Author(s): Luming Song; Bin Lian; Lili Huang; Tong Li; Jijun Zhao

Linked Author(s):

Keywords: Rainfall intensity; CML; RF; LSTM

Abstract: Accurate rainfall intensity estimation is of paramount importance for water resource management and flood disaster mitigation. Commercial microwave link (CML) -based rainfall inversion has been recognized as an effective method to provide more detailed rainfall information. Our research presents a hybrid model for CML-based rainfall measurement that employs long short-term memory (LSTM) with regularization and dropout, expressed as LSTM-RD, and random forest (RF). This model aims at probing into the correlation between CML attenuation and rainfall intensity and consequently improving the precision of rainfall intensity estimation. Regarding the estimation of medium-term and short-term rainfall events, our hybrid model yields a correlation coefficient of 0.84,0. 96, and a mean squared error of 0.07,0. 09. These results exhibited notable superiority over performance indices presented by the standalone LSTM-RD and RF models. The results obtained from our model indicated a tremendous potential in the overall accuracy and stability of rainfall intensity estimates based on CML.

DOI: https://doi.org/10.3850/iahr-hic2483430201-297

Year: 2024