Reconstructing Reservoir Storage of Indian Large Dams using Hydrological Model and Deep Learning Algorithms

Abstract

Reservoirs play a pivotal role in mitigating floods and regulating river flows, particularly during irregular flows. In India, there are 6,138 large dams, yet only 119 are monitored by the Central Water Commission (CWC). In addition, the observations are available only from the year 2000 onwards. The long-term reservoir storage data are used in water management planning and flood control operations. It can also be used in calibrating hydrological models. However, long-term reservoir storage observations for the large dams in India are still lacking. We used hydrologic and hydrodynamic model framework to simulate the daily long-term reservoir storage data. We included more than 150 dams within the state-of-the-art Catchment-based Macroscale (CaMa) Flood model and generated the model simulated reservoir storage data. Further, we used the Long Short Term Memory (LSTM) algorithm to improve the model simulations using observations from CWC. We used the Global Reservoir Storage (GRS) data as observations for the dams not monitored by CWC. We intend to assess the application of the combined framework of the hydrological model and deep learning technique in simulating reservoir storage. Furthermore, we intend to analyse the long-term changes in the basin hydrology and the reservoir seasonal cycle. Long-term reservoir storage data can be utilized to plan water management and adaptation to climate change.