基于基于多源遥感数据的安徽皖南山区森林蓄积量估测研究
Study on the estimation of forest stock in the mountainous area of South Anhui Province based on multi-source remote sensing data
我国在2020年气候峰会上提出"到2030年中国森林蓄积量要比2005年增加60亿立方米"的目标，有效结合多源遥感数据与地面样地调查数据建立森林蓄积量估测模型、提高森林蓄积量估测精度是目前遥感蓄积量热点。本研究利用机载Lidar数据、Sentinel-2光学遥感数据与样地调查数据对安徽省皖南山区进行蓄积量估测研究，以黄山区与绩溪县为例，寻找最佳蓄积量估测模型。研究结果表明：协同机载Lidar数据与Setinel-2光学遥感数据的S-CHM变量集能有效提高森林蓄积量估测精度，相较于仅基于CHM数据的B-CHM变量集建模，整体线性回归验证R2提高0.12，RMSE降低7.58m3/ha，RMSEr降低4.83%。基于B-CHM与S-CHM变量集的分层贝叶斯的建模均有效提高整体建模精度，分层因素包括森林类型、坡度、坡向，基于S-CHM变量集的以森林类型分层的分层贝叶斯模型拥有最佳模型精度。不同森林类型单独建模中，多源遥感数据同样比单一遥感数据表现好，杉木、马尾松、阔叶混交最佳模型的RMSEr分别降低8.52%、8.02%、1.49%。分层贝叶斯有效缓解了不同客观条件下（森林类型、坡度、坡向）对样本数量的需求，提高森林蓄积量估测精度，多源遥感数据的结合弥补单一遥感数据的不足，成为遥感估测森林蓄积量的关键手段。
t the 2020 Climate Summit, China proposed the goal of "China\'s forest stock should be increased by 6 billion cubic meters by 2030 compared with 2005", and it is now a hotspot for remote sensing stock to effectively combine multi-source remote sensing data and ground sample survey data to establish a forest stock estimation model and improve the accuracy of forest stock estimation. In this study, airborne Lidar data, Sentinel-2 optical remote sensing data and sample plot survey data were utilized to estimate the forest stock in the southern Anhui Province, and the best stock estimation model was found in Huangshan District and Jixi County as an example. The results of the study showed that the S-CHM variable set with airborne Lidar data and Setinel-2 optical remote sensing data could effectively improve the accuracy of forest stock estimation, and the overall linear regression validation improved the R2 by 0.12, reduced the RMSE by 7.58 m3/ha, and reduced the RMSEr by 4.83% compared with the B-CHM variable set modeling based on CHM data only. The modeling of stratified Bayes based on both B-CHM and S-CHM variable sets was effective in improving the overall modeling accuracy, and the stratification factors included forest type, slope, and slope direction, and the stratified Bayesian model stratified by forest type based on the S-CHM variable set had the best modeling accuracy. In the separate modeling of different forest types, multi-source remote sensing data also performed better than single remote sensing data, and the RMSEr of the best model for fir, pony pine, and broadleaf mixed were reduced by 8.52%, 8.02%, and 1.49%, respectively. Hierarchical Bayes effectively alleviates the demand for sample number under different objective conditions (forest type, slope, slope direction), improves the accuracy of forest stock estimation, and the combination of multi-source remote sensing data makes up for the shortcomings of single-source remote sensing data, and becomes a key means of remote sensing estimation of forest stock.???????
李桂英、杨雯林
测绘学环境科学技术现状工程设计、工程测绘
遥感技术与应用蓄积量多源遥感机载Lidar
remote sensing technology and applications volume multi-source remote sensing airborne Lidar
李桂英,杨雯林.基于基于多源遥感数据的安徽皖南山区森林蓄积量估测研究[EB/OL].(2024-02-28)[2025-08-02].http://www.paper.edu.cn/releasepaper/content/202402-105.点此复制
评论