基于多源遥感数据与随机森林模型的辽宁省作物生长季干旱监测研究  

Drought Monitoring in Crop Growing Season in Liaoning Province Based on Multi-source Remote Sensing Data and Random Forest Model

作　　者：王荣莉 Wang Rongli(School of Geographical Sciences,Liaoning Normal University,Dalian 116029,China)

机构地区：[1]辽宁师范大学地理科学学院,辽宁大连116029

出　　处：《黑龙江科学》2023年第18期16-19,共4页Heilongjiang Science

摘　　要：以辽宁省为研究区,基于MODIS、TRMM等遥感数据及气象站点数据,利用随机森林算法,构建作物生长季干旱监测模型,并验证监测结果的准确性。结果表明,模型模拟值与实测值基本一致,根据实测值与模拟值对干旱类别进行统计分析,干旱等级一致率为70.96%。各干旱表征因子中,作物水分胁迫指数(CWSI)与降水状态指数(PCI)的相对重要性最高,其次是土壤湿度状况指数(SMCI)、植被状态指数(VCI)及温度状态指数(TCI)。利用此模型对辽宁省2014年与2017年的干旱事件进行监测,结果与实际干旱发生程度及空间分布基本一致,表明此模型能很好地反映作物干旱的空间分布与演变过程,可应用于区域干旱监测。Through taking Liaoning Province as the study area,based on MODIS,TRMM and other remote sensing data and meteorological station data,the study constructs a drought monitoring model for crop growing season through random forest algorithm,and verifies the accuracy of the model monitoring results.The results show that the simulated values of the model are basically consistent with the measured values.According to the statistical analysis of drought categories based on the measured and simulated values,the consistency rate of drought levels is 70.96%.Among the various drought characterization factors,the relative importance of crop water stress index(CWSI)and precipitation state index(PCI)is the highest,followed by soil moisture state index(SMCI),vegetation state index(VCI),and temperature state index(TCI).This model is used to monitor drought events in 2014 and 2017 in Liaoning Province,and the results are basically consistent with the actual drought occurrence degree and spatial distribution,indicating that the model could well reflect the spatial distribution and evolution process of crop drought,and could be applied to regional drought monitoring.

关 键 词：干旱 生长季 多源数据 随机森林 辽宁省 

分 类 号：K903[历史地理—人文地理学]