عنوان مقاله [English]
One of the challenges facing meteorologists in recent years is to improve the quality and accuracy of weather nowcasting for limited areas and in this regard various methods based on the statistics principles, such as data assimilation and ensemble forecasting methods, have been used in numerical weather prediction models. In the data assimilation methods, by transferring and collecting different data including atmospheric measurements by observational stations, satellites and radars, the process of rectifying the results of numerical models is performed statistically. The aim of this research is to investigate and address the question of whether it is possible to facilitate a cycle of numerical weather prediction and improve the prediction accuracy using remote sensing data, without involving very large computational effort required in the data assimilation techniques.
To reach the objectives of this research, we first designed and developed a software for radar data analysis. Second, based on both the output data provided by the radar model and the innovative changes in the relevant part of the main model for controlling convection, the Weather Research and Forecasting (WRF) model was modified in such a way that the best convection scheme is chosen during the execution of the model. In fact, the appropriate convection scheme is chosen automatically by the capability of the radar system in detecting convection within the execution of the model. To evaluate the results, the modified model was used for a region of Iran in such a way that the site of Tehran weather radar was located in the center of the simulation domain. Before carrying out the simulations, two necessary actions were taken. First, the sensitivity of the results provided by the WRF model to the initial input data was examined. In this stage, using two reanalysis datasets of the NCEP-FNL (Final) on 1°×1° grid prepared operationally every six hours and the ECMWF dataset gridded to a horizontal resolution of approximately 80 km at 6-h intervals, a selected case was studied and the results were compared with the observational data. Then, the processing algorithms necessary to identify and remove radio electromagnetic interference (RFI) noise from radar returns were prepared.
The modified WRF model was run for 12 hours to evaluate its ability and the quality of prediction for very short time periods. In total, forty experiments were carried out using eight configurations of physical parameterization schemes as well as inclusion of the radar data. In addition, the modified model was implemented for a severe squall line that occurred in Tehran’s area at 2330UTC on the 30th of March 2009 and was detected by the Tehran weather radar.
Results for the root mean square error index and correlation between the forecasted and the observed precipitation showed that the accuracy of precipitation forecast in the study area for very short time, e.g. 6 hours, was increased when the modified model was carried out. The comparison between the forecasted precipitation patterns and the observations confirms higher consistency for the modified model’s results. Also, evaluating the results by the statistical methods, it is seen that the correlation between the forecasted precipitation and the observed values is increased significantly and the root mean square error is decreased. In addition, the time series of the precipitation data for Mehrabad synoptic station in Tehran was investigated for which the root mean square error in the precipitation time series was decreased when the Tehran radar data was included in the working of the WRF model.