| 摘 要: 干旱在中国是一种频繁发生的灾情现象。近年的研究发现,相较于传统因子,日光诱导叶绿素荧光(SIF)能够更加快速、准确地反映农业干旱状况。该研究以SIF数据集和实际灾情数据资料为基础,综合运用归一化、随机森林回归预测及优化的随机森林算法等方法,构建干旱预测模型,并对比各模型对干旱事件的预测准确性,同时分析了预测结果与帕默尔干旱指数(PDSI)的相关性。研究结果表明,优化后的随机森林预测模型获得的最优拟合(R2)为0.9202、最优均方误差(MSE)为0.0042、均方根误差(RMSE)为0.064;SIF与PDSI之间存在显著的正相关性,并且预测结果与干旱等级高度吻合。由此表明,基于优化监测模型预测的SIF数据可以很好地应用到无气象站点资料的区域,为农业干旱胁迫的早期预警和监测提供科学依据。 |
| 关键词: 日光诱导叶绿素荧光 干旱监测 随机森林 优化模型 皮尔逊相关系数 |
|
中图分类号: TP407
文献标识码: A
|
| 基金项目: 国家自然科学基金项目(42361060);甘肃省博士后科学研究基金项目(BSH2024002);干旱基金(IAM202420) |
|
| Study on Agricultural Drought Monitoringin Typical Arid Regions of Northwest China Based on Random Fores |
|
ZHANG Jifang1, GUO Jifu1,2, DAI Yongqiang1, ZHAO Funian2
|
(1.College of Information Science and Technology, Gansu Agricultural University, Lanzhou 730070, China;
2.Institute of Arid Meteorology, CMA, Lanzhou 730020, China)
2742325969@qq.com; guojf@gsau.edu.cn; dyq@gsau.edu.cn; zfn0622@163.com
|
| Abstract: Drought is a frequently occurring disaster in China. Recent studies indicate that Sola-r Induced chlorophyll Fluorescence (SIF) can reflect agricultural drought more rapidly and accurately than traditional indicators.This study integrates SIF datasets and actual disaster records to construct drought monitoring models using normalization methods, random forest regression prediction, and an optimized random forest algorithm. The results demonstrate that the optimized random forest prediction model achieved the best fitting performance, with an R2 of 0.920 2, a minimum mean squared error (MSE) of 0.004 2, and a root mean squared error (RMSE) of 0.064. SIF exhibits a significant positive correlation with PDSI, and the predictions align closely with drought severity levels. This confirms that the optimized model enables effective application of SIF data in areas lacking meteorological station records, providing a scientific basis for early warning and monitoring of agricultural drought stress. |
| Keywords: Sola-r Induced Chlorophyll Fluorescence drought monitoring random forest optimized model Pearson correlation coefficient |
