摘要
为探索漓江流域喀斯特森林固碳功能的量化方式,明确漓江流域森林碳储量的空间分布特征,以漓江流域喀斯特森林为研究对象,借助野外调查实测数据和遥感影像纹理特征信息,建立了多元线性回归模型和基于机器学习的随机森林模型来预测流域喀斯特森林碳储量与遥感参数之间的关系,并利用In VEST模型和土地利用数据模拟了漓江流域2005、2010、2015、2020年的森林碳储量分布情况。结果表明,相比多元线性回归模型,随机森林模型能够更好地模拟样地实际碳储量,表现出较高的校正系数(0.726)和较低的均方根误差(8.742),说明碳储量与多个遥感参数之间存在较为复杂的非线性关系。漓江流域森林平均碳密度为27.48 t/hm^(2),最低2.43 t/hm^(2),最高94.35 t/hm^(2),高值区主要分布在流域的北部、西北部和沿流域轮廓的东部区域,中值区主要分布在流域的中游和南部的旅游景区,低值区主要分布在城市建成区。漓江流域碳储量高值区域大多分布着亚热带常绿阔叶林,中值区域大多分布着亚热带针叶林、常绿果树园和亚热带经济林。可见植被类型和林型对漓江流域森林碳储量的空间分布有影响。由于流域在10年间耕地和林地向建筑用地转变,碳储量高值区向碳储量低值区的变化也逐步加剧。该研究验证了遥感技术结合实地调查在森林碳储量估测方面具有优势与可行性。对漓江流域的喀斯特森林而言,多元回归模型能较好地拟合其碳储量与植被指数、纹理特征的定量关系,而随机森林模型拟合的适配度更高。
To explore methods for quantifying the carbon sequestration capabilities of karst forests in the Lijiang River Basin and clarify the spatial distribution characteristics of forest carbon stocks within the basin,taking the karst forests in the Lijiang River Basin as the research object,a multiple linear regression model and a random forest model based on machine learning were established to predict the relationship between forest carbon stocks and remote sensing parameters with the help of field survey data and texture characteristics of remote sensing images.Additionally,the InVEST model and land use data were employed to simulate carbon stock scenarios for the years 2005,2010,2015,and 2020.The results showed that the random forest model outperformed the multiple linear regression model in simulating actual carbon stocks,exhibiting a higher coefficient of determination(0.726)and a lower root mean square error(8.742),indicating that there was a complex nonlinear relationship between carbon stocks and various remote sensing parameters.The average forest carbon density in the Lijiang River Basin was 27.48 t/hm^(2),with a minimum of 2.43 t/hm^(2)and a maximum of 94.35 t/hm^(2).The high carbon density areas were mainly distributed in the northern,northwestern,and eastern contour regions of the basin,while medium-density areas were primarily located in the midstream and southern tourist regions.Low-density areas were mainly distributed in urban built-up areas.The high carbon stock regions are predominantly covered by subtropical evergreen broad-leaved forests,while medium-density regions mostly consist of subtropical coni-ferous forests,evergreen orchards,and subtropical economic forests.This indicated that vegetation type and forest type significantly influence the spatial distribution of forest carbon stocks in the Lijiang River Basin.Over the past decade,due to the conversion of cropland and forestland to construction land,the change from high-value carbon storage area to low-value carbon storage area has gradually intensified.Therefore,it was verified the advantages and feasibility of combining remote sensing technology with field surveys in estimating carbon storage.For the karst forests in the Li River Basin,the multivariate regression model could effectively capture the quantitative relationship between carbon storage and vegetation index,as well as texture features,while the random forest model demonstrates had even higher fitting accuracy.
作者
康佳琦
李林
储小雪
赵毅
刘佳润
谭一波
KANG Jiaqi;LI Lin;CHU Xiaoxue;ZHAO Yi;LIU Jiarun;TAN Yibo(Guangxi Key Laboratory of Superior Timber Trees Resource Cultivation,Nanning 530002,China;School of Life and Environmental Sciences,Guilin University of Electronic Technology,Guilin 541004,Guangxi,China;Guangxi Lijiangyuan Forest Ecosystem Research Station,Guilin Xing’an Lijiangyuan Forest Ecosystem Observation and Research Station of Guangxi,Guilin 541316,Guangxi,China)
出处
《热带亚热带植物学报》
北大核心
2025年第4期417-427,共11页
Journal of Tropical and Subtropical Botany
基金
广西自然科学基金青年项目(2021GXNSFBA196052)
广西优良用材林资源培育重点实验室课题(23-B-04-02)
桂林兴安漓江源森林生态系统广西野外科学观测研究站科研能力建设项目(桂科22-035-130-02)资助。
关键词
碳储量
喀斯特森林
遥感技术
随机森林模型
多元回归分析
Carbon stock
Karst forest
Remote sensing technology
Random forest model
Multiple regression analysis
