Climate warming is constantly causing hydro-meteorological perturbations,especially in high-altitude mountainous regions,which lead to the occurrences of landslides.The impact of climatic variables(i.e.,precipitation ...Climate warming is constantly causing hydro-meteorological perturbations,especially in high-altitude mountainous regions,which lead to the occurrences of landslides.The impact of climatic variables(i.e.,precipitation and temperature)on the distribution of landslides in the eastern regions of the Himalayas is poorly understood.To address this,the current study analyzes the relationship between the spatial distribution of landslide characteristics and climatic variables from 2013 to 2021.Google Earth Engine(GEE)was employed to make landslide inventories using satellite data.The results show that 2163,6927,and 9601 landslides were heterogeneously distributed across the study area in 2013,2017,and 2021,respectively.The maximum annual temperature was positively correlated with the distribution of landslides,whereas precipitation was found to have a non-significant impact on the landslide distribution.Spatially,most of the landslides occurred in areas with maximum annual precipitation ranging from 800 to 1600 mm and maximum annual temperature above 15℃.However,in certain regions,earthquake disruptions marginally affected the occurrence of landslides.Landslides were highly distributed in areas with elevations ranging between 3000 and 5000 m above sea level,and many landslides occurred near the lower permafrost limit and close to glaciers.The latter indicates that temperature change-induced freeze-thaw action influences landslides in the region.Temperature changes have shown a positive correlation with the number of landslides within elevations,indicating that temperature affects their spatial distribution.Various climate projections suggest that the region will experience further warming,which will increase the likelihood of landslides in the future.Thus,it is crucial to enhance ground observation capabilities and climate datasets to effectively monitor and mitigate landslide risks.展开更多
针对大数据环境下随机森林算法存在冗余与不相关特征过多、特征子空间信息含量不足以及并行化效率低等问题,提出了结合增益率与堆叠自编码器的并行随机森林算法PRFGRSAE(parallel random forest algorithm combining gain ratio and sta...针对大数据环境下随机森林算法存在冗余与不相关特征过多、特征子空间信息含量不足以及并行化效率低等问题,提出了结合增益率与堆叠自编码器的并行随机森林算法PRFGRSAE(parallel random forest algorithm combining gain ratio and stacked auto encoders)。首先,提出了结合非线性归一化增益率和堆叠自编码器的降维策略DRNGRSAE(dimension reduction combining nonlinear normalization gain ratio and stacked auto encoders),通过过滤特征集中的冗余和不相关特征,并利用堆叠自编码器提取特征,有效减少了冗余以及不相关特征数;其次,提出了结合拉丁超立方抽样与归一化相关度的子空间选择策略SSLF(subspace selection strategy combining Latin hypercube sampling and feature class correlation),通过对特征集进行多层划分抽样,形成空间表达度较高的特征子空间,有效保证了特征子空间的信息含量;最后,提出结合可变动作学习自动机的reducer分配策略DSVLA(distribution strategy based on variable-action learning automata),使每个数据簇均匀分配到reducer进行处理,有效提高了并行化效率。实验结果表明,PRFGRSAE算法的加速比与准确度较IMRF、KSMRF和GAPRF算法都有显著提升,因此该算法应用于大数据处理,特别对包含较多特征的数据集有更高的精准度和并行效率。展开更多
基金supported by the Second Tibetan Plateau Scientific Expedition and Research(STEP)(2019QZKK0903)the National Natural Science Foundation of China(No.42071017)the science and technology research program of the Chinese Academy of Sciences'Institute of Mountain Hazards and Environment(No.IMHE-ZDRW-03).
文摘Climate warming is constantly causing hydro-meteorological perturbations,especially in high-altitude mountainous regions,which lead to the occurrences of landslides.The impact of climatic variables(i.e.,precipitation and temperature)on the distribution of landslides in the eastern regions of the Himalayas is poorly understood.To address this,the current study analyzes the relationship between the spatial distribution of landslide characteristics and climatic variables from 2013 to 2021.Google Earth Engine(GEE)was employed to make landslide inventories using satellite data.The results show that 2163,6927,and 9601 landslides were heterogeneously distributed across the study area in 2013,2017,and 2021,respectively.The maximum annual temperature was positively correlated with the distribution of landslides,whereas precipitation was found to have a non-significant impact on the landslide distribution.Spatially,most of the landslides occurred in areas with maximum annual precipitation ranging from 800 to 1600 mm and maximum annual temperature above 15℃.However,in certain regions,earthquake disruptions marginally affected the occurrence of landslides.Landslides were highly distributed in areas with elevations ranging between 3000 and 5000 m above sea level,and many landslides occurred near the lower permafrost limit and close to glaciers.The latter indicates that temperature change-induced freeze-thaw action influences landslides in the region.Temperature changes have shown a positive correlation with the number of landslides within elevations,indicating that temperature affects their spatial distribution.Various climate projections suggest that the region will experience further warming,which will increase the likelihood of landslides in the future.Thus,it is crucial to enhance ground observation capabilities and climate datasets to effectively monitor and mitigate landslide risks.
文摘针对大数据环境下随机森林算法存在冗余与不相关特征过多、特征子空间信息含量不足以及并行化效率低等问题,提出了结合增益率与堆叠自编码器的并行随机森林算法PRFGRSAE(parallel random forest algorithm combining gain ratio and stacked auto encoders)。首先,提出了结合非线性归一化增益率和堆叠自编码器的降维策略DRNGRSAE(dimension reduction combining nonlinear normalization gain ratio and stacked auto encoders),通过过滤特征集中的冗余和不相关特征,并利用堆叠自编码器提取特征,有效减少了冗余以及不相关特征数;其次,提出了结合拉丁超立方抽样与归一化相关度的子空间选择策略SSLF(subspace selection strategy combining Latin hypercube sampling and feature class correlation),通过对特征集进行多层划分抽样,形成空间表达度较高的特征子空间,有效保证了特征子空间的信息含量;最后,提出结合可变动作学习自动机的reducer分配策略DSVLA(distribution strategy based on variable-action learning automata),使每个数据簇均匀分配到reducer进行处理,有效提高了并行化效率。实验结果表明,PRFGRSAE算法的加速比与准确度较IMRF、KSMRF和GAPRF算法都有显著提升,因此该算法应用于大数据处理,特别对包含较多特征的数据集有更高的精准度和并行效率。