Remote sensing inversion of above-ground biomass of grassland in the Tibet Autonomous Region

PDF HTML阅读 XML下载 导出引用 引用提醒 西藏自治区草地地上生物量遥感反演研究 DOI: 10.5846/stxb202206061593 作者: 作者单位: 作者简介: 通讯作者: 中图分类号: 基金项目: 湖南省教育厅重点项目（17A225）；国家自然科学基金面上项目（31971578）；湖南省林业科技创新专项（XLK201986）；长沙市科技计划项目（kq2004095） Remote sensing inversion of above-ground biomass grassland in the Tibet Autonomous Region Author: Affiliation: Fund Project: 摘要 | 图/表 访问统计 参考文献 相似文献 引证文献 资源附件 文章评论 摘要:草地地上生物量（Aboveground Biomass，AGB）是反映草地生态系统功能和质量的关键指标，大尺度地准确估算草地AGB对草地生态系统经营管理至关重要。研究以MODIS影像为数据源，提取反射率、植被指数和植被产品三种不同类型的特征变量，结合野外实测样地草地AGB数据，构建以多元线性逐步回归为代表的参数模型以及随机森林、支持向量机和kNN等非参数模型进行西藏自治区草地AGB估测及空间分布制图。结果表明：（1）多元线性逐步回归、随机森林、支持向量机和kNN模型在加入植被产品特征变量后，RMSE分别降低了15.8%、13.5%、4.1%和17.3%，表明植被产品作为建模变量用于草地AGB估测可有效提高模型精度；（2）三种组合变量构建的草地AGB估测模型中，反射率、植被指数、植被产品组合构建的模型效果最佳，其中kNN模型估测精度最高，R2达到0.60，RMSE和MAE分别为0.43、0.34 t/hm2；（3）草地AGB空间分布呈现出西北地区较低、中部地区较高且分布形态较破碎和东部地区较高的变化特征；（4）利用MODIS植被产品结合kNN模型的预测值与草地实测的AGB空间分布趋势基本一致。综上，MODIS植被产品结合kNN模型可作为大尺度区域草地AGB遥感估测的有效参考。 Abstract:Aboveground Biomass (AGB) is a key indicator reflecting function and quality ecosystems. Estimating AGB on large scale crucial for management In this study, MODIS images were used as data source to extract three different types characteristic variables, such reflectance, vegetation index products. Combined with field measured data, parametric models represented by Multiple Linear Stepwise (MLR) regression nonparametric which Random Forest (RF), Support Vector Machine (SVM) k Nearest Neighbor (kNN) constructed respectively estimate map Region. The results showed that:(1) multiple linear stepwise regression, random forest, support vector machine model kNN reduced Root Mean Squared Error (RMSE) 15.8%, 13.5%, 4.1% 17.3% after inclusion products feature that modeling variables estimation could effectively improve accuracy model. (2) Among from combinations combination consisting index, product worked best, having highest R2 0.60, RMSE 0.43 t/hm2, Absolute (MAE) 0.34 t/hm2. (3) spatial distribution variation characteristics lower northwest region, higher central region more fragmented pattern eastern region. (4) predicted values combined basically consistent observed trend grassland. conclusion, can be an effective reference remote estimation.