A novel approach for estimating the capacity of ungauged small reservoirs using remote sensing and DEM

The estimation of small reservoir capacity is of great significance to water resources management. However, many widely distributed small reservoirs lack the capacity information because of the high costs of field measurements. This study proposed a novel approach to estimate the small reservoir capacity in the hilly area by using remote sensing and DEM. The basic idea of this approach is to explore the relationship between influential factors (i.e., topographic and geomorphic parameters) and measured reservoirs' capacity to establish a machine learning model based on particle swarm optimization-extreme learning machine (PSO-ELM) to estimate the capacity. The Mihe River basin in northern China is selected as a case study, 111 measured reservoirs, and six optional influential factors are selected to develop and test this model. The results show that the five influential factors (i.e., the area of sub-catchment, the water surface area, the longest flow path of sub-catchment, the average slope of sub-catchment, and the average slope of buffer area) are the optimal combination with the lowest difference between the measured and the estimated reservoir capacities. The results demonstrate that the proposed approach is a robust tool for estimating the capacity of small reservoirs in the hilly area.