Reservoir Storage Monitoring from Satellite Imageries and Digital Elevation Models in South Asia

Near realtime reservoir storage information is essential for optimal water resources utilization and flood mitigation. However, South Asia is dominated by international river basins where communications among neighboring countries about reservoir storage is extremely limited. The existing remote sensing reservoir dataset covers only 21 reservoirs (28% of total capacity) because of the use of altimeters which have insufficient coverage. This has inhibited applications which require denser, remotely sensed information about the reservoir networks. In order to extend the spatial coverage, a new algorithm is developed to estimate reservoir storage by using water surface area from Moderate Resolution Imaging Spectroradiometer (MODIS) imageries and surface elevation from Digital Elevation Model (DEM) data collected by the Shuttle Radar Topography Mission (SRTM). The approach contains three steps. First, the largest MODIS area (for each reservoir) obtained from 2000 to 2015 is overlaid onto the SRTM DEM map to delineate the region in which the water area-elevation relationship is derived. Then, the area-elevation relationship is obtained by regressing the cumulative area against the elevation values within the delineated reservoir region. Finally, the storage change values are calculated over the entire study period by applying the area-elevation relationship to the area estimations. Using the SRTM based method, the spatial coverage of the South Asian reservoir dataset can be extended significantly. Nine additional South Asian reservoirs (each with areas larger than 65 km2 at capacity) will allow the new product to represent more than 50% of the overall storage capacity in South Asia. This approach has been tested over five reservoirs where gauge observations are accessible. The storage estimates obtained are highly correlated with observations, with coefficients of determination ranging from 0.85 to 0.98 and normalized root mean square errors (NRMSE) ranging from 12.78% to 29.59%.