Comparison of monthly rainfall generated from dynamical and statistical downscaling methods: a case study of the Heihe River Basin in China

Abstract

Monthly rainfall in the Heihe River Basin (HRB) was simulated by the dynamical downscaling model (DDM) and statistical downscaling model (SDM). The rainy-season rainfall in the HRB obtained by SDM and DDM was compared with the observed datasets (OBS) over the period of 2003–2012. The results showed the following: (1) Both methods reasonably reproduced the spatial pattern of rainy-season rainfall in the HRB with a high-level skill. Rainfall simulated by DDM was better than that by SDM in the upstream, with biases of −12.09 and −13.59 %, respectively; rainfall simulated by SDM was better than that by DDM in the midstream, with biases of 3.91 and −23.22 %, respectively; there was little difference between the rainfall simulated by SDM and DDM in the downstream, with biases of −10.89 and −9.50 %, respectively. (2) Both methods reasonably reproduced monthly rainfall in rainy season in different subregions. Rainfall simulated by DDM was better than that by SDM in May and July in the upstream, whereas rainfall simulated by SDM was closer to OBS except August in the midstream and except August and September in the downstream. (3) For multi-year mean rainy-season rainfall in different stations, there was a little difference between the rainfall simulated by DDM and SDM in Tuole station in the upstream, with biases of −13.16 and −12.40 %, respectively; rainfall in Zhangye station simulated by SDM was overestimated with bias of 14.02 %, and rainfall simulated by DDM was underestimated with bias of −14.60 %; rainfall in Dingxin station simulated by DDM was reproduced better than that by SDM, with biases of −19.34 and −32.75 %, respectively.