摘要
基于大数据分析,研究中国降水量和土地利用/土地覆被类型的时空演化特征,并构建基于XGBoost的地形气候交互模型,预测暴雨灾害的发生及成灾风险.数据来源于中国气象局、国家地理信息数据中心等公开数据集,涵盖1990至2020年的降水、气温、地形等多项环境指标.研究结果表明:1)1990至2020年间,中国年均降水量表现出显著波动性,整体趋势未呈现单调增加或减少,但空间分布上呈现出明显的南北差异,东南沿海地区降水量显著高于西北内陆地区,南北降水差异最大可达50%;2)土地利用变化方面,耕地面积从1990年的58.2%下降至2020年的53.6%,林地面积从19.4%上升至23.7%,湿地面积从0.9%减少至0.6%,反映了退耕还林等环保政策的实施效果。基尼系数分析表明,耕地的空间集中度逐年上升,而林地的分布趋于均匀;3)通过XGBoost模型分析,降水量是影响暴雨形成的最重要因素,其权重占比为32%,地形高度和坡度也对暴雨的形成有显著影响。模型在测试集上的准确率达到92.4%,验证了其预测暴雨事件的可靠性.
Based on big data analysis,this paper studies the spatio-temporal evolution char-acteristics of precipitation and land use/land cover types in China,and constructs a terrain climate interaction model based on XGBoost to predict the occurrence and risk of rainstorm disasters.The data is sourced from publicly available datasets such as the China Meteorolog-ical Administration and the National Geographic Information Data Center,covering multiple environmental indicators such as precipitation,temperature,and terrain from 1990 to 2020.The research results show that:1)From 1990 to 2020,the average annual precipitation in China showed significant fuctuations,and the overall trend did not show a monotonic increase or decrease,but there were significant north-south differences in spatial distribution.The pre-cipitation in the southeast coastal area was significantly higher than that in the northwest inland area,and the maximum north-south precipitation difference could reach 50%;2)In terms of land use change,the cultivated land area decreased from 58.2%in 1990 to 53.6%in 2020,the forest land area increased from 19.4%to 23.7%,and the wetland area decreased from 0.9%to 0.6%,reflecting the implementation effect of environmental protection policies such as returning farmland to forests.The Gini coefficient analysis shows that the spatial concentration of cultivated land has been increasing year by year,while the distribution of forest land tends to be uniform;3)Through XGBoost model analysis,precipitation is the most important factor affecting the formation of rainstorm,with its weight accounting for 32%.Terrain height and slope also have a significant impact on the formation of rainstorm.The accuracy of the model in the test set reaches 92.4%,which verifies the reliability of its prediction of rainstorm events.
作者
李鸿霄
张怀念
杨洁
王浩懿
LI Hong-xiao;ZHANG Huai-nian;YANG Jie;WANG Hao-yi(School of Safety Engineering,Beijing Institute of Petrochemical Technology,Beijing 102627,China;Zhiyuan College,Beijing Institute of Petrochemical Technology,Beijing 102627,China;School of Information Engineering,Beijing Institute of Petrochemical Technology,Beijing 102627,China;School of Mechanical Engineering,Beijing Institute of Petrochemical Technology,Beijing 102627,China)
出处
《数学的实践与认识》
北大核心
2025年第10期98-108,共11页
Mathematics in Practice and Theory
关键词
基尼系数
极端天气预测
暴雨成灾
XGBoost模型
时空演化分析
gini coefficient
extreme weather forecast
rainstorm disaster
XGBoost model
spatiotemporal evolution analysis
