A Study Of Ensemble-sensitivity-based Initial Condition Perturbation Methods For Convection-permitting Ensemble Forecasts

Based on the Weather Research and Forecast(WRF)model and two convective heavy precipitation cases,this study applied Ensemble-based Sensitivity Analysis(ESA)method to explore the sensitivity of the convective precipitation forecast to previous environment conditions at different times,and tried to introduce the sensitivity patterns into the initial perturbation.A Sensitivity-based Breeding Growth Mode(Se BGM)method is proposed to generate the initial perturbations that combined with sensitivity.This research contains four procedure.First,a hindcast experiment is performed to simulate the weather situation and environmental conditions before the occurrence of two convection cases;Then,based on the results of the hindcast,sensitivity patterns of the convection initiation to the previous weather situations and thermodynamic conditions at various times are obtained by performing ESA;Next,the Se BGM method is designed to continuously introduce the sensitivity patterns at different times into the fast-growing perturbations through the breeding cycle.The two convective cases are predicted and the prediction results are evaluated.Finally,the Barnes filter method is used to filter the sensitivity patterns into three groups by wavelength(WL).The filtered sensitivity patterns are introduced into initial perturbations by the same method.The development of forecast errors with change of perturbation scales is explored to analysis the response mechanism of model to the sensitivity-based perturbation.The research mainly reached the following conclusions:(1)ESA can reveal the relationship between the development of convection and the evolution of previous weather conditions.The sensitivity shows different structures with the evolution of weather situation,reflecting the impact of the weather system's movement and intensity development on precipitation.It is proved that ESA method can indicate the factors that play a crucial role in the occurrence of precipitation,and provide valuable information for subsequent forecasts.(2)The Se BGM method proposed in this paper has better ability to capture convective precipitation,improving the forecast skills efficiently.When the perturbation magnitude is changed,the traditional BGM method and the Gaussian-based perturbations show large fluctuations of the forecast error in different cases,while the error of Se BGM method shows a relatively regular development trend with the change of the perturbation magnitude,and it has a relatively lower error across perturbation magnitudes.In addtion,the Se BGM method has a systematic adjustment of the near-surface situation so that the speed and position forecasts of the convective systems are improved,which makes the trajectory of the rain band more in line with reality.(3)The perturbations of different wavelengths show different characteristics in the prediction.The main difference is that the small-and meso-scale sensitivity-based perturbations focus on the local adjustment of precipitation and convection intensity,while large-scale sensitivity-based perturbations show excessive systemic adjustments.However,three groups of perturbations based on filtered sensitivity patterns have certain limitations in representing the uncertainty of convection forecasting at different scale,so that the perturbation based on original ESA patterns can better adapt to different weather processes and have a greater improvement effect on forecasting.This study combines the sensitivity analysis and the initial perturbation technique to target the crucial areas in the forecasts.This concept is beneficial for the future research and operation of convection-permitting ensemble system to improve the forecast skill.