LAUSR

It is common knowledge that drought is considered one of the most damaging natural disasters in terms of economic costs, societal problems, and ecological impacts. In this study, we selected 53 Karst drainage basins in South China as research areas and automatically extracted the characteristics of water system based on 30 m DEM data using GIS technology. The surface confluence and runoff process of atmospheric precipitation were simulated by BP neural network, and we analyzed space coupling of drainage characteristics and studied their driving mechanisms for hydrological drought. Results show that (1) basin shape index, river network density, and main channel’s longitudinal slope respond positively to atmospheric precipitation, while other drainage characteristics respond negatively. This shows that drainage characteristics’ response to atmospheric precipitation gradually decreases from input layer to output layer, which means that watershed confluence capability gradually weakens. (2) Driving of drainage characteristics to runoff process is positive in the first hidden layer with its driving effect order arranged from small to big as follows: river network density < main channel length < fractal dimension index of drainage < basin shape index < longitudinal slope of main channel < average drainage elevation, and negative in the second hidden layer with its driving effect order arranged as follows: 3 < 6 < 9 < 12 months. (3) The driving mechanisms of drainage characteristics to hydrologic droughts are positive in first hidden layer and negative in second hidden layer. Fitting effects of the driving mechanism models are good (R2 > 0.9), with excellent model significance (Sig. = 0.0).