草业科学2025，Vol.42Issue(9)：2153-2165,13.DOI:10.11829/j.issn.1001-0629.2024-0276
基于支持向量回归改进的温带草原总初级生产力估算
Improvement of GPP estimation in temperate grasslands based on the SVR algorithm
摘要
Abstract
Vegetation productivity is an important indicator of the ecosystem's response to climate change and human activity.The primary objective of this study is to improve the accuracy of gross primary productivity(GPP)estimations to address the urgent need for accurate ecological monitoring.This study developed an optimized GPP estimation model,CASA_SVR,based on the light-use efficiency model,which incorporates the support vector regression(SVR)technique to improve estimation of the fraction of absorbed photosynthetically active radiation(FPAR)using the normalized difference vegetation index(NDVI)and modified soil adjusted vegetation index(MSAVI),which is essential for accurate calculation of the GPP.The results showed that:1)The CASA_SVR model,through its innovative use of SVR for FPAR calculation,demonstrated a substantial improvement in GPP estimation accuracy,as evidenced by an R2 of 0.71(P＜0.001),marking a 0.19 increase over the traditional CASA(Carnegie-Ames-Stanford approach)model.This enhancement was further supported by significant reductions in both the RMSE and MAE.Similarly,the VPM_SVR model showed a significant improvement compared to the standard vegetation photosynthesis model(VPM),with an increase of 0.25 in the R2 value and a reduction in both the RMSE and MAE,which highlights the efficacy of incorporating SVR into the GPP estimation model and the introduction of MSAVI to optimize the FPAR calculation.2)The RMSE values of the CASA_SVR model for estimating the GPP values of meadow,typical,and desert grasslands increased by 7.11,2.31 and 10.41 g·(m2·month)-1,respectively,in comparison with the CASA model.The CASA_SVR model significantly improved GPP estimation for grasslands,with a significant reduction in RMSE compared to traditional models and existing(moderate-resolution imaging spectroradiometer,MODIS)products,and showed wider applicability and higher accuracy in different grassland types.关键词
温带草原/总初级生产力/陆地资源卫星/支持向量回归/光合有效辐射占比/土壤植被指数/光能利用效率模型Key words
temperate grasslands/gross primary productivity/Landsat 8-OLI/support vector regression/fraction of photosynthetically active radiation/soil vegetation index/light-use efficiency model引用本文复制引用
曹云刚,赵俊,杜蒲英,曾雅慧,张成利,杨秀春..基于支持向量回归改进的温带草原总初级生产力估算[J].草业科学,2025,42(9):2153-2165,13.基金项目
第三次新疆综合科学考察项目(2022xjkk0402) （2022xjkk0402）
四川省青年科技创新研究团队项目(2020JDTD0003) （2020JDTD0003）
