Study on the localization of tropical cyclone rainfall climatology and persistence model R-CLIPER

The Tropical Cyclone (TC) Rainfall Climatology and Persistence Model (R-CLIPER) is a statistical parametric model for TC precipitation based on observations of North Atlantic TCs or global TCs. It boasts simple inputs and quick calculations, enabling climate-scale TC precipitation simulations, thus providing technical support for TC precipitation forecasting and risk assessments. This study focuses on TCs in the Western North Pacific region and performs localization of the R-CLIPER model. Firstly, based on precipitation data from Tropical Rainfall Measuring Mission (TRMM) satellite and Fengyun satellite (FY2C/2E) precipitation data, the radial average profiles of precipitation from TCs of 6 different intensity levels (i.e., super typhoon, strong typhoon, typhoon, severe tropical storm, tropical storm, tropical depression) are extracted. Secondly, combined with the R-CLIPER model framework, a globally optimal parameterization fitting method for TC precipitation profiles is developed. Two models are constructed, including the R-CLIPER for the Western North Pacific Ocean based on TRMM data/FY data, namely TRMM-R-CLIPER-WNP (Model 1) and FY-R-CLIPER-WNP (Model 2). Finally, based on the 62 tropical cyclones in the Northwest Pacific from 2012 to 2013 and the 26 cyclones that affected Zhejiang from 2009 to 2021 the models are evaluated in terms of fitting error and precipitation falling area. The results are as follows. (1) The TC rainfall profiles derived from TRMM data are relatively sharp near the maximum rainfall, while those from FY2C/2E satellite data are smoother and more detailed due to its higher spatial resolution. (2) The RMSE of TRMM parameterized profiles is 0.28 mm·h^-1, and that of FY2C/2E profiles is 0.51 mm·h^-1. (3) Based on the ground-based station data evaluation, when the rainfall threshold is less than 3 mm·h^-1, the ETS score of Model 1 is better than that of Model 2. But when the threshold is greater than 3 mm·h^-1, Model 2 is better. Both models perform better when the TC is centered over the sea than over land, indicating higher accuracy of TC central rainfall intensity prediction over the ocean than over land. In conclusion, using the parameterized fitting method, the accuracy of the R-CLIPER model in simulating TC precipitation in the Northwest Pacific region can be improved by 5.5%. Additionally, it was revealed that the error in TC precipitation simulations is influenced by factors such as model data sources, model framework, and parameterization schemes.