Sensitivity Test Of A Heavy Rainfall And Ensemble Forecast Study By WRF Model

In recent years, the horizontal resolution and the physical process of numerical models have greatly improved, thereby enabling the operational use of numerical forecasts with higher resolution and more detailed description of topography. Their performance in scientific research and operational forecasting is getting better. The scientists have designed sensitivity tests in order to study rainstorm. Nevertheless, it’s still difficult to get the satisfactory results about the severe weather. The limitation of numerical prediction makes people realize that there are many uncertainties in traditional numerical predictions made with a single model. Now the ensemble forecasts have been a very popular technology of weather forecast. It appears to be a good choice in improving the short-range numeric prediction.The typical rainstorm process from 12 to 13 July 2013 was simulated by WRF V3.5 meso-scale model. Two sensitivity tests were designed to explore the effect of terrain and water vapor on the rainstorm. Finally, several experimental short-range ensemble forecast systems are constructed. The main results are:(1)Two sensitivity tests,which dropped the elevation and reduced the water vapor content in initial field.were designed to explore the effect of terrain and water vapor on the rainstorm.The results of topographic sensitivity test show that the terrain can significantly influence the extent and the maximum amount of precipitation. Terrain can affect the atmospheric convergence and divergence. It influences speed of warm air. The results of water vapor sensitivity test show that the water vapor transferred from the rainfall area and the south have significant effects on the amount of rainfall. The collection of water vapor in the rainstorm process plays an important role in rainfall.(2)The horizontal filtering technique is employed to separate meso-scale filed from large scale field. The results show the obvious differences are existed in sub-synoptic scale characteristics of ERA-interim and FNL initial fields. Case experiments are conducted with the initial fields and lateral boundaries provided with ERA-interim and FNL(called exp_ERA and exp_FNL). The WRF simulated results have obvious differences. And also the different WRF simulated results have contrast in area、intensity and rainfall of precipitation. Exp_ERA has a better result than exp_FNL. Namely the difference in initial fields cause the contrast in simulated results. After adding perturbations to the initial fields of physical variables (temperature, humidity, wind, lateral boundaries layer),the authors analyze the impacts of perturbations of different physical variable fields on the precipitation forecast. For 24-hour precipitation forecast, the perturbation in humidity field has the greatest impact on precipitation, closely followed by that in wind and temperature perturbation fields, the perturbation in lateral boundaries layer has the least impact.(3)The simulation of a heavy rainfall event was conducted to examine the effects of perturbations of the model uncertainty, the initial value uncertainty and the lateral boundary uncertainty in the ensemble forecast. The effects from the model uncertainty, the initial value uncertainty and the lateral boundary uncertainty on precipitation forecast are different. The effect from the model uncertainty has the greatest impact on precipitation. During the early period of model integration, the effect from the initial value uncertainty is obvious. Later, this effect becomes smaller and smaller. During the early period of model integration, the effect from the lateral boundary uncertainty is small. Later, this effect becomes larger and larger. After some time, the effect is as large as the effect from the model uncertainty. The ensemble mean precipitation appeals that the rain has been made up for the shortcomings of the control test. The average RMSE of precipitation have been reduced. The improvement rate of the lateral boundary perturbation is the largest, followed by the model perturbation. That of the initial perturbation is the smallest. Comparing the three kinds of uncertainty with each other, we found that the ensemble forecast which contains perturbations of the model uncertainty is the best, and the ensemble forecast which contains perturbations of the initial value uncertainty is the worst. Overall, the ensemble forecast which contains the three kinds of uncertainty is the best and it improves the forecast of precipitation.