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

WRI introduces a deep learning model, the Bayesian long short-term memory sequence-to-sequence-to-sequence model, to provide near-real-time forecasts of reservoir water volumes for a 90-day period in India. The pilot project focused on 11 reservoirs supplying thermal coal power generation, achieving a coefficient of determination score of 92% for short-term (1-14 day) forecasts and 56% for long-term (15-90 day) forecasts. The tool is intended to flag drought-like conditions and support the transition to water-prudent power production, though it is not intended for reservoir management operations.

• WRI developed a near-real-time forecasting tool for reservoir water volumes over a 90-day horizon using a Bayesian long short-term memory sequence-to-sequence-to-sequence deep learning framework and global meteorological data.
• A pilot study of 11 Indian reservoirs that provide water for thermal coal power generation showed the model's accuracy varied by timeframe, with a coefficient of determination score of 92% for 1-14 day forecasts and 56% for 15-90 day forecasts.
• The forecast is intended to help avoid acute power outages caused by insufficient cooling water supply and to support water-prudent policies, but it is explicitly not for use in reservoir management operations.

APA

World Resources Institute (n.d.). Using a Deep Learning Framework to Forecast Reservoir Water Availability in India. https://www.wri.org/research/using-deep-learning-framework-forecast-reservoir-water-availability-india

Chicago

World Resources Institute. Using a Deep Learning Framework to Forecast Reservoir Water Availability in India. n.d. https://www.wri.org/research/using-deep-learning-framework-forecast-reservoir-water-availability-india.

Wikipedia

{{cite report |author=World Resources Institute |title=Using a Deep Learning Framework to Forecast Reservoir Water Availability in India |url=https://www.wri.org/research/using-deep-learning-framework-forecast-reservoir-water-availability-india |access-date=17 August 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{worldresourcesinstitutendusing, author = {{World Resources Institute}}, title = {{Using a Deep Learning Framework to Forecast Reservoir Water Availability in India}}, institution = {World Resources Institute}, url = {https://www.wri.org/research/using-deep-learning-framework-forecast-reservoir-water-availability-india}, urldate = {2026-08-17}, note = {Indexed by Climate Insights Directory} }