Cultivating Resilience: Sentinel-2 Remote Sensing for Precision Flood Detection and Susceptibility Mapping in Agricultural Landscapes

Abstract

In agricultural areas, flooding is a frequent and destructive natural calamity that seriously harms crops, infrastructure, and property. Early detection and prediction of floods is crucial for disaster management and response efforts in agricultural areas. In this study, we present a novel method for Sentinel-2 remote sensing data-based flood detection and susceptibility mapping in agricultural farms, based on the application of different unsupervised clustering methods and supervised machinelearning algorithms. The approach involves pre-processing the Sentinel-2 images to extract features, such as spectral reflectance values, and transforming the images into feature vectors. These feature vectors are then used as inputs to several differentalgorithms to identify regions that represent flooded areas. This work presents two approaches namely, unsupervised clustering approach, where 5 different algorithms are used and supervised machine learning approach, where 3 different algorithms areused in identifying the flooded regions under different conditions. Accuracy and water percentage increase of both approaches are compared. The study’s findings demonstrate that the suggested strategy is efficient for detecting and mapping flood susceptibility in agricultural areas using Sentinel-2 remote sensing data, with an overall accuracy of up to 92% for the Support Vector machine(SVM) algorithm in the supervised algorithms category and DBScan algorithm with an accuracy of 84% in the unsupervised alorithms category.