A Deep Learning Approach for Rainfall Estimation from Dual-Polarization Radar and Gauges Measurements for Landfall Typhoon in South China

At present, the raindrop size distribution (DSD) fitting method and traditional neural network method are the main quantitative precipitation estimation (QPE) algorithms. The former is not enough to express the spatiotemporal variability of the DSDs using generalized
$Z-R$
parameter relationship, and the latter is limited by the number of network layers, network structure and computing power. Their performance cannot both meet actual needs. In this research, we propose an alternative approach to dual-polarization radar (DPR) QPE. In particular, a radar-gauges dataset (RGD) is constructed for QPE based deep learning using radar raw data and gauges measurements for landfall typhoon in South China, and a deep learning model is designed and trained using this dataset. The model is applied to radar data to produce rainfall estimates. Preliminary results show the promising performance of this novel method compared to traditional QPE estimators.