适合西藏地区的归一化植被指数预测模型构建及验证  

Construction and Validation of Normalized Difference Vegetation Index Prediction Model Suitable for the Xizang Region

作　　者：孟慧美[1,4] 吴凌霄 宣越健 米玛旺堆[1] MENG Huimei;WU Lingxiao;XUAN Yuejian;Migmar Wangdwei(College of Ecology and Environment,Tibet University,Lhasa 850000;College of Intelligent Systems Science and Engineering,Hubei Minzu University,Enshi,Hubei Province 445000;Key Laboratory of Middle Atmosphere and Global Environment Observation,Institute of Atmospheric Physics,Chinese Academy of Sciences,Beijing 100029;Fuyang Preschool Teachers College,Fuyang,Anhui Province 236000)

机构地区：[1]西藏大学生态环境学院,拉萨850000 [2]湖北民族大学智能科学与工程学院,湖北恩施445000 [3]中国科学院大气物理研究所中层大气和全球环境探测重点实验室,北京100029 [4]阜阳幼儿师范高等专科学校,安徽阜阳236000

出　　处：《气候与环境研究》2025年第2期199-211,共13页Climatic and Environmental Research

基　　金：安徽省高等学校科学研究重点项目,2024AH051477,国家自然科学基金项目,32360269。

摘　　要：基于差分自回归移动平均(ARIMA)方法、随机森林(RF)方法、Prophet方法构建适合西藏地区的归一化植被指数(Normalized Difference Vegetation Index,NDVI)预测模型,利用羊八井地区2000~2021年MODIS遥感NDVI数据进行了验证,结果表明:该地区植被覆盖率总体呈现不明显减少趋势;3个预测模型中,RF预测精度最高,其归一化均方根误差、平均绝对百分比误差、决定系数,分别达到了6.92%、4.04%、0.9;小波变换方法能有效提高模型预测精度;组合模型可以提高预测精度,其中误差倒数权重组合模型优于平均权重和方差倒数加权组合模型。因此可以利用RF等机器学习方法结合小波变换、组合模型在西藏地区进行NDVI预测,为生态环境保护和农牧业生产决策提供科学指导。Suitable Normalized Difference Vegetation Index(NDVI)prediction models were developed for the Xizang region using AutoRegressive Integrated Moving Average(ARIMA),Random Forest(RF),and prophet methods.The validation was conducted using MODIS remote sensing NDVI data for the Yangbajing area from 2000 to 2021.The results show no significant decrease in the overall vegetation coverage in this region.Among the three prediction models,RF demonstrates the highest prediction accuracy,with normalized root-mean-squared error,mean absolute percentage error,and coefficient of determination of 6.92%,4.04%,and 0.9,respectively.The wavelet transformation method efficiently enhances the prediction accuracy of the models.The combined models improve prediction accuracy,and the reciprocal of error weights combined model outperforms the average weight and inverse variance weighted combined models.Therefore,machine learning methods such as RF,when combined with wavelet transformation and the reciprocal of error weights model,can be effectively utilized for NDVI prediction in the Xizang region.This approach provides scientific guidance for ecological protection and agricultural decision-making.

关 键 词：归一化植被指数(NDVI)预测模型 随机森林(RF)方法 差分自回归移动平均(ARIMA)方法 Prophet方法 小波变换 

分 类 号：Q948[生物学—植物学]