基于Landsat5 TM遥感影像估算江山市公益林生物量

Landsat5 TM-based Biomass Estimation of Public-welfare Forest of Jiangshan City

本研究基于Landsat5 TM遥感影像数据和样地调查数据,利用多元逐步回归、偏最小二乘回归和随机森林回归3种方法,建立江山市公益林生物量估算模型,分析和比较3种模型的精度结果,探究随机森林回归模型在估算生物量方面的应用,为提高估算森林生物量的精度提供参考。结果表明,多元逐步回归模型的预测精度为58.31%、均方根误差为31.02 t/hm2,偏最小二乘回归模型分别为60.84%、30.72 t/hm2,随机森林回归模型为70.02%,22.18 t/hm2。由此可得,随机森林回归模型的预测精度优于其他2种模型,随机森林算法能提高估算森林生物量的精度。

By using Landsat5 TM data and forest inventory data, muhi-stepwise regression model, partial least square regression model and random forest regression model were buih to estimate forest biomass in Jiangshan City, and the accuracy of these three models were analyzed and compared to study the application of regression models in forest biomass estimation. The results showed that the precisions and root mean square errors of muhi-stepwise, partial least square regression and random forest were 58.31%, and 31.02 t/hm2, 60.84 % and 30. 72 t/hm2, 70. 02 % and 22. 18 t/hm2 respectively. Therefore random forest regression model is better than the other two models, and it could improve the accuracy of forest biomass estimation.