生态学报2024，Vol.44Issue(6)：2464-2478,15.DOI:10.20103/j.stxb.202304070700
三峡库区消落带植被NPP估算
Estimation of vegetation carbon sink in fluctuation zone of Three Gorges Reservoir Area based on CASA Model optimized by machine learning
摘要
Abstract
After the storage of the Three Gorges Reservoir,its ecological impact has received great attention.As an important indicator of ecosystem health in the reservoir area,carbon sequestration of vegetation in fluctuating zone is of great significance to carbon cycle and ecological purification in the reservoir area.Due to the differences in the time of receiving sunlight for vegetation at different elevations in the ebb and flow zone,and the impact of changes in river water levels,the traditional CASA model has problems such as inaccurate calculation of the light energy utilization rate of plants when calculating the carbon sequestration amount of vegetation in the ebb and flow zone.In this paper,a new method(RBF-CASA)for coupling the Radial Basis Function Neural Network(RBFNN)model with the Carnegie Ames Stanford approach(CASA)model is proposed based on the study area of the steep slope fluctuation zone of the Xiangxi River in the Three Gorges Reservoir Area.We have established an environmental impact factor model based on RBFNN.This model utilizes features such as elevation data and vegetation index to calculate environmental impact factors suitable for the region,and combines them with temperature and water stress factors to calculate environmental stress factors,in order to improve the accuracy of estimating vegetation net primary productivity(NPP)at the pixel scale.The results showed that the R2 between the estimated value and the observed value of the RBF-CASA model was 0.730(P＜0.01,n=32).Compared with the CASA model,the MAE decreased by 10.991,RRMSE decreased by 5.10％,and MAPE decreased by 1.12％.Using the RBF-CASA model to estimate carbon sequestration in the typical steep slope and falling zone of the Three Gorges Reservoir area,the monthly average NPP of the study area in July,the most suitable month for vegetation growth,was between 66.234 and 134.144g C/m2.In the area of the ebb and flow zone,NPP fluctuated with the increase of elevation,with the total amount of NPP reaching a peak value between 150 and 155 m,and the area with the average value of NPP above 170 m had the highest value.In September 2021,the average NPP of vegetation was 35.883g C/m2,while in September 2022,the average NPP of vegetation was 25.964g C/m2.Due to the decrease in rainfall and the decline in the water level of the Yangtze River,the vegetation restoration situation was poor between 2021 and 2022.Therefore,this research can provide scientific basis for carbon cycle,ecological purification and ecological restoration decision-making.关键词
基于过程的遥感模型(CASA)/机器学习/植被净初级生产力(NPP)/无人机/环境影响因子模型Key words
CASA model/machine learning/net primary productivity/unmanned aerial vehicle/environmental impact factor model引用本文复制引用
靳专,胥焘,黄应平,肖敏,张家璇,周爽爽,席颖,熊彪..三峡库区消落带植被NPP估算[J].生态学报,2024,44(6):2464-2478,15.基金项目
国家自然科学基金(22136003,42177397) （22136003,42177397）
长江水环境教育部重点实验室(同济大学)开放基金项目(YRWEF202103) （同济大学）
湖北省教育厅优秀中青年人才项目(Q20211205) （Q20211205）
