. india u.,;.ing remote
sensing and gis tcc-hniqucs:· At>J)Ue,I £ro/of{l' mu! Em irrmmen10/ Re-•
1

.(eorc/1. vol. 15. pp. 221- 239. 01 2017.

PSI Z.-H. Zhou. 1\1'1chine /e,1mhl,:. Springer Nature. 202 1.
191 L Gupta and J. Dixit. ..A gis-ba.,;.ed ltood risk mapping of iw.am. india.
using the mcdil-ahp appro..1ch at ti).' region.11 and ndministrati\'e level: ·
Geocano lmemmionol. vol. 37. no. 26. pp. 11867-1 1 b'99. 2022.
l l 0I ).,I. Gomathi. M. G . Priya. C. C. Gowda. and 0 . Krishn.weni. ··f lood
inundiltion JU.1Pl>ing for using senti™"l• I S.'lt dat..1 for assam during 20 18: ·
Re:t Re-\·. J. S1N1ce Sci. Tec/mQ/. \101. 8. pp. 16-25. 2019.
l 111 S. Summpudi nnd K. Yamikula. ··Mapping aOO assessing s1>:itial cxk'nt of
lloods from muhitcmpoml synthetic npctlurc radar images: a CILSC' slUdy
on bmhm.1pulm river in nss..,m state. india: · Em•itrmmemol Science a,id
R)/1/llion Re.,ealY'lt. ,•ol. 27. no. 2. pp. 152 1-1 532. 2020.
l 121 S. B. Borah. T. Si\'a.,;.ankar. M. Ramya. nnd P. Raju. ··Rood inund.'l·
tion m.1P1>ing and monitoring in ka:,.imng;a national p.uk. assam lL-.ing
sentinel-I sar data: · Em•irrmmemol 11umiwring oll(/ a.J.(e.b'IIU!nl. vol. 190.
pp. 1-1 1. 2018.
l 131 S. Shivaprns.ad S h.1m1.1. I~ S. Roy. V. Ch.1kmvnnhi. and G. Srinivas..1 Rao.
..Flood risk assessment lL,;.ing multi-criteria amlysi.,;.: n case s.tudy from
kopili river b:isin. a.<;..,;.am. irxtin:· Gemm11ic.1. No111ml Ho:.,mt, ,md Rfrk.
vol. 9. no. I. pp. 79-93. 2018.
L) 41 A Agnihotri. A. Sahoo. and M. K. Diwakar. ..Flood prediction lL-.ing
hybrid anfis-nco model: a ca.,;.e s tudy: · in hn·en1i11e Comp111mio11 amt
lnformotion Tec/mQ/ogie.c: Proct>e,lhl,:., Qj /CICff 2011. Springer. 2022.
pp. 169-180.

Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on October 04,2024 at 15:31:52 UTC from IEEE Xplore. Restrictions apply.