Assessment of the accuracy of numerical forecasts of meteorological conditions in the region of Tomsk using WRF model

The paper describes the results of errors assessment of major meteorological values forecast using the predictive WRF (Weather Research & Forecasting, version 3.4.1) system in its local variant. Hourly meteorological prognostic values for the period from 6 p.m. of June 15, 2012 to 12 p.m. of June 17, 2012 and the actual weather during this period were the material for our research. The research was made by repeated launching of the WRF modeling system with changing parameterization of humidity microphysics while parameterization of other processes remained the same. The calculations were performed in the three embedded areas with a common centre in the city of Tomsk. The size of the local area was 50 x 50 km with 1 km step grid spacing. Assessment of forecast quality was made for this area. The accuracy of temperature, pressure, humidity, wind speed and direction, rainfall forecasts were assessed by the following criteria: mean absolute, mean squared and mean arithmetic errors. To assess the rainfall forecast other characteristics were also calculated. We received the following results: 1. The implementation of the local WRF model, version 3.4.1, of meteorological values and phenomena forecast for Tomsk showed good opportunities and prospects for its use in the assessment of local weather conditions. 2. Characteristics of the model's accuracy is satisfactory: the accuracy of the air temperature forecast corresponds to the achieved level for other modern models (e.g., COSMO-RU); dew point temperature and wind speed forecasts were slightly better than those of well-known models; rainfall forecast by the number and fact of their existence is comparable to the used mezo-models and in some characteristics was better; the forecast of the cloud base height and their capacity proved to be completely realistic, but the cloud base height was overstated by 1-2 km. Additional attraction of prognostic data on relative humidity of air allows to clarify the height of cloud base and at high humidity (over 90%) to also diagnose the presence of fog or haze. Air pressure forecast at the station level requires further elaboration. 3. Different parameterization of microphysical processes in the model resulted in different accuracy of basic meteorological values forecast, but in general they are comparable. WSM5 scheme and ETA gave the best results for the forecast of rain. Lin system gave the worst results. In general, greater detail is required on the cloud model parameters for the lower 1-2-km layer, which is particularly important for the local weather forecast. Exploring the capacities of the WRF model and other models to determine the best and its further implementation in forecasting should continue. The work is of interest for the improvement of local and mesoscale numerical weather prediction.

Keywords

Authors

References