针对基于机器学习的滑坡易发性评价中非滑坡样本选取不规范导致的分类精度较低问题,本文提出联合基于密度的噪声应用空间聚类(Density-Based Spatial Clustering of Applications with Noise,DBSCAN)采样策略和支持向量机(Support Vector Machine,SVM)分类方法的DBSCAN-SVM滑坡易发性评价模型。首先,基于DBSCAN聚类和空间分析选取非滑坡样本;然后,将样本数据代入SVM分类模型进行训练与验证,预测并提取SVM分类中属于滑坡的概率,获得滑坡易发性;最后,以四川省绵阳市为试验区,预测滑坡易发性概率,基于滑坡易发性精度与分级结果等要素,与传统非滑坡样本采集策略的SVM滑坡易发性评价模型进行对比,并结合实际情况对DBSCAN-SVM模型评价结果进行分析。研究结果表明,相比传统SVM滑坡易发性评价模型,本文提出的DBSCAN-SVM滑坡易发性评价模型在高易发区和极高易发区中包含的滑坡样本数量较多,准确率、召回率、AUC、F1分数均得到提高,精度较高。 相似文献
A simple grid cell‐based distributed hydrologic model was developed to provide spatial information on hydrologic components for determining hydrologically based critical source areas. The model represents the critical process (soil moisture variation) to run‐off generation accounting for both local and global water balance. In this way, it simulates both infiltration excess run‐off and saturation excess run‐off. The model was tested by multisite and multivariable evaluation on the 50‐km2 Little River Experimental Watershed I in Georgia, U.S. and 2 smaller nested subwatersheds. Water balance, hydrograph, and soil moisture were simulated and compared to observed data. For streamflow calibration, the daily Nash‐Sutcliffe coefficient was 0.78 at the watershed outlet and 0.56 and 0.75 at the 2 nested subwatersheds. For the validation period, the Nash‐Sutcliffe coefficients were 0.79 at the watershed outlet and 0.85 and 0.83 at the 2 subwatersheds. The per cent bias was less than 15% for all sites. For soil moisture, the model also predicted the rising and declining trends at 4 of the 5 measurement sites. The spatial distribution of surface run‐off simulated by the model was mainly controlled by local characteristics (precipitation, soil properties, and land cover) on dry days and by global watershed characteristics (relative position within the watershed and hydrologic connectivity) on wet days when saturation excess run‐off was simulated. The spatial details of run‐off generation and travel time along flow paths provided by the model are helpful for watershed managers to further identify critical source areas of non‐point source pollution and develop best management practices. 相似文献
A series of confirmed and suspected dammed palaeo‐lake sedimentary successions is scattered within the middle Yarlung Tsangpo valley in Tibet. However, the chronology, the genesis of the dam and its location, the water level of the dammed lake, the process of dam failure and the spatiotemporal relationships between the sedimentary successions remain controversial. Here, we focus on one sedimentary succession of the suspected dammed palaeo‐lake at Xigazê. We measured the grain‐size distribution, magnetic susceptibility, organic and inorganic carbon content, and δ13Corg and δ15Ntotal ratios of the sediments. In addition, we measured the δ18Oshell and δ13Cshell values of modern and fossil Radix sp. shells, and the δ18Owater and δ13CDIC values of the ambient water with different hydrological regimes. The results indicate that the δ18Oshell values of modern Radix sp. and the δ18Owater of the ambient water body significantly depend on its hydrological status. In addition, a strong positive relationship was observed between δ18Oshell values of modern Radix sp. shells and the δ18Owater of the ambient water on the Tibetan Plateau. According to this correlation, the δ18Owater values of the palaeo‐water body are reconstructed using the δ18Oshell values of Radix sp. fossil shells in the Xigazê section. Further, based on the δ18Oshell values of fossil Radix sp., the reconstructed δ18Owater of the palaeo‐water body and the specific habitats of Radix sp., we infer that the sedimentary succession in the Xigazê broad valley was mainly formed within the backwater terminal zone of a dammed palaeo‐lake and that the elevation of the water level of the lake was approximately 3811 m a.s.l. AMS 14C dating indicates that the deposits of the dammed palaeo‐lake were formed at about 33–22 cal. ka BP. Finally, the presence of Radix sp. fossil shells within the Xigazê section suggests that Radix sp. survived the late Last Glacial Period on the Tibetan Plateau. 相似文献
Regarded as an effective method for treating the global warming problem, carbon emissions abatement (CEA) allocation has become a hot research topic and has drawn great attention recently. However, the traditional CEA allocation methods generally set efficient targets for the decision-making units (DMUs) using the farthest targets, which neglects the DMUs’ unwillingness to maximize (minimize) some of their inputs (outputs). In addition, the total CEA level is usually subjectively determined without any consideration of the current carbon emission situations of the DMUs. To surmount these deficiencies, we incorporate data envelopment analysis and its closest target technique into the CEA allocation problem. Firstly, a two-stage approach is proposed for setting the optimal total CEA level for the DMUs. Then, another two-stage approach is given for allocating the identified optimal total CEA among the DMUs. Our approach provides more flexibility when setting new input and output targets for the DMUs in CEA allocation. Finally, the proposed approaches are applied for CEA target setting and allocation for 20 Asia-Pacific Economic Cooperation economies.