Sensitivity of WRF microphysics schemes to Convective Permitting Simulation of January 2017 Heavy Rainfall Events in Southern Thailand

Abstract

<p>Heavy rainfall events are devastating, could trigger flash urban and rural river floods in some environmental settings. If we can better predict those floods, we will better protect the communities and vulnerable people. Here, we present some high-resolution regional climate model simulations to reproduce the January 2017 heavy rainfall events that occurred in Southern Thailand, causing major flash and river floods with high death tolls and significant socioeconomic impacts in the region of Krabi and Nakhon Si Thammarat provinces.  High rainfall events persisted in the region starting from January 4, 2017, peaking around January 6, and lasted until January 10. To reproduce the detailed timeline and spatial changes of heavy rainfall events, we have run the community Weather Research Forecasting (WRF) model with different combinations of cloud microphysical schemes.  The initial and lateral boundary conditions for our WRF simulations are based on the ERA5 reanalysis. The model simulations cover the period from 1st to 18th January 2017 using two nested domains with horizontal resolutions at 3-km and 9-km for the inner and outer domains, respectively. Four simulations were conducted with different cloud microphysics (WSM-6 scheme, Goddard scheme, Thompson scheme, and Purdue-Lin scheme) while keeping all other model configurations the same. In addition to these four experiments, we have also carried out one further experiment with a single domain at 3-km horizontal resolution. WRF simulations are compared using two satellite-derived measurements: 1) NASA Global Precipitation Measurement (GPM) Integrated Multi-satellitE Retrievals for GPM (IMERG) and 2) Climate Data Prediction Morphing (CMORPH). </p><p> </p><p>Our WRF simulations have reproduced the spatial distribution of this particular rainfall event, but the rainfall magnitude (intensity) is underestimated as compared with observations. Furthermore, different microphysical schemes have resulted in varying magnitudes of rainfall intensities with WSM-6 and Purdue-Lin schemes performing much better as compared with both Goddard and Thompson schemes. We also found the 3-km single domain run including spectral nudging has the best result of rainfall magnitude and spatial distribution as compared with the four nested runs. This paper re-emphasizes that with some careful selection of model configurations, WRF can reproduce detailed regional atmospheric processes. The best WRF model configuration can then be used in our dynamical downscaling of the IPCC AR5 CESM model run under the RCP6.0 from 2080 to 2100.</p>