【作者】 王玮；

【导师】 梁天刚；

【作者基本信息】 兰州大学 ， 草业科学， 2014， 博士

【摘要】 青藏高原地区不仅是我国三大积雪分布区之一,也是我国的主要牧区。由于该地区生产方式相对落后,经营粗放,基础设施极为薄弱,抵抗自然灾害的能力非常有限,从而导致冬春季大量的降雪经常引发区域性的灾害,严重制约着当地草地畜牧业的可持续发展。因此,利用遥感和GIS技术准确监测青藏高原牧区的积雪动态变化,深入开展雪灾预警及风险评价研究,对防灾减灾,维持草地畜牧业的可持续发展都具有极其重要的意义。本研究对2003～2010年青藏高原无云积雪制图算法、积雪覆盖率算法改进、雪深数据比较与反演、积雪对气候变化的响应、雪灾预警和风险评价进行了系统的研究。研究结果表明：1)本研究提出一种去云积雪图像合成算法,由此生成的青藏高原地区无云积雪分类图像(MA)在各种天气状况下积雪分类精度和总精度分别达到80.75%和97.52%。该图像不仅具有MODIS较高空间分辨率和AMSR-E不受天气状况影响的特点,而且可有效地提高积雪监测的时空分辨率。因此,MA积雪合成图像具有准确监测研究区内积雪覆盖范围的能力,这对深入研究雪灾频发地区的积雪动态变化,提供了有效可靠的数据支持。2)提出一种改进的积雪面积比例算法。与TM雪盖图相比,改进算法提取的雪盖面积标准误差由原来的0.35降低到0.22,平均绝对误差由0.25降到0.18,相关系数由0.74提高到0.85以上。3)加拿大气候中心发布的雪深数据在不同积雪深度条件下误差较高(RMSE达47.70cm),不适用于青藏高原地区积雪深度的监测。基于被动微波资料SSM/I和AMSR-E模拟的雪深不仅与台站实测数据之间的误差较小,而且与青藏高原地区积雪空间分布之间的一致性较好。4)2003～2010年青藏高原年平均温度、年降水量和积雪覆盖面积呈现出增加的趋势,年平均温度增加了0.72℃,年平均降水增加6.85mm,积雪覆盖面积增加5.75%。8年间永久积雪面积和雪深则呈现出减少的趋势,其中永久积雪面积以每年0.35%的速率在减少,8年共计减少2.80%;雪深减少约2.40%,雪水当量下降4.16%。5)8年间整个青藏高原地区有35.3%的区域积雪覆盖天数(SCD)和34.3%的区域雪水当量(SWE)呈现下降的趋势。温度和降水的空间分布对青藏高原地区SCD和SWE的空间分布具有明显的影响。SCD和SWE与年均温或年降水量存在空间相关关系的区域均在50%以上,在海拔6300m以下时相关系数均达到0.6以上。6)影响青藏高原牧区雪灾发生的关键因子有年雪灾概率、积雪覆盖天数、载畜力、日均温<-10。C的低温天数、草地掩埋指数、草地积雪覆盖率及畜均GDP。依据受灾程度及积雪对放牧牲畜采食影响情况,本项研究构建出一种牧区雪灾危害等级预警模型,制定出青藏高原地区雪灾预警分级标准,并提出一种基于格网单元的雪灾风险评价方法。根据青藏高原近3年(2008-2010年)积雪季(10-12月和翌年1-3月)各县(市)旬雪灾危害等级预警反演结果显示,雪灾危害等级预警模型总精度可达85.64%。7)在以上研究的基础上,从系统设计体系结构、数据库建设、系统功能模块设计等方面出发,设计并开发了基于ArcGIS Server和Flex技术的青藏高原牧区积雪监测与雪灾预警系统(http://snow.ecograss.com.cn/)。更多还原

【Abstract】 Tibetan Plateau (TP) is not only one of the three major snowfall regions, but also an important pastoral area in China. However, because of the undeveloped agriculture infrastructure in the region, heavy snow in winter or spring often causes death of a large number of livestock due to cold weather and forage shortage. This has a great negative influence on the sustainable development of grassland animal husbandry. Therefore, it is extremely important to monitor snow dynamic variation for improving the ability of preventing disaster and maintaining the sustainable development of grassland animal husbandry in pastoral areas.In this study, snow cover mapping algorithm of cloud removal, algorithm improvement of fractional snow cover, comparison and inversion of snow depth, the response of snow to climate change, early warning of snow-caused disaster and risk assessment were analyzed systematically on TP from2003to2010. The results show that:1) This study presents an improved cloud removal algorithm to produce a daily cloud-free snow cover product (MA). The MA composite images not only have the advantages of AMSR-E (i.e., unaffected by weather conditions) and MODIS (i.e., relatively higher resolution), but also the high snow and overall accuracies (i.e.,80.75%and97.52%, respectively), much higher than those of existing daily snow cover products in all sky conditions. Therefore, MA has the ability to accurately monitor the daily snow cover dynamics in the study area, which is extremely important to study snow-caused disasters and offer snow-covered data.2) An improved algorithm of fractional snow cover is presented. Compared with the TM-based snow cover map, the standard error and the mean absolute error of the improved algorithm are reduced from0.35to0.22, and0.25to0.18, respectively. The correlation coefficient is increased from0.74to0.85.3) The snow depth product issued by the Canadian Meteorological Centre is not suitable to monitor snow depth because of higher errors (RMSE=47.70cm) on the different snow depth condition in the TP region. The snow depth data based on the SSM/I and AMSR-E products are able to accurately monitor daily snow depth in the study area.4) The mean annual temperature(MAT), mean annual precipitation(MAP) and snow cover area(SCA) during2003-2010in the TP region have increasing trend, in which the MAT, MAP and SCA increased0.72℃,6.85mm and5.75%, respectively. The annual permanent SCA and snow depth have decreasing trend, in which the annual permanent SCA decreased by the rate of0.35%, and2.80%totally from2003to2010; the average snow depth and snow water equivalant decreased about2.40%and4.16%.5) The snow-covered days(SCD) and snow water equivalent(SWE) have decreasing trend from2003to2010, and the decreasing areas take up35.3%and34.3%of the total area in the TP region respectively. It shows that spatial distribution of temperature and precipitation has correlation relationship with SCD and SWE. About50%of study area shows correlation relationship among SCD, SWE, mean annual temperature and mean annual precipitation. The correlation coefficient is up to0.6on the areas where the elevation is below6300m.6) There are seven crucial factors for early warning of snow disasters on TP. They are mean annual probability of snow disaster, SCD, livestock stocking rate, continual days of mean daily temperature blow-10℃, grassland burial index, rate of snow-covered grassland and per livestock GDP. Based on snow-caused disaster magnitude and snow influence on grazing of livestock, this study develops a model for early warning of snow disasters on county basis and proposes a method of risk assessment of snow disasters at500meter resolution for pastoral areas of TP. We choose411cases from2008to2010to validate the predicting results from the developed early warning model. An overall mean accuracy of85.64%is reached in classifying snow disaster and no disaster.7) On the basis of the existed studies, considering the system structure, data organization and system function design, a management information system for snow monitoring and early warning of snow disasters in the TP region is designed and developed by use of ArcGIS Server and Flex.更多还原