基于MODIS数据的决策树分类方法研究与应用

介绍了目前国际上流行的两种决策树算法———CART算法与C4·5算法,并引入了两种机器学习领域里的分类新技术———boosting和bagging技术,为探究这些决策树分类算法与新技术在遥感影像分类方面的潜力,以中国华北地区MODIS250m分辨率影像进行了土地覆盖决策树分类试验与分析。研究结果表明决策树在满足充分训练样本的条件下,相对于传统方法如最大似然法(MLC)能明显提高分类精度,而在样本量不足下决策树分类表现差于MLC;并发现在单一决策树生成中,分类回归树CART算法表现较C4·5算法具有分类精度和树结构优势,分类精度的提高取决于树结构的合理构建与剪枝处理;另外在决策树CART中引入boosting技术,能明显提高那些较难识别类别的分类准确率18·5%到25·6%。     

基于分类回归树分析的遥感影像土地利用／覆被分类研究

以专家知识和经验为基础,综合影像光谱信息和其他辅助信息进行分类的基于知识的遥感影像解译方法,是提高遥感影像分类精度,实现自动解译的有效途径之一。然而,知识的获取一直是其得以广泛应用的“瓶颈”问题。以江苏省江宁试验区土地利用/覆被分类为例,利用分类回归树分析(CART)从训练样本数据集中发现分类规则,集成遥感影像的光谱特征、纹理特征和空间分布特征进行分类实验,并与传统的监督分类和逻辑通道分类方法进行比较。结果表明,基于CART的分类方法的精度基本在80%以上,与另两种方法相比,有了较大的提高,而且该算法复杂性低,效率高。由此说明,利用CART算法构建决策树获取的分类规则是合理的。它可以快速、有效地获取大量分类规则,是促进基于知识的遥感影像分类方法在土地利用/覆被分类中广泛应用的一项有效手段。     

New snow metrics for a warming world

Snow is Earth's most climatically sensitive land cover type. Traditional snow metrics may not be able to adequately capture the changing nature of snow cover. For example, April 1 snow water equivalent (SWE) has been an effective index for streamflow forecasting, but it cannot express the effects of midwinter melt events, now expected in warming snow climates, nor can we assume that station-based measurements will be representative of snow conditions in future decades. Remote sensing and climate model data provide capacity for a suite of multi-use snow metrics from local to global scales. Such indicators need to be simple enough to “tell the story” of snowpack changes over space and time, but not overly simplistic or overly complicated in their interpretation. We describe a suite of spatially explicit, multi-temporal snow metrics based on global satellite data from NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) and downscaled climate model output for the U.S. We describe and provide examples for Snow Cover Frequency (SCF), Snow Disappearance Date (SDD), At-Risk Snow (ARS), and Frequency of a Warm Winter (FWW). Using these retrospective and prospective snow metrics, we assess the current and future snow-related conditions in three hydroclimatically different U.S. watersheds: the Truckee, Colorado Headwaters, and Upper Connecticut. In the two western U.S. watersheds, SCF and SDD show greater sensitivity to annual differences in snow cover compared with data from the ground-based Snow Telemetry (SNOTEL) network. The eastern U.S. watershed does not have a ground-based network of data, so these MODIS-derived metrics provide uniquely valuable snow information. The ARS and FWW metrics show that the Truckee Watershed is highly vulnerable to conversion from snowfall to rainfall (ARS) and midwinter melt events (FWW) throughout the seasonal snow zone. In comparison, the Colorado Headwaters and Upper Connecticut Watersheds are colder and much less vulnerable through mid- and late-century.     

煤田火区特征的土地覆盖分类方法 ——以乌达煤田火区为例

土地覆盖变化是土地分析与评价和生态环境变化预测的重要科学基础, 通过精确的土地覆盖分类方法 获取高精度的土地覆盖图是研究煤田火区生态环境变化的必要手段。本文以最大似然法、光谱角度法、面向对象 分类法和基于复合分区的分层分类法进行乌达煤田火区土地覆盖分类的方法研究。研究结果表明, 基于复合分区的 分层分类方法分类精度较高, 总体分类精度为92.97%, kappa 系数为0.9155。该方法通过基于地表热辐射特征、热 异常状况、地貌类型, 以及对生态系统扰动状况等的划分, 减少了地物信息的混淆度, 即通过提     

Snow cover monitoring in the Northern Patagonia Icefield using MODIS satellite images (2000–2006)

The snow cover of the Northern Patagonia Icefield (NPI) was monitored after applying the Normalized Difference Snow Index (NDSI) and the Red/NIR band ratio to 134 Moderate Resolution Imaging Spectroradiometer (MODIS) images captured between 2000 and 2006. The final results show that the snow cover extent of the NPI fluctuates a lot in winter, in addition to its seasonal behaviour. The minimum snow cover extent of the period (3600 km²) was observed in March 2000 and the maximum (11,623 km²) in August 2001. We found that temperature accounts for approximately 76% of the variation of the snow cover extent over the entire icefield. We also show two different regimes of winter snow cover fluctuations corresponding to the eastern and the western sides of the icefield. The seasonality of the snow cover on the western side was determined by temperature rather than precipitation, while on the east side the seasonality of the snow cover was influenced by the seasonal behaviour of both temperature and precipitation. This difference can be explained by the two distinct climates: coastal and continental. The fluctuations in the winter snow cover extent were more pronounced and less controlled by temperature on the western side than on the eastern side of the icefield. Snow cover extent was correlated with temperature R² = 0.75 and R² = 0.74 for the western and eastern sides, respectively. Since limited meteorological data are available in this region, our investigation confirmed that the change in snow cover is an interesting climatic indicator over the NPI providing important insights in mass balance comprehension. Since snow and ice were distinguished snow cover fluctuations can be associated to fluctuations in the snow accumulation area of the NPI. In addition, days with minimum snow covers of summer season can be associated to the period in which Equilibrium Line Altitude (ELA) is the highest.     

河西地区土地利用/覆盖变化驱动力研究

河西地区的土地利用／覆盖格局2000多年来发生了很大的变化。参考大量的数据与资料分析了河西地区土地利用覆盖变化的驱动力,结果表明,自然机制与人为机制同时改变着该地区的土地利用／覆盖状况,自然机制的驱动是缓慢的,长期的,而人为机制的驱动,尤其到了近现代,则是剧烈的,集中的,人口变化是驱动土地利用／覆盖变化的非常活跃的因素。     

江陵凹陷北部高速玄武岩下二维广角地震初探

江汉油田江陵凹陷北部高速玄武岩发育,利用常规地震资料很难得到高速玄武岩下的地震反射。分析研究表明,广角地震有利于玄武岩下的地震成像。通过综合正演和实际地震数据完成了广角地震资料的处理,获得了江陵凹陷北部高速玄武岩下的反射地震信息。在开展广角地震波场的识别和利用研究过程中,从正演模型出发,探讨了在不同地质条件下高速层对地震波场的影响,并针对靶区资料讨论了如何识别广角地震信息。在上述基础上,通过对实际地震资料做精细的分析和处理,最终完成了高速玄武岩下地震资料的成像。     

辽宁省积雪对自动气象站观测地气温差的影响

利用1971—2016年辽宁省61个气象站气温、地表温度、积雪日数和积雪深度资料,分析了积雪的保温作用及其对地气温差的影响。结果表明：更换自动站前后地表温度观测方式的差异导致地气温差显著增大,地气温差的增大程度受所在区域积雪日数、积雪深度的影响显著。在积雪期较长、积雪较厚的地区,积雪引起反照率增大,使得雪面温度降低,导致雪气温差减小,而雪的保温作用使得地气温差显著增大。因此,更换自动站前地（雪）气温差与积雪日数呈显著负相关,而更换自动站后地气温差与积雪日数呈显著正相关。各台站之间地气温差随积雪深度的变化系数差异较大,为0.045~0.858 ℃?cm^-1,在年平均积雪日数＜40 d、年平均极端积雪深度＜10 cm的区域,积雪的保温作用随积雪深度增大而显著增大;在年平均积雪日数＞40 d、年平均极端积雪深度＞10 cm的区域,10 cm以下的积雪对土壤保温作用随积雪深度增大显著,当积雪深度＞10 cm后,其保温作用随积雪深度增大的幅度明显减小。     

青藏高原积雪范围和雪深/雪水当量遥感反演研究进展及挑战

积雪是地表特征的重要参数,对辐射收支、气候和长期天气变化均有重要影响。雪本身又是一个重要的天气现象和水文气象参数,过量的降雪也会带来严重的雪灾,如牧区雪灾、雪崩和融雪洪水灾害等。因此对积雪的监测,尤其是对山区的积雪监测,具有多方面的意义。利用卫星遥感技术监测积雪已有50余年的历史,并已形成了系列业务产品。青藏高原平均海拔超过4 000 m,该地区的积雪具有重要的水文、气候和生态环境意义。由于地形复杂,人迹罕至,地面观测站点稀少,受较强太阳辐射的影响,积雪消融迅速、区域差异消融以及风吹雪等因素导致积雪分布破碎化严重,对使用遥感资料监测该地区的积雪造成的极大的困难和不确定性。随着国内外传感器技术的不断发展,光学和被动微波遥感数据的同步获取技术已经非常成熟,综合利用光学遥感数据高空间分辨率和被动微波数据不受云干扰的特点,结合机器学习、无人机等技术,将环境参数加入反演模型中,有助于提高青藏高原积雪参数反演精度。     

滇东南建水地区高镁火山岩包体的成因和构造背景

在滇东南建水地区发现产于峨眉山玄武岩中的高镁火山岩包体,这对于地幔柱的形成演化具有重要研究意义.对这些包体进行了锆石U-Pb年代学、地球化学和岩矿分析.高镁火山岩包体具斑状结构,致密块状构造,斑晶主要为贵橄榄石和透辉石.13颗锆石U-Pb LA-ICP-MS加权平均年龄为259±2Ma(MSWD=1.9),显示与寄主岩石同期形成.包体岩石具有高镁(Mg~#=68～75)、低硅(SiO_2=45.11%～45.93%)特征,轻稀土元素(LREE)、高场强元素(HFSE)富集而重稀土元素(HREE)亏损,属于亚碱性、拉斑玄武岩系列,具有板内玄武岩(IPB)特征.火山岩包体的原始岩浆起源于石榴子石二辉橄榄岩低程度部分熔融的产物,岩浆演化过程中发生了橄榄石和单斜辉石的分离结晶作用,在侵位上升过程中未受明显的地壳混染作用.该高镁火山岩的存在,显示地幔柱除了垂直上升运动外,在地球深部不同的边界还有多次侧向扩展移动,表明滇东南晚二叠世存在峨眉山地幔柱的一个分支-地幔枝活动.