基于多类支持向量机的化学物质生态危害分类研究

应用多类支持向量机(M-SVMs)方法研究了化学物质生态危害程度的分类,以提高分类的准确性和效率.对采集到的61种环境优先污染物的环境行为和生物毒性方面的7项指标进行相关性分析,建立了M-SVMs分类模型并对数据集进行10折交叉验证以评价模型的分类能力,运用所建模型对7种化学物质的生态危害进行预测.结果表明,去除与鱼毒有信息重叠的溞毒指标,选取鱼毒、藻毒、降解性、蓄积性、分配系数和吸附系数6项指标用于构建M-SVMs分类模型;M-SVMs模型识别率较高,交叉验证平均分类正确率达86.89％;对7种化学物质生态危害的预测结果与实际情况基本相符.     

荒地的概念、分类及其生态功能解析

随着我国对生态环境保护日益重视,荒地作为生态用地的一种重要类型,其保护与开发已成为重要的研究课题。论文在梳理国内外荒地概念、厘清荒地与相关概念的关系以及系统分析中国、美国、印度土地分类的基础上,重新定义了我国荒地的概念,并针对我国国情将荒地划分为三种类型：荒野地、未利用地和废弃地。论文还首次提出荒地具有防风固沙、保育土壤、固碳释氧、涵养水源、保育生物多样性等多种生态功能的理念,这对我国荒地以及生态用地的开发与保护有重要意义。最后提出建议在我国现行土地分类体系中直接增列一类“生态用地”,包括林地、草地、荒地、水域、滩涂等;在未来开发与保护荒地过程中,从观念上重视荒地的价值,建立荒地产权制度,编制自然资源资产负债表,加强对荒地的分类保护与开发。     

松嫩平原西部湿地自然保护区保护有效性遥感分析

以松嫩平原西部国家级湿地自然保护区为研究对象,以Landsat MSS/TM/OLI遥感影像为基础信息源,综合运用最小均方差、变异系数、景观指数、信息熵模型、地理信息系统空间分析和景观发展强度指数,研究保护区保护有效性变化及其影响因素.结果表明:松嫩平原西部国家级湿地自然保护区保护有效性大致呈递减趋势;1987~2013年人类活动对保护区的干扰强度不断增加,且与保护区保护有效性呈显著负相关;耕地和草地为湿地转出的主要类型,分别占总转出面积的63.71%和21.69%;湿地自然保护区保护有效性的降低是自然环境变化和人类活动影响共同作用的结果,自然环境变化是内因,而人为活动干扰,尤其是开垦造田和水利修建加速了湿地退化.最后,论文就研究中存在的不足以及保护区管理建设中应关注的问题进行了阐述.     

利用TM图像逐级分类提取非点源信息

探讨了运用逐级分类从TM图像中提取与非点源污染有关的土地利用信息的方法.研究表明,逐级分类与一般的最大似然度分类相比效果更佳,具有简化复杂问题、突出研究重点和方法灵活等优点,能够快速有效地提取所需要的信息.     

深度学习在裸地扬尘源监测中的应用研究

裸地是扬尘的重要来源,施工建设过程中形成的裸地极易在大风天气作用下造成扬尘污染。因此,快速、有效地定位裸地位置,并确认其管控措施落实情况,对于开展裸地扬尘源监管具有重要意义。基于高分辨率遥感监测数据,结合人工解译裸地扬尘源数据集,以北京市大兴区为例,利用深度学习方法对裸地和防尘网覆盖裸地进行分类识别。同时,利用颜色匹配法对大兴区防尘网覆盖裸地进行识别,横向评估深度学习方法的识别精度。结果显示：深度学习方法对防尘网覆盖裸地的识别精度达97%,对裸地的识别精度达61%;颜色匹配法对防尘网覆盖裸地的识别精度达85%。防尘网覆盖裸地的颜色特征鲜明,深度学习方法和颜色匹配法对防尘网覆盖裸地的识别精度都在85%以上。深度学习方法对于面积大于2 000 m²的图斑有着较好的识别精度。深度学习方法可以提高裸地遥感解译的效率,实现规范化图像识别,可以作为人工判读的辅助手段。在实际应用中,可通过进一步积累样本来增强模型性能。深度学习方法适用于裸地扬尘源线索快速发现、工地防尘网措施落实情况快速检测等场景。     

城市生活垃圾智能分类技术研究与应用进展

城市生活垃圾清运量逐年增加,由此带来的环境问题日益严峻.不同垃圾组分有其各自适宜的处理与利用方式,因此,有效的垃圾分类技术具有重要意义.我国现有的以人工为主的垃圾分类方式存在效率低、成本高、准确性差等弊端,而垃圾智能分类技术是解决相关瓶颈的重要手段.本文简述了中国代表性城市生活垃圾分布和近年变化趋势,从密度、电、磁、声...     

Hydrologic Landscape Characterization for the Pacific Northwest,USA

We update the Wigington et al. (2013) hydrologic landscape (HL) approach to make it more broadly applicable and apply the revised approach to the Pacific Northwest (PNW; i.e., Oregon, Washington, and Idaho). Specific changes incorporated are the use of assessment units based on National Hydrography Dataset Plus V2 catchments, a modified snowmelt model validated over a broader area, an aquifer permeability index that does not require preexisting aquifer permeability maps, and aquifer and soil permeability classes based on uniform criteria. Comparison of Oregon results for the revised and original approaches found fewer and larger assessment units, loss of summer seasonality, and changes in rankings and proportions of aquifer and soil permeability classes. Differences could be explained by three factors: an increased assessment unit size, a reduced number of permeability classes, and use of smaller cutoff values for the permeability classes. The distributions of the revised HLs in five groups of Oregon rivers were similar to the original HLs but less variable. The improvements reported here should allow the revised HL approach to be applied more often in situations requiring hydrologic classification and allow greater confidence in results. We also apply the map results to the development of hydrologic landscape regions.     

基于多源遥感影像融合的武汉市土地利用分类方法研究

准确高效的获取土地利用信息,对于合理利用和开发土地资源具有十分重要的意义。在快速城镇化地区,土地利用活动频繁且密集,土地利用格局演变十分剧烈,增加了城市土地利用精准分类的不确定性;且受环境气候和云雨天气影响增加了有效光学影像获取的难度。为提高城市土地分类精度,该文选取武汉市中心城区为研究案例,以Sentinel-1A和Landsat8 OLI影像为数据源,采用Gram-Schmidt变换方法进行影像融合,选取最大似然、支持向量机、CART决策树、BP神经网络等4种分类方法对融合的影像进行分类,提取了研究区土地利用信息,并对其进行分析。进一步,通过与光学影像的分类结果对比,探究了Sentinel-1A和Landsat8 OLI融合影像在土地利用信息提取方面是否具有优势。研究结果表明:(1)对比其他3种方法,CART决策树分类方法对于融合后的影像分类精度最高,总体分类精度和Kappa系数分别达到88.55%和0.841 4;(2)与光学影像相比,Sentinel-1A和Landsat8 OLI融合影像可以更有效地获取高精度城市土地利用信息;(3)基于多源遥感影像融合的CART决策树分类方法是获取研究区高精度土地利用信息的一种行之有效的技术手段。研究成果可为快速城镇化区域的土地利用分类提供参考。     

Revealing the Diversity of Natural Hydrologic Regimes in California with Relevance for Environmental Flows Applications

Alterations to flow regimes for water management objectives have degraded river ecosystems worldwide. These alterations are particularly profound in Mediterranean climate regions such as California with strong climatic variability and riverine species highly adapted to the resulting flooding and drought disturbances. However, defining environmental flow targets for Mediterranean rivers is complicated by extreme hydrologic variability and often intensive water management legacies. Improved understanding of the diversity of natural streamflow patterns and their spatial arrangement across Mediterranean regions is needed to support the future development of effective flow targets at appropriate scales for management applications with minimal resource and data requirements. Our study addresses this need through the development of a spatially explicit reach‐scale hydrologic classification for California. Dominant hydrologic regimes and their physio‐climatic controls are revealed, using available unimpaired and naturalized streamflow time‐series and generally publicly available geospatial datasets. This methodology identifies eight natural flow classes representing distinct flow sources, hydrologic characteristics, and catchment controls over rainfall‐runoff response. The study provides a broad‐scale hydrologic framework upon which flow‐ecology relationships could subsequently be established towards reach‐scale environmental flows applications in a complex, highly altered Mediterranean region.     

Food alert patterns for metal contamination analyses in seafoods: Longitudinal and geographical perspectives

The aims of this study are to explore metal contamination in food with respect to variations in: i) the levels found over a period of eleven months and/or ii) by food types; and iii) to report patterns in reporting between countries. The frequencies of reported cases of metal ion contamination were ≤ 25% of total food alerts with notable increases between June and October. The alerted food products were dominated by seafood, where the majority were due to Mercury contamination and were made by EU member states with Italy making 57 of the 102 reports. The contaminated seafoods originate in thirty two countries with only Spain producing over 10 contaminated seafood products with a total of 47. In conclusion, this analysis highlights the main detector and transgressor nations along with the seafood and metal types that lead to alerts owing to high levels of contamination.