Impact of climate variability on streamflow and suspended sediment load and its implications on reservoir storage capacity in the Indian tropics

Increasing hydroclimatic extremes driven by climate change complicate the estimate of streamflow and suspended sediment load (SSL), while accelerating reservoir sedimentation threatens long-term water security. This study employs a Physics-Informed Machine Learning (PIML) framework, integrating the Soil Water Assessment Tool (SWAT) with a Long Short-Term Memory (LSTM) network, to project future streamflow and SSL in two major tropical river basins: the Narmada River Basin (NRB) and the Tapi River Basin (TRB) of the Indian tropics. Climate projections from four CMIP6 General Circulation Models (GCMs) were used under optimistic (SSP1–2.6) and high-emission (SSP5–8.5) scenarios for the future period 2015–2100 relative to a historical baseline (1952–2014). The PIML model demonstrated superior predictive skill, with NSE exceeding 0.85 for streamflow calibration and 0.55 for SSL. Ensemble projections indicate mean annual SSL will increase by more than 25% (SSP1–2.6) and 30% (SSP5–8.5) relative to the baseline for both basins. These increases are projected to reduce the storage capacity of the Indira Sagar reservoir (NRB) by more than 25%, and the Ukai reservoir (TRB) by more than 50%, by the year 2100. Concurrently, mean annual streamflow is projected to rise by 17–22% in the NRB and 39–52% in the TRB, significantly elevating flood risk, particularly for downstream communities in Surat city. This study underscores the urgency of adopting proactive sediment management strategies and climate-resilient reservoir operations to safeguard water security and mitigate future flood hazards in tropical river basins.