多元信息耦合的致灾山洪降雨预报方法

Abstract

雷达预估信息(0~2 h)、数值天气预报产品(0~72 h)及基于高空和地面大气探测资料的综合预报信息(0~24 h)在不同预见期的降雨预报各有差异，预报效果上各有优劣，而针对山洪的预报存在准确率低、预警有效时间短等问题，研究面向山洪灾害防治区的数值预报模式的选取、模式最优物理参数化组合方案、同化方案、降水偏差订正技术以及基于大数据的自分型雷达估测降水最优化算法。以湖南临湘市为试验区进行验证，结果表明：WRF中尺度数值模式适用于山洪灾害降雨预报，其WSM6云微物理过程、Grell-Devenyi ensemble对流参数化方案和YSU边界层参数方案对致洪山洪暴雨过程模拟较好，基于频率(或面积)匹配的降水偏差订正方法能显著改善模式降水预报中雨量和雨区范围的系统性偏差，大数据分析方法应用于雷达估测降雨能显著提高准确度。在此研究基础上，提出了基于高空和地面大气探测、数值预报模式的0~72 h短期定量降雨预报和基于雷达的0~2 h临近定量降雨预报的多元信息耦合预报方法，将对致灾山洪的预报时间由2 h延长至72 h，并可显著提高了山洪预报的精度。 The precipitation forecasts from radar and satellite cloud pictures (with a lead time of 0 - 2 h), from upper air and ground surface atmospheric sounding (with a lead time of 0 - 24 h) and from numerical forecasting mode (with a lead time of 0 - 72 h) have different forecast effect in different forecast pe-riods. Considering the short lead time and low accuracy forecast for early warning of mountain tor-rents, study on the numerical forecasting model, the model’s optimized combination of parameters, data assimilation, correcting the errors on forecasted rainfall in flash-flood affected region and an op-timized calculating method based on big data for radar monitoring precipitation. Linxiang City of Hu-nan Province was selected as experimental area and the results show that the WRF numerical forecast is more suitable for the forecast of disasters including mountain torrents. Among them, WSM6 cloud microphysical process, Grell-Devenyi ensemble convective parameterization scheme and YSU boun-dary layer parameter configuration scheme can greatly simulate the process of mountain torrents and storm causing a flood, precipitation deviation correction method based on frequency (or area) match can considerably improve the systematic deviation in the rainfall and scope of rain area in the preci-pitation forecast, the big data analysis method used in radar rainfall forecast can remarkably improve the accuracy. A forecasting disastrous flash floods method has been proposed by coupling upper air and ground surface atmospheric sounding, the 0 - 72 h short-term quantitative precipitation forecasts (QPF) from numerical forecasting mode and the 0 - 2 h nowcast QPF based on radar, which helps to increase the lead time of flash flooding forecasts from 2 h to 72 h and enhance significantly the fore-casting accuracy.