Storm Surge Susceptibility Mapping and Prediction of Tropical Cyclones using Machine Learning along the Eastern Coastline of India

Storm Surge Susceptibility Mapping and Prediction of Tropical Cyclones using Machine Learning along the Eastern Coastline of India | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Storm Surge Susceptibility Mapping and Prediction of Tropical Cyclones using Machine Learning along the Eastern Coastline of India Subhra Mukherjee This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9505599/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract Storm Surge is an unusual upliftment of the sea water and is measured in terms of the height of the water predicted above astronomical tide. Primary cause of such surge is the strong wind during the landfall of cyclone which forces the sea water beyond the coast line. The Bay of Bengal is considered for the genesis of maximum number of tropical cyclones particularly due to its higher sea surface temperature during the pre-monsoon and post-monsoon period. Additionally, the Bay of Bengal is a shallow, partially enclosed body of water and is funnel-shaped which triggers the genesis and intensification of cyclone. The states of Odisha and Andhra Pradesh lie along the eastern coastline of India which stretch for roughly around 574.71 km and 1053.07 km respectively. Tropical Cyclones mostly make landfall along these coastal districts making these highly susceptible to storm surge particularly due to their low lying terrain, presence of River deltas, and degradation of natural barriers. Present research aims to map storm surge susceptibility for six coastal districts of Odisha and twelve coastal districts of Andhra Pradesh using Light Gradient Boosting Machine (LightGBM), Random Forest and XG Boost (Extreme Gradient Boosting), the machine learning models. Area Under Curve (AUC), calculated using test data, indicated that XGBoost, Random Forest and LightGBM which map the susceptibility with acceptable accuracy of0.980, 0.977 and 0.978 respectively. Additionally, this research implements couple of XAI techniques such as SHapley Additive exPlanations (SHAP) and permutation importance which revealed Rainfall and Elevation having highest contribution. Storm surge prediction was done with Multi-Layer Perceptron regressor mechanism where some important meteorological parameters were considered. These factors were trained with Storm Surge levels from the period spanning from 2007–2023 which achieved MAE at 0.18 and MSE at 0.26. Cyclone parameters of 2024 were given as an input in the algorithm and the resultant output was Storm Surge levels in meters which was also verified with the INCOIS predicted report and IMD reported surge levels along the tidal gauges. Storm Surge Eastern Coastline of India Cyclone Machine Learning Explainable AI (XAI) Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 1. Introduction Tropical cyclone are among the most disastrous meteorological phenomenon, causing devastation in the coastal areas, mainly due to the ferocious wind, heavy rainfall, lightning, thunderstorm and, most critically, the storm surge (Varma et al., 2023 ).. Storm surge is the unusual upliftment of the sea surface with respect to the predicted astronomical tide influenced by changes in atmospheric pressure or by powerful winds pushing water towards the coast (Gregory et al., 2019 ). The magnitude of storm surge in a shallow coastal region depends on coastline geometry, bathymetry, wind speed and size (Lockwood et al., 2022 ). The Indian Peninsula bounded by north Indian Ocean is a hotspot region for the genesis of tropical cyclone (Varma et al., 2023 ). Around 2% of the India’s GDP and 12% of the revenue of Government of India is spent every year for aftermath of the cyclones hitting the coastal belt of India (National Cyclone Risk Mitigation Project, https://ncrmp.gov.in/cyclones-their-impact-in-india/ ). Primarily, the eastern coastline of India, especially Odisha and Andhra Pradesh coast is severely affected by the cyclone-induced flood causing massive destruction and loss of life (Sarkhel et al., 2019 ; Basheer Ahammed & Pandey, 2019 ;Varma et al., 2023 ). There is a bimodal type of distribution of tropical cyclone, mainly forming during pre-monsoon (April to June) and post monsoon (September to December) seasons (Varma et al., 2023 ). Formation, intensification, and movement of tropical cyclone are influenced by both seasonal and intraseasonal fluctuation in temperature, moisture content in the atmosphere and low-level potential vorticity in the ocean and atmosphere (Xiao-ting et al., 2020 ). Ocean hosts the breeding ground for the genesis of cyclone therefore understanding the atmospheric and oceanic phenomenon becomes crucial to determine the formation and intensification of cyclone (Emanuel, 2005 ). Strong force of tropical cyclone tends to result in two type of dynamic reaction that is upwelling and turbulent mixing which results in lower sea surface temperature (Pan et al., 2018 ). Storm surge accompanied with land subsidence exacerbate inland flooding, salt water intrusion and erosion of shorelines (Wang et al., 2012 ; Xianwu et al., 2021 ). In the past fifty years, storm surge has claimed lives of the people worldwide rather than by combined effect of earthquakes, tornadoes, freshwater flood and lightning (Emanuel, 2005 ; Sahana et al., 2021 ). Sarkhel et al. ( 2019 ) states that nearly one third proportion of the total population of the nation inhabited in the coastal area are at the risk of cyclones, storms and heavy rainfall. The harsh winds accompanied with the cyclone destroys infrastructure, houses especially causing loss to marginal section of the society (Sarkhel et al., 2019 ). Odisha is considered as extremely vulnerable to cyclone compared to other coastal regions of the India. The coastal districts of Odisha are vulnerable for around 35% of all cyclone crossed Indian Coast (Sarkhel et al., 2019 . In 1737, Odisha was hit by a super cyclone, followed by another led to the death of around 75000 lives (Sarkhel et al., 2019 ). The vulnerability atlas of India has categorised Balasore, Bhadrak, Kendrapara, Jagatsinghpur, Puri and Ganjam into high risk zones in Odisha (Sarkhel et al., 2019 ). Disaster response force is often hindered by mobility and infrastructure facilities (Sarkhel et al., 2019 ). The impact of climate change will increase the levels of vulnerability (Sarkhel et al., 2019 ). Andhra Pradesh coastal belt comprises of mostly low elevated topography thus making it highly sensitive to coastal flooding and saltwater intrusion (Basheer Ahammed & Pandey, 2019 ). Three large rivers of Andhra Pradesh make Andhra Pradesh susceptible to storm surge penetration namely Godavari, Krishna and Pennar rivers (Rao et.al,2006). There might be loss of residential and agricultural land due to the predicted sea level rise and increase in temperatures and there will be an upsurge in salt concentration and decline in water quality especially in agricultural field by the ravages of time (Kantamaneni et al., 2020 ). Advancement in computational power of machine learning techniques have led to developments in data driven modelling of storm surge (Bruneau et al., 2020 ; Chen et al., 2021 ; Ramos-Valle et al., 2021 ). The results derived from earlier studies suggests that machine learning algorithm can model storm surge levels with high accuracy compared to physics based algorithms (Lockwood et al., 2022 ). In low latitude regions one may derive large number of errors after implementing machine learning algorithm which is particularly linked to tracks of hurricane as the dataset available for training and testing are limited (Bruneau et al., 2020 ; Tadesse et al., 2020 ; Tiggeloven et al., 2021 ). The surge level increases as a cyclone approaches towards the coast, and surge levels decreases with the increase in the hurricane speed (Lockwood et al., 2022 ). There is a projected estimate that by 2100 the sea level will rise in the range of 0.26–0.98 m (Bittermann et al., 2013 ). In contemporary studies Machine Larning and Deep Learning stands as a novel approach (Alshayeb et al., 2024 ). Deep Learning is a subfield of Machine Learning serves as important architect in flood modelling and susceptibility mapping (Alshayeb et al., 2024 ). It can handle large complex geospatial dataset but the condition is that the proper factors should be prepared after Multicollinearity analysis (Alshayeb et al., 2024 ). Thus, it can manifest the spatio-temporal pattern in a susceptibility map (Alshayeb et al., 2024 ). Present research aims to develop a storm surge susceptibility map to identify areas highly vulnerable to storm surge events along the Odisha-Andhra Pradesh coastline. Additionally, it attempts to predict storm surge triggered by tropical cyclones along the Odisha-Andhra Pradesh for the year 2024. Such mapping and prediction will be carried out using several machine learning approaches to test the potential and efficacy of ML in storm surge susceptibility analysis and prediction modelling. 2. Study area, Data used, and Methodology 2.1 Study Area The coastal districts of Odisha and Andhra Pradesh have been chosen as the study region extending from 21.49°N to 13.63°N and 86.93°E to 80.02°E with total area of 84,765.98 km 2 (Fig. 1). Among eighteen coastal districts Balasore, Bhadrak, Kendrapara, Jagatsinghpur, Puri and Ganjam districts belong to Odisha while, Srikakulam, Vizianagaram, Visakhapatnam, Anakapalli, Kakinada, Dr. B.R. Ambedkar Konaseema, West Godavari, Krishna, Bapatla, Prakasam, Sri Potti Sriramulu Nellore and Tirupati belong to Andhra Pradesh... Present study area is strategically important for trade and commerce, fisheries, marine economy, and biodiversity hotspots. Figure (1) Geographical overview of the study area representing the coastal districts of Odisha and Andhra Pradesh 2.2 Data used Details of the data used in the present study are presented in Table 1. To solve the first objective we obtain data like Aspect, Distance to the River, Drainage Density, Elevation, LULC, NDVI, NDWI, Rainfall, Slope, TWI refer Fig. 3 & Cyclone tracks. Every layer was obtained. For finding out the elevation dataset we chose Shuttle Radar Topographic Mission (SRTM) 1 Arc-Second Global elevation data has been used to derive elevation, slope and aspect in Google Earth Engine Platform. Furthermore, Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) – Daily data has been used to for rainfall, due to it’s high frequency. High-resolution multispectral imagerySentinel-2 Level-1C (Top-of-Atmosphere Reflectance) dataset provided by Copernicus has been obtained for calculating NDWI and NDVI representing water abundance and vegetation density. Global land cover dataset has been provided by European Space Agency. From the HydroSHEDS ( https://www.hydrosheds.org/ ) has been obtained to calculate Distance to the river and drainage density. To obtain flood inundation we used Sentinel-1 SAR data, NDWI generated from LANDSAT 5 TM and LANDSAT 7 ETM+. The cyclone tracks were obtained from IMD best track ( https://rsmcnewdelhi.imd.gov.in/report.php?internal_menu=MzM= ) filtering out tracks from 1995–2024. To fulfil the second objective we obtained IBTrACS (International Best Track Archive for Climate Stewardship) dataset ( https://www.ncei.noaa.gov/products/international-best-track-archive ) from where we have taken important factor like latitude, longitude, Radius of the outer storm, wind speed, Bearing and translation speed. These factors were trained with Storm Surge levels from the period spanning from 1995–2023. Table. (1) Details of the data used in the present study Sl. No Provider Product Name Spatial Resolution Temporal Resolution Section 1 : Data used for susceptibility mapping 1 USGS Shuttle Radar Topography Mission (SRTM) Global 1 arc-second 30 meters — 2 CHIRPS Precipitation (rainfall) 5.5 km Daily 3 Copernicus Sentinel-2 Level-1C (Top-of-Atmosphere Reflectance) 10 meters 5 days 4 ESA Global land cover dataset 10 meters — 5 HydroSHEDS Free-Flowing Rivers — — 6 Copernicus Sentinel-1 Synthetic Aperture Radar 10 meters 12 days 7 USGS LANDSAT 5 Thematic Mapper 30 meters 16 days 8 USGS Landsat 7 Enhanced Thematic Mapper Plus (ETM+) 30 meters 16 days Section 2 : Data used for prediction modelling 9 IMD Best Track (1995–2024) — — 10 NOAA IBTrACS (International Best Track Archive for Climate Stewardship) — — 2.3 Storm Surge Inventory Storm Surge inventory dataset useful for model training, testing and validation has been used for generating storm susceptibility map and for prediction of storm surge. The dataset is therefore. The inventory was formed using a comprehensive approach, that includes IMD best track data forming into Cyclonic Storm from 1995–2024 (Table 2) that made landfall in Odisha & Andhra Coast. INCOIS predicted surge location, height and IMD reported surge levels for 16 years with location of landfall. Flood inundation information have been obtained for different cyclones using Sentinel 1 SAR Interferometric Wide Swath (IW) dataset Table (2): List of Cyclones considered for study Basin CYCLONE Category Estimated Central Pressure Maximum sustained wind Pressure Drop Date of Formation Date of Landfall Year Bay of Bengal UNNAMED VSCS 978 77 30 07-11-1995 09-11-1995 1995 Bay of Bengal UNNAMED VSCS 982 77 29 13-11-1998 15-11-1998 1995 Bay of Bengal UNNAMED CS 992 45 - 12-06-1996 16-06-1996 1996 Bay of Bengal UNNAMED VSCS 988 77 - 04-11-1996 06-11-1996 1996 Bay of Bengal UNNAMED ESCS 968 90 40 15-10-1999 17-10-1999 1999 Bay of Bengal UNNAMED SUCS 912 140 98 25-10-1999 29-10-1999 1999 Bay of Bengal UNNAMED CS 1000 35 8 14-10-2001 16-10-2001 2001 Bay of Bengal UNNAMED SCS 992 55 14 11-12-2003 15-12-2003 2003 Bay of Bengal PYARR CS 990 35 6 17-09-2005 19-09-2005 2005 Bay of Bengal OGNI CS 998 35 6 29-10-2006 30-10-2006 2006 Bay of Bengal KHAIMUK CS 994 40 8 13-11-2008 15-11-2008 2008 Bay of Bengal LAILA SCS 986 55 15 17-05-2010 20-05-2010 2010 Bay of Bengal HELEN SCS 990 55 17 19-11-2013 22-11-2013 2013 Bay of Bengal LEHAR VSCS 980 75 26 23-11-2013 28-11-2013 2013 Bay of Bengal PHAILIN ESCS 940 115 66 08-10-2013 12-10-2013 2013 Bay of Bengal HUDHUD ESCS 950 100 54 07-10-2014 12-10-2014 2014 Bay of Bengal DAYE CS 992 35 8 19-09-2018 20-09-2018 2018 Bay of Bengal PHETHAI SCS 992 55 15 13-12-2018 17-12-2018 2018 Bay of Bengal TITLI VSCS 972 80 32 08-10-2018 10-10-2018 2018 Bay of Bengal Fani ESCS 932 115 66 26-04-2019 03-05-2019 2019 Bay of Bengal GULAB CS 992 45 10 24-09-2021 26-09-2021 2021 Bay of Bengal YAAS VSCS 970 75 28 23-05-2021 26-05-2021 2021 Bay of Bengal ASANI SCS 982 55 16 07-05-2022 11-05-2022 2022 Bay of Bengal MICHAUNG SCS 986 55 16 01-12-2023 05-12-2023 2023 Bay of Bengal DANA SCS 986 60 18 22-10-2024 24-10-2024 2024 2.3 Methodology Entire methodology is divided in to two parts (Fig. 2 ). First part deals with the dividing the entire study area into four zones (very high, high, moderate and low) based on its susceptibility to storm surge events. Numerous driving factors e.g. aspect, distance to the river, drainage density, elevation, LULC, NDVI, NDWI, rainfall, slope, TWI & Cyclone tracks from 1995–2024 have been considered for deriving the rate of susceptibility. The Deep depression which had intensified into cyclonic storms were considered for this study. The tracks were laid over the layers in GEE environment and flooded areas were identified by using SAR data of Sentinel 1. The tropical cyclones which were formed on or before 2014 for that NDWI was generated using LANDSAT 7 ETM+. For the flood inundation studies prior to 1999, we used NDWI generated from LANDSAT 5 TM. 12,932 samples were taken for training and testing in ML models. The entire dataset was split into training and testing subset i.e 75:25 ratio.. While selecting the Machine Learning algorithm we carefully considered the potential of handling large complex geospatial datasets (Alshayeb et al., 2024 ). Specifically, we opted for LightGBM, XGBoost & Random Forest due to it’s effectiveness in handling large and complex geospatial datasets. This diverse selection was mainly aligned with our objective to systematically compare different output of the 18 coastal districts. In the second part, ibtracs dataset of NOAA has been used. We mainly considered the latitude, longitude, Radius of the outer storm (Vo), Translation speed, Bearing in degrees and wind speed in knot (Lockwood et al., 2022 ). A Multi-Layer Perceptron deep learning model was chosen for this task the model was trained for the period of 2007–2023 with a training and testing subset ratio of 75:25due to its efficiency in capturing non-linear relationship in the data giving it a strong foundation for prediction of storm surge for the year 2024. 2.3.1 Preparation of driving factors for storm surge susceptibility mapping On the basis of meteorology, hydrology, topography and land use datasets, driving factors that have been selected influences the extent and severity of storm surges. Meteorological Parameters are selected to analyse the potential impact on coastal areas, Wind Speed (kt) is responsible to drive the surge along the coastline and it is directly proportional with surge levels, i.e higher the wind speed higher is the surge level(Characteristics of the Hurricane Storm Surge - D. Lee Harris, United States. Weather Bureau.) Radius of the outer storm is considered as the distance from cyclone center to the outermost closed isobar(Chavas & Lin, 2016 ). The translation speed is generally influenced by steering wind of the environment governed by their motion (Chavas & Lin, 2016 ). The direction of the storm from one point to other leads to coastal water to displace leading to surge impact (Lin & Chavas, 2012 ). Hydrological Parameters influence how surge water interacts with the coastal terrain. Heavy rainfall tends to exacerbates surge impacts by in increasing flooding along the coast (Lin & Chavas, 2012 ). The presence of Coastal rivers can result in amplification of storm surge flooding due to it’s backwater affect. The area which is having high drainage density may witness prolonged inundation. Topographical parameters are responsible to determine how surge waters inundates inland. Aspect signifies the direction of a slope face and the slope exposed onshore are vulnerable to wave effect and surge penetration. A recent data has revealed that wetland inundation is affected by storm surge(Cahoon et al., 2006 ). The slope angle is inversely related to occurrence of water stagnation (Cahoon et al., 2006 ). Topographic Wetness Index determines flood prone areas on the basis of accumulation of surface water potential (Beven et al., 1984 ). LULC is the controlling factor for surface run off and infiltration (Chilagane et al., 2021 ).The forest area allows infiltration but the built up area does not allow infiltration (Abdel Hamid et al., 2020 ). The NDVI and NDWI shows the interplay of water presence and vegetation, the decrease in vegetation cover and increase in water presence make an area susceptible to flooding. There is more or less influence on surge heights due to change in track orientation along western and eastern coast respectively (Azam et al., 2004 ) 2.3.2. Data pre-processing for Flood inundation studies In the cloud computing platform different time data were given for each cyclone to derive the flood inundation. The pre cyclonic data and post cyclonic data is an important key to conduct this study. Sentinel-1 SAR images are chosen because it emits Microwave radiation of the Electromagnetic Spectrum which can penetrate cloud cover. Interferometric Wide Swath (IW) mode was used with VH Polarisation. The significance of VH polarisation to prefer it for this study is because it is sensitive to surface roughness, Less Affected by Double-Bounce Effects& capable of Better Vegetation Penetration. The backscatter coefficient of VH polarization is expressed as σVH, dB​=10log10​(σVH, linear​)(Halder & Bandyopadhyay, 2022 ). The decrease in VH backscatter for pre and post flood image shows inundation. The spatial resolution this dataset is 10m with a swath of 250 Km. Global Surface water dataset has been considered as an important layer which removes the permanent water bodies like lakes and river. The values of VH polarization backscatter (σ) is converted into logarithmic scale (dB) by using σdB​=10log10​ (σlinear​) which helps to compute median backscatter values in segments. The perform the change detection studies the difference was found out between pre-flood and post-flood images mathematically it can be defined as ΔσdB​=σdB, pre​−σdB, post​. Refined Lee filter is used to remove the speckle noise effect which helps to preserve edges while it can remove noise. The terrain slope was calculated to exclude steep slopes where water does not accumulate by using Slope = tan−1(Δh/Δd​). The NDWI from LANDSAT 7 ETM + was generated by using Band 2 (Green) and Band 4 (NIR) and the NDWI from LANDSAT 5 TM was generated by using Band 3 (Green) and Band 4 (NIR). INCOIS predicted storm surge report was obtained and IMD reported surge event was referred. The cyclones which were forming into cyclonic storm and having landfall on Odisha and Andhra coast were considered for the study thus the track was considered as a layer for this study. For inundation studies training samples were taken in cloud computing platform and coastal area was considered as more vulnerable than the outskirts. The training samples were collected in terms of very high, high, moderate and low. SAR data was considered for those Cyclones which had landfall on or before 2014. NDWI of LANDSAT 7 ETM + was considered for those Cyclones having landfall on or before 2014. NDWI of LANDSAT 5 TM was considered for those Cyclones having landfall before 1999. 2.3.3. Multicollinearity analysis For generation of susceptibility map various predictor variables have been generated consisting of Aspect, Distance to the River, Drainage Density, Elevation, LULC, NDVI, NDWI, Rainfall, Slope, TWI & Cyclone tracks. LULC, Aspect, Cyclone tracks, Distance to the River, Drainage Density, NDVI, Slope and Rainfall has less Variance Inflation Factor (VIF) in terms of Multicollinearity, values thus make it as important predictor variables for storm surge susceptibility mapping (Alshayeb et al., 2024 ).Elevation, NDWI, and TWI has showed higher VIF values exacerbating VIF values. However, their contribution can’t be entirely neglected on susceptibility due to their scientific importance and considering the extent of the study area. Elevation, slope, curvature, aspect, SPI, TWI, STI, LULC, rainfall, river width, TWI, and soil types were having less Multi-collinearity while doing flood inundation studies (Rahman et al., 2021 ) 2.3.4.. Machine Learning Models Random Forest is a statistical learning model and an ensemble learning approach (Rahman et al., 2019), it is the aggregate of decision trees used for regression and classification founded by Brieman, 2001. Thus, it laid the blocks to build multiple decision trees and combines the result to prevent overfitting and to improve training accuracy. 100 decision trees were trained on a randomly generated subset of the data. Random split was kept at 42. The resultant prediction from multiple decision trees are averaged in case of regression analysis or voted on for generation of classification. There were some missing values in the training dataset to handle the same Simple imputation of mean were used which ensures that model is not affected by missing data. Ten features were termed as X and were considered for the study which has been derived from previous studies having less VIF value in terms of multicollinearity analysis. Y is considered as the target variable that is actually the training samples. Feature scaling was performed to assign equal weightage to all predictor variables. The train test data ratio was split into 75 and 25 respectively. Thus, the training process includes Bootstrapping, Splitting and Aggregation. The test data is considered as the unseen data that has been actually derived from X and the trained model is used to make predictions on X_test. The mathematical formula Random Forest Regression Formula after Breiman, L. (2001): $$\:\widehat{y}=\frac{1}{B}\sum\:_{b=1}^{B}{f}_{b}\left(X\right)\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(1\right)$$ The Mean Absolute error is a key to accuracy assessment measure of the model which finds out the average absolute difference between the actual and predicted values. Lower MAE is a sign of better performance of the model. $$\:\text{MAE}=\frac{1}{N}\sum\:_{i=1}^{N}{\left({y}_{i}-\stackrel{ˉ}{y}\right)}^{2}\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:..\dots\:..\left(2\right)$$ The MAE of Random Forest algorithm is 0.2 which shows that on an average the predictions deviate from the actual values by 0.2 units. LightGBM uses the mechanism of Gradient-boosted decision trees, having an efficient and scalable techniques for storm surge susceptibility modelling. This algorithm uses histogram based learning method, this gives it a foundation to handle large no. of datasets more efficiently .The model takes into account the parameters selected after multicollinearity analysis from the previous studies. The loss function is optimized and each tree in the ensemble learning is constructed (Iban and Bilgilioglu 2023 ). The algorithm gets to learn by correcting errors made by the earlier trees, beside this it can handle missing values which makes it quite adaptable to handle a large no of datasets. LightGBM can be an attractive choice for storm surge susceptibility mapping due to it’s effective traits of interpretability with fast training and prediction (Alshayeb et.al., 2024 ). The model builds 100 decision trees where the learning rate has been set to 0.1 for more accurate learning, to prevent it’s overfitting max depth was chosen as 5. Random state ensured the reproducibility of results. The mathematical formula for this gradient boosting algorithm after Friedman, J. H. (2001) which minimize loss using gradient boosting is. $$\:{h}_{m}\left(x\right)=-\sum\:_{i=1}^{n}\frac{\partial\:L\left({y}_{i},{F}_{m-1}\left({x}_{i}\right)\right)}{\partial\:{F}_{m-1}\left({x}_{i}\right)}\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(3\right)$$ The trees that are add sequentially by the mathematical formula given below: $$\:{F}_{m}\left(x\right)={F}_{m-1}\left(x\right)+\eta\:\cdot\:{h}_{m}\left(x\right)\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:..\left(4\right)$$ The Mean Absolute error is a key to accuracy assessment measure of the model which finds out the average absolute difference between the actual and predicted values. Lower MAE is a sign of better performance of the model. $$\:\text{MAE}=\frac{1}{N}\sum\:_{i=1}^{N}{\left({y}_{i}-\stackrel{ˉ}{y}\right)}^{2}\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(5\right)$$ The MAE of Random Forest algorithm is 0.3 which shows that on an average the predictions deviate from the actual values by 0.3 units. XGBoost is a gradient boosting framework which incorporates regularization technique. The basic idea is to grow trees gradually by fitting the residuals that the preceding tree predicted by the earlier trees by performing feature splits. The final prediction result is obtained by adding the scores of each leaf node throughout the trees. This algorithm employs first and second derivatives to perform Taylor expansion for it’s loss function. Thus model accuracy with it’s complexity can both be managed at a time (Wang et.al., 2012 ). The regressor model have created 100 trees sequentially the learning rate has been set to 0.1 to control how much each tree contribute to final prediction. The max depth is 5 to control overfitting. Each tree corrects the error from previous iterations after Friedman, J. H. (2001 by the following formula: $$\:{F}_{m}\left(x\right)={F}_{m-1}\left(x\right)+\eta\:\cdot\:{h}_{m}\left(x\right)\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(6\right)$$ For Mean Squared Error (MSE), the loss function is: $$\:L\left({y}_{i},{\widehat{y}}_{i}\right)=\frac{1}{2}{\left({y}_{i}-{\widehat{y}}_{i}\right)}^{2}\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(7\right)$$ The Mean Absolute error is a key to accuracy assessment measure of the model which finds out the average absolute difference between the actual and predicted values. Lower MAE is a sign of better performance of the model. $$\:\text{MAE}=\frac{1}{N}\sum\:_{i=1}^{N}{\left({y}_{i}-\stackrel{ˉ}{y}\right)}^{2}\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(8\right)$$ The MAE of XGBoost algorithm is 0.2 which shows that on an average the predictions deviate from the actual values by 0.2 units. MLP stands for Multilayer Perceptron is one type of Neural network or more specifically Artificial Neural Network. This algorithm doesn’t make any prior assumption about the data distribution. Extremely Non linear functions can be trained to appropriately generalise them. Thus, when developing numerical models and also when choosing between other statistical learning algorithm MLP becomes an attractive choice for it’s application in atmospheric science. The weights and output signals are determined by the addition of the input of the nodes and it is changed by a straightforward non-linear function. As from it’s name signifies it may consist multiple hidden layers with an output layer and finally an output layer. one node is interconnected to every other node in the former and next layer. During training the weights in the network are adjusted until the desired input—output mapping is obtained. The code is trained with input features obtained from ibtracs dataset of NOAA, the variables that are selected are latitude, longitude, wind speed (Vmax), radius of maximum wind (Ro), bearing, and translation speed. For the given study the model was trained with 6 hidden layers with 200 neurons with ReLU Activation function. For each hidden layer output the mathematical formula McCulloch, W. S., & Pitts, W. (1943 is $$\:{z}_{j}=\sum\:_{i=1}^{n}{w}_{ji}{x}_{i}+{b}_{j\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(9\right)}$$ $$\:{h}_{j}=f\left({z}_{j}\right)\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(10\right)$$ To compute output layer the mathematical formula is: $$\:\widehat{y}=\sum\:_{j=1}^{H}{v}_{j}{h}_{j}+c\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(11\right)$$ MLP updates weights using gradient descent and backpropagation by using the given formula after Rumelhart, Hinton, and Williams (1986) $$\:{w}_{ji}\left(t+1\right)={w}_{ji}\left(t\right)-\eta\:\frac{\partial\:L}{\partial\:{w}_{ji}}\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(12\right)$$ $$\:{v}_{j}\left(t+1\right)={v}_{j}\left(t\right)-\eta\:\frac{\partial\:L}{\partial\:{v}_{j}}\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(13\right)$$ 3. Results and Discussion 3.1 Implementation of Explainable AI (XAI) techniques The XAI techniques like SHAP analysis and Permutation importance serves as a Sophisticated approach for the explanation of DNN model(Alshayeb et al., 2024 ). SHAP analysis is a game theoretical methodology which analysis the prediction made by Machine learning algorithms (Alshayeb et al., 2024 ).Our study was subjected to SHAP analysis to derive feature importance it can help the researcher to get an idea how each factor is contributing to model prediction. SHAP value gives us an insight how each predictor variable influences the model output. Red represents high value whereas blue represents low value, the vertical spread shows that density and how varied the impact is. The Fig. (4),(5),(6) shows Elevation and Rainfall are the top contributors, since low lying and high rainfall areas are surge prone. In Random Forest, XGBoost and LightGBM High elevation and rainfall increases the susceptibility. In terms of LULC, probably the agricultural areas or built up areas are at higher risk, the areas closer to river are susceptible beside this higher NDWI is a potential indicator of flooding, these has been depicted in SHAP analysis of Random Forest in Fig. (4). In case of XGBoost, Increase in NDWI, closer proximity to Distance to River exacerbates surge induced flooding Fig. (5). LULC and NDVI has moderate impact and TWI, Aspect has minor impact. In case of LightGBM Fig. (6), higher NDVI shows less susceptibility, Gentle slope can contribute more, TWI has less importance compared to RF, Drainage Density and Aspect is having almost negligible influence. This analysis serves as a significant measure to frame storm management strategies, this helps to sort and prioritize the factors that influence the surge events(Alshayeb et al., 2024 ). Permutation importance measures global feature importance by checking the model performance when a single feature’s value is shuffled. This XAI technique was used for RF, XGBoost and LightGBM models Fig. (7),(8) & (9). Rainfall and Elevation emerged as the most influential factors in the three models thus influencing the dynamics of storm surge, LULC served as moderately important in all models indicating that the wetlands, vegetation cover area and wetlands have a meaningful impact on water retention. NDWI, and Distance to River have lower importance. NDVI, Slope, Topographic Wetness Index (TWI), Drainage Density, and Aspect consistently showed low to negligible importance scores. Collectively SHAP and Permutation importance gives an insight and assign importance to different factors in storm surge induced flood management(Alshayeb et al., 2024 ). 3.2 Storm Surge Prediction The ANN model was trained with a limited 25 datasets refer Fig. (10). The dataset was selected for the year spanning from 2007–2023. It included storm surge corresponding location collected from IMD for every selected cyclonic storms and surge. The input layer for the neural network includes six parameters with regard to the cyclone’s physical properties(Ramos-Valle et al., 2021 b). The dataset was obtained from ibtracs of NOAA and was sorted, the depression which formed into Cyclone Storms and those data of the cyclones were trained whose Storm surge data from IMD were available for studies. Input parameters consists of latitude and longitude of Storm Surge locations, maximum wind speed (Vmax), Radius of the outer storm (Ro), bearing, translation speed. The Vmax was considered as a metric to assess cyclone intensity rather than it’s central pressure because it barely contributes 15% to the magnitude of storm surge (Horsburgh et al., 2011). Ro is derived from a lognormal distribution therefore it is uncorrelated with other parameters. Ro was considered as the metric to derive cyclone size rather than Radius of maximum wind speed (Rmax) as it has positive correlation with Vmax (Chavas & Lin, 2016 ). While training an ANN model we provide both the input and output to get relationship between them. The model consists of multiple interconnected neurons that are capable to extract linear non-linear features, the ANN architecture is built with an input layer and a no. of intermediate layers that is denoted as hidden layers. In a hidden layer the relationship between an input and an output neuron can be denoted after (Ramos-Valle et al., 2021 ) as: $$\:AF={b}_{i}+\sum\:_{j=1}^{\text{n}}{x}_{{\text{in}}_{j}}{w}_{j\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(14\right)}$$ $$\:{x}_{\text{out}}=f\left(AF\right)\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(15\right)$$ Xinj and xouti are the input and output data of neuron I, AF plays the input to the activation function, the total no. of neurons is denoted by n and it is from the previous hidden layer connected to neuron I in the current hidden layer(Lockwood et al., 2022 ). The hidden layer consists of ReLU activation function that has been derived from Scikit Learn package. There are total 6 hidden layer consists of 200 neurons per layer, maximum training iterations, was limited to 1,000. The dataset given as an input were divided into 75% for training and 25 % for testing. The Mea Square Error and Mean Absolute Error serves as an evaluation metrics and it came as 0.1 and 0.2 respectively. X was having 5 parameters Latitude (°)', 'Longitude (°)', 'Vmax (kt)', 'Ro (km)', 'Bearing (°)', 'Translation Speed (km/h) and Y was Storm Surge in meters. The output was Predicted Storm surge for every classes. During Validation we used the same X parameters for the year 2024 and the output was Storm Surge in meters. The predicted report is a given in Table (3) Figure (10) Predicted output by ANN Table 3 Predicted Storm Surge Level by MLP Regressor (ANN) Basin Landfall Cyclone Category Hour Latitude Longitude Vmax (kt) Ro Translation Speed Bearing Neural Network Predicted Storm Surge (m) INCOIS Forecasted IMD Reported BoB Sagar islands and Khepupara Remal SCS 1800 21.75 89.2 54 139 8 5 1.47 1 to 2 m 1 to 2 m BoB Bhitarkanika and Dhamara Dana SCS 2100 20.5 87.1 62 231 7 335 2.1 1 to 2 m 1 to 2 m 3.3 Storm Surge Height Estimation Using Inverse Distance Weighting (IDW) Interpolation This section tries to generate a spatial distribution map of storm surge heights along the coasts of West Bengal, Odisha, and Andhra Pradesh as showed in Fig. (11), the dataset obtained from IMD tcintensity. The dataset was obtained for the year 1995–2024 which formed into Cyclonic Storm, it includes storm surge heights and their corresponding locations. The tropical cyclones that were considered include Amphan, Yaas, Fani, Gulab, Titli, Phailin, Hudhud, Laila, Aila, Sidr, and other significant storms affecting the region. Point data was generated in ArcGIS Pro environment where X and Y was Latitude and longitude respectively and Z was Storm Surge level in meter. A continuous spatial representation was generated using Storm Surge heights along the coast using Inverse Distance Weighting (IDW) that stands as an interpolation method. Thus, the unknown values were estimated using the value based on nearby known points. Districts having close proximity with Bay of Bengal such as South 24 Parganas, Bhadrak, and Balasore in West Bengal and Odisha recorded higher surge levels due to major Super Cyclonic storm like Amphan. The Andhra Pradesh coast experienced moderate surge levels with Laila, Michaung, and Hudhud contributing to surge variations. The visual pattern closely aligns with historical storm surge data. The IDW formula after (Watson, 1992; Burrough & McDonnell, 1998 is given below $$\:Z\left(s\right)=\frac{\sum\:_{i=1}^{N}{w}_{i}{Z}_{i}}{\sum\:_{i=1}^{N}{w}_{i}}\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:\dots\:.\left(16\right)$$ Figure (11) Simulation of Storm Surge using IDW 3.4 Accuracy Assessment and Validation The accuracy assessment was done from the published reports of IMD & INCOIS. The locations which are susceptible to storm surge induced flood inundation were collected from IMD reports and laid on ArcGIS Pro environment. The areas are Balasore, Bhadrak, Paradeep, Jagatsinghpur, Khurda, Puri, Chilika Lake, Ganjam, Gopalpur, Palassa, Tekkali, Srikakulam, Vishakapatnam, Kakinada, Konaseema, West Godavari, Machilipatnam, Krishna, Bapatla, Prakasham. This disricts are very highly susceptible to storm surge induced flood inundation as per IMD tcintensity report. XGB model (Fig. 12) shows that Balasore, Bhadrak, Jagatsinghpur, Palassa, Tekkali, Srikakulam, Vishakapatnam, Kakinada, Konaseema, West Godavari, Prakasham falls under very highly susceptible zone, Paradeep, Khurda, Puri, Chilika Lake, Ganjam, Gopalpur, Bapatla falls under High susceptible zone. LGB model (Fig. 13) shows that Balasore, Bhadrak, Paradeep, Jagatsinghpur, Chilika Lake, Ganjam, Palassa, Tekkali, Srikakulam, Vishakapatnam, Konaseema, West Godavari, Machilipatnam, Krishna, Bapatla falls under very high susceptible zone, Puri, Gopalpur falls under High susceptible zone Khurda, Kakinada, Prakasham falls under moderately vulnerable. RF model (Fig. 14) shows that Balasore, Bhadrak, Jagatsinghpur, Palassa, Tekkali, Srikakulam, Vishakapatnam, Bapatla, Prakasham falls under very highly susceptible zone, Paradeep, Konaseema, West Godavari, Machilipatnam, Krishna High susceptible zone, Khurdah, Puri, Chilika lake, Ganjam, Gopalpur, Kakinada have low susceptibility. The AUC was prepared using test data which showed higher percentage of accuracy of XG Boost i.e 98.04% followed by LightGBM 97.88% and Random Forest 97.77% refer Fig. (15). The ANN that was trained on the basis of historical data matches with IMD reported surge level. This model is having Mean square Error of 0.26, Mean Absolute Error of 0.18 and the classification accuracy is 96% which shows a good performance of the Neural Network. Figure (12) Storm Surge Susceptibility map from XGBoost Figure (13) Storm Surge Susceptibility map from LightGBM Figure (14) Storm Surge Susceptibility map from Random Forest Figure (15) AUC for RF, XGBoost & LightGBM In this study we focused particularly on handling storm surge induced flooding in Odisha and Andhra Coast as well our map also shows the flooding and compares the output of the three models. The state of art machine learning models which has been used to fulfil the first objective are LightGBM, XGBoost and Random Forest along with it the researcher derives AUC curve and our second objective aims to predict the Storm Surge in 2024 by training an ANN model from 2007–2023 and validate with the data of 2024. To prepare the susceptibility map we drew the insights from study conducted by (Alshayeb et al., 2024 ).Study conducted by (Halder & Bandyopadhyay, 2022 ) states that Baleshwar, Bhadrak, Cuttack, Dhenkanal, Ganjam, Jagatsinghapur, Jajapur, Kendrapara, Kendujhar, Khordha, Mayurbhanj, Nayagarh and Puri are some of the floods affected areas induced by cyclone. Our findings reveals that Udaipur, Gambharia, Talsari, Kharibil, Chaumukh, Chandamani, kasafal, Bagda, Gudu, Kusumuli, Khandia, Talapada, Maharudrapur, Kheranga, Padhuan, Kasia, Badahabelisahi, Bijaypatana, Karanapali, Dhanakuta, Rabindranagar, Sabitirisarai, Ekkakula, Hental Bana, Govindapur, Chinchiri, Sankuji, Pinchha Jungle, Paunsiapal, Hukitala, Saralikud, Paradip, Mahala, Jatadhartanda, Mira Sea Beach, Kalibedi, Nadiakhia, Dhanuhar Belari, Daluakani Beach, Tandahar Beach, Penthakata, Chilika Lake, Bateswar Beach, Gopalpur, Sonapur are the highly vulnerable areas which falls in Odisha. Battivanipelam, Hanuman Sagar Beach,Manchineeellapeta, Bandaruvanipeta, Vatsavalasa Beach, PD Paleem Beach, Kona Forest, Gaddipeta Beach, Amaravalli, Subbampeta, Kakinada, Uppalanka, Coringa Forest, Gongulalanka, Balusthippa, Nillarevu, Chirra Yanam, Vasalaitippa, Komaragiripatnam Beach, Turpupalem, Shankaraguptham Beach, Antervedi Pallipalem, Vemuladeevi Beach, KP Palem Beach, Mollaparru Beach, Pedapatnam, Machilipatnam, Hamsaladeevi Beach, Vadamudibba Mini Forest, Krishna wildlife Sanctuary, Sanjeevanagar, Bapatla, Katamvari Palem Beach, Kothapatnam, Chaipartha Beach, Vatturupallepalem, Penna Sangamam of Andhra Pradesh are highly susceptible to storm surge induced flooding. These areas are vulnerable because of several factors, particularly the funnel shaped coastline of northern Bay of Bengal especially Odisha (Das et al., 2016 ). Odisha and Andhra Pradesh are the most affected Cyclone prone states (Raghavan & Rajesh, 2003 ). Other major factors are low elevation especially areas like Chilika lake, Paradip coast is having low elevation. Mangrove Forest like Odisha’s Bhitarkanika and Coringa mangroves of Andhra Pradesh acts as a natural barrier but loss of natural barrier make the areas vulnerable to storm surge induced flooding (Kathiresan & Rajendran, 2005 ). Some of the studies have revealed in respect to broader regions like Gopalpur, Chilika and Paradip were found to be vulnerable using Coastal Vulnerability Index (Kumar et al., 2010 ). (Murali et al., 2018 )Conducted vulnerability studies using AHP based approach they found that the regions like Paradip and nearby areas like Kendrapara and Puri are vulnerable. Some studies focus on Chilika Lake including Penthakata, Bateswar, Satapada. Krishna, Guntur, Prakasham and East Godavari regions comes under risk of cyclonic hazard (Basheer Ahammed & Pandey, 2019 ). Kakinada and Machilipatnam and Krishna Delta demarcated as vulnerable area as per IMD reports. Three models were employed in the Study which concluded with very good MAE and MSE result. Whereas many studies have been conducted using Machine Learning and Deep learning algorithms like SVM, CNN, RM. A study was conducted on snowmelt flood susceptibility mapping CNN achieved remarkable performance on test data i.e AUC 0.97 followed by DNN securing 0.96 AUC (Yang et al. 2022 ). Alshayeb et al ( 2024 ) conducted a study on storm surge susceptibility mapping of Sagar Island where they achieved an outstanding performance on their finely tuned DNN, CNN and LightGBM model. They achieved accuracy for DNN, CNN and LightGBM at 97.75%, 97.5% and 97.5% respectively. In terms of accuracy assessment our model secured Mean Absolute Error for Random Forest at 0.28, XGBoost at 0.3 and LightGBM at 0.3. The AUC for the three models on the test data showed an unbelievable performance for XGBoost at 0.98, Random Forest at 0.97 and LightGBM at 0.97. The area has been calculated for each model in terms of four classes. For Balasore district the mainly the western coast and North western part particularly the areas that are lying in proximity to Subarnekha River are considered as very highly susceptible in Random Forest it covers approximately 39.14 Sq. Km area. LightGBM model that has been used for Balasore focuses on the same areas like Random Forest and classifies the susceptible area in two parts 15.39sq.km was considered as high and 0.18 sq. km area was considered as Very High. Same area is depicted in XGBoost it considers 58.93 Sq.Km as Highly susceptible followed by 1.38 Sq.Km as very high susceptibility. The area lies in proximity to Subarnarekha rivers gets drained due to storm surge induced flooding. The Southernmost part of Bhadrak and the entire coast except the coastline sharing boundary with Balasore comes under Very high to high susceptibility class. Around 33.46, 33.98 and 34.92 Sq.km area comes under high susceptibility in RF, LGB and XGB respectively. 7.53, 0.03,1.47 Sq. Km area was considered in the models as very highly susceptible in the in RF, LGB and XGB respectively, Bitharkanika National Park (IMD report) and the estuary of Baitarani river comes under susceptibility. Kendrapara district is having 26.82, 15.34 and 45.63 Sq.km area as highly susceptible and 5.34, 31.91 and 11.69 Sq. Km area is considered as Very Highly Susceptible in RF, LGB and XGB model respectively. Areas lying in proximity with Baitarani river, Brahmani, Mahanadi and the North eastern coastline falls under susceptibility class. In case of Jagatsinghpur 31.62, 29.84, 34.65 Sq.Km area is considered as highly susceptible and 5.41 Sq.km, 22.66 Sq.Km, and 13.83 Sq.km RF, LGB and XGB model respectively. Areas lying in proximity to Mahanadi river, and Jatadhar sea beach is very highly susceptible. For puri district 2.47, 1.68, 1.61 Sq. Km area is considered as very highly susceptible, presence of Chilika lake make the central coastal area susceptible. Ganjam district is having 4.43 sq.km, 2.28 sq.km, 0.23 Sq.km area that is considered as very highly susceptible in RF, LGB and XGB model respectively. In Andhra Pradesh Srikakulam district is having 52.31 is considered as very highly susceptible mainly the eastern coastline, Vizianagaram is having 13.56 sq.km as very highly susceptible towards the central coastline and Vishakapatnam is having 13.56 Sq.km area as very highly susceptible towards South east coastline. RF, LGB and XGB model shows that 9.50 Sq.km, 10.59 Sq.km, 11.32 Sq.km area is considered as very highly susceptibility. The South eastern coast of Kakinada district is considered as very highly susceptible 12.95 Sq.km, 14 Sq.km and 14.66 Sq. km area comes under that class in RF, LGB and XGB respectively. The coastal side of Konaseema district comes under very high susceptibility. 34.08, 29.83 Sq.km, 11.35 Sq.km area is very highly susceptible class in RF, LGB and XGB model respectively. The presence of Godavari river exacerbates storm surge induced floods. The South east side of Godavari district is very highly susceptible, 28.24 Sq.km, 35.21 Sq.km, and 27.18 Sq.km area comes under very susceptibility zone in RF, LGB and XGB model respectively. The southern facing side of Krishna district is very highly susceptible, 37.01 Sq.km, 35.98 Sq.km, 15.78 Sq.km area comes under very high susceptibility class in RF, LGB and XGB model respectively. The South eastern side of Baptala district is having 30.73 Sq.km, 37.17 Sq.km and 21.62 Sq.km area comes under very high susceptibility in RF, LGB and XGB model respectively. Baptala district having 30.73, 37.17, and 27.32 Sq.km comes under very high susceptibility class. Nellore district is having 0.02, 15.72, 4.09 Sq.km area comes under in Very High susceptibility class. Tirupati district doesn’t not comes under very high and high susceptibility class. To fulfil the second objective the researcher has used Multi-Layer Perceptron which is a part of ANN model. In terms of Mean Absolute Error, Mean Squared Error and the classification accuracy 0.18, 0.26, and 96% classification accuracy came for this neural network. The neural network was trained using various parameters inferred from the study conducted by (Lockwood et al., 2022 ).. The model was trained with X and Y. X was trained using 5 factors: Cyclone, Latitude, Longitude, Wind Speed (Vmax Km/s), Radius of the outer storm (Ro), Bearing in degrees and Translation Speed. Y was having Storm Surge in meter. The Cyclones forming in Bay of Bengal which is categorised into Cyclonic Storm and having landfall in the coast of Odisha and Andhra Pradesh were chosen for study. The dataset was obtained from 2007–2024 and it was split for training, testing and validation. For training and testing the dataset was chosen from 2007–2023 and for Validation we opt for 2024 dataset and it was verified with IMD reports. 6 hidden layers were there each consisting of 200 neuron, train test split ratio was 75:25. Historical hurricane events and physics-based hurricane surge modelling have showed that wind intensity is not a prime parameter that can influence surge levels (Irish et al., 2008 ; Peng et al., 2004 ). Increase or decrease in Translation speed may increase or decrease surge levels (Lockwood et al., 2022 ). Coastal configuration can influence surge level (Thomas et al., 2019 ). Lockwood et al. ( 2022 ) found that semi enclosed regions produce largest surge levels in slow paced hurricanes than the faster moving storms. Some of the Semi closed site in the eastern coastline of India are the largest brackish water lagoon is Chilika lake located in Odisha, a shallow brackish lagoon is Pulicat lake located in Andhra Pradesh–Tamil Nadu, Godavari Delta is a semi-enclosed area located in Andhra Pradesh, semi-enclosed site surrounded by mangrove vegetation is Pichavaram Mangroves, Tamil Nadu. The above mentioned sites may be at greater hazard. (Irish et al., 2008 ) stated that Numerical modelling like ADCIRC revealed that maximum wind speed plays an intensive role in surge generation on mild sloping regions. 4. Conclusions This research primary checks the efficiency of LightGBM, XGBoost and Random Forest and compares the result obtained from the algorithm within the domain of hydro-meteorological phenomena. The default optimization of the model was set to improve the accuracy of the result. As done in the previous study conducted by (Alshayeb et al., 2024 ) we use XAI and techniques like SHAP (SHapley Additive exPlanations) and permutation importance methods to check the data driven factors contributing to prepare susceptibility map. These methods the black box nature of deep learning models (Alshayeb et al., 2024 ). The influential factors can aid the decision makers to reduce the impact of surge induced flooding effectively. Our study contributed to develop storm surge susceptibility map for Odisha and Andhra Pradesh using Random Forest, XGBoost and LightGBM on the other hand we used an Artificial Neural Network based Multi-Layer Perceptron to predict storm surge for the year 2024. Our model used a wide range of parameters derived from previous studies to prepare the susceptibility map the parameters are Elevation, Rainfall, LULC, and Distance to River, Drainage Density, Normalized Difference Wetness Index (NDWI), Slope, Topographic Wetness Index (TWI), Normalized Difference Vegetation Index (NDVI) and Aspect. The model secures more than 90% accuracy on test data thus enhancing the model performance. To develop the neural network architecture for Prediction of Storm Surge Height in meters we consider Cyclone, Latitude, Longitude, Wind Speed (Vmax Km/s), Radius of the outer storm (Ro), Bearing in degrees, Translation Speed Storm Surge in meter. Our findings should encourage further research on employing machine learning, neural network algorithms and XAI techniques in hazard susceptibility modelling and prediction for other coastal districts of India. Thus, the integrated approaches is beneficial of academics, risk mitigation and policy-making. Due to limited ground truth data on storm surge heights, model validation relies on a combination of reported data and satellite-derived inundation extents, which may introduce uncertainties.Secondly, while the performance metrics were high, it is important to validate these models using real-world events for robust assessment. The future studies should focus on incorporation of real time data on availability from IMD, INCOIS and NCMRWF Declarations Author Contribution Subhra Mukherjee: Conceptualization, Methodology, Data curation, Software, Formal analysis, Writing – original draft. Ethics Approval and Consent to Participate: This study did not involve human participants, human data, or animals. Therefore, ethical approval and consent to participate were not required. Consent for Publication: Not applicable. Data Availability Data is available upon request No funding has been taken for this study References Abdel Hamid, H. T., Wenlong, W., & Qiaomin, L. (2020). Environmental sensitivity of flash flood hazard using geospatial techniques. Global Journal of Environmental Science and Management , 6 (1), 31–46. https://doi.org/10.22034/GJESM.2020.01.03 Alshayeb, M. J., Hang, H. T., Shohan, A. A. A., & Bindajam, A. A. (2024). Novel optimized deep learning algorithms and explainable artificial intelligence for storm surge susceptibility modeling and management in a flood-prone island. Natural Hazards , 120 (6), 5099–5128. https://doi.org/10.1007/s11069-024-06414-6 Azam, M. H., Samad, M. A., & Mahboob-Ul-Kabir. (2004). Effect of cyclone track and landfall angle on the magnitude of storm surges along the coast of Bangladesh in the Northern Bay of Bengal. Coastal Engineering Journal , 46 (3), 269–290. https://doi.org/10.1142/S0578563404001051 Basheer Ahammed, K. K., & Pandey, A. C. (2019). Coastal Social Vulnerability and Risk Analysis for Cyclone Hazard Along the Andhra Pradesh, East Coast of India. KN - Journal of Cartography and Geographic Information , 69 (4), 285–303. https://doi.org/10.1007/S42489-019-00029-9 Beven, K. J., Kirkby, M. J., Schofield, N., & Tagg, A. F. (1984). Testing a physically-based flood forecasting model (TOPMODEL) for three U.K. catchments. Journal of Hydrology , 69 (1–4), 119–143. https://doi.org/10.1016/0022-1694(84)90159-8 Bittermann, K., Rahmstorf, S., Perrette, M., & Vermeer, M. (2013). Predictability of twentieth century sea-level rise from past data. Environmental Research Letters , 8 (1). https://doi.org/10.1088/1748-9326/8/1/014013 Bruneau, N., Polton, J., Williams, J., & Holt, J. (2020). Estimation of global coastal sea level extremes using neural networks. Environmental Research Letters , 15 (7), 074030. https://doi.org/10.1088/1748-9326/AB89D6 Cahoon, D. R., Hensel, P. F., Spencer, T., Reed, D. J., McKee, K. L., & Saintilan, N. (2006). Coastal Wetland Vulnerability to Relative Sea-Level Rise: Wetland Elevation Trends and Process Controls. Wetlands and Natural Resource Management , 271–292. https://doi.org/10.1007/978-3-540-33187-2_12 Characteristics of the Hurricane Storm Surge - D. Lee Harris, United States. Weather Bureau Chavas, D. R., & Lin, N. (2016). A Model for the Complete Radial Structure of the Tropical Cyclone Wind Field. Part II: Wind Field Variability. Journal of the Atmospheric Sciences , 73 (8), 3093–3113. https://doi.org/10.1175/JAS-D-15-0185.1 Chen, K., Kuang, C., Wang, L., Chen, K., Han, X., & Fan, J. (2021). Storm Surge Prediction Based on Long Short-Term Memory Neural Network in the East China Sea. Applied Sciences 2022, Vol. 12, Page 181 , 12 (1), 181. https://doi.org/10.3390/APP12010181 Chilagane, N. A., Kashaigili, J. J., Mutayoba, E., Lyimo, P., Munishi, P., Tam, C., Burgess, N., Chilagane, N. A., Kashaigili, J. J., Mutayoba, E., Lyimo, P., Munishi, P., Tam, C., & Burgess, N. (2021). Impact of Land Use and Land Cover Changes on Surface Runoff and Sediment Yield in the Little Ruaha River Catchment. Open Journal of Modern Hydrology , 11 (3), 54–74. https://doi.org/10.4236/OJMH.2021.113004 Das, Y., Mohanty, U. C., & Jain, I. (2016). Development of tropical cyclone wind field for simulation of storm surge/sea surface height using numerical ocean model. Modeling Earth Systems and Environment , 2 (1), 1–22. https://doi.org/10.1007/S40808-015-0067-5/FIGURES/11 Emanuel, K. (2005). Increasing destructiveness of tropical cyclones over the past 30 years. Nature 2005 436:7051 , 436 (7051), 686–688. https://doi.org/10.1038/nature03906 Gregory, J. M., Griffies, S. M., Hughes, C. W., Lowe, J. A., Church, J. A., Fukimori, I., Gomez, N., Kopp, R. E., Landerer, F., Cozannet, G. Le, Ponte, R. M., Stammer, D., Tamisiea, M. E., & van de Wal, R. S. W. (2019). Concepts and Terminology for Sea Level: Mean, Variability and Change, Both Local and Global. Surveys in Geophysics , 40 (6), 1251–1289. https://doi.org/10.1007/S10712-019-09525-Z/FIGURES/3 Halder, B., & Bandyopadhyay, J. (2022). Monitoring the tropical cyclone ‘Yass’ and ‘Amphan’ affected flood inundation using Sentinel-1/2 data and Google Earth Engine. Modeling Earth Systems and Environment , 8 (3), 4317–4332. https://doi.org/10.1007/s40808-022-01359-w Horsburgh, K., De Vries, & H., E. (2011). Guide to storm surge forecasting . Irish, J. L., Resio, D. T., & Ratcliff, J. J. (2008). The Influence of Storm Size on Hurricane Surge. Journal of Physical Oceanography , 38 (9), 2003–2013. https://doi.org/10.1175/2008JPO3727.1 Kantamaneni, K., Rice, L., Yenneti, K., & Campos, L. C. (2020). Assessing the Vulnerability of Agriculture Systems to Climate Change in Coastal Areas: A Novel Index. Sustainability 2020, Vol. 12, Page 4771 , 12 (11), 4771. https://doi.org/10.3390/SU12114771 Kathiresan, K., & Rajendran, N. (2005). Coastal mangrove forests mitigated tsunami. Estuarine, Coastal and Shelf Science , 65 (3), 601–606. https://doi.org/10.1016/J.ECSS.2005.06.022 Kumar, T. S., Mahendra, R. S., Nayak, S., Radhakrishnan, K., & Sahu, K. C. (2010). Coastal vulnerability assessment for Orissa State, East Coast of India. Journal of Coastal Research , 26 (3), 523–534. https://doi.org/10.2112/09-1186.1 Lin, N., & Chavas, D. (2012). On hurricane parametric wind and applications in storm surge modeling. Journal of Geophysical Research: Atmospheres , 117 (D9), 9120. https://doi.org/10.1029/2011JD017126 Lockwood, J. W., Lin, N., Oppenheimer, M., & Lai, C. Y. (2022). Using Neural Networks to Predict Hurricane Storm Surge and to Assess the Sensitivity of Surge to Storm Characteristics. Journal of Geophysical Research: Atmospheres , 127 (24). https://doi.org/10.1029/2022JD037617 Murakami, H., Vecchi, G. A., & Underwood, S. (2017). Increasing frequency of extremely severe cyclonic storms over the Arabian Sea. Nature Climate Change . https://doi.org/10.1038/s41558-017-0008-6 Murali, R. M., Ankita, M., & Vethamony, P. (2018). Author Version of. Journal of Coastal Conservation , 22 (4), 799–819. Pan, J., Huang, L., Devlin, A. T., & Lin, H. (2018). Quantification of Typhoon-Induced Phytoplankton Blooms Using Satellite Multi-Sensor Data. Remote Sensing 2018, Vol. 10, Page 318 , 10 (2), 318. https://doi.org/10.3390/RS10020318 Peng, M., Xie, L., & Pietrafesa, L. J. (2004). A numerical study of storm surge and inundation in the Croatan-Albemarle- Pamlico Estuary System. Estuarine, Coastal and Shelf Science , 59 (1), 121–137. https://doi.org/10.1016/J.ECSS.2003.07.010 Raghavan, S., & Rajesh, S. (2003). Trends in tropical cyclone impact: A study in Andhra Pradesh, India. Bulletin of the American Meteorological Society , 84 (5), 635-644+549. https://doi.org/10.1175/BAMS-84-5-635 Rahman, M., Ningsheng, C., Mahmud, G. I., Islam, M. M., Pourghasemi, H. R., Ahmad, H., Habumugisha, J. M., Washakh, R. M. A., Alam, M., Liu, E., Han, Z., Ni, H., Shufeng, T., & Dewan, A. (2021). Flooding and its relationship with land cover change, population growth, and road density. Geoscience Frontiers , 12 (6), 101224. https://doi.org/10.1016/J.GSF.2021.101224 Ramos-Valle, A. N., Curchitser, E. N., Bruyère, C. L., & McOwen, S. (2021). Implementation of an Artificial Neural Network for Storm Surge Forecasting. Journal of Geophysical Research: Atmospheres , 126 (13), e2020JD033266. https://doi.org/10.1029/2020JD033266 Sahana, M., Rehman, S., Ahmed, R., & Sajjad, H. (2021). Assessing losses from multi-hazard coastal events using Poisson regression: empirical evidence from Sundarban Biosphere Reserve (SBR), India. Journal of Coastal Conservation , 25 (1). https://doi.org/10.1007/S11852-021-00804-9 Sarkhel, P., Biswas, D., & Swain, S. S. (2019). A review of cyclone and its impact on the coastal belts of Odisha . International Journal of Engineering Research & Technology (IJERT). Retrieved from http://www.ijert.org Shi, X., Chen, B., Liang, Y., Zhang, B., & Ye, T. (2021). Inundation simulation of different return periods of storm surge based on a numerical model and observational data. Stochastic Environmental Research and Risk Assessment , 35 (10), 2093–2103. https://doi.org/10.1007/S00477-021-02010-3 Tadesse, M., Wahl, T., & Cid, A. (2020). Data-Driven Modeling of Global Storm Surges. Frontiers in Marine Science , 7 , 512653. https://doi.org/10.3389/FMARS.2020.00260/BIBTEX Thomas, A., Dietrich, J., Asher, T., Bell, M., Blanton, B., Copeland, J., Cox, A., Dawson, C., Fleming, J., Luettich, R., Thomas, A., Dietrich, J., Asher, T., Bell, M., Blanton, B., Copeland, J., Cox, A., Dawson, C., Fleming, J., & Luettich, R. (2019). Influence of storm timing and forward speed on tides and storm surge during Hurricane Matthew. OcMod , 137 , 1–19. https://doi.org/10.1016/J.OCEMOD.2019.03.004 Tiggeloven, T., Couasnon, A., van Straaten, C., Muis, S., & Ward, P. J. (2021). Exploring deep learning capabilities for surge predictions in coastal areas. Scientific Reports 2021 11:1 , 11 (1), 1–15. https://doi.org/10.1038/s41598-021-96674-0 Varma, A. K., Jaiswal, N., Das, A., Kumar, M., Lele, N. V., Tripathy, R., Maity, S., Pandya, M., Bhattacharya, B., Mandal, A. K., Jishad, M., Seemanth, M., Sahay, A., Ganguly, D., Bhowmick, S. A., Sarangi, R. K., Agarwal, N., Raman, M., Sharma, R., Desai, N. M. (2023). A pathway for multi-stage cyclone-induced hazard tracking—case study for Yaas. Natural Hazards , 117 (1), 1035–1067. https://doi.org/10.1007/s11069-023-05893-3 Wang, J., Gao, W., Xu, S., & Yu, L. (2012). Evaluation of the combined risk of sea level rise, land subsidence, and storm surges on the coastal areas of Shanghai, China. Climatic Change , 115 (3–4), 537–558. https://doi.org/10.1007/S10584-012-0468-7/TABLES/9 Xianwu, S., Bingrui, C., Jufei, Q., Xing, K., & Tao, Y. (2021). Simulation of inundation caused by typhoon-induced probable maximum storm surge based on numerical modeling and observational data. Stochastic Environmental Research and Risk Assessment , 35 (11), 2273–2286. https://doi.org/10.1007/S00477-021-02034-9 Xiao-ting, F., Ying, L., Ai-min, L., Long-sheng, L., Xiao-ting, F., Ying, L., Ai-min, L., & Long-sheng, L. (2020). Statistical and Comparative Analysis of Tropical Cyclone Activity over the Arabian Sea and Bay of Bengal (1977-2018). Journal of Tropical Meteorology, 2020, Vol. 26, Issue 4, Pages: 441-452 , 26 (4), 441–452. https://doi.org/10.46267/J.1006-8775.2020.038 National Cyclone Risk Mitigation Project. Cyclones & their impact in India . https://ncrmp.gov.in/cyclones-their-impact-in-india/ Rao, A. D., Chittibabu, P., Murty, T. S., Dube, S. K., & Mohanty, U. C. (2007). Vulnerability from storm surges and cyclone wind fields on the coast of Andhra Pradesh, India. Natural Hazards , 41(3), 429–446. https://doi.org/10.1007/s11069-006-9041-z IBAN, M.C., BILGILIOGLU, S.S. Snow avalanche susceptibility mapping using novel tree-based machine learning algorithms (XGBoost, NGBoost, and LightGBM) with eXplainable Artificial Intelligence (XAI) approach. Stoch Environ Res Risk Assess 37, 2243–2270 (2023). https://doi.org/10.1007/s00477-023-02392-6 Yang, R., Zheng, G., Hu, P., Liu, Y., Xu, W., & Bao, A. (2022). Snowmelt flood susceptibility assessment in Kunlun Mountains based on the Swin Transformer deep learning method . Remote Sensing, 14(24), 6360. https://doi.org/10.3390/rs14246360 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 18 May, 2026 Reviewers agreed at journal 18 May, 2026 Reviewers invited by journal 18 May, 2026 Editor assigned by journal 30 Apr, 2026 Submission checks completed at journal 30 Apr, 2026 First submitted to journal 23 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9505599","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":629673519,"identity":"b72b8f89-442d-489e-b6c2-f3904bedb9fe","order_by":0,"name":"Subhra Mukherjee","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA60lEQVRIiWNgGAWjYFACHjY48wBDBZBkZm4gUgsbSMsZkBZGErQwMLaBWAS0yM/IPfa4ooIhj1++gfFw4bzaaP52oJYfFdtwajG4kZdueOYMQ7Ek0ILDM7cdz51xmLGBsefMbdxaJHLMJBvbGBI3HANq4d12LLcBqIWZsQ23FvkZUC37wVrmHMudT0gLww2YLWwgLQ01uRsIaTE48y5NsuGMROKMY4kNh3mOHcjdCNRyEJ9f5Ntzj0k2VNgk9jcfPvyZp6Yud975wwcf/KjA4zAIkGCARsdhMPcAIfXIoI4UxaNgFIyCUTBCAAB5MFiHL5dCawAAAABJRU5ErkJggg==","orcid":"","institution":"Central University of Karnataka","correspondingAuthor":true,"prefix":"","firstName":"Subhra","middleName":"","lastName":"Mukherjee","suffix":""}],"badges":[],"createdAt":"2026-04-23 10:38:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9505599/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9505599/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108006684,"identity":"33eb84e9-d2dc-4192-8913-7f3467ba89ea","added_by":"auto","created_at":"2026-04-28 12:56:27","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":377235,"visible":true,"origin":"","legend":"\u003cp\u003eGeographical overview of the study area representing the coastal districts of Odisha and Andhra Pradesh\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/eaa25fd58c84917ef091661a.jpeg"},{"id":107858480,"identity":"13d4e49b-feb6-4196-bf0a-1a36acba5dd0","added_by":"auto","created_at":"2026-04-27 04:40:03","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":271558,"visible":true,"origin":"","legend":"\u003cp\u003eSchematic diagram of the methodology formulation\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/09ff623ebde73cfba85c749c.jpeg"},{"id":107858478,"identity":"ba36ec4a-a809-433a-9095-20b1f50c3555","added_by":"auto","created_at":"2026-04-27 04:40:03","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":411055,"visible":true,"origin":"","legend":"\u003cp\u003eContributing factors for storm surge; (a)ASPECT, (b) DISTANCE TO RIVER, (c) DRAINAGE DENSITY, (d) ELEVATION, (e) LULC, (f) NDWI, (g)RAINFALL,(h) SLOPE, (i) TWI\u0026amp; (j) NDVI\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/3383e8312141a8a8d30fc957.jpeg"},{"id":107858487,"identity":"5d5e8d46-30ee-40df-87d7-9f6c96274bd5","added_by":"auto","created_at":"2026-04-27 04:40:04","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":85163,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP analysis for Random Forest\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/6e0a3c6ceba4d81e330f9e0e.png"},{"id":107858485,"identity":"8c1fff17-2838-4c51-b58e-59e4e95f9067","added_by":"auto","created_at":"2026-04-27 04:40:04","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":86649,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP analysis for XGBoost\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/e991bfbe132c10473d3954b9.png"},{"id":107858486,"identity":"4183b53f-f018-42d7-a9fa-877b34423706","added_by":"auto","created_at":"2026-04-27 04:40:04","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":85523,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP analysis for LightGBM\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/30d5bb74a80b2e855ad5485f.png"},{"id":107858479,"identity":"d730f793-f454-426c-a7fb-02b3d6e23147","added_by":"auto","created_at":"2026-04-27 04:40:03","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":37948,"visible":true,"origin":"","legend":"\u003cp\u003ePermutation importance for Random Forest\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/0d451471edbbafb127ddf5c9.png"},{"id":107858491,"identity":"9b8859e0-fb73-453a-854c-b71d30d9b17d","added_by":"auto","created_at":"2026-04-27 04:40:05","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":36157,"visible":true,"origin":"","legend":"\u003cp\u003ePermutation importance for XGBoost\u003c/p\u003e","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/d8a2f2767da71a364b5ac9e8.png"},{"id":107858489,"identity":"7854b2b8-8cdb-4826-a69c-2c159e6dc99f","added_by":"auto","created_at":"2026-04-27 04:40:05","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":35902,"visible":true,"origin":"","legend":"\u003cp\u003ePermutation importance for LightGBM\u003c/p\u003e","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/b04f693c76590f279e74ce06.png"},{"id":107858492,"identity":"25e4c9ad-58f5-4692-82b4-923511a261b0","added_by":"auto","created_at":"2026-04-27 04:40:05","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":36967,"visible":true,"origin":"","legend":"\u003cp\u003ePredicted output by ANN\u003c/p\u003e","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/e7a0c99838ea3fe7e2c623fe.png"},{"id":107858482,"identity":"dc0aaade-80ab-4fef-aa3e-b4f0be6aff8b","added_by":"auto","created_at":"2026-04-27 04:40:03","extension":"jpeg","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":240298,"visible":true,"origin":"","legend":"\u003cp\u003eSimulation of Storm Surge using IDW\u003c/p\u003e","description":"","filename":"floatimage11.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/b62d8b30c2b1def1a4f0b517.jpeg"},{"id":107858490,"identity":"40942e99-93a0-4c8b-b2fc-2125e86ca2f3","added_by":"auto","created_at":"2026-04-27 04:40:05","extension":"jpeg","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":283181,"visible":true,"origin":"","legend":"\u003cp\u003eStorm Surge Susceptibility map from XGBoost\u003c/p\u003e","description":"","filename":"floatimage12.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/622b2533dad4765557dfa8be.jpeg"},{"id":107858484,"identity":"ba17a028-028b-45cd-887a-f321e7eba657","added_by":"auto","created_at":"2026-04-27 04:40:03","extension":"jpeg","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":285597,"visible":true,"origin":"","legend":"\u003cp\u003eStorm Surge Susceptibility map from LightGBM\u003c/p\u003e","description":"","filename":"floatimage13.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/f59cd634f514c5933e6de353.jpeg"},{"id":107858488,"identity":"0bbeb8ac-764c-4145-94cb-f83a2c8f934a","added_by":"auto","created_at":"2026-04-27 04:40:05","extension":"jpeg","order_by":14,"title":"Figure 14","display":"","copyAsset":false,"role":"figure","size":282391,"visible":true,"origin":"","legend":"\u003cp\u003eStorm Surge Susceptibility map from Random Forest\u003c/p\u003e","description":"","filename":"floatimage14.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/ba093fe4c8d53382889484de.jpeg"},{"id":107870341,"identity":"0e144b26-0d4e-44e2-883a-d67398c8e94a","added_by":"auto","created_at":"2026-04-27 07:39:26","extension":"png","order_by":15,"title":"Figure 15","display":"","copyAsset":false,"role":"figure","size":43992,"visible":true,"origin":"","legend":"\u003cp\u003eAUC for RF, XGBoost \u0026amp; LightGBM\u003c/p\u003e","description":"","filename":"floatimage15.png","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/75d83b5b9885c808d1135417.png"},{"id":108008657,"identity":"13f7901e-ebc6-4faa-a08c-1d496e03631a","added_by":"auto","created_at":"2026-04-28 13:07:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3077338,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9505599/v1/b37bbc82-379b-4edb-92e0-3828a27cdf5c.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Storm Surge Susceptibility Mapping and Prediction of Tropical Cyclones using Machine Learning along the Eastern Coastline of India","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eTropical cyclone are among the most disastrous meteorological phenomenon, causing devastation in the coastal areas, mainly due to the ferocious wind, heavy rainfall, lightning, thunderstorm and, most critically, the storm surge (Varma et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).. Storm surge is the unusual upliftment of the sea surface with respect to the predicted astronomical tide influenced by changes in atmospheric pressure or by powerful winds pushing water towards the coast (Gregory et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The magnitude of storm surge in a shallow coastal region depends on coastline geometry, bathymetry, wind speed and size (Lockwood et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The Indian Peninsula bounded by north Indian Ocean is a hotspot region for the genesis of tropical cyclone (Varma et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Around 2% of the India\u0026rsquo;s GDP and 12% of the revenue of Government of India is spent every year for aftermath of the cyclones hitting the coastal belt of India (National Cyclone Risk Mitigation Project, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ncrmp.gov.in/cyclones-their-impact-in-india/\u003c/span\u003e\u003cspan address=\"https://ncrmp.gov.in/cyclones-their-impact-in-india/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Primarily, the eastern coastline of India, especially Odisha and Andhra Pradesh coast is severely affected by the cyclone-induced flood causing massive destruction and loss of life (Sarkhel et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Basheer Ahammed \u0026amp; Pandey, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2019\u003c/span\u003e;Varma et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). There is a bimodal type of distribution of tropical cyclone, mainly forming during pre-monsoon (April to June) and post monsoon (September to December) seasons (Varma et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Formation, intensification, and movement of tropical cyclone are influenced by both seasonal and intraseasonal fluctuation in temperature, moisture content in the atmosphere and low-level potential vorticity in the ocean and atmosphere (Xiao-ting et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Ocean hosts the breeding ground for the genesis of cyclone therefore understanding the atmospheric and oceanic phenomenon becomes crucial to determine the formation and intensification of cyclone (Emanuel, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). Strong force of tropical cyclone tends to result in two type of dynamic reaction that is upwelling and turbulent mixing which results in lower sea surface temperature (Pan et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eStorm surge accompanied with land subsidence exacerbate inland flooding, salt water intrusion and erosion of shorelines (Wang et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Xianwu et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In the past fifty years, storm surge has claimed lives of the people worldwide rather than by combined effect of earthquakes, tornadoes, freshwater flood and lightning (Emanuel, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Sahana et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Sarkhel et al. (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) states that nearly one third proportion of the total population of the nation inhabited in the coastal area are at the risk of cyclones, storms and heavy rainfall. The harsh winds accompanied with the cyclone destroys infrastructure, houses especially causing loss to marginal section of the society (Sarkhel et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Odisha is considered as extremely vulnerable to cyclone compared to other coastal regions of the India. The coastal districts of Odisha are vulnerable for around 35% of all cyclone crossed Indian Coast (Sarkhel et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e. In 1737, Odisha was hit by a super cyclone, followed by another led to the death of around 75000 lives (Sarkhel et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The vulnerability atlas of India has categorised Balasore, Bhadrak, Kendrapara, Jagatsinghpur, Puri and Ganjam into high risk zones in Odisha (Sarkhel et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Disaster response force is often hindered by mobility and infrastructure facilities (Sarkhel et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The impact of climate change will increase the levels of vulnerability (Sarkhel et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Andhra Pradesh coastal belt comprises of mostly low elevated topography thus making it highly sensitive to coastal flooding and saltwater intrusion (Basheer Ahammed \u0026amp; Pandey, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Three large rivers of Andhra Pradesh make Andhra Pradesh susceptible to storm surge penetration namely Godavari, Krishna and Pennar rivers (Rao et.al,2006). There might be loss of residential and agricultural land due to the predicted sea level rise and increase in temperatures and there will be an upsurge in salt concentration and decline in water quality especially in agricultural field by the ravages of time (Kantamaneni et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAdvancement in computational power of machine learning techniques have led to developments in data driven modelling of storm surge (Bruneau et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Ramos-Valle et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The results derived from earlier studies suggests that machine learning algorithm can model storm surge levels with high accuracy compared to physics based algorithms (Lockwood et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). In low latitude regions one may derive large number of errors after implementing machine learning algorithm which is particularly linked to tracks of hurricane as the dataset available for training and testing are limited (Bruneau et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Tadesse et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Tiggeloven et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The surge level increases as a cyclone approaches towards the coast, and surge levels decreases with the increase in the hurricane speed (Lockwood et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). There is a projected estimate that by 2100 the sea level will rise in the range of 0.26\u0026ndash;0.98 m (Bittermann et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). In contemporary studies Machine Larning and Deep Learning stands as a novel approach (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Deep Learning is a subfield of Machine Learning serves as important architect in flood modelling and susceptibility mapping (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). It can handle large complex geospatial dataset but the condition is that the proper factors should be prepared after Multicollinearity analysis (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Thus, it can manifest the spatio-temporal pattern in a susceptibility map (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003ePresent research aims to develop a storm surge susceptibility map to identify areas highly vulnerable to storm surge events along the Odisha-Andhra Pradesh coastline. Additionally, it attempts to predict storm surge triggered by tropical cyclones along the Odisha-Andhra Pradesh for the year 2024. Such mapping and prediction will be carried out using several machine learning approaches to test the potential and efficacy of ML in storm surge susceptibility analysis and prediction modelling.\u003c/p\u003e"},{"header":"2. Study area, Data used, and Methodology","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Study Area\u003c/h2\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe coastal districts of Odisha and Andhra Pradesh have been chosen as the study region extending from 21.49\u0026deg;N to 13.63\u0026deg;N and 86.93\u0026deg;E to 80.02\u0026deg;E with total area of 84,765.98 km\u003csup\u003e2\u003c/sup\u003e (Fig.\u0026nbsp;1). Among eighteen coastal districts Balasore, Bhadrak, Kendrapara, Jagatsinghpur, Puri and Ganjam districts belong to Odisha while, Srikakulam, Vizianagaram, Visakhapatnam, Anakapalli, Kakinada, Dr. B.R. Ambedkar Konaseema, West Godavari, Krishna, Bapatla, Prakasam, Sri Potti Sriramulu Nellore and Tirupati belong to Andhra Pradesh... Present study area is strategically important for trade and commerce, fisheries, marine economy, and biodiversity hotspots.\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;(1) Geographical overview of the study area representing the coastal districts of Odisha and Andhra Pradesh\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Data used\u003c/h2\u003e \u003cp\u003eDetails of the data used in the present study are presented in Table\u0026nbsp;1. To solve the first objective we obtain data like Aspect, Distance to the River, Drainage Density, Elevation, LULC, NDVI, NDWI, Rainfall, Slope, TWI refer Fig.\u0026nbsp;3 \u0026amp; Cyclone tracks. Every layer was obtained. For finding out the elevation dataset we chose Shuttle Radar Topographic Mission (SRTM) 1 Arc-Second Global elevation data has been used to derive elevation, slope and aspect in Google Earth Engine Platform. Furthermore, Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) \u0026ndash; Daily data has been used to for rainfall, due to it\u0026rsquo;s high frequency. High-resolution multispectral imagerySentinel-2 Level-1C (Top-of-Atmosphere Reflectance) dataset provided by Copernicus has been obtained for calculating NDWI and NDVI representing water abundance and vegetation density. Global land cover dataset has been provided by European Space Agency. From the HydroSHEDS (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.hydrosheds.org/\u003c/span\u003e\u003cspan address=\"https://www.hydrosheds.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) has been obtained to calculate Distance to the river and drainage density. To obtain flood inundation we used Sentinel-1 SAR data, NDWI generated from LANDSAT 5 TM and LANDSAT 7 ETM+. The cyclone tracks were obtained from IMD best track (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://rsmcnewdelhi.imd.gov.in/report.php?internal_menu=MzM=\u003c/span\u003e\u003cspan address=\"https://rsmcnewdelhi.imd.gov.in/report.php?internal_menu=MzM=\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) filtering out tracks from 1995\u0026ndash;2024. To fulfil the second objective we obtained IBTrACS (International Best Track Archive for Climate Stewardship) dataset (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncei.noaa.gov/products/international-best-track-archive\u003c/span\u003e\u003cspan address=\"https://www.ncei.noaa.gov/products/international-best-track-archive\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) from where we have taken important factor like latitude, longitude, Radius of the outer storm, wind speed, Bearing and translation speed. These factors were trained with Storm Surge levels from the period spanning from 1995\u0026ndash;2023.\u003c/p\u003e \u003cp\u003eTable. (1) Details of the data used in the present study\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Taba\" border=\"1\"\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSl. No\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProvider\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eProduct Name\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpatial Resolution\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTemporal Resolution\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003eSection \u003cspan refid=\"Sec1\" class=\"InternalRef\"\u003e1\u003c/span\u003e: Data used for susceptibility mapping\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUSGS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eShuttle Radar Topography Mission (SRTM) Global 1 arc-second\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e30 meters\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCHIRPS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePrecipitation (rainfall)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5.5 km\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDaily\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCopernicus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSentinel-2 Level-1C (Top-of-Atmosphere Reflectance)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10 meters\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5 days\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eESA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGlobal land cover dataset\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10 meters\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHydroSHEDS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFree-Flowing Rivers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCopernicus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSentinel-1 Synthetic Aperture Radar\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10 meters\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e12 days\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUSGS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLANDSAT 5 Thematic Mapper\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e30 meters\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e16 days\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUSGS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLandsat 7 Enhanced Thematic Mapper Plus (ETM+)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e30 meters\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e16 days\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c4\" namest=\"c1\"\u003e \u003cp\u003eSection \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e: \u003cb\u003eData used for prediction modelling\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIMD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBest Track (1995\u0026ndash;2024)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNOAA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eIBTrACS (International Best Track Archive for Climate Stewardship)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Storm Surge Inventory\u003c/h2\u003e \u003cp\u003eStorm Surge inventory dataset useful for model training, testing and validation has been used for generating storm susceptibility map and for prediction of storm surge. The dataset is therefore. The inventory was formed using a comprehensive approach, that includes IMD best track data forming into Cyclonic Storm from 1995\u0026ndash;2024 (Table\u0026nbsp;2) that made landfall in Odisha \u0026amp; Andhra Coast. INCOIS predicted surge location, height and IMD reported surge levels for 16 years with location of landfall. Flood inundation information have been obtained for different cyclones using Sentinel 1 SAR Interferometric Wide Swath (IW) dataset\u003c/p\u003e \u003cp\u003eTable\u0026nbsp;(2): List of Cyclones considered for study\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabb\" border=\"1\"\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026minus;\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026minus;\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBasin\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCYCLONE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCategory\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eEstimated Central Pressure\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMaximum sustained wind\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePressure Drop\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eDate of Formation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eDate of Landfall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eYear\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNNAMED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e978\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e07-11-1995\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e09-11-1995\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1995\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNNAMED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e982\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e29\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e13-11-1998\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e15-11-1998\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1995\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNNAMED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e992\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e12-06-1996\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e16-06-1996\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1996\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNNAMED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e988\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e04-11-1996\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e06-11-1996\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1996\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNNAMED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eESCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e968\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e15-10-1999\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e17-10-1999\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1999\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNNAMED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSUCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e912\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e140\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e25-10-1999\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e29-10-1999\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1999\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNNAMED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e14-10-2001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e16-10-2001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNNAMED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e992\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e11-12-2003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e15-12-2003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2003\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePYARR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e990\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e17-09-2005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e19-09-2005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2005\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOGNI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e998\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e29-10-2006\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e30-10-2006\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2006\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eKHAIMUK\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e994\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e13-11-2008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e15-11-2008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2008\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLAILA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e986\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e17-05-2010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e20-05-2010\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2010\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHELEN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e990\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e19-11-2013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e22-11-2013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2013\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLEHAR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e980\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e26\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e23-11-2013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e28-11-2013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2013\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePHAILIN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eESCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e940\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e115\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e08-10-2013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e12-10-2013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2013\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHUDHUD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eESCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e950\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e07-10-2014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e12-10-2014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2014\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDAYE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e992\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e19-09-2018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e20-09-2018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2018\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePHETHAI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e992\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e13-12-2018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e17-12-2018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2018\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTITLI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e972\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e08-10-2018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e10-10-2018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2018\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFani\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eESCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e932\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e115\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e26-04-2019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e03-05-2019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2019\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGULAB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e992\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e24-09-2021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e26-09-2021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2021\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYAAS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e970\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e23-05-2021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e26-05-2021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2021\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eASANI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e982\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e07-05-2022\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e11-05-2022\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2022\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMICHAUNG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e986\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e01-12-2023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e05-12-2023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2023\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBay of Bengal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDANA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e986\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c7\"\u003e \u003cp\u003e22-10-2024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c8\"\u003e \u003cp\u003e24-10-2024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2024\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Methodology\u003c/h2\u003e \u003cp\u003eEntire methodology is divided in to two parts (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003e). First part deals with the dividing the entire study area into four zones (very high, high, moderate and low) based on its susceptibility to storm surge events. Numerous driving factors e.g. aspect, distance to the river, drainage density, elevation, LULC, NDVI, NDWI, rainfall, slope, TWI \u0026amp; Cyclone tracks from 1995\u0026ndash;2024 have been considered for deriving the rate of susceptibility. The Deep depression which had intensified into cyclonic storms were considered for this study. The tracks were laid over the layers in GEE environment and flooded areas were identified by using SAR data of Sentinel 1. The tropical cyclones which were formed on or before 2014 for that NDWI was generated using LANDSAT 7 ETM+. For the flood inundation studies prior to 1999, we used NDWI generated from LANDSAT 5 TM. 12,932 samples were taken for training and testing in ML models. The entire dataset was split into training and testing subset i.e 75:25 ratio.. While selecting the Machine Learning algorithm we carefully considered the potential of handling large complex geospatial datasets (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Specifically, we opted for LightGBM, XGBoost \u0026amp; Random Forest due to it\u0026rsquo;s effectiveness in handling large and complex geospatial datasets. This diverse selection was mainly aligned with our objective to systematically compare different output of the 18 coastal districts.\u003c/p\u003e \u003cp\u003eIn the second part, ibtracs dataset of NOAA has been used. We mainly considered the latitude, longitude, Radius of the outer storm (Vo), Translation speed, Bearing in degrees and wind speed in knot (Lockwood et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). A Multi-Layer Perceptron deep learning model was chosen for this task the model was trained for the period of 2007\u0026ndash;2023 with a training and testing subset ratio of 75:25due to its efficiency in capturing non-linear relationship in the data giving it a strong foundation for prediction of storm surge for the year 2024.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.3.1 Preparation of driving factors for storm surge susceptibility mapping\u003c/h2\u003e \u003cp\u003eOn the basis of meteorology, hydrology, topography and land use datasets, driving factors that have been selected influences the extent and severity of storm surges. Meteorological Parameters are selected to analyse the potential impact on coastal areas, Wind Speed (kt) is responsible to drive the surge along the coastline and it is directly proportional with surge levels, i.e higher the wind speed higher is the surge level(Characteristics of the Hurricane Storm Surge - D. Lee Harris, United States. Weather Bureau.) Radius of the outer storm is considered as the distance from cyclone center to the outermost closed isobar(Chavas \u0026amp; Lin, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). The translation speed is generally influenced by steering wind of the environment governed by their motion (Chavas \u0026amp; Lin, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). The direction of the storm from one point to other leads to coastal water to displace leading to surge impact (Lin \u0026amp; Chavas, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Hydrological Parameters influence how surge water interacts with the coastal terrain. Heavy rainfall tends to exacerbates surge impacts by in increasing flooding along the coast (Lin \u0026amp; Chavas, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). The presence of Coastal rivers can result in amplification of storm surge flooding due to it\u0026rsquo;s backwater affect. The area which is having high drainage density may witness prolonged inundation. Topographical parameters are responsible to determine how surge waters inundates inland. Aspect signifies the direction of a slope face and the slope exposed onshore are vulnerable to wave effect and surge penetration. A recent data has revealed that wetland inundation is affected by storm surge(Cahoon et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). The slope angle is inversely related to occurrence of water stagnation (Cahoon et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). Topographic Wetness Index determines flood prone areas on the basis of accumulation of surface water potential (Beven et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e1984\u003c/span\u003e). LULC is the controlling factor for surface run off and infiltration (Chilagane et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).The forest area allows infiltration but the built up area does not allow infiltration (Abdel Hamid et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The NDVI and NDWI shows the interplay of water presence and vegetation, the decrease in vegetation cover and increase in water presence make an area susceptible to flooding. There is more or less influence on surge heights due to change in track orientation along western and eastern coast respectively (Azam et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2004\u003c/span\u003e)\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e2.3.2. Data pre-processing for Flood inundation studies\u003c/h2\u003e \u003cp\u003eIn the cloud computing platform different time data were given for each cyclone to derive the flood inundation. The pre cyclonic data and post cyclonic data is an important key to conduct this study. Sentinel-1 SAR images are chosen because it emits Microwave radiation of the Electromagnetic Spectrum which can penetrate cloud cover. Interferometric Wide Swath (IW) mode was used with VH Polarisation. The significance of VH polarisation to prefer it for this study is because it is sensitive to surface roughness, Less Affected by Double-Bounce Effects\u0026amp; capable of Better Vegetation Penetration. The backscatter coefficient of VH polarization is expressed as σVH, dB​=10log10​(σVH, linear​)(Halder \u0026amp; Bandyopadhyay, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The decrease in VH backscatter for pre and post flood image shows inundation. The spatial resolution this dataset is 10m with a swath of 250 Km. Global Surface water dataset has been considered as an important layer which removes the permanent water bodies like lakes and river. The values of VH polarization backscatter (σ) is converted into logarithmic scale (dB) by using σdB​=10log10​ (σlinear​) which helps to compute median backscatter values in segments. The perform the change detection studies the difference was found out between pre-flood and post-flood images mathematically it can be defined as ΔσdB​=σdB, pre​\u0026minus;σdB, post​. Refined Lee filter is used to remove the speckle noise effect which helps to preserve edges while it can remove noise. The terrain slope was calculated to exclude steep slopes where water does not accumulate by using Slope\u0026thinsp;=\u0026thinsp;tan\u0026minus;1(Δh/Δd​). The NDWI from LANDSAT 7 ETM\u0026thinsp;+\u0026thinsp;was generated by using Band 2 (Green) and Band 4 (NIR) and the NDWI from LANDSAT 5 TM was generated by using Band 3 (Green) and Band 4 (NIR). INCOIS predicted storm surge report was obtained and IMD reported surge event was referred. The cyclones which were forming into cyclonic storm and having landfall on Odisha and Andhra coast were considered for the study thus the track was considered as a layer for this study. For inundation studies training samples were taken in cloud computing platform and coastal area was considered as more vulnerable than the outskirts. The training samples were collected in terms of very high, high, moderate and low. SAR data was considered for those Cyclones which had landfall on or before 2014. NDWI of LANDSAT 7 ETM\u0026thinsp;+\u0026thinsp;was considered for those Cyclones having landfall on or before 2014. NDWI of LANDSAT 5 TM was considered for those Cyclones having landfall before 1999.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.3.3. Multicollinearity analysis\u003c/h2\u003e \u003cp\u003eFor generation of susceptibility map various predictor variables have been generated consisting of Aspect, Distance to the River, Drainage Density, Elevation, LULC, NDVI, NDWI, Rainfall, Slope, TWI \u0026amp; Cyclone tracks. LULC, Aspect, Cyclone tracks, Distance to the River, Drainage Density, NDVI, Slope and Rainfall has less Variance Inflation Factor (VIF) in terms of Multicollinearity, values thus make it as important predictor variables for storm surge susceptibility mapping (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).Elevation, NDWI, and TWI has showed higher VIF values exacerbating VIF values. However, their contribution can\u0026rsquo;t be entirely neglected on susceptibility due to their scientific importance and considering the extent of the study area. Elevation, slope, curvature, aspect, SPI, TWI, STI, LULC, rainfall, river width, TWI, and soil types were having less Multi-collinearity while doing flood inundation studies (Rahman et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003e2.3.4.. Machine Learning Models\u003c/h2\u003e \u003cp\u003eRandom Forest is a statistical learning model and an ensemble learning approach (Rahman et al., 2019), it is the aggregate of decision trees used for regression and classification founded by Brieman, 2001. Thus, it laid the blocks to build multiple decision trees and combines the result to prevent overfitting and to improve training accuracy. 100 decision trees were trained on a randomly generated subset of the data. Random split was kept at 42. The resultant prediction from multiple decision trees are averaged in case of regression analysis or voted on for generation of classification. There were some missing values in the training dataset to handle the same Simple imputation of mean were used which ensures that model is not affected by missing data. Ten features were termed as X and were considered for the study which has been derived from previous studies having less VIF value in terms of multicollinearity analysis. Y is considered as the target variable that is actually the training samples. Feature scaling was performed to assign equal weightage to all predictor variables. The train test data ratio was split into 75 and 25 respectively. Thus, the training process includes Bootstrapping, Splitting and Aggregation. The test data is considered as the unseen data that has been actually derived from X and the trained model is used to make predictions on X_test.\u003c/p\u003e \u003cp\u003eThe mathematical formula Random Forest Regression Formula after Breiman, L. (2001):\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\widehat{y}=\\frac{1}{B}\\sum\\:_{b=1}^{B}{f}_{b}\\left(X\\right)\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(1\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe Mean Absolute error is a key to accuracy assessment measure of the model which finds out the average absolute difference between the actual and predicted values. Lower MAE is a sign of better performance of the model.\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:\\text{MAE}=\\frac{1}{N}\\sum\\:_{i=1}^{N}{\\left({y}_{i}-\\stackrel{ˉ}{y}\\right)}^{2}\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:..\\dots\\:..\\left(2\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe MAE of Random Forest algorithm is 0.2 which shows that on an average the predictions deviate from the actual values by 0.2 units.\u003c/p\u003e \u003cp\u003eLightGBM uses the mechanism of Gradient-boosted decision trees, having an efficient and scalable techniques for storm surge susceptibility modelling. This algorithm uses histogram based learning method, this gives it a foundation to handle large no. of datasets more efficiently .The model takes into account the parameters selected after multicollinearity analysis from the previous studies. The loss function is optimized and each tree in the ensemble learning is constructed (Iban and Bilgilioglu \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The algorithm gets to learn by correcting errors made by the earlier trees, beside this it can handle missing values which makes it quite adaptable to handle a large no of datasets. LightGBM can be an attractive choice for storm surge susceptibility mapping due to it\u0026rsquo;s effective traits of interpretability with fast training and prediction (Alshayeb et.al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The model builds 100 decision trees where the learning rate has been set to 0.1 for more accurate learning, to prevent it\u0026rsquo;s overfitting max depth was chosen as 5. Random state ensured the reproducibility of results.\u003c/p\u003e \u003cp\u003eThe mathematical formula for this gradient boosting algorithm after Friedman, J. H. (2001) which minimize loss using gradient boosting is.\u003cdiv id=\"Equc\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equc\" name=\"EquationSource\"\u003e\n$$\\:{h}_{m}\\left(x\\right)=-\\sum\\:_{i=1}^{n}\\frac{\\partial\\:L\\left({y}_{i},{F}_{m-1}\\left({x}_{i}\\right)\\right)}{\\partial\\:{F}_{m-1}\\left({x}_{i}\\right)}\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(3\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe trees that are add sequentially by the mathematical formula given below:\u003cdiv id=\"Equd\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equd\" name=\"EquationSource\"\u003e\n$$\\:{F}_{m}\\left(x\\right)={F}_{m-1}\\left(x\\right)+\\eta\\:\\cdot\\:{h}_{m}\\left(x\\right)\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:..\\left(4\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe Mean Absolute error is a key to accuracy assessment measure of the model which finds out the average absolute difference between the actual and predicted values. Lower MAE is a sign of better performance of the model.\u003cdiv id=\"Eque\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Eque\" name=\"EquationSource\"\u003e\n$$\\:\\text{MAE}=\\frac{1}{N}\\sum\\:_{i=1}^{N}{\\left({y}_{i}-\\stackrel{ˉ}{y}\\right)}^{2}\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(5\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe MAE of Random Forest algorithm is 0.3 which shows that on an average the predictions deviate from the actual values by 0.3 units.\u003c/p\u003e \u003cp\u003eXGBoost is a gradient boosting framework which incorporates regularization technique. The basic idea is to grow trees gradually by fitting the residuals that the preceding tree predicted by the earlier trees by performing feature splits. The final prediction result is obtained by adding the scores of each leaf node throughout the trees. This algorithm employs first and second derivatives to perform Taylor expansion for it\u0026rsquo;s loss function. Thus model accuracy with it\u0026rsquo;s complexity can both be managed at a time (Wang et.al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). The regressor model have created 100 trees sequentially the learning rate has been set to 0.1 to control how much each tree contribute to final prediction. The max depth is 5 to control overfitting.\u003c/p\u003e \u003cp\u003eEach tree corrects the error from previous iterations after Friedman, J. H. (2001 by the following formula:\u003cdiv id=\"Equf\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equf\" name=\"EquationSource\"\u003e\n$$\\:{F}_{m}\\left(x\\right)={F}_{m-1}\\left(x\\right)+\\eta\\:\\cdot\\:{h}_{m}\\left(x\\right)\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(6\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eFor Mean Squared Error (MSE), the loss function is:\u003cdiv id=\"Equg\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equg\" name=\"EquationSource\"\u003e\n$$\\:L\\left({y}_{i},{\\widehat{y}}_{i}\\right)=\\frac{1}{2}{\\left({y}_{i}-{\\widehat{y}}_{i}\\right)}^{2}\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(7\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe Mean Absolute error is a key to accuracy assessment measure of the model which finds out the average absolute difference between the actual and predicted values. Lower MAE is a sign of better performance of the model.\u003cdiv id=\"Equh\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equh\" name=\"EquationSource\"\u003e\n$$\\:\\text{MAE}=\\frac{1}{N}\\sum\\:_{i=1}^{N}{\\left({y}_{i}-\\stackrel{ˉ}{y}\\right)}^{2}\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(8\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe MAE of XGBoost algorithm is 0.2 which shows that on an average the predictions deviate from the actual values by 0.2 units.\u003c/p\u003e \u003cp\u003eMLP stands for Multilayer Perceptron is one type of Neural network or more specifically Artificial Neural Network. This algorithm doesn\u0026rsquo;t make any prior assumption about the data distribution. Extremely Non linear functions can be trained to appropriately generalise them. Thus, when developing numerical models and also when choosing between other statistical learning algorithm MLP becomes an attractive choice for it\u0026rsquo;s application in atmospheric science. The weights and output signals are determined by the addition of the input of the nodes and it is changed by a straightforward non-linear function. As from it\u0026rsquo;s name signifies it may consist multiple hidden layers with an output layer and finally an output layer. one node is interconnected to every other node in the former and next layer. During training the weights in the network are adjusted until the desired input\u0026mdash;output mapping is obtained. The code is trained with input features obtained from ibtracs dataset of NOAA, the variables that are selected are latitude, longitude, wind speed (Vmax), radius of maximum wind (Ro), bearing, and translation speed. For the given study the model was trained with 6 hidden layers with 200 neurons with ReLU Activation function. For each hidden layer output the mathematical formula McCulloch, W. S., \u0026amp; Pitts, W. (1943 is\u003cdiv id=\"Equi\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equi\" name=\"EquationSource\"\u003e\n$$\\:{z}_{j}=\\sum\\:_{i=1}^{n}{w}_{ji}{x}_{i}+{b}_{j\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(9\\right)}$$\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equj\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equj\" name=\"EquationSource\"\u003e\n$$\\:{h}_{j}=f\\left({z}_{j}\\right)\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(10\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eTo compute output layer the mathematical formula is:\u003cdiv id=\"Equk\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equk\" name=\"EquationSource\"\u003e\n$$\\:\\widehat{y}=\\sum\\:_{j=1}^{H}{v}_{j}{h}_{j}+c\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(11\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eMLP updates weights using gradient descent and backpropagation by using the given formula after Rumelhart, Hinton, and Williams (1986)\u003cdiv id=\"Equl\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equl\" name=\"EquationSource\"\u003e\n$$\\:{w}_{ji}\\left(t+1\\right)={w}_{ji}\\left(t\\right)-\\eta\\:\\frac{\\partial\\:L}{\\partial\\:{w}_{ji}}\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(12\\right)$$\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equm\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equm\" name=\"EquationSource\"\u003e\n$$\\:{v}_{j}\\left(t+1\\right)={v}_{j}\\left(t\\right)-\\eta\\:\\frac{\\partial\\:L}{\\partial\\:{v}_{j}}\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(13\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"3. Results and Discussion","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Implementation of Explainable AI (XAI) techniques\u003c/h2\u003e \u003cp\u003eThe XAI techniques like SHAP analysis and Permutation importance serves as a Sophisticated approach for the explanation of DNN model(Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). SHAP analysis is a game theoretical methodology which analysis the prediction made by Machine learning algorithms (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).Our study was subjected to SHAP analysis to derive feature importance it can help the researcher to get an idea how each factor is contributing to model prediction. SHAP value gives us an insight how each predictor variable influences the model output. Red represents high value whereas blue represents low value, the vertical spread shows that density and how varied the impact is. The Fig.\u0026nbsp;(4),(5),(6) shows Elevation and Rainfall are the top contributors, since low lying and high rainfall areas are surge prone. In Random Forest, XGBoost and LightGBM High elevation and rainfall increases the susceptibility. In terms of LULC, probably the agricultural areas or built up areas are at higher risk, the areas closer to river are susceptible beside this higher NDWI is a potential indicator of flooding, these has been depicted in SHAP analysis of Random Forest in Fig.\u0026nbsp;(4). In case of XGBoost, Increase in NDWI, closer proximity to Distance to River exacerbates surge induced flooding Fig.\u0026nbsp;(5). LULC and NDVI has moderate impact and TWI, Aspect has minor impact. In case of LightGBM Fig.\u0026nbsp;(6), higher NDVI shows less susceptibility, Gentle slope can contribute more, TWI has less importance compared to RF, Drainage Density and Aspect is having almost negligible influence. This analysis serves as a significant measure to frame storm management strategies, this helps to sort and prioritize the factors that influence the surge events(Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003ePermutation importance measures global feature importance by checking the model performance when a single feature\u0026rsquo;s value is shuffled. This XAI technique was used for RF, XGBoost and LightGBM models Fig.\u0026nbsp;(7),(8) \u0026amp; (9). Rainfall and Elevation emerged as the most influential factors in the three models thus influencing the dynamics of storm surge, LULC served as moderately important in all models indicating that the wetlands, vegetation cover area and wetlands have a meaningful impact on water retention. NDWI, and Distance to River have lower importance. NDVI, Slope, Topographic Wetness Index (TWI), Drainage Density, and Aspect consistently showed low to negligible importance scores.\u003c/p\u003e \u003cp\u003eCollectively SHAP and Permutation importance gives an insight and assign importance to different factors in storm surge induced flood management(Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Storm Surge Prediction\u003c/h2\u003e \u003cp\u003eThe ANN model was trained with a limited 25 datasets refer Fig.\u0026nbsp;(10). The dataset was selected for the year spanning from 2007\u0026ndash;2023. It included storm surge corresponding location collected from IMD for every selected cyclonic storms and surge. The input layer for the neural network includes six parameters with regard to the cyclone\u0026rsquo;s physical properties(Ramos-Valle et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2021\u003c/span\u003eb). The dataset was obtained from ibtracs of NOAA and was sorted, the depression which formed into Cyclone Storms and those data of the cyclones were trained whose Storm surge data from IMD were available for studies. Input parameters consists of latitude and longitude of Storm Surge locations, maximum wind speed (Vmax), Radius of the outer storm (Ro), bearing, translation speed. The Vmax was considered as a metric to assess cyclone intensity rather than it\u0026rsquo;s central pressure because it barely contributes 15% to the magnitude of storm surge (Horsburgh et al., 2011). Ro is derived from a lognormal distribution therefore it is uncorrelated with other parameters. Ro was considered as the metric to derive cyclone size rather than Radius of maximum wind speed (Rmax) as it has positive correlation with Vmax (Chavas \u0026amp; Lin, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). While training an ANN model we provide both the input and output to get relationship between them. The model consists of multiple interconnected neurons that are capable to extract linear non-linear features, the ANN architecture is built with an input layer and a no. of intermediate layers that is denoted as hidden layers. In a hidden layer the relationship between an input and an output neuron can be denoted after (Ramos-Valle et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) as:\u003cdiv id=\"Equn\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equn\" name=\"EquationSource\"\u003e\n$$\\:AF={b}_{i}+\\sum\\:_{j=1}^{\\text{n}}{x}_{{\\text{in}}_{j}}{w}_{j\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(14\\right)}$$\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equo\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equo\" name=\"EquationSource\"\u003e\n$$\\:{x}_{\\text{out}}=f\\left(AF\\right)\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(15\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eXinj and xouti are the input and output data of neuron I, AF plays the input to the activation function, the total no. of neurons is denoted by n and it is from the previous hidden layer connected to neuron I in the current hidden layer(Lockwood et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The hidden layer consists of ReLU activation function that has been derived from Scikit Learn package. There are total 6 hidden layer consists of 200 neurons per layer, maximum training iterations, was limited to 1,000. The dataset given as an input were divided into 75% for training and 25 % for testing. The Mea Square Error and Mean Absolute Error serves as an evaluation metrics and it came as 0.1 and 0.2 respectively. X was having 5 parameters Latitude (\u0026deg;)', 'Longitude (\u0026deg;)', 'Vmax (kt)', 'Ro (km)', 'Bearing (\u0026deg;)', 'Translation Speed (km/h) and Y was Storm Surge in meters. The output was Predicted Storm surge for every classes. During Validation we used the same X parameters for the year 2024 and the output was Storm Surge in meters. The predicted report is a given in Table\u0026nbsp;(3)\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;(10) Predicted output by ANN\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePredicted Storm Surge Level by MLP Regressor (ANN)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"14\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c12\" colnum=\"12\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c13\" colnum=\"13\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c14\" colnum=\"14\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBasin\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLandfall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCyclone\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCategory\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eHour\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eLatitude\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eLongitude\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eVmax (kt)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eRo\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c10\"\u003e \u003cp\u003eTranslation Speed\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c11\"\u003e \u003cp\u003eBearing\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c12\"\u003e \u003cp\u003eNeural Network Predicted Storm Surge (m)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c13\"\u003e \u003cp\u003eINCOIS Forecasted\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c14\"\u003e \u003cp\u003eIMD Reported\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBoB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSagar islands and Khepupara\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRemal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1800\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e21.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e89.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e139\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e1.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e1 to 2 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e1 to 2 m\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBoB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBhitarkanika and Dhamara\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDana\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSCS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e20.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e87.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e231\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e335\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c12\"\u003e \u003cp\u003e2.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e1 to 2 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e1 to 2 m\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Storm Surge Height Estimation Using Inverse Distance Weighting (IDW) Interpolation\u003c/h2\u003e \u003cp\u003eThis section tries to generate a spatial distribution map of storm surge heights along the coasts of West Bengal, Odisha, and Andhra Pradesh as showed in Fig.\u0026nbsp;(11), the dataset obtained from IMD tcintensity. The dataset was obtained for the year 1995\u0026ndash;2024 which formed into Cyclonic Storm, it includes storm surge heights and their corresponding locations. The tropical cyclones that were considered include Amphan, Yaas, Fani, Gulab, Titli, Phailin, Hudhud, Laila, Aila, Sidr, and other significant storms affecting the region. Point data was generated in ArcGIS Pro environment where X and Y was Latitude and longitude respectively and Z was Storm Surge level in meter. A continuous spatial representation was generated using Storm Surge heights along the coast using Inverse Distance Weighting (IDW) that stands as an interpolation method. Thus, the unknown values were estimated using the value based on nearby known points. Districts having close proximity with Bay of Bengal such as South 24 Parganas, Bhadrak, and Balasore in West Bengal and Odisha recorded higher surge levels due to major Super Cyclonic storm like Amphan. The Andhra Pradesh coast experienced moderate surge levels with Laila, Michaung, and Hudhud contributing to surge variations. The visual pattern closely aligns with historical storm surge data.\u003c/p\u003e \u003cp\u003eThe IDW formula after (Watson, 1992; Burrough \u0026amp; McDonnell, 1998 is given below\u003cdiv id=\"Equp\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equp\" name=\"EquationSource\"\u003e\n$$\\:Z\\left(s\\right)=\\frac{\\sum\\:_{i=1}^{N}{w}_{i}{Z}_{i}}{\\sum\\:_{i=1}^{N}{w}_{i}}\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:\\dots\\:.\\left(16\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;(11) Simulation of Storm Surge using IDW\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Accuracy Assessment and Validation\u003c/h2\u003e \u003cp\u003eThe accuracy assessment was done from the published reports of IMD \u0026amp; INCOIS. The locations which are susceptible to storm surge induced flood inundation were collected from IMD reports and laid on ArcGIS Pro environment. The areas are Balasore, Bhadrak, Paradeep, Jagatsinghpur, Khurda, Puri, Chilika Lake, Ganjam, Gopalpur, Palassa, Tekkali, Srikakulam, Vishakapatnam, Kakinada, Konaseema, West Godavari, Machilipatnam, Krishna, Bapatla, Prakasham. This disricts are very highly susceptible to storm surge induced flood inundation as per IMD tcintensity report. XGB model (Fig.\u0026nbsp;12) shows that Balasore, Bhadrak, Jagatsinghpur, Palassa, Tekkali, Srikakulam, Vishakapatnam, Kakinada, Konaseema, West Godavari, Prakasham falls under very highly susceptible zone, Paradeep, Khurda, Puri, Chilika Lake, Ganjam, Gopalpur, Bapatla falls under High susceptible zone. LGB model (Fig.\u0026nbsp;13) shows that Balasore, Bhadrak, Paradeep, Jagatsinghpur, Chilika Lake, Ganjam, Palassa, Tekkali, Srikakulam, Vishakapatnam, Konaseema, West Godavari, Machilipatnam, Krishna, Bapatla falls under very high susceptible zone, Puri, Gopalpur falls under High susceptible zone Khurda, Kakinada, Prakasham falls under moderately vulnerable. RF model (Fig.\u0026nbsp;14) shows that Balasore, Bhadrak, Jagatsinghpur, Palassa, Tekkali, Srikakulam, Vishakapatnam, Bapatla, Prakasham falls under very highly susceptible zone, Paradeep, Konaseema, West Godavari, Machilipatnam, Krishna High susceptible zone, Khurdah, Puri, Chilika lake, Ganjam, Gopalpur, Kakinada have low susceptibility. The AUC was prepared using test data which showed higher percentage of accuracy of XG Boost i.e 98.04% followed by LightGBM 97.88% and Random Forest 97.77% refer Fig.\u0026nbsp;(15). The ANN that was trained on the basis of historical data matches with IMD reported surge level. This model is having Mean square Error of 0.26, Mean Absolute Error of 0.18 and the classification accuracy is 96% which shows a good performance of the Neural Network.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;(12) Storm Surge Susceptibility map from XGBoost\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;(13) Storm Surge Susceptibility map from LightGBM\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;(14) Storm Surge Susceptibility map from Random Forest\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;(15) AUC for RF, XGBoost \u0026amp; LightGBM\u003c/p\u003e \u003cp\u003eIn this study we focused particularly on handling storm surge induced flooding in Odisha and Andhra Coast as well our map also shows the flooding and compares the output of the three models. The state of art machine learning models which has been used to fulfil the first objective are LightGBM, XGBoost and Random Forest along with it the researcher derives AUC curve and our second objective aims to predict the Storm Surge in 2024 by training an ANN model from 2007\u0026ndash;2023 and validate with the data of 2024. To prepare the susceptibility map we drew the insights from study conducted by (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).Study conducted by (Halder \u0026amp; Bandyopadhyay, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) states that Baleshwar, Bhadrak, Cuttack, Dhenkanal, Ganjam, Jagatsinghapur, Jajapur, Kendrapara, Kendujhar, Khordha, Mayurbhanj, Nayagarh and Puri are some of the floods affected areas induced by cyclone. Our findings reveals that Udaipur, Gambharia, Talsari, Kharibil, Chaumukh, Chandamani, kasafal, Bagda, Gudu, Kusumuli, Khandia, Talapada, Maharudrapur, Kheranga, Padhuan, Kasia, Badahabelisahi, Bijaypatana, Karanapali, Dhanakuta, Rabindranagar, Sabitirisarai, Ekkakula, Hental Bana, Govindapur, Chinchiri, Sankuji, Pinchha Jungle, Paunsiapal, Hukitala, Saralikud, Paradip, Mahala, Jatadhartanda, Mira Sea Beach, Kalibedi, Nadiakhia, Dhanuhar Belari, Daluakani Beach, Tandahar Beach, Penthakata, Chilika Lake, Bateswar Beach, Gopalpur, Sonapur are the highly vulnerable areas which falls in Odisha. Battivanipelam, Hanuman Sagar Beach,Manchineeellapeta, Bandaruvanipeta, Vatsavalasa Beach, PD Paleem Beach, Kona Forest, Gaddipeta Beach, Amaravalli, Subbampeta, Kakinada, Uppalanka, Coringa Forest, Gongulalanka, Balusthippa, Nillarevu, Chirra Yanam, Vasalaitippa, Komaragiripatnam Beach, Turpupalem, Shankaraguptham Beach, Antervedi Pallipalem, Vemuladeevi Beach, KP Palem Beach, Mollaparru Beach, Pedapatnam, Machilipatnam, Hamsaladeevi Beach, Vadamudibba Mini Forest, Krishna wildlife Sanctuary, Sanjeevanagar, Bapatla, Katamvari Palem Beach, Kothapatnam, Chaipartha Beach, Vatturupallepalem, Penna Sangamam of Andhra Pradesh are highly susceptible to storm surge induced flooding. These areas are vulnerable because of several factors, particularly the funnel shaped coastline of northern Bay of Bengal especially Odisha (Das et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Odisha and Andhra Pradesh are the most affected Cyclone prone states (Raghavan \u0026amp; Rajesh, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). Other major factors are low elevation especially areas like Chilika lake, Paradip coast is having low elevation. Mangrove Forest like Odisha\u0026rsquo;s Bhitarkanika and Coringa mangroves of Andhra Pradesh acts as a natural barrier but loss of natural barrier make the areas vulnerable to storm surge induced flooding (Kathiresan \u0026amp; Rajendran, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). Some of the studies have revealed in respect to broader regions like Gopalpur, Chilika and Paradip were found to be vulnerable using Coastal Vulnerability Index (Kumar et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). (Murali et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2018\u003c/span\u003e)Conducted vulnerability studies using AHP based approach they found that the regions like Paradip and nearby areas like Kendrapara and Puri are vulnerable. Some studies focus on Chilika Lake including Penthakata, Bateswar, Satapada. Krishna, Guntur, Prakasham and East Godavari regions comes under risk of cyclonic hazard (Basheer Ahammed \u0026amp; Pandey, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Kakinada and Machilipatnam and Krishna Delta demarcated as vulnerable area as per IMD reports. Three models were employed in the Study which concluded with very good MAE and MSE result. Whereas many studies have been conducted using Machine Learning and Deep learning algorithms like SVM, CNN, RM. A study was conducted on snowmelt flood susceptibility mapping CNN achieved remarkable performance on test data i.e AUC 0.97 followed by DNN securing 0.96 AUC (Yang et al. \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Alshayeb et al (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) conducted a study on storm surge susceptibility mapping of Sagar Island where they achieved an outstanding performance on their finely tuned DNN, CNN and LightGBM model. They achieved accuracy for DNN, CNN and LightGBM at 97.75%, 97.5% and 97.5% respectively. In terms of accuracy assessment our model secured Mean Absolute Error for Random Forest at 0.28, XGBoost at 0.3 and LightGBM at 0.3. The AUC for the three models on the test data showed an unbelievable performance for XGBoost at 0.98, Random Forest at 0.97 and LightGBM at 0.97. The area has been calculated for each model in terms of four classes. For Balasore district the mainly the western coast and North western part particularly the areas that are lying in proximity to Subarnekha River are considered as very highly susceptible in Random Forest it covers approximately 39.14 Sq. Km area. LightGBM model that has been used for Balasore focuses on the same areas like Random Forest and classifies the susceptible area in two parts 15.39sq.km was considered as high and 0.18 sq. km area was considered as Very High. Same area is depicted in XGBoost it considers 58.93 Sq.Km as Highly susceptible followed by 1.38 Sq.Km as very high susceptibility. The area lies in proximity to Subarnarekha rivers gets drained due to storm surge induced flooding. The Southernmost part of Bhadrak and the entire coast except the coastline sharing boundary with Balasore comes under Very high to high susceptibility class. Around 33.46, 33.98 and 34.92 Sq.km area comes under high susceptibility in RF, LGB and XGB respectively. 7.53, 0.03,1.47 Sq. Km area was considered in the models as very highly susceptible in the in RF, LGB and XGB respectively, Bitharkanika National Park (IMD report) and the estuary of Baitarani river comes under susceptibility. Kendrapara district is having 26.82, 15.34 and 45.63 Sq.km area as highly susceptible and 5.34, 31.91 and 11.69 Sq. Km area is considered as Very Highly Susceptible in RF, LGB and XGB model respectively. Areas lying in proximity with Baitarani river, Brahmani, Mahanadi and the North eastern coastline falls under susceptibility class. In case of Jagatsinghpur 31.62, 29.84, 34.65 Sq.Km area is considered as highly susceptible and 5.41 Sq.km, 22.66 Sq.Km, and 13.83 Sq.km RF, LGB and XGB model respectively. Areas lying in proximity to Mahanadi river, and Jatadhar sea beach is very highly susceptible. For puri district 2.47, 1.68, 1.61 Sq. Km area is considered as very highly susceptible, presence of Chilika lake make the central coastal area susceptible. Ganjam district is having 4.43 sq.km, 2.28 sq.km, 0.23 Sq.km area that is considered as very highly susceptible in RF, LGB and XGB model respectively. In Andhra Pradesh Srikakulam district is having 52.31 is considered as very highly susceptible mainly the eastern coastline, Vizianagaram is having 13.56 sq.km as very highly susceptible towards the central coastline and Vishakapatnam is having 13.56 Sq.km area as very highly susceptible towards South east coastline. RF, LGB and XGB model shows that 9.50 Sq.km, 10.59 Sq.km, 11.32 Sq.km area is considered as very highly susceptibility. The South eastern coast of Kakinada district is considered as very highly susceptible 12.95 Sq.km, 14 Sq.km and 14.66 Sq. km area comes under that class in RF, LGB and XGB respectively. The coastal side of Konaseema district comes under very high susceptibility. 34.08, 29.83 Sq.km, 11.35 Sq.km area is very highly susceptible class in RF, LGB and XGB model respectively. The presence of Godavari river exacerbates storm surge induced floods. The South east side of Godavari district is very highly susceptible, 28.24 Sq.km, 35.21 Sq.km, and 27.18 Sq.km area comes under very susceptibility zone in RF, LGB and XGB model respectively. The southern facing side of Krishna district is very highly susceptible, 37.01 Sq.km, 35.98 Sq.km, 15.78 Sq.km area comes under very high susceptibility class in RF, LGB and XGB model respectively. The South eastern side of Baptala district is having 30.73 Sq.km, 37.17 Sq.km and 21.62 Sq.km area comes under very high susceptibility in RF, LGB and XGB model respectively. Baptala district having 30.73, 37.17, and 27.32 Sq.km comes under very high susceptibility class. Nellore district is having 0.02, 15.72, 4.09 Sq.km area comes under in Very High susceptibility class. Tirupati district doesn\u0026rsquo;t not comes under very high and high susceptibility class.\u003c/p\u003e \u003cp\u003eTo fulfil the second objective the researcher has used Multi-Layer Perceptron which is a part of ANN model. In terms of Mean Absolute Error, Mean Squared Error and the classification accuracy 0.18, 0.26, and 96% classification accuracy came for this neural network. The neural network was trained using various parameters inferred from the study conducted by (Lockwood et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).. The model was trained with X and Y. X was trained using 5 factors: Cyclone, Latitude, Longitude, Wind Speed (Vmax Km/s), Radius of the outer storm (Ro), Bearing in degrees and Translation Speed. Y was having Storm Surge in meter. The Cyclones forming in Bay of Bengal which is categorised into Cyclonic Storm and having landfall in the coast of Odisha and Andhra Pradesh were chosen for study. The dataset was obtained from 2007\u0026ndash;2024 and it was split for training, testing and validation. For training and testing the dataset was chosen from 2007\u0026ndash;2023 and for Validation we opt for 2024 dataset and it was verified with IMD reports. 6 hidden layers were there each consisting of 200 neuron, train test split ratio was 75:25. Historical hurricane events and physics-based hurricane surge modelling have showed that wind intensity is not a prime parameter that can influence surge levels (Irish et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Peng et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2004\u003c/span\u003e). Increase or decrease in Translation speed may increase or decrease surge levels (Lockwood et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Coastal configuration can influence surge level (Thomas et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Lockwood et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) found that semi enclosed regions produce largest surge levels in slow paced hurricanes than the faster moving storms. Some of the Semi closed site in the eastern coastline of India are the largest brackish water lagoon is Chilika lake located in Odisha, a shallow brackish lagoon is Pulicat lake located in Andhra Pradesh\u0026ndash;Tamil Nadu, Godavari Delta is a semi-enclosed area located in Andhra Pradesh, semi-enclosed site surrounded by mangrove vegetation is Pichavaram Mangroves, Tamil Nadu. The above mentioned sites may be at greater hazard. (Irish et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2008\u003c/span\u003e) stated that Numerical modelling like ADCIRC revealed that maximum wind speed plays an intensive role in surge generation on mild sloping regions.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Conclusions","content":"\u003cp\u003eThis research primary checks the efficiency of LightGBM, XGBoost and Random Forest and compares the result obtained from the algorithm within the domain of hydro-meteorological phenomena. The default optimization of the model was set to improve the accuracy of the result. As done in the previous study conducted by (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) we use XAI and techniques like SHAP (SHapley Additive exPlanations) and permutation importance methods to check the data driven factors contributing to prepare susceptibility map. These methods the black box nature of deep learning models (Alshayeb et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The influential factors can aid the decision makers to reduce the impact of surge induced flooding effectively. Our study contributed to develop storm surge susceptibility map for Odisha and Andhra Pradesh using Random Forest, XGBoost and LightGBM on the other hand we used an Artificial Neural Network based Multi-Layer Perceptron to predict storm surge for the year 2024. Our model used a wide range of parameters derived from previous studies to prepare the susceptibility map the parameters are Elevation, Rainfall, LULC, and Distance to River, Drainage Density, Normalized Difference Wetness Index (NDWI), Slope, Topographic Wetness Index (TWI), Normalized Difference Vegetation Index (NDVI) and Aspect. The model secures more than 90% accuracy on test data thus enhancing the model performance. To develop the neural network architecture for Prediction of Storm Surge Height in meters we consider Cyclone, Latitude, Longitude, Wind Speed (Vmax Km/s), Radius of the outer storm (Ro), Bearing in degrees, Translation Speed Storm Surge in meter.\u003c/p\u003e \u003cp\u003eOur findings should encourage further research on employing machine learning, neural network algorithms and XAI techniques in hazard susceptibility modelling and prediction for other coastal districts of India. Thus, the integrated approaches is beneficial of academics, risk mitigation and policy-making.\u003c/p\u003e \u003cp\u003eDue to limited ground truth data on storm surge heights, model validation relies on a combination of reported data and satellite-derived inundation extents, which may introduce uncertainties.Secondly, while the performance metrics were high, it is important to validate these models using real-world events for robust assessment.\u003c/p\u003e \u003cp\u003eThe future studies should focus on incorporation of real time data on availability from IMD, INCOIS and NCMRWF\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eAuthor Contribution\u003c/p\u003e\n\u003cp\u003eSubhra Mukherjee: Conceptualization, Methodology, Data curation, Software, Formal analysis, Writing \u0026ndash; original draft.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Approval and Consent to Participate:\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;This study did not involve human participants, human data, or animals. Therefore, ethical approval and consent to participate were not required.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for Publication:\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Not applicable.\u003c/p\u003e\n\u003cp\u003eData Availability\u003c/p\u003e\n\u003cp\u003eData is available upon request\u003c/p\u003e\n\u003cp\u003eNo funding has been taken for this study\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAbdel Hamid, H. T., Wenlong, W., \u0026amp; Qiaomin, L. (2020). Environmental sensitivity of flash flood hazard using geospatial techniques. \u003cem\u003eGlobal Journal of Environmental Science and Management\u003c/em\u003e, \u003cem\u003e6\u003c/em\u003e(1), 31\u0026ndash;46. https://doi.org/10.22034/GJESM.2020.01.03\u003c/li\u003e\n\u003cli\u003eAlshayeb, M. J., Hang, H. T., Shohan, A. A. A., \u0026amp; Bindajam, A. A. (2024). Novel optimized deep learning algorithms and explainable artificial intelligence for storm surge susceptibility modeling and management in a flood-prone island. \u003cem\u003eNatural Hazards\u003c/em\u003e, \u003cem\u003e120\u003c/em\u003e(6), 5099\u0026ndash;5128. https://doi.org/10.1007/s11069-024-06414-6\u003c/li\u003e\n\u003cli\u003eAzam, M. H., Samad, M. A., \u0026amp; Mahboob-Ul-Kabir. (2004). Effect of cyclone track and landfall angle on the magnitude of storm surges along the coast of Bangladesh in the Northern Bay of Bengal. \u003cem\u003eCoastal Engineering Journal\u003c/em\u003e, \u003cem\u003e46\u003c/em\u003e(3), 269\u0026ndash;290. https://doi.org/10.1142/S0578563404001051\u003c/li\u003e\n\u003cli\u003eBasheer Ahammed, K. K., \u0026amp; Pandey, A. C. (2019). Coastal Social Vulnerability and Risk Analysis for Cyclone Hazard Along the Andhra Pradesh, East Coast of India. \u003cem\u003eKN - Journal of Cartography and Geographic Information\u003c/em\u003e, \u003cem\u003e69\u003c/em\u003e(4), 285\u0026ndash;303. https://doi.org/10.1007/S42489-019-00029-9\u003c/li\u003e\n\u003cli\u003eBeven, K. J., Kirkby, M. J., Schofield, N., \u0026amp; Tagg, A. F. (1984). Testing a physically-based flood forecasting model (TOPMODEL) for three U.K. catchments. \u003cem\u003eJournal of Hydrology\u003c/em\u003e, \u003cem\u003e69\u003c/em\u003e(1\u0026ndash;4), 119\u0026ndash;143. https://doi.org/10.1016/0022-1694(84)90159-8\u003c/li\u003e\n\u003cli\u003eBittermann, K., Rahmstorf, S., Perrette, M., \u0026amp; Vermeer, M. (2013). Predictability of twentieth century sea-level rise from past data. \u003cem\u003eEnvironmental Research Letters\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e(1). https://doi.org/10.1088/1748-9326/8/1/014013\u003c/li\u003e\n\u003cli\u003eBruneau, N., Polton, J., Williams, J., \u0026amp; Holt, J. (2020). Estimation of global coastal sea level extremes using neural networks. \u003cem\u003eEnvironmental Research Letters\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(7), 074030. https://doi.org/10.1088/1748-9326/AB89D6\u003c/li\u003e\n\u003cli\u003eCahoon, D. R., Hensel, P. F., Spencer, T., Reed, D. J., McKee, K. L., \u0026amp; Saintilan, N. (2006). Coastal Wetland Vulnerability to Relative Sea-Level Rise: Wetland Elevation Trends and Process Controls. \u003cem\u003eWetlands and Natural Resource Management\u003c/em\u003e, 271\u0026ndash;292. https://doi.org/10.1007/978-3-540-33187-2_12\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eCharacteristics of the Hurricane Storm Surge - D. Lee Harris, United States. Weather Bureau\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eChavas, D. R., \u0026amp; Lin, N. (2016). A Model for the Complete Radial Structure of the Tropical Cyclone Wind Field. Part II: Wind Field Variability. \u003cem\u003eJournal of the Atmospheric Sciences\u003c/em\u003e, \u003cem\u003e73\u003c/em\u003e(8), 3093\u0026ndash;3113. https://doi.org/10.1175/JAS-D-15-0185.1\u003c/li\u003e\n\u003cli\u003eChen, K., Kuang, C., Wang, L., Chen, K., Han, X., \u0026amp; Fan, J. (2021). Storm Surge Prediction Based on Long Short-Term Memory Neural Network in the East China Sea. \u003cem\u003eApplied Sciences 2022, Vol. 12, Page 181\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(1), 181. https://doi.org/10.3390/APP12010181\u003c/li\u003e\n\u003cli\u003eChilagane, N. A., Kashaigili, J. J., Mutayoba, E., Lyimo, P., Munishi, P., Tam, C., Burgess, N., Chilagane, N. A., Kashaigili, J. J., Mutayoba, E., Lyimo, P., Munishi, P., Tam, C., \u0026amp; Burgess, N. (2021). Impact of Land Use and Land Cover Changes on Surface Runoff and Sediment Yield in the Little Ruaha River Catchment. \u003cem\u003eOpen Journal of Modern Hydrology\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(3), 54\u0026ndash;74. https://doi.org/10.4236/OJMH.2021.113004\u003c/li\u003e\n\u003cli\u003eDas, Y., Mohanty, U. C., \u0026amp; Jain, I. (2016). Development of tropical cyclone wind field for simulation of storm surge/sea surface height using numerical ocean model. \u003cem\u003eModeling Earth Systems and Environment\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(1), 1\u0026ndash;22. https://doi.org/10.1007/S40808-015-0067-5/FIGURES/11\u003c/li\u003e\n\u003cli\u003eEmanuel, K. (2005). Increasing destructiveness of tropical cyclones over the past 30\u0026thinsp;years. \u003cem\u003eNature 2005 436:7051\u003c/em\u003e, \u003cem\u003e436\u003c/em\u003e(7051), 686\u0026ndash;688. https://doi.org/10.1038/nature03906\u003c/li\u003e\n\u003cli\u003eGregory, J. M., Griffies, S. M., Hughes, C. W., Lowe, J. A., Church, J. A., Fukimori, I., Gomez, N., Kopp, R. E., Landerer, F., Cozannet, G. Le, Ponte, R. M., Stammer, D., Tamisiea, M. E., \u0026amp; van de Wal, R. S. W. (2019). Concepts and Terminology for Sea Level: Mean, Variability and Change, Both Local and Global. \u003cem\u003eSurveys in Geophysics\u003c/em\u003e, \u003cem\u003e40\u003c/em\u003e(6), 1251\u0026ndash;1289. https://doi.org/10.1007/S10712-019-09525-Z/FIGURES/3\u003c/li\u003e\n\u003cli\u003eHalder, B., \u0026amp; Bandyopadhyay, J. (2022). Monitoring the tropical cyclone \u0026lsquo;Yass\u0026rsquo; and \u0026lsquo;Amphan\u0026rsquo; affected flood inundation using Sentinel-1/2 data and Google Earth Engine. \u003cem\u003eModeling Earth Systems and Environment\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e(3), 4317\u0026ndash;4332. https://doi.org/10.1007/s40808-022-01359-w\u003c/li\u003e\n\u003cli\u003eHorsburgh, K., De Vries, \u0026amp; H., E. (2011). \u003cem\u003eGuide to storm surge forecasting\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eIrish, J. L., Resio, D. T., \u0026amp; Ratcliff, J. J. (2008). The Influence of Storm Size on Hurricane Surge. \u003cem\u003eJournal of Physical Oceanography\u003c/em\u003e, \u003cem\u003e38\u003c/em\u003e(9), 2003\u0026ndash;2013. https://doi.org/10.1175/2008JPO3727.1\u003c/li\u003e\n\u003cli\u003eKantamaneni, K., Rice, L., Yenneti, K., \u0026amp; Campos, L. C. (2020). Assessing the Vulnerability of Agriculture Systems to Climate Change in Coastal Areas: A Novel Index. \u003cem\u003eSustainability 2020, Vol. 12, Page 4771\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(11), 4771. https://doi.org/10.3390/SU12114771\u003c/li\u003e\n\u003cli\u003eKathiresan, K., \u0026amp; Rajendran, N. (2005). Coastal mangrove forests mitigated tsunami. \u003cem\u003eEstuarine, Coastal and Shelf Science\u003c/em\u003e, \u003cem\u003e65\u003c/em\u003e(3), 601\u0026ndash;606. https://doi.org/10.1016/J.ECSS.2005.06.022\u003c/li\u003e\n\u003cli\u003eKumar, T. S., Mahendra, R. S., Nayak, S., Radhakrishnan, K., \u0026amp; Sahu, K. C. (2010). Coastal vulnerability assessment for Orissa State, East Coast of India. \u003cem\u003eJournal of Coastal Research\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(3), 523\u0026ndash;534. https://doi.org/10.2112/09-1186.1\u003c/li\u003e\n\u003cli\u003eLin, N., \u0026amp; Chavas, D. (2012). On hurricane parametric wind and applications in storm surge modeling. \u003cem\u003eJournal of Geophysical Research: Atmospheres\u003c/em\u003e, \u003cem\u003e117\u003c/em\u003e(D9), 9120. https://doi.org/10.1029/2011JD017126\u003c/li\u003e\n\u003cli\u003eLockwood, J. W., Lin, N., Oppenheimer, M., \u0026amp; Lai, C. Y. (2022). Using Neural Networks to Predict Hurricane Storm Surge and to Assess the Sensitivity of Surge to Storm Characteristics. \u003cem\u003eJournal of Geophysical Research: Atmospheres\u003c/em\u003e, \u003cem\u003e127\u003c/em\u003e(24). https://doi.org/10.1029/2022JD037617\u003c/li\u003e\n\u003cli\u003eMurakami, H., Vecchi, G. A., \u0026amp; Underwood, S. (2017). Increasing frequency of extremely severe cyclonic storms over the Arabian Sea. \u003cem\u003eNature Climate Change\u003c/em\u003e. https://doi.org/10.1038/s41558-017-0008-6\u003c/li\u003e\n\u003cli\u003eMurali, R. M., Ankita, M., \u0026amp; Vethamony, P. (2018). Author Version of. \u003cem\u003eJournal of Coastal Conservation\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(4), 799\u0026ndash;819.\u003c/li\u003e\n\u003cli\u003ePan, J., Huang, L., Devlin, A. T., \u0026amp; Lin, H. (2018). Quantification of Typhoon-Induced Phytoplankton Blooms Using Satellite Multi-Sensor Data. \u003cem\u003eRemote Sensing 2018, Vol. 10, Page 318\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(2), 318. https://doi.org/10.3390/RS10020318\u003c/li\u003e\n\u003cli\u003ePeng, M., Xie, L., \u0026amp; Pietrafesa, L. J. (2004). A numerical study of storm surge and inundation in the Croatan-Albemarle- Pamlico Estuary System. \u003cem\u003eEstuarine, Coastal and Shelf Science\u003c/em\u003e, \u003cem\u003e59\u003c/em\u003e(1), 121\u0026ndash;137. https://doi.org/10.1016/J.ECSS.2003.07.010\u003c/li\u003e\n\u003cli\u003eRaghavan, S., \u0026amp; Rajesh, S. (2003). Trends in tropical cyclone impact: A study in Andhra Pradesh, India. \u003cem\u003eBulletin of the American Meteorological Society\u003c/em\u003e, \u003cem\u003e84\u003c/em\u003e(5), 635-644+549. https://doi.org/10.1175/BAMS-84-5-635\u003c/li\u003e\n\u003cli\u003eRahman, M., Ningsheng, C., Mahmud, G. I., Islam, M. M., Pourghasemi, H. R., Ahmad, H., Habumugisha, J. M., Washakh, R. M. A., Alam, M., Liu, E., Han, Z., Ni, H., Shufeng, T., \u0026amp; Dewan, A. (2021). Flooding and its relationship with land cover change, population growth, and road density. \u003cem\u003eGeoscience Frontiers\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(6), 101224. https://doi.org/10.1016/J.GSF.2021.101224\u003c/li\u003e\n\u003cli\u003eRamos-Valle, A. N., Curchitser, E. N., Bruy\u0026egrave;re, C. L., \u0026amp; McOwen, S. (2021). Implementation of an Artificial Neural Network for Storm Surge Forecasting. \u003cem\u003eJournal of Geophysical Research: Atmospheres\u003c/em\u003e, \u003cem\u003e126\u003c/em\u003e(13), e2020JD033266. https://doi.org/10.1029/2020JD033266\u003c/li\u003e\n\u003cli\u003eSahana, M., Rehman, S., Ahmed, R., \u0026amp; Sajjad, H. (2021). Assessing losses from multi-hazard coastal events using Poisson regression: empirical evidence from Sundarban Biosphere Reserve (SBR), India. \u003cem\u003eJournal of Coastal Conservation\u003c/em\u003e, \u003cem\u003e25\u003c/em\u003e(1). https://doi.org/10.1007/S11852-021-00804-9\u003c/li\u003e\n\u003cli\u003eSarkhel, P., Biswas, D., \u0026amp; Swain, S. S. (2019). \u003cem\u003eA review of cyclone and its impact on the coastal belts of Odisha\u003c/em\u003e. International Journal of Engineering Research \u0026amp; Technology (IJERT). Retrieved from http://www.ijert.org\u003c/li\u003e\n\u003cli\u003eShi, X., Chen, B., Liang, Y., Zhang, B., \u0026amp; Ye, T. (2021). Inundation simulation of different return periods of storm surge based on a numerical model and observational data. \u003cem\u003eStochastic Environmental Research and Risk Assessment\u003c/em\u003e, \u003cem\u003e35\u003c/em\u003e(10), 2093\u0026ndash;2103. https://doi.org/10.1007/S00477-021-02010-3\u003c/li\u003e\n\u003cli\u003eTadesse, M., Wahl, T., \u0026amp; Cid, A. (2020). Data-Driven Modeling of Global Storm Surges. \u003cem\u003eFrontiers in Marine Science\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e, 512653. https://doi.org/10.3389/FMARS.2020.00260/BIBTEX\u003c/li\u003e\n\u003cli\u003eThomas, A., Dietrich, J., Asher, T., Bell, M., Blanton, B., Copeland, J., Cox, A., Dawson, C., Fleming, J., Luettich, R., Thomas, A., Dietrich, J., Asher, T., Bell, M., Blanton, B., Copeland, J., Cox, A., Dawson, C., Fleming, J., \u0026amp; Luettich, R. (2019). Influence of storm timing and forward speed on tides and storm surge during Hurricane Matthew. \u003cem\u003eOcMod\u003c/em\u003e, \u003cem\u003e137\u003c/em\u003e, 1\u0026ndash;19. https://doi.org/10.1016/J.OCEMOD.2019.03.004\u003c/li\u003e\n\u003cli\u003eTiggeloven, T., Couasnon, A., van Straaten, C., Muis, S., \u0026amp; Ward, P. J. (2021). Exploring deep learning capabilities for surge predictions in coastal areas. \u003cem\u003eScientific Reports 2021 11:1\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(1), 1\u0026ndash;15. https://doi.org/10.1038/s41598-021-96674-0\u003c/li\u003e\n\u003cli\u003eVarma, A. K., Jaiswal, N., Das, A., Kumar, M., Lele, N. V., Tripathy, R., Maity, S., Pandya, M., Bhattacharya, B., Mandal, A. K., Jishad, M., Seemanth, M., Sahay, A., Ganguly, D., Bhowmick, S. A., Sarangi, R. K., Agarwal, N., Raman, M., Sharma, R., Desai, N. M. (2023). A pathway for multi-stage cyclone-induced hazard tracking\u0026mdash;case study for Yaas. \u003cem\u003eNatural Hazards\u003c/em\u003e, \u003cem\u003e117\u003c/em\u003e(1), 1035\u0026ndash;1067. https://doi.org/10.1007/s11069-023-05893-3\u003c/li\u003e\n\u003cli\u003eWang, J., Gao, W., Xu, S., \u0026amp; Yu, L. (2012). Evaluation of the combined risk of sea level rise, land subsidence, and storm surges on the coastal areas of Shanghai, China. \u003cem\u003eClimatic Change\u003c/em\u003e, \u003cem\u003e115\u003c/em\u003e(3\u0026ndash;4), 537\u0026ndash;558. https://doi.org/10.1007/S10584-012-0468-7/TABLES/9\u003c/li\u003e\n\u003cli\u003eXianwu, S., Bingrui, C., Jufei, Q., Xing, K., \u0026amp; Tao, Y. (2021). Simulation of inundation caused by typhoon-induced probable maximum storm surge based on numerical modeling and observational data. \u003cem\u003eStochastic Environmental Research and Risk Assessment\u003c/em\u003e, \u003cem\u003e35\u003c/em\u003e(11), 2273\u0026ndash;2286. https://doi.org/10.1007/S00477-021-02034-9\u003c/li\u003e\n\u003cli\u003eXiao-ting, F., Ying, L., Ai-min, L., Long-sheng, L., Xiao-ting, F., Ying, L., Ai-min, L., \u0026amp; Long-sheng, L. (2020). Statistical and Comparative Analysis of Tropical Cyclone Activity over the Arabian Sea and Bay of Bengal (1977-2018). \u003cem\u003eJournal of Tropical Meteorology, 2020, Vol. 26, Issue 4, Pages: 441-452\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(4), 441\u0026ndash;452. https://doi.org/10.46267/J.1006-8775.2020.038\u003c/li\u003e\n\u003cli\u003eNational Cyclone Risk Mitigation Project. \u003cem\u003eCyclones \u0026amp; their impact in India\u003c/em\u003e. https://ncrmp.gov.in/cyclones-their-impact-in-india/\u003c/li\u003e\n\u003cli\u003eRao, A. D., Chittibabu, P., Murty, T. S., Dube, S. K., \u0026amp; Mohanty, U. C. (2007). Vulnerability from storm surges and cyclone wind fields on the coast of Andhra Pradesh, India. \u003cem\u003eNatural Hazards\u003c/em\u003e, 41(3), 429\u0026ndash;446. https://doi.org/10.1007/s11069-006-9041-z\u003c/li\u003e\n\u003cli\u003eIBAN, M.C., BILGILIOGLU, S.S. Snow avalanche susceptibility mapping using novel tree-based machine learning algorithms (XGBoost, NGBoost, and LightGBM) with eXplainable Artificial Intelligence (XAI) approach. \u003cem\u003eStoch Environ Res Risk Assess\u003c/em\u003e 37, 2243\u0026ndash;2270 (2023). https://doi.org/10.1007/s00477-023-02392-6\u003c/li\u003e\n\u003cli\u003eYang, R., Zheng, G., Hu, P., Liu, Y., Xu, W., \u0026amp; Bao, A. (2022). \u003cem\u003eSnowmelt flood susceptibility assessment in Kunlun Mountains based on the Swin Transformer deep learning method\u003c/em\u003e. Remote Sensing, 14(24), 6360. https://doi.org/10.3390/rs14246360\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-geoscience","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Geoscience](https://www.springer.com/journal/44288)","snPcode":"44288","submissionUrl":"https://submission.nature.com/new-submission/44288","title":"Discover Geoscience","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Storm Surge, Eastern Coastline of India, Cyclone, Machine Learning, Explainable AI (XAI)","lastPublishedDoi":"10.21203/rs.3.rs-9505599/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9505599/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eStorm Surge is an unusual upliftment of the sea water and is measured in terms of the height of the water predicted above astronomical tide. Primary cause of such surge is the strong wind during the landfall of cyclone which forces the sea water beyond the coast line. The Bay of Bengal is considered for the genesis of maximum number of tropical cyclones particularly due to its higher sea surface temperature during the pre-monsoon and post-monsoon period. Additionally, the Bay of Bengal is a shallow, partially enclosed body of water and is funnel-shaped which triggers the genesis and intensification of cyclone. The states of Odisha and Andhra Pradesh lie along the eastern coastline of India which stretch for roughly around 574.71 km and 1053.07 km respectively. Tropical Cyclones mostly make landfall along these coastal districts making these highly susceptible to storm surge particularly due to their low lying terrain, presence of River deltas, and degradation of natural barriers. Present research aims to map storm surge susceptibility for six coastal districts of Odisha and twelve coastal districts of Andhra Pradesh using Light Gradient Boosting Machine (LightGBM), Random Forest and XG Boost (Extreme Gradient Boosting), the machine learning models. Area Under Curve (AUC), calculated using test data, indicated that XGBoost, Random Forest and LightGBM which map the susceptibility with acceptable accuracy of0.980, 0.977 and 0.978 respectively. Additionally, this research implements couple of XAI techniques such as SHapley Additive exPlanations (SHAP) and permutation importance which revealed Rainfall and Elevation having highest contribution. Storm surge prediction was done with Multi-Layer Perceptron regressor mechanism where some important meteorological parameters were considered. These factors were trained with Storm Surge levels from the period spanning from 2007\u0026ndash;2023 which achieved MAE at 0.18 and MSE at 0.26. Cyclone parameters of 2024 were given as an input in the algorithm and the resultant output was Storm Surge levels in meters which was also verified with the INCOIS predicted report and IMD reported surge levels along the tidal gauges.\u003c/p\u003e","manuscriptTitle":"Storm Surge Susceptibility Mapping and Prediction of Tropical Cyclones using Machine Learning along the Eastern Coastline of India","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-27 04:39:42","doi":"10.21203/rs.3.rs-9505599/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"171363943424335829429049360466031246674","date":"2026-05-18T22:36:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"200328403623828516668999301553715693499","date":"2026-05-18T17:12:28+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-05-18T16:52:14+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-30T12:54:55+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-30T12:54:14+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Geoscience","date":"2026-04-23T10:20:07+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-geoscience","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Geoscience](https://www.springer.com/journal/44288)","snPcode":"44288","submissionUrl":"https://submission.nature.com/new-submission/44288","title":"Discover Geoscience","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"257252e2-1b3d-47c2-af84-22837de93415","owner":[],"postedDate":"April 27th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"171363943424335829429049360466031246674","date":"2026-05-18T22:36:40+00:00","index":16,"fulltext":""},{"type":"reviewerAgreed","content":"200328403623828516668999301553715693499","date":"2026-05-18T17:12:28+00:00","index":15,"fulltext":""},{"type":"reviewersInvited","content":"3","date":"2026-05-18T16:52:14+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-30T12:54:55+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-30T12:54:14+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-18T17:08:10+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-27 04:39:42","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9505599","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9505599","identity":"rs-9505599","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}