摘要
【目的】科学估测阿尔泰山天然林生物量,提高山区天然林生物量反演模型精度,同时为阿尔泰山天然乔木林的科学管护提供理论基础支持。【方法】基于6月6日、7月8日、7月24日和9月10日等4期Landsat 8遥感影像和新疆第九次全国森林资源连续清查数据,以岭回归、主成分分析、偏最小二乘法为建模方法分别构建阿尔泰山天然乔木林生物量反演模型,并验证其精度,从中选取最优模型,根据所选模型测算研究区天然乔木林生物量并分析其空间分布。【结果】从7月8日遥感影像所提取的与生物量显著相关的特征变量在数量和质量上都优于其他3期影像,最适宜用于研究区生物量反演;岭回归构建的生物量模型拟合出的预测值确定系数(R^(2))为0.918,预测值标准差(SEE)为16.70 t/hm^(2),平均系统误差(MSE)为5.39%,拟合精度(P)为85.55%,各项模型精度指标均优于主成分分析(R^(2):0.824、SEE:20.19 t/hm^(2)、MSE:8.37%、P:80.88%)和偏最小二乘法(R^(2):0.626、SEE:44.77 t/hm^(2)、MSE:-13.79%、P:70.21%)构建的模型。【结论】基于7月8日Landsat 8遥感影像,采用岭回归构建的模型最适用于研究区天然乔木林生物量估测。研究区天然乔木林生物量整体呈现南部低北部高的趋势,随海拔的升高先升高后降低,阴坡、半阴坡大于阳坡、半阳坡,随坡度的增大先升高后降低。精准估算山区天然乔木林生物量对于研究森林生态系统固碳能力、生产力以及评估天然乔木林的质量和生态效益具有重要意义。
【Objective】The aim of this work was to scientifically estimate the biomass of natural arbor forests in the Altay Mountains,and improve the accuracy of the inversion model of natural forest biomass in mountain areas,providing a theoretical basis for the scientific management and protection of natural forests in the Altay Mountains.【Method】Based on 4 periods of Landsat 8 remote sensing images on June 6,July 8,July 24 and September 10 and the 90 data sets of the ninth national forest resources inventory in Xinjiang,the modelling methods including ridge regression,principal component analysis and partial least square method were used.The inversion models of natural forest biomass in the Altay Mountains were constructed,which accuracy was then verified,and the optimal model was selected.Taking advantage of the selected model,the biomass of natural arbor forests in the study area was calculated and its spatial distribution was analyzed.【Result】The characteristic variables significantly related to biomass extracted from the remote sensing images on July 8 were better than the other three images in terms of quantity and quality,which was the most suitable for biomass inversion in the study area.The determination coefficient(R2)of the predicted value of the biomass model constructed by ridge regression was 0.918,the standard deviation of the predicted value(SEE)was 16.70 t/hm^(2),the average systematic error(MSE)was 5.39%,and the fitting accuracy(P)was 85.55%.The accuracy index of ridge regression was better than that of principal component analysis(R2:0.824,SEE:20.19 t/hm^(2),MSE:8.37%,P:80.88%)and partial least square method(R2:0.626,SEE:44.77 t/hm^(2),MSE:-13.79%,P:70.21%).【Conclusion】Based on the Landsat 8 remote sensing images on July 8,the model constructed by ridge regression is the most suitable to estimate the biomass of natural arbor forests in the study area.The biomass of natural arbor forests in the study area showed a low content in the south and a high accumulation level in the north,which i
作者
张景路
朱雅丽
张绘芳
刘建
地力夏提·包尔汉
ZHANG Jinglu;ZHU Yali;ZHANG Huifang;LIU Jian;Dilixiati Baoerhan(Modern Forestry Research Institute of Xinjiang Academy of Forestry,Urumqi 830000,Xinjiang,China;Altay branch of Altay State-owned Forest Management Bureau,Altay 836500,Xinjiang,China)
出处
《中南林业科技大学学报》
CAS
CSCD
北大核心
2022年第6期33-44,共12页
Journal of Central South University of Forestry & Technology
基金
新疆维吾尔自治区公益性科研院所基本科研业务专项(KY2019043,KY2020019)
中央财政专项(新林规字〔2021〕476号)。
关键词
天然乔木林
Landsat
8遥感影像
生物量
岭回归
主成分分析
偏最小二乘法
natural arbor forests
Landsat 8 remote sensing images
biomass
ridge regression
principal component analysis
partial least square method
