Using a Deep Learning Framework to Forecast Reservoir Water Availability in India

Abstract

This paper introduces a machine learning-based model to forecast reservoir water volumes in India. In areas with high water stress, having access to timely information on forecasted water availability could help decision-makers avoid the risk of acute water-driven power outages and advocate for long-term, water-prudent policies and management. This forecast can flag when drought-like conditions threaten water supply, but should not be used to inform reservoir management operations.