Dynamics of waterlogging and drainage congestion on the Kosi Fan, Himalayan Foreland

Abstract

This study assesses the dynamics of waterlogging using time series optical satellite images from 1987 to 2021 on alluvial fan of the Kosi River in the Himalayan Foreland. We classified the satellite images to extract waterlogging patches by hybridising Simple Non-Iterative Clustering (SNIC) segmentation and Random Forest (RF) algorithms. This hybrid framework can classify the waterlogged patches from satellite images with overall accuracy ranging between 75%–90%. We observed that the waterlogging patches during the pre-monsoon period show a significantly increasing trend (4km2/year) from 1987 to 2021. During the post-monsoon period, this trend is not statistically significant to the 95% confidence level. We used these classified waterlogged images to understand the dynamics of waterlogging. We observed a high probability of waterlogging in areas adjacent to the Fan margin. Further, we assessed the likelihood of waterlogging in the vicinity road-rail network. The concentration of waterlogged patches is relatively high within a one-kilometer buffer of the road-rail network. This study is a step towards understanding the impact of anthropogenic intervensions on the dynamics of waterlogging and drainage congestion.