Optimising sentinel-1 backscatter thresholds for surface water mapping in small irrigation tanks: a google earth engine web application for tank water management

Abstract

Small irrigation tanks are essential structures for water management in the dry zone of Sri Lanka, yet consistent time-series observations of their surface-water extent remain scarce. This study develops and assesses optimised Sentinel-1 backscatter thresholds for surface-water mapping, implemented on the Google Earth Engine platform and demonstrating its application at Eru Wewa in the Malwathu Oya River Basin. The model was calibrated using reference points collected on 30 March 2021 by testing VV, VH, and VV−VH threshold combinations and selecting the combination with the highest overall accuracy and Kappa coefficient. The optimal thresholds were negative 17 dB for VV, negative 35 dB for VH, and 14 dB for the VV minus VH difference. The calibrated model achieved an overall accuracy of 95% and a Kappa coefficient of 0.894. Independent reference points from 3 December 2021 and 22 August 2025 gave overall accuracies of 90% and 90% with Kappa coefficients of 0.807 and 0.800, respectively. The estimated water-surface areas were validated on three dates, with a consistent underestimation of between 6% and 15%. The water-surface-area time series from 1 January 2020 to 30 November 2024 was analysed alongside seasonal rainfall to assess the temporal response of the reservoir. The workflow was further developed into a user-friendly web application that generates water-extent time series and converts them into storage based on the area-capacity curve, following the depth-area-storage capacity relationships for village tanks lacking surveyed bathymetric data. The application demonstrates practical reservoir monitoring at Eru Wewa, while broader applicability requires multi-site validation.