顾及GF-3影像纹理特征的洪涝范围提取方法

洪涝灾害给社会、经济造成巨大损失,及时、快速监测洪涝范围在抗灾救灾中具有重要意义。合成孔径雷达(SAR)由于其主动式微波成像的机理,可为全天时、全天候、大范围洪涝灾害监测提供支持。本文首先以高分三号(GF-3)卫星影像为数据源,基于灰度共生矩阵(GLCM)、局部二值模式(LBP)等6种纹理描述方法提取138个SAR影像纹理特征;然后利用随机森林(RF)指标重要性评估功能,筛选出重要性得分较高的纹理特征进行水体信息提取;最后结合数学形态学对初始水体提取结果进行后处理,评估安徽巢湖附近区域洪涝灾害。试验表明,本文方法的水体提取精度优于传统阈值法(Otsu)及分类算法(KNN和SVM),可有效提取洪涝灾害的影响范围,为选取合适的SAR影像纹理特征进行洪涝范围快速监测提供参考。     

基于随机森林模型的岸线不透水面提取方法研究

以南京长江沿岸保护区域为研究对象,利用2016年和2019年两期Sentinel-2影像构造多种遥感指数;再利用随机森林模型进行不透水面信息提取。结果表明,2016年的总体精度为88.25%,Kappa系数为0.76;2019年的总体精度为90.51%,Kappa系数为0.80,说明该方法具有较好的精度,能准确反映不透水面的空间分布。结合提取结果进行时空变化分析发现,2016-2019年南京长江沿线不透水面整体呈减少趋势,总面积减少2.784 km²,年均减少0.928 km²,其中大部分是由工矿用地转化为绿地以及港口码头消失导致的。长江沿线的不透水面提取和变化检测分析有助于全面、细致和真实地监测河道岸线情况,对于南京市开展长江岸线研究以及开发长江岸线资源和可持续利用具有举足轻重的意义。     

青藏高原及横断山区地质灾害易发区空间格局及驱动因子

青藏高原及横断山区位中国西南部,地貌类型复杂,地质灾害频发,严重威胁到西南地区人民生命财产安全,尤其影响进藏交通线的通达度。系统研究该区地质灾害易发区空间格局及驱动因子,可为青藏高原及横断山区地质灾害风险防范及地质灾害的防灾减灾救灾提供重要理论支撑。研究结果表明,基于随机森林构建的判识模型对于灾害点的判识精度均高于80%,甚至达到91%,可准确模拟与预测研究区各分区的地质灾害点。研究区地质灾害易发点主要分布在横断山区南部与东北部以及青藏高原中南部地区,且以小型及中型规模地质灾害为主（占比为87%）。综合分区Ⅰ~Ⅲ的地质灾害易发区面积分别为17.5万km²、17.4万km²与27.5万km²。各综合分区地质灾害驱动因子研究表明,横断山区南部区域（综合分区Ⅰ内）小型及中型地质灾害的主要驱动力为道路建设导致的沿途坡面稳定性变化（贡献率为20.2%）;横断山区东北部区域及青藏高原地区（综合分区Ⅱ~Ⅲ）小型及中型地质灾害的主要驱动力为植被覆盖状况的变化对于坡面稳定性的影响（贡献率分别为23.6%与27.3%）。此外,综合分区Ⅱ~Ⅲ内影响小型及中型地质灾害空间格局的第二个驱动因子为道路建设导致的沿途坡面稳定性的变化（贡献率为15.7%）与河流对于周边坡面的侵蚀作用（贡献率为17%）。     

基于随机森林的山西省柳林县黄土滑坡空间敏感性评价

基于随机森林模型,以GF-6影像和ALOS DEM数据为基本信息源,结合高程,地形起伏度及地形湿度等11项因子,对山西省柳林县进行滑坡敏感性空间区划。模型精度评价表明：随机森林模型精度为0.75,支持向量机模型精度为0.7,表明随机森林更适合柳林县的滑坡敏感性评价。指标重要性分析结果表明：高程、坡度、距道路距离以及距河流距离,是影响柳林县滑坡发育的主要因素。敏感性空间区划结果表明：高度敏感区约占柳林县总面积的28%,主要分布在三川河流域的南北边界及邻近区域内,其中贾家垣乡分布面积最广。从时间成本、训练难度、稳定度以及精确度考虑,随机森林模型更适合滑坡敏感性评价这类非线性计算问题。     

基于随机森林算法的草原地上生物量遥感估算方法研究

草原是我国面积最大的陆地生态系统,生物量是反映生态系统质量和功能的关键指标,准确地掌握草原生物量对草原资源合理利用、生态修复、畜牧业高质量发展都具有重要的意义和作用。本研究以内蒙古锡林郭勒盟为研究区,利用高分一号遥感卫星影像,结合216个野外样本数据,采用随机森林算法（Random Forest,RF）对草原地上生物量（Aboveground Biomass,AGB）遥感估算进行了适用性分析与应用。在运用随机森林算法的过程中,进行了K-折交叉验证、多元共线性诊断、偏效应等一系列分析,完成了随机森林模型的构建,同时,将建模结果与其它模型进行了对比,最终实现了锡林郭勒盟草原AGB的反演估算。结果表明：① 随机森林算法能够较好地规避生物量建模中自变量多元共线性的问题;② 随机森林模型在草原AGB估算中较其它模型具有更好的适用性,模型精度达85%,RMSE为202.13 kg/hm²;③ 应用构建的随机森林算法估算了研究区2017年草原AGB,从结果来看,其空间分布上呈现为自东向西逐渐递减的趋势;从草地类型上看,山地草甸类AGB单产最高,温性草原类总产量最高。研究结果将对草原生态系统监测评估和草原宏观管理具有一定的参考价值。     

ANALYSIS OF BIOLOGICAL GEOCHEMISTRY OFCHEMICAL ELEMENTS IN Betula ermanii FORESTIN CHANGBAI MOUNTAINS,CHINA

匠he 乙hangbal Mountains Nature Reserve Is a al，1982ZHUANGetal，1989；HUANGetal，1982；state－grade nature rese。e，also Is a component part of LIU，1993；LIU，1988； ZHU et al，1991）．Most ofreserve net of International blosphere．its area Is large，their research works focus on the chemical structure andthe natural environment and ecosystem lsprotected material cycling In alpine tundra zone andcompletely．the natural elements and ecological envl－broadleatKorean pine fores…     

Transit time evaluation using a chloride concentration input step shift after forest cutting in a Japanese headwater catchment

Mean transit times were estimated for a small headwater catchment in Japan (the Fukuroyamasawa Experimental Watershed) using the step shift in input chloride (Cl^?) concentrations that occurred immediately after an episode of forest clear‐cutting. Measured Cl^? concentrations in stream water began to decrease immediately after clear‐cutting, and this trend continued for 6 years. Before clear‐cutting, the input Cl^? concentrations were controlled by wet and dry deposition processes, and most of the dry Cl^? deposition was collected by the forest canopy and reached the ground as throughfall and stemflow. After clear‐cutting, dry deposition was no longer collected by the canopy in this way, thus causing a sharp decrease in input Cl^? concentrations. By comparing measured Cl^? concentrations in stream water with estimates based on the input and evaporative Cl^? concentrations, it was shown that the decrease in stream water Cl^? concentrations was caused mainly by this step shift in the Cl^? input. It was proposed that the change in Cl^? concentrations after forest cutting could be used to represent the replacement of ‘old’ water that existed before cutting by ‘new’ water that was supplied after cutting. The breakthrough curve for the new water fraction gave an approximately exponential distribution of transit times in flow‐corrected time. The mean flow‐corrected transit time was estimated as 1068 days (runoff: 3497 mm). It was therefore concluded that the step change in input Cl^? concentrations immediately following forest clear‐cutting could be successfully used to estimate transit times for the entire catchment. Copyright © 2009 John Wiley & Sons, Ltd.     

RS AND GIS-BASED FOREST FIRE RISK ZONE MAPPING IN DA HINGGAN MOUNTAINS

1INTRODUCTIONForests are one of the major natural resources that per-form important environmental and recreational func-tions (COSTANZA and GROOT, 1997). It is well known that forests can absorb atmospheric carbon, maintain a certain degree of humidity in atmosphere, regulate rainfall, moderate temperature, and restrain soil erosion, etc. (AURELIA, 2003). So, the health of a forest in any given area is a very important indicator of the ecological conditions. But fire is the great…     

Streamflow response to Pinus plantation harvesting: Canobolas State forest,southeastern Australia

Streamflows were measured in two Pinus radiata plantation catchments and one native eucalypt forest catchment in Canobolas State forest from 1999 to 2007. In 2002/2003, clearfall harvesting of 43·2 and 40·3% of two plantation catchments occurred, respectively. Water yields increased by 54 mm (52%), 71 mm (35%) and 50 mm (19%) in the first three years post‐harvest in treated catchment A and by 103 mm (118%), 157 mm (82%) and 119 mm (48%) in treated catchment B relative to the native forest control catchment. In the fourth post‐harvest water year annual rainfall was only 488 mm, which resulted in negligible run‐off in all catchments, regardless of forest cover. In both plantation catchments, monthly streamflows increased significantly (p = 0·01, p < 0·001) due to a significant increase in baseflows (p < 0·001) after harvesting. Monthly stormflows were not significantly affected by harvesting. Flow duration curve analyses indicated a variable response between the two plantation catchments. Treated catchment A was converted from an ephemeral stream flowing 42% of the time pre‐harvest to a temporary stream flowing 82% of the time post‐harvest. These changes occurred throughout all seasons of the year but were most pronounced during summer and autumn when baseflows were maintained post‐harvest but were not observed under native forest or mature pine plantations. By contrast, flow duration increased in treated catchment B from 12% of the time pre‐harvest to 38% of the time post‐harvest with the greatest changes measured during the winter and spring months when streamflow would normally occur under native forest conditions. These observations have important implications for the development of models of plantation water use to be utilized in water resource planning in Australia. Copyright © 2009 John Wiley & Sons, Ltd.     

Soil partitioning and surface store controls on spring runoff from a boreal forest peatland basin in north‐central Manitoba,Canada

The mid‐ to high‐boreal forest in Canada occupies the discontinuous permafrost zone, and is often underlain by glaciolacustrine sediments mantled by a highly porous organic mat. The result is a poorly drained landscape dominated by wetlands. Frost‐table dynamics and surface storage conditions help to control runoff contributions from various landscape elements, hydrological linkages between these elements, and basin streamflow during spring snowmelt. Runoff components and pathways in a forested peatland basin were assessed during two spring snowmelts with contrasting input and basin conditions. Runoff from relatively intense melt (up to 16 mm day^?1) on slopes with limited soil thawing combined with large pre‐melt storage in surface depressions to produce high flows composed primarily of meltwater (78% of the 0·29 m³ s^?1 peak discharge) routed over wetland surfaces and through permeable upper peat layers. Melt intensity was less in the subsequent year (maximum of 10 mm day^?1) and active layer development was relatively greater (0·2 m deeper at the end of spring melt), resulting in less slope runoff. Coupling of reduced slope contributions with lower storage levels in basin wetlands led to relatively subdued streamflows dominated by older water (73% of the 0·09 m³ s^?1 peak discharge) routed through less‐permeable deeper peat layers and mineral soil. Interannual differences in runoff conditions provide important insight for the development of distributed hydrological models for boreal forest basins and into potential influences on biogeochemical cycling in this landscape under a warming climate. Copyright © 2001 John Wiley & Sons, Ltd.