Assessment of the performance of cumulus and boundary layer schemes in the WRF-NMM model in simulation of heavy rainfalls over the Bushehr Province during 2000-2020

The mesoscale numerical weather prediction system of Weather Research and Forecasting (WRF), with two cores of ARW and NMM, has been used for atmospheric research, operational forecasting, and dynamical downscaling of Global Climate Models. Many parameterizations for each physics option can be accessed in this model. It is noteworthy that the performance of the model depends on the selected configuration and varies in different areas. Therefore, choosing a configuration with the lowest error for each terrain is mandatory. Here, the performances of various physics schemes, including cumulus and boundary layer schemes of the WRF-NMM model, were examined to simulate twelve heaviest extreme rainfall events in the southwest of Iran, the Bushehr Province, during 2000-2020. These events lasted for eighteen days. Three domains with 27, 9, and 3 km resolution were used in the model configurations, with no cumulus option for the smallest one. The initial and boundary conditions were used from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) datasets. One hundred and eight simulations were done using six cumulus schemes of KF, BMJ, SAS, oldSAS, NSAS, and TiedTKE, and seventy-two runs were done to evaluate the boundary layer schemes of MRF, MYJ, QNSE, and YSU. The simulated precipitation patterns were assessed using two observational data sets, including (I) in-situ measured data from eleven automatic weather stations and (II) grid point data from Global Precipitation Measurement (GPM) satellite with 0.1-degree horizontal resolution. Four statistic indices of Root Mean Square Error, Correlation Coefficient, Standard Deviation, and Bias were applied in the evaluation process. The evaluation process with the data measured at 11 automatic weather stations was done using outputs of the third domain. The outputs of the second domain were used for evaluation basis on GPM data at grid points. For a comprehensive analysis, the assessment process was performed separately for rainfall events (March-April and November-December events) in coastal and non-coastal stations. Comparison of precipitation from simulations of various cumulus schemes with the eleven in-situ data showed that the schemes from SAS family well performed at March-April events at coastal and noncoastal stations. While, the KF scheme produced the least error at coastal and noncoastal stations during the November-December events. The precipitation data from 1271 GPM grid-point data revealed that the oldSAS scheme generated the least error for the March-April and November-December events. According to the number of GPM grid-point data, the oldSAS scheme opted as the cumulus option for the next runs. Evaluation of WRF-NMM simulations using different boundary layer physics with the in-situ data indicated that MRF scheme produced the minor error at coastal and noncoastal stations for both March-April and November-December events. Using the 1271 GPM grid-point data illustrated that the QNSE and MRF (MYJ and MRF) options did the best performance for March-April (November-December) events. In conclusion, based on the number of GPM grid-point data compared with in-situ measured data, it is suggested that the oldSAS cumulus scheme and MRF boundary layer scheme can be chosen with some robustness in predicting the amount and pattern of the heavy rainfall precipitation in Bushehr Province of Iran. It is also notable that the default options introduced by the model for cumulus scheme and boundary layer scheme in the WRF-NMM model produce the largest error and are not appropriate for the selected area. This reveals the importance of adequately selecting physics options for this area.