基于Landsat影像数据的香格里拉市优势树种蓄积量估测模型构建

王飞平; 张加龙; 申茂华; 薛雯芸; 李坤美

文章摘要

王飞平,张加龙,申茂华,薛雯芸,李坤美.基于Landsat影像数据的香格里拉市优势树种蓄积量估测模型构建[J].林业调查规划,2023,48(2):26-31

基于Landsat影像数据的香格里拉市优势树种蓄积量估测模型构建

Estimation Model Construction of Dominant Tree Species Volume in Shangri-La City Based on Landsat Image Data

DOI：

中文关键词: 优势树种蓄积量估测模型构建 landsat影像数据香格里拉市

英文关键词: dominant tree species estimation of volume model construction Landsat image data Shangri-La City

基金项目:国家自然科学基金（31860207）；2020年云南省高层次人才培养支持计划“青年拔尖人才”专项（81210468）；西南林业大学科研启动基金(111932）.

作者	单位
王飞平	西南林业大学西南地区生物多样性保育国家林业局重点实验室，云南昆明 650224 西南林业大学林学院，云南昆明 650224
张加龙	西南林业大学西南地区生物多样性保育国家林业局重点实验室，云南昆明 650224
申茂华	西南林业大学林学院，云南昆明 650224
薛雯芸	西南林业大学林学院，云南昆明 650224
李坤美	西南林业大学林学院，云南昆明 650224

摘要点击次数: 57

全文下载次数: 59

中文摘要:

以香格里拉市优势树种高山栎、高山松、云冷杉及云南松为研究对象，构建基于遥感的森林蓄积量估测模型。采用2006年和2016年的Landsat TM/OLI影像及二类调查数据，随机从每期数据中各选取100个小班，提取其影响因子，利用相关性强的因子构建RF模型和MLR模型。结果表明，2006年各树种蓄积量RF模型的拟合R²在0.779~0.862范围，预测精度P值为80.17%~92.16%；2016年RF模型的拟合R²为0.761~0.865，预测精度P值为81.61%~95.53%。建立MLR模型后，2006年树种蓄积量估测模型的拟合R²为0.214~0.336，预测精度P值为34.12%~47.16%；2016年模型的拟合R²为0.238~0.391，预测精度P值为34.82%~52.57%。估测结果与二类调查数据的误差为：高山栎、高山松、云冷杉、云南松分别增加了8.54×10⁵m³、3.65×10⁶m³、4.12×10⁶m³、3.96×10⁴m³，误差分别为0.36%、0.78%、0.44%、0.62%。对比二种模型估测结果表明，随机森林模型能更精确地估测优势树种蓄积量。

英文摘要:

The remote sensing based forest volume estimation model was constructed in Shangri-La City with the research objects of dominant tree species of Quercus aquifolioides, Pinus densata, spruce fir and Pinus yunnanensis. Based on Landsat TM/OLI images in 2006 and 2016 and forest management inventory data, this paper randomly selected 100 sub-compartments from each period of data, extracted the impact factor, and constructed an RF model and MLR model by highly correlated factors. The results showed that the fitting R² of the RF model for the stock volume of various tree species in 2006 ranged from 0.779 to 0.862, with a prediction accuracy P-value of 80.17% to 92.16%; the fitting R² of the RF model in 2016 was from 0.761 to 0.865, and the prediction accuracy P-value was from 81.61% to 95.53%. After establishing the MLR model, the fitting R² of the model was from 0.214 to 0.336, and the prediction accuracy P-value was from 34.12% to 47.16%; the fitting R² in 2016 was from 0.238 to 0.391, and the prediction accuracy P-value was from 34.82% to 52.57%. The volume of Quercus aquifolioides, Pinus densata, spruce fir and Pinus yunnanensis increased by 8.54×105m³, 3.65×10⁶m³, 4.12×10⁶m³ and 3.96×10⁴m³ respectively, and the errors were 0.36%, 0.78%, 0.44% and 0.62% respectively. The comparison between the two models showed that the random forest model could more accurately estimate the volume of dominant tree species.

查看全文查看/发表评论下载PDF阅读器

关闭