青藏高原牧区雪灾预警机制研究

The Tibetan Plateau (TP) region in China is not only a main pastoral area, but is one of China's three snow distribution centers as well. In winter and spring seasons the snow-caused disasers usually happen frequently, distribute widely and result in huge losses. Meanwhile, it is the most serious natural hazard in the TP region. Early warning is not only a difficult point in the snow hazard research, but also a key to snow disaster prevention and reduction. Conducting early warning research of snow disasters in pastoral areas on the plateau is not only an important scientific problem, but also is a great strategic requirement of national disaster prevention and reduction, which has a wide application prospect in the TP region. Therefore, taking the pastoral area in the south of Qinghai Province as a study area, in accordance to the characteristics of animal husbandry production and snow disaster prevention, by use of "3S" technologies (RS:Remote Sensing, GIS:Geographic Information System, GPS:Global Positioning System) and ecology, grassland science, animal science, catastrophology and other interdisciplinary theories and methods, focusing on the meteorological hazard factors, hazard-affected bodies and pastoral preventention capability, this project is mainly to: (1) study their interaction relationships and their correponding processes with early warning factors to clarify the early warning mechanism of pastoral snow disasters; (2) screen out the key factors for early warning of snow disasters, establish early warning identification model and classification standard at county level and risk diagnosis mode and classification standard on 1km grid cell scale combining with short- and medium-term weather forecast and social economic information; and (3) develop an early warning system based on "3S" and network technologies, as well as multi-sources information integration, and put forward snow disaster prevention and mitigation plans, which provides a scientific basis for decision-making of snow disaster early warning management in the pastoral areas of the TP region.

青臧高原地区是我国重点牧区，也是我国三大积雪分布中心之一。该区雪灾频发，分布广泛，损失惨重，雪灾是冬春季最为严重的自然灾害；雪灾预警是积雪危害研究的难点和防灾减灾的关键。开展高原牧区雪灾预警研究，不仅是一个重大的科学问题，而且是国家防灾减灾的重大战略需求，具有广阔的应用前景。因此，本项研究以青海南部牧区作为研究区，针对草地畜牧业生产与雪灾防御特点，利用3S技术与生态学、草地学、畜牧学、灾害学等学科交叉的理论和方法，围绕气象致灾力、承灾体和牧区抗灾力，研究三者间的互作关系及其与雪灾预警因子之间的联系过程，阐明牧区雪灾预警机理；结合短中期天气预报和社会经济等信息，筛选关键预警指标，构建基于县域尺度的牧区雪灾预警判别模型和标准，以及基于格网单元空间尺度的风险度诊断模式和分级标准；研发基于3S和网络等技术和多源信息集成的牧区雪灾预警系统，提出防灾减灾决策预案，为牧区雪灾管理决策提供科学依据。

雪灾是青臧高原牧区冬春季最为严重的自然灾害，而雪灾预警则是积雪危害研究的难点和防灾减灾的关键。因此，开展青藏高原牧区雪灾预警研究，具有重要的意义。本项研究利用3S技术与生态学等学科交叉的理论和方法，主要开展了雪灾监测与关键气象致灾因子、致灾力和承灾体及抗灾力互作关系、雪灾预警机制等3个方面的研究。在6个方面取得重要进展：1）创建了一种高精度的基于机器学习（Machine learning）算法的多因素草地生物量动态反演模型； 2）提出一种基于无人机（UAV）技术和MODIS数据的大范围草地植被盖度遥感监测方法；3）系统研究了近10多年以来青南牧区草地载畜力及草畜平衡变化动态；4）发展了一种逐日去云积雪覆盖范围（SCA）卫星遥感图像合成算法，并分析了近10多年以来青藏高原积雪时空变化特征；5）构建出一种基于Logistic回归分析算法的雪灾综合风险评价方法，创建了一种基于多因素（平均温度、春季雪灾概率、雪灾综合风险率、均温<-5℃的持续天数和积雪覆盖率）和反向传播（Back Propagation，BP）人工神经网络（Artificial Neural Networks，ANN）的雪灾预警方法和流程；6）研发出包括基础数据、致灾力、承灾体、抗灾力、雪灾预警、抗灾预案等模块的基于WebGIS技术的青海牧区积雪监测与雪灾预警系统。发表相关学术论文29篇，其中SCI论文12篇（1区4篇，2区6篇），EI论文1篇，国际会议论文1篇，获实用新型专利授权4项，获发明专利授权1项，获计算机软件著作权2项，出版专著1部。这些研究成果，可以为青海牧区大范围草地植被生长状况、草地载畜力及草畜平衡状况的准确评估，冬春季家畜出栏、饲料贮备等草畜管理的科学决策，以及高时效性的积雪空间变化监测及雪灾预警提供科学基础，为积雪危害的科学防治提供基本的信息服务。