暴雨数值预报模式热启动关键技术研究

Now not neglected is the model SPIN-UP in numerical forecast, which still exist even with explicit cloud microphysics schemes used in high-resolution numerical models. Due to observational limitations, initial cloud hydrometeors, like cloud water, cloud ice, rain water, snow, and among others, are set to be zero at the beginning of model integration, which is called model cold start. Numerical errors are usually very large in the first several hours of model integration because cloud hydrometeors need several hours to be generated in the model integration. According to the particular geography and weather background in China, this program is planned to carry out studies on the cloud-analysis theory and method, with heavy rainfall processes of special cloud features in South China. Multiple data, like satellite remote sensing, radar detections, GPS water vapor, micrometers, surface and sounding observations and among others, will be used to test the macro- and micro- features of clouds. Besides, the contributions and roles of different observational data will be evaluated in the cloud analysis, so as to obtain cloud analyzed products more reasonably and scientifically. Furthermore, research on the hot start methodology of heavy-rain numerical model will be conducted. Four-dimensional variational method will be adopted to adjust model initial filed, including dynamic, thermodynamic and cloud fields, and to make them in concert with each other during the model integration. Lastly, efforts will be tried to eliminate the quick attenuation of hot start impacts in model integration due to the inharmonious among model dynamic, thermodynamic and cloud fields, to diminish the model SPIN-UP, and to improve short-time numerical forecast of clouds and precipitation.

目前制约数值预报的SPIN-UP现象不容忽视。尽管现代高分辨数值模式大多已采用显式云方案，但SPIN-UP现象仍存在，其原因在于受观测限制，大气云水、云冰、雨水、雪等凝结水物质含量大都采用零初值法（即冷启动），冷启动产生出符合实际的云需要一定的时间，使得模式积分初期云和降水误差很大。本项目根据中国特定地理区域和天气气候背景，利用卫星、雷达、GPS、微波辐射计、地面、探空等多源探测资料检验云的宏、微观特征，开展云分析方案在我国南方暴雨过程中的适用性分析和参数优化，弄清不同探测资料在云分析中的贡献和作用，以便更加科学合理地获得云分析产品；同时开展暴雨数值模式热启动方法研究，采用四维变分同化方法调整模式动力、热力和云初值场使其相互协调，尝试解决由于模式动力、热力和云初值场不协调带来的热启动影响很快被抹掉的问题，以减轻SPIN-UP现象，提高云和降水短时临近数值预报准确率。

针对制约数值预报的spin-up现象，本项目开展了模式热启动技术研究；为提高热启动预报效果，又针对背景误差协方差（B）开展了统计和试验研究；最后进行了资料同化试验，主要完成了以下研究内容：（1）LAPS云分析产品在我国南方典型暴雨过程中的适应性分析；（2）云信息在数值模式初值中的应用研究；（3）变分调整和RUC技术在动力、热力和云初值场同时调整的热启动技术中的应用；四维变分同化技术研究；（4）B矩阵的统计与预报检验；（5）冷、热启动不同初值方案模拟的背景场误差样本对B矩阵及其同化预报效果的影响研究；（6）微波辐射计、GPS PW、双雷达反演风场三维变分同化研究。主要研究进展如下：（1）LAPS云分析产品适用于我国南方典型暴雨过程，强度略偏强；（2）初值中增加云水、雨水等信息，模式积分初始阶段就产生出与实况相近的强降水，有效缓解了模式spin-up问题，其降水预报改进最明显的时段在前6 h；（3）采用变分调整方案获得相互协调的动力、热力和云的初值场，在降水TS评分上未表现出优势，但降水偏差改善明显。采用RUC技术获得相互协调的动力、热力和云的初值场，对提高强降水短临数值预报水平有重要作用。建立了多源资料GSI-DRP4DVar四维变分同化系统，为今后进一步开展动力、热力和云初值场同时调整的热启动技术研究奠定了基础。（4）不同B对同化影响显著，RUC系统采用本地化B后分析的初值更趋合理，因而显著改进了降水预报。（5）由冷、热启动两种不同初值方案模拟的背景场误差样本统计得到的B存在较大差异，并对同化和预报产生较显著的影响。RUC系统采用由热启动方案模拟的背景场误差样本统计的B，对“苏迪罗”台风路径和降水预报效果较好，其中对台风路径的影响主要在积分24 h之后。（6）微波辐射计、GPS PW、双雷达反演风场资料的三维变分同化均能改进降水数值预报效果。本项目取得的主要研究成果为：在国内外学术期刊上发表相关科技论文10篇，其中SCI论文2篇，二级核心期刊论文6篇，国内会议论文2篇。