基于天气系统自动识别的新疆牧区雪灾遥感监测与预警研究

Weather system recognition is very important to weather forecasting service, and is the theory prerequisite to the study of meteorological disaster early-warning system. However, the existing methods of weather system recognition were mainly based on ground weather observations data and expertise, and it's difficult to meet the timeliness, objectivity and accuracy demand of weather forecasting and disaster warning systems. Aiming at this problem, in this project, we will develop a novel weather system automatic recognition method based on satellite remote sensing, by using the 3S technique and machine learning method, and on this basis, develop a remote sensing monitoring and early warning system of snow disaster in pastoral area. The main studies include: (1) Using the Fengyun-4 remote sensing, ground weather observations data, numerical weather forecast and the CLDAS forcing data to achieve weather system automatic recognition with high spatial and temporal resolution. (2) Develop a novel remote sensing monitoring system of snow disaster, which can been using for dynamic monitoring the disaster environment, disaster drivers and disaster bearers of snow disaster. (3)Develop a multistage early warning model for snow disaster, and improve its effectiveness and accuracy by fusion of the weather system recognition, medium-range numerical weather prediction, short-range weather forecasting and the geography and society-economy background of Xinjiang pastureland. For the ultimate goal, this project will provide theoretical support for snow disaster monitoring and early warning in Xinjiang pastureland, improve the level of disaster prevention and mitigation, and gradually extended to other regions.

天气系统识别对天气预报业务具有重要意义，是研究气象灾害及其预警模型的理论前提。然而，目前的天气系统识别主要依靠气象站观测数据和专家经验，较难满足天气预报与灾害预警对时效性、客观性和准确性的要求。针对该问题，本项目结合3S技术与机器学习方法，实现基于卫星遥感的天气系统自动识别，并在此基础上发展高时空分辨率的牧区雪灾遥感监测与预警模型。内容包括：（1）利用风云四号卫星遥感资料、地面气象观测资料、数值预报以及CLDAS大气驱动场资料，实现天气系统遥感自动识别。（2）建立高时空分辨率雪灾遥感监测平台，动态监测雪灾相关致灾因子、承灾体以及孕灾环境。（3）结合新疆牧区自然地理与社会经济背景，根据天气系统、中期数值预报以及临近天气预报，建立分阶段雪灾遥感预警模型，从而提高牧区雪灾预警时效性。最终可为新疆牧区雪灾监测与预警工作提供理论支撑与应用示范，切实提高新疆牧区雪灾的防灾减灾能力，并向其他地区推广。

开展新疆典型牧区的积雪监测及雪灾预警研究能切实提高该区域的防灾减灾能力。本项目围绕雪灾遥感监测与预警问题，利用多源卫星观测资料，重点开展高时空分辨率积雪参数监测模型研究，和山区降水预报研究。项目研究期间，同时在新疆北部地区收集和开展地面积雪辐射观测、积雪参数的测量等工作，获取野外实测数据作为有效验证数据。本项目的主要研究成果如下：1)以雪粒径在流域尺度的分布为基础，在现有积雪遥感识别技术基础上，发展了基于雪粒径填补的积雪去云算法，提升了雪盖的空间分辨率；2)结合光学与微波遥感的观测优势，发展适用于北疆地区的降尺度雪深反演算法，将山区雪深数据集的空间分辨率提升到500 m，3)发展了基于雪粒径的降雪频次估计算法，提升对于高海拔地区降雪观测的能力；4)发展了基于静止卫星的降水预报算法，为山区雪灾预警提供高时间分辨率的数据参考；5)结合环准噶尔盆地积雪实测资料和北疆气象站历史资料，构建了较为完整的新疆北部积雪参数数据库；6)基于站点、遥感数据，从站点、区域和全国的尺度研判了积雪的时空分布及趋势变化。研究成果提高了区域积雪监测和雪灾预警的能力，为新疆牧区雪灾监测与预警工作提供理论支撑与应用示范。