【作者】 黄晓东；

【导师】 任继周； 梁天刚；

【作者基本信息】 兰州大学 ， 草业科学， 2009， 博士

【摘要】 北疆地区是我国三大积雪分布中心之一，也是新疆主要的畜牧业基地。由于区内地形复杂，海拔高、气候寒冷潮湿，冬、春季雪灾频繁，大批牲畜因雪灾而死亡，积雪灾害成为主要的自然灾害之一，严重影响着草地畜牧业的可持续发展。而遥感技术是监测积雪覆盖范围的最有效的方法。本论文的设计就是利用遥感与地理信息系统技术，利用MODIS和AMSR-E资料，分析北疆地区积雪覆盖范围和积雪深度的时空变化特征，建立北疆地区雪灾监测与预警基础数据库，对准确监测北疆地区积雪时空动态变化状况，评价牧区雪灾受灾程度，快速提供救援对策，减少灾区经济损失，具有重要的科学意义和实际应用价值。本论文以北疆为研究区，利用气象台站记录的雪情数据和云量数据，验证了每日积雪产品MOD10Al和每8日积雪合成产品MOD10A2的分类精度；利用MOD10Al产品开发出用户自定义积雪覆盖范围合成新算法及产品，并对新产品进行了精度验证；研发出基于MODIS每日雪被数据MOD10Al和AMSR-E／Aqua每日雪水当量产品AE_DySno新的合成积雪产品MOD-AE1，并验证了其精度；建立了北疆地区基于AMSR-E亮温数据的雪深反演模型，并对模型的精度进行了评价；利用用户自定义旬积雪合成产品，分析了北疆地区积雪覆盖范围的时空动态变化、积雪覆盖范围与海拔以及气温的关系以及不同积雪覆盖面积的天数，研究积雪覆盖面积的变化特征，更好地解释了北疆地区积雪分布的规律。研究结果表明：1)在晴空的状态下，MOD10Al积雪产品的分类精度达到94.6％，总精度为95.5％。说明MODIS积雪制图算法具有很高的积雪分类精度，但受到云的严重制约；积雪深度和土地覆盖类型是影响积雪分类精度的两个重要因素，其中雪深是最重要的因子。MOD10A2产品可较好地消除云层对地表积雪分类精度的影响，平均积雪识别率达87.5％，可较好地反映地表积雪的分布状况。虽然8日积雪合成产品可以有效地去除大部分云的污染，但是仍然无法去除研究区所有的云像元。2) MOD10A2产品基本上可以满足从事冰雪及相关学科的较大空间尺度上的科学研究，但较长的合成时期、固定的合成起算时间和依MOD10Al获取的时间顺序进行合成的方法，不利于对区域积雪灾害事件进行快速有效的实时监测和评价。在本项研究中，以旬为基本合成周期，对90个时相的MOD10Al积雪产品用2～11天分别进行了合成分析，发现合成时间越长，合成图像的积雪识别精度越高，如以旬为时间段，合成图像的精度介于70～100％，平均精度可达87.2％；在旬内8日合成图像的精度也介于70～100％，平均精度为80.6％。3)基于MOD10Al和AE_DySno的新合成算法充分结合了MOD10Al有较高空间分辨率和AE_DySno数据具有全天候成像的优点，能够有效地改进每日积雪范围监测的精度。合成图像MOD-AE1的总一致性为76.1％，积雪分类精度达75.4％。同AE_DySno产品相比较，合成图像的积雪多测误差略有增加，漏测误差则减少约8％，平均总精度和积雪分类精度分别提高6.6％和8.8％。温度和雪深是影响积雪覆盖率变化的主导因素。不同时期的每日合成图像MOD-AE1可以用于分析季节性积雪的不同特征，合成图像适合北疆牧区雪灾期间每日雪被动态监测的需要。4)雪深反演模型受气温、融雪、降雨、湿雪、深霜层等因素的严重影响。对大于2.5 cm的积雪深度与AMSR-E亮温数据的相关性分析表明，积雪深度与垂直极化18GHz和36GHz波段的亮温差之间具有很强的线性相关性，相关系数达0.65，雪深回归模型公式为SD=0.49(Tb18V-Tb36V)+8.72。5)积雪面积的增减随气温的变化而变化，气温降低，积雪覆盖面积开始增加，气温升高，积雪面积开始减少，气温与积雪面积之间存在着很强的相关性，但是当气温低于零下15摄氏度时，由于此时积雪几乎覆盖了研究区，此时气温的降低对雪盖面积的增加贡献较少。更多还原

【Abstract】 In general, the northern Xinjiang is one of three major snow distribution regions, andis also an important pastoral area in China. Massive snow accumulation frequently causesdisasters such as frost-bite and death of a large number of grazing animals, and destroytraffic and telecommunication devices. Therefore, monitoring snow-covered extentprecisely plays a significant role in the dynamic studies and preventing of snow disastersin pastoral areas.In this Ph.D. dissertation, an great effort was made to systematically study theMODIS snow mapping algorithm, the MODIS snow cover composite products, the snowdepth model based on passive microwave remote sensing data AMSR-E, and the temporaland spatial variation of snow cover area and snow depth in Northern Xinjiang.The main results may be concluded as follows:1) The snow mapping agreement between MODIS daily snow maps and surfaceobservations is high at 94.6% over the four snow seasons under clear sky conditions. Theomission errors mainly determined by snow depth and land cover types, especially whensnow depth is less than 3 cm, the MODIS snow cover mapping algorithm intends tomisclassify thin and patchy snow as land. The cloud agreement is 95.9%, andapproximately 4.1% cloud is misclassified as snow when the sky view at climate stationswas completely covered by clouds.2) Basically, MOD10A2 products can satisfy snow and related subject researches on alarge spatial scale. However, the sequential composite approach in terms of the timeseries of receiving MODIS data, with a longer composite period and a given compositestarting date, lacks of flexibilities, which is not advantageous to the efficient and effectivemonitoring and estimates for regional snow-caused disasters. In this study, wecomposited a new 2-11 days composite snow products, and the snow classificationaccuracy is between 70～100%, the average accuracy reach to 87.2%.3) A new daily snow cover product was developed through combining MODIS dailysnow cover data and AMSR-E daily snow water equivalent (SWE) data. By takingadvantage of both high spatial resolution of optical data and cloud transparency ofpassive microwave data, the new daily snow cover product greatly complements thedeficiency of MODIS product when cloud cover is present especially for snow coverproduct on a daily basis and effectively improves daily snow detection accuracy. In ourexample, the daily snow agreement of the new product with the in situ measurements at20 stations is 75.4%, which is much higher than the 33.7% of the MODIS daily product in all weather conditions, even a little higher than the 71% of the MODIS 8-day product(cloud cover of～5%). The new snow cover product can better and effectively capturedaily SCA dynamics during the snow seasons, which plays a significant role in reduction,mitigation, and prevention of snow-caused disasters in pastoral areas.4) Through regression analysis of horizontal, vertical polarization brightnesstemperature difference of 18 GHz and 36 GHz band and snow depth value, the snowdepth model was established based on the AMSR-E brightness temperature data innorthern Xinjiang. At the same time, the accuracy of the model was evaluated. The resultsindicated that the remote sensing model is impacted seriously by temperature, snowmelt,rain, wet snow and deep frost layers. And there is a good correlation between snow depth(y) over 2.5 cm and the vertical polarization brightness temperature difference of 18GHzand 36GHz. The equation is SD=0.49(Tb18V-Tb36V)+8.72, and the correlationcoefficient is up to 0.65. However, the accuracy of the model is lower when the surface iscovered by fallow or deep snow. Basically, the model can reflect the trend of snow depthvariation in Northern Xinjiang, but it has a low accuracy, and needs to be improved in thefuture.5) The air temperature and elevation play important roles in the fractional snow coveredarea and the spatial distribution of snow cover differed greatly in varied areas. It showedmore snow accumulation in the mountainous areas than that in the plain areas, and themountainous areas had a longer snow period than the plain area.更多还原