【作者】 张学通；

【导师】 陈全功； 梁天刚；

【作者基本信息】 兰州大学 ， 草业科学， 2010， 博士

【摘要】 季节性雪被对水文过程和气候具有重要的作用。积雪覆盖面积的动态变化对水体和能量循环以及社会经济和生态环境均具有重大的影响。而且积雪融水是干旱、半干旱地区生态系统的重要水源。冬春季降雪是制约我国牧区畜牧业发展的重要因素。积雪不仅会掩埋牧草,造成畜牧草料供应不足,而且在没有饲草储备或储备不足的牧区,造成大批家畜因冻饿而死亡的情况,从而发生“雪灾”。青海草原辽阔,草原面积3647×104ha,其中可利用草地面积3161×104ha,是我国六大牧区之一。雪灾是青海牧区冬春季节的主要自然灾害,每年10月至次年4月这一时段,青海牧区玉树、果洛、黄南南部、海南南部、祁连山等地区极易出现局地或区域的强降雪天气过程,加之气温较低,积雪难以融化,时常造成大雪封山、冻死、饿死牲畜,使牧区人民生命财产遭受巨大损失。应用遥感技术动态监测积雪的覆盖面积和反演积雪深度对牧区雪灾监测与评价具有重要意义。本文首先利用可见光MODIS每日积雪产品具有较高空间分辨率(500m)和被动微波AMSR-E SWE不受云影响的特点,合成具有较高积雪识别率的用户自定义每日积雪产品,用于监测青海积雪覆盖面积；其次用被动微波AMSR-E BT每日产品,结合气象地面台站数据,反演青海积雪深度；然后利用MODIS EVI数据结合地面实测,反演青南牧区草地地上生物量；最后依据前面研究得到的积雪覆盖面积、积雪深度、草地载畜量结合青南牧区其他数据库,对建立青南牧区雪灾预警及风险评估模型作方法性研究,以期为青海积雪监测和青南牧区雪灾预警提供科学依据。通过本文的研究,可以得出如下结论：1)以青海省为研究区,用MODIS/Terra和MODIS/Aqua的每日积雪产品合成新的积雪影像(MOYD10A1),该影像的图像云量比MOD10A1、MYD10A1减少了大约14%。青海用户自定义合成的积雪影像MOYD10A1在各种天气条件下的积雪分类精度为43.7%(MOD10A1的积雪分类精度为43.5%,MYD10A1的积雪分类精度为27.7%)。用MOYD10A1和AMSR-E SWE用户自定义合成积雪影像MAl0A1的积雪分类精度为54.5%,陆地分类精度为89.2%,总分类精度(雪+陆地)为85.5%。2)利用2003-2008年5个积雪季(10-3月)降轨AMSR-E 18和36 GHz波段的水平和垂直极化亮温数据,结合43个气象台站的实测雪深、日最高温等气象数据,分析青海省影响雪深反演的因子,发现湿雪、融雪、大型水体、深霜层等因素严重影响雪深模型的建立,其中深霜层的影响最大。根据剔除不合理数据后的639对有效样本,用不同的亮温差和实测雪深值作回归分析,建立青海省基于AMSR.E亮度温度数据的雪深反演模型SD=0.43(Tb18V—Tb36V)+2.06,用2002-2003年10-3月104对雪深大于3cm的样本,对得到的线性回归模型的反演精度进行评价。结果表明,公式反演的雪深值和实测值之间的各项误差均比较小,可以用于监测研究区雪深。3)以青南牧区为研究区,利用2006年8月和2007年8月的植被指数(MODIS-EVI)和草地地面样方实测值,建立草地地上生物量遥感反演模型,y=297.15e4.9492x(R2=0.5626,N=396)。用2008年草地地面样方实测值代入该公式得出的估测值对比,平均误差为21.06%,其精度为78.94%。依据陈全功的关键场理论,计算出2006年-2008年青南牧区14个县的草地地上生物量、采食牧草总量、理论载畜量、冬春场载畜量和夏秋场载畜量。4)从草地畜牧业角度考虑青南牧区雪灾情况,利用遥感和地理信息技术结合本文第二章、第三章、第四章的结果及研究区统计资料,对青南牧区的积雪、草地、家畜三者之间的关联性进行分析,建立雪灾预警分级指标,最后通过对雪灾发生前的草地抗灾力和家畜承灾体的监测,建立青南牧区的雪灾预警判别模型和风险评估模型,对青南牧区的雪灾预警判别模型和风险评估模型作了方法性研究。更多还原

【Abstract】 Seasonal snow cover has an important meaning for hydrological process and climate change. The dynamic change of snow cover area influences aquatic system, energy cycle, social economy and ecological system significantly. In addition, snowmelt is primary water source for arid and semiarid ecological system. Winter-spring snowfall is an important restrictive factor of animal husbandry development. Snow cover has some harmful effects, for example, burying forage, shortage of forage. Moreover, snow cover maybe cause snow disaster, because of freezing a large number of livestock to death in the pastoral area which have no or lack of forage storage. In this research, the study area is Qinghai, holding a large rangeland with 3647×104 ha and available pasture with 3161×104 ha, which made the pasture of Qinghai be one of the five six largest pastures in China. Snow disaster is the main natural hazards. In the period from October to April of the next year, because of the series snowfall and low temperature, heavy snow cover and freezing and starving of the livestock to death always happen in Yushu, Guoluo, the south part of Huangnan and Qilianshan, which make local herdsmen lost a lot property or even life.Snow cover area monitor and snow depth inversion have important implications for snow hazard monitor and evaluation in the pasture. In this study, the combination of MODIS daily images for snow product (with the resolution of 500m) and passive microwave images (AMSR-E SWE) which are not affected by clouds are combined into user-defined daily snow images with higher resolution firstly. Then, based on the passive microwave daily images (AMSR-E TB) and climate data, snow depth of Qinghai is inverted. Finally, using the combination of snow cover area and depth which is obtained in format researches, grazing capacity of the pasture and some other databases, the snow disast early warming and risk evaluation models were built. This study provided a scientifical basis for snow cover and snow disaster monitoring and estimation.The results of this study indicated that:1) As the study area of Qinghai Province, the cloud cover of the combined image (MOYD10A1) of the MODIS/Terra and MODIS/Aqua daily image for snow product is about 14 percent less than that of MOD10A1 image or MYD10A1 image. For all the climate, the classification accuracy of the user-defined images (MOYD10A1) is 43.7%, compared with that of MOD10A1 (43.5%) and that of MYD10A1 (27.7%). The snow classification accuracy of MA10A1, composited by MOYD10A and AMSR-E SWE, is 54.5%, and the land classification accuracy and overall classification accuracy is 89.2% and 85.5%, respectively.2) The factors which affect the snow depth inversion is analyzed base on the AMSR-E 18 and 36 GHz data during the snow seasons (October to March, from 2003 to 2008) and the sampling data (including snow depth and daily maximum temperature) of 43 meteorological stations. The result shows that the factors like snowmelt, large water body and especially deep frost layer influence the building of snow depth model seriously. According to the 639 pairs data excluding the unreasonable records, the snow depth inversion model (SD= 0.43(Tb18V-Tb36V)+ 2.06) was built based on the regressional analysis of various brightness temperature difference data and snow depth samples. Then the 104 pairs data (the snow depth is over 3cm) from October to March in 2002-2003 was used to evaluate the linear regression model. The result of this evaluation reveals that the obtained inversion model can be used to monitor snow depth of the study area Based on MODIS-EVI and aboveground biomass samples of August 2006 and August 2007 in the south part of Qinghai, the aboveground biomass inversion model (y= 297.15e4.9492x, R2=0.5626, N=396) was built. According to the aboveground biomass samples of 2008, average error of the model is 21.06%. Based on the key pasture theory, the aboveground biomass, total consuming forage, theoretical capacity, winter-spring pasture capacity and summer-autumn capacity of 14 counties are calculated and analylized through 2006 to 2008 in the southern pasture of Qinghai.4) Based on remote sensing, GIS, statistics of the study area and the results of chapter 2, chapter 3 and chapter 4, the relationship among snow cover, rangeland and livestock was analyzed. And the snow disaster warning grade index was put forward. Finally, the snow disaster warning model and the risk assessment model were built based on monitoring the anti-calamity ability of pasture and livestock.更多还原