雪盖制图在气象、水文和全球变化研究以及流域管理应用中发挥着重要的作用。在2008年中国南方大雪期间，由于云层遮盖使可见光-近红外遥感失效。被动微波能够穿透云层获得地面积雪信息，然而常规的被动微波监测积雪算法仅适用于干雪，而中国南方地区以湿雪、浅雪为主。因此，本文采用被动微波遥感技术，通过使用HUT (Helsinki University of Technology)积雪辐射模型模拟积雪亮温，使用SNTHERM (SNow THermal Model)积雪过程模型模拟积雪参数变化，并与河北栾城站的实验观测进行比较，研究了湿雪辐射特征以及冻融循环造成的亮温昼夜变化特征。采用2008年1月1日-2月20日大雪期间卫星观测的AMSR-E (Advanced Microwave Scanning Radiometer - EOS)被动微波遥感数据开展了积雪和无雪覆盖地物的辐射特征分析。根据中国南方气象站点的雪深和地面温度观测、IMS (Interactive Multi-sensor Snow and Ice Mapping System)雪盖、MODIS (Moderate Resolution Imaging Spectroradiometer)雪盖比例以及中国土地利用图提取了低矮植被区和森林区的融土、冻土、湿雪、干雪共八个地物类别，研究了不同地物类别的单波段亮温、频率差、极化差和昼夜亮温差特征。此研究的基础上，建立了积雪判识决策树算法。算法首先以积雪散射强烈的降轨数据(当地时间夜晚过境)开展判识，然后使用升轨数据(当地时间白天过境)做进一步判识。由于考虑到89 GHz亮温受大气影响，建立了使用和不使用89GHz亮温的两种算法。在决策树的每一步采用SPSS (Statistical Product and Service Solutions)统计软件的分类回归树(CRT, Classification Regression Tree)模块帮助选择分类的亮温指标及其阈值。新算法使用2008年站点观测雪深、站点所在AMSR-E像元内的IMS和MODIS雪盖比例判断积雪状况(有雪或无雪)一致的所有数据做点尺度上的验证，使用IMS雪盖做区域上的验证。结果表明：从点上的验证结果看，使用89GHz亮温的算法在低矮植被区和森林区的总体判识精度分别为91.3%和88.6%，不使用89GHz亮温的算法分别为87.4%和85.7%。低矮植被区的积雪判识效果优于森林区。从区域上与IMS雪盖比较的结果看，两种算法的平均总体精度为94%左右，最小精度达到85%以上。其中，使用89GHz亮温的算法能够在研究区偏南方的区域判出更多的积雪。为了挖掘算法在后续应用中的潜力，文章还尝试在积雪判识前采用前一天的数据对当天AMSR-E亮温的缺失做分升降轨道的插补，提高了积雪判识范围的连续性，增大了积雪判出率。本文的研究工作在完善训练数据集和地面真值上还有改进的空间。

Snow mapping is of great importance in meteorological, hydrology and global change researches and watershed management applications. In the heavy snow event in the south of China in 2008, the visible-near infrared remote sensing failed to map snow cover due to the existence of thick clouds. Passive microwave can penetrate the clouds to acquire ground snow information. However, traditional passive microwave snow detection algorithm could only be used for dry snow, and the snowpacks in the south of China are mostly wet, shallow snow.Therefore, the passive microwave remote sensing technology is used in this paper. By using the HUT (Helsinki University of Technology) snow emission model to simulate snow brightness temperatures and the SNTHERM (SNow THermal Model) snow process model to predict snow parameters, and comparing the simulations with experimental observations measured at Luancheng, Hebei Province, the wet snow emission properties and the day-night change of brightness temperatures caused by frozen-thaw circles are studied.The satellite-observed AMSR-E (Advanced Microwave Scanning Radiometer-EOS) passive microwave brightness temperatures from January 1st to February 20th is used to study the emission properties of the snow-covered and snow-free terrain during the snow event in 2008. Eight land surface types such as the thaw soil, frozen soil, wet snow and dry snow in the sparse-vegetated region and the forest-covered region, respectively, are extracted from the daily meteorological site dataset, according to the site-observed snow depth (SD) and ground temperatures, the IMS (Interactive Multi-sensor Snow and Ice Mapping System) snow cover, the MODIS (Moderate Resolution Imaging Spectroradiometer) snow fraction and the Chinese land use map. The single band brightness temperatures, the frequency differences, the polarization differences and the day-night brightness temperature differences of each type are analyzed. Based on this study, snow detection decision tree algorithms are built. The algorithms first use the brightness temperature in descending orbit (with local overpass time at night) when the scattering of snowpacks is stronger to do snow detection, then use the data in ascending orbit (with local overpass time in daytime) to map snow. Considering the atmospheric effect on the brightness temperature at 89 GHz, both the algorithms using and not using the 89 GHz brightness temperatures are established. At each step of the decision tree, the CRT (Classification Regression Tree) module in the SPSS (Statistical Product and Service Solutions) software is used to help choose the best classification brightness temperature criteria and its threshold.The new algorithms use all meteo-site data in 2008 which have the same snow-covered or snow-free conditions judged by the site-observed snow depth, IMS and MODIS snow fraction in the AMSR-E pixel on the sites to do validation on the scale of point, and use the IMS snowcover to do validation on the scale of region. Results show that, on the scale of point, the algorithm using the 89 GHz brightness temperatures has a total accuracy of 91.3% and 88.6% in the sparse-vegetated region and the forest-covered region, respectively, while the total accuracies of the algorithm not using the 89 GHz brightness temperatures for the two regions are 87.4% and 85.7%, respectively. The detection performance in the sparse-vegetated region is better than the forest-covered region. Comparing with the IMS snowcover on the scale of region, the average total accuracies of the two algorithms are both about 94%, and the lowest accuracies are above 85%. Of the two algorithms, the algorithm using the 89 GHz brightness temperatures can extract more snow cover in the southern part of the study region.In order to expand the application potential of the algorithm, the paper also tries to fill the missing of the brightness temperatures in the descending or the ascending orbit with the AMSR-E data on the previous day in the same type of orbit before snow detection. Results show that continuity of the detected snow region is improved and the snow found-out rate is increased. The research work in this paper can be improved by increasing the representativeness of the training dataset and the accuracy of the ground-truths in future studies.