| 基于机器学习模型的关中地区GPMIMERG降水数据订正方法 |
| |
| 引用本文： | 谢祥洲,刘军龙,霍斐斐,张相春.基于机器学习模型的关中地区GPMIMERG降水数据订正方法[J].水电能源科学,2022(2):6-9. |
| |
| 作者姓名： | 谢祥洲 刘军龙 霍斐斐 张相春 |
| |
| 作者单位： | 电子科技大学机械与电气工程学院;重庆城市管理职业学院;遵义师范学院资源与环境学院;遵义师范学院生物与农业科技学院（食品科技学院） |
| |
| 基金项目： | 重庆城市管理职业学院科研创新团队资助项目(KYTD202002)。 |
| |
| 摘 要： | 针对GPMIMERG降水数据存在系统误差的问题，以关中地区为例，在筛选海陆位置、地形、植被指数（NDVI）变量的基础上，运用支持向量机（SVM）、随机森林（RF）、高斯过程回归（GPR）模型对IMERG月降水数据进行订正，并通过34个站点数据验证订正模型。结果表明，关中地区IMERG数据具有良好的可替代性，其决定系数R2达0.76,平均绝对误差MMAE、均方根误差RRMSE分别为6.94、9.77 mm;经机器学习模型订正后星地数据之间的R2提升了2.05%～58.33%,RRMSE、MMAE分别降低了0.85%～71.23%、0.10%～73.47%;与GPR、RF模型相比，SVM模型的RRMSE、MMAE分别减小16.76%、9.76%和24.73%、14.11%,对IMERG数据订正具有更好的适用性。
|
| 关 键 词： | IMERG数据 订正 精度检验 机器学习 |
Correction Method of GPM;MERG Precipitation Data in Guanzhong Region Using Machine Learning Models |
| |
| XIE Xiang-zhou,LIU Jun-long,HUO Fei-fei,ZHANG Xiang-chun.Correction Method of GPM;MERG Precipitation Data in Guanzhong Region Using Machine Learning Models[J].International Journal Hydroelectric Energy,2022(2):6-9. |
| |
| Authors: | XIE Xiang-zhou LIU Jun-long HUO Fei-fei ZHANG Xiang-chun |
| |
| Affiliation: | (School of Mechanical and Electrical Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China;Chongqing City Management College,Chongqing 401331,China;College of Resource and Environment,Zunyi Normal University,Zunyi 563006,China;College of Biology and Griculture(College of Food Science and Technology),Zunyi Normal University,Zunyi 563006,China) |
| |
| Abstract: | Aiming at the systematic errors of the GPM_IMERG satellite product, on the basis of screening variables such as sea-land location, topography, and vegetation index(NDVI), the monthly IMERG precipitation data were revised by using support vector machine(SVM), random forest(RF), and Gaussian process regression(GPR) models in Guanzhong region. The comparison of the results of different correction models was validated by the measured data from 34 meteorological stations. The results show that the IMERG data in Guanzhong has good substitutability, with R2reaching 0.76, MMAEand RRMSEof 6.94 and 9.77 mm, respectively;The machine learning-based correction models incorporating multiple factors showed reliability, and the R2between the revised IMERG data and ground data improved by 2.05%-58.33%, and the RRMSEand MMAEdecreased by 0.85%-71.23% and 0.10%-73.47%, respectively;Among the three correction models, the RRMSEand MMAEare reduced by 16.76%, 9.76% and 24.73%, 14.11% for the SVM model compared with the GPR and RF models, which have better applicability to the IMERG data correction in the Guanzhong region. |
| |
| Keywords: | IMERG data correction verification of accuracy machine learning |
| 本文献已被 维普 等数据库收录！ |
|
