Sensitivity of the WRF Regional Meteorological Model to Input Datasets and Surface Parameters for the Kanto Plain on Fine Summer Days

The Weather Research and Forecasting (WRF) model has been developed as the next-generation model after the widely used the Fifth-Generation Penn State/NCAR Mesoscale Model (MM5). The number of climatologists who use the WRF for local climate study has recently been increasing. Therefore, it is important to investigate the performance of the WRF with the default settings and to estimate how the forecast accuracy of the WRF can be improved by resetting the parameters and altering the input data. However, these sensitivities have not been compared synthetically in past studies. In this study, we quantitatively estimated the sensitivity of the WRF for surface air temperature to the input datasets and surface parameters. Furthermore, we determined the most sensitive factor. The results showed that with the default settings, the WRF tends to estimate surface air temperature to be lower than observations throughout the day, particularly in the northwestern parts of the Kanto plain. The results of the sensitivity experiment indicated that the atmospheric datasets and land-use datasets have the most impact on simulated surface air temperature. Sensitivities of the WRF for surface air temperature to the surface parameter and other datasets are less than those to the above two datasets. These results indicate that further studies integrated with land-use management and other disciplines within the field of human geography will be important for improving the WRF repetitive precision and increasing accuracy in local climate prediction.