欢迎访问林业科学,今天是

林业科学 ›› 2020, Vol. 56 ›› Issue (3): 38-47.doi: 10.11707/j.1001-7488.20200305

• 论文与研究报告 • 上一篇    下一篇

基于空间回归模型的思茅松林生物量遥感估测及光饱和点确定

周律,欧光龙,王俊峰,胥辉*   

  1. 西南林业大学 西南地区生物多样性保育国家林业和草原局重点实验室 昆明 650224
  • 收稿日期:2019-08-22 出版日期:2020-03-25 发布日期:2020-04-08
  • 通讯作者: 胥辉
  • 基金资助:
    国家自然科学基金项目(31760206);国家自然科学基金项目(31770677);国家自然科学基金项目(31660202);云南省王广兴专家工作站(2018IC100);云南省万人计划青年拔尖人才专项(YNWR-QNBJ-2018-184)

Light Saturation Point Determination and Biomass Remote Sensing Estimation of Pinus kesiya var. langbianensis Forest Based on Spatial Regression Models

Lü Zhou,Guanglong Ou,Junfeng Wang,Hui Xu*   

  1. Key Laboratory of National Forestry and Grassland Administration for Biodiversity Conservation in Southwest China, Southwest Forestry University Kunming 650224
  • Received:2019-08-22 Online:2020-03-25 Published:2020-04-08
  • Contact: Hui Xu

摘要:

目的: 确定思茅松林生物量遥感估测的光饱和点,构建空间全局和局域遥感信息模型反演思茅松林生物量,为思茅松林生物量遥感估测提供参考。方法: 以云南省普洱市思茅松林为研究对象,基于Landsat8 OLI遥感影像数据和森林资源二类调查数据,运用二次项函数和幂函数求解思茅松林生物量光饱和点,采用普通最小二乘模型(OLS)、空间滞后模型(SLM)、空间误差模型(SEM)和地理加权回归模型(GWR)构建遥感信息模型,估测思茅松林地上生物量。结果: 1)普洱市思茅松林Landsat8 OLI遥感估测地上生物量的光饱和点为106.3 t·hm-2;2)空间回归模型拟合精度较高,尤其是GWR模型具有最高的R2(0.373)和最小的AIC(4 577.8),其拟合精度显著高于OLS、SLM和SEM模型;3)独立性样本检验结果表明,GWR模型的预估精度最高,且通过刀切法检验可知GWR模型在高值阶段(≥100 t·hm-2)和低值阶段(0~50 t·hm-2)的生物量估测能力强于OLS、SLM和SEM模型;4)GWR模型反演计算结果表明,思茅松林单位面积地上生物量为66.496 t·hm-2,与实测值偏差23.511%,估测误差低于OLS、SLM和SEM模型。结论: 对普洱市思茅松林生物量进行遥感估测时,GWR模型优于OLS模型和其他空间全局回归模型,且GWR模型可在一定程度上解决高值低估和低值高估问题,减小光饱和点对遥感估测精度的影响。

关键词: 地上生物量, 光饱和点, 空间回归模型, 二类调查数据, 思茅松林

Abstract:

Objective: This paper aims to determinate the light saturation point of biomass estimation using remote sensing data of Simao pine(Pinus kesiya var. langbianensis)forest, and construct the spatial global model and the local remote sensing information model, then inverse the biomass of Simao pine forest. The results will provide a reference for estimating biomass of Simao pine using remote sensing data. Method: This study calculated the light saturation point of Simao pine forest biomass using the quadratic term function and power function model based on Landsat8 OLI images, and the forest resource inventory data in 2016 of Pu'er city, Yunnan Province, were used to construct the remote sensing information model with the ordinary least square(OLS), spatial lag model(SLM), spatial error model(SEM)and geographically weighed regression(GWR), and then the aboveground biomass of Simao pine forest was estimated. Result: 1) The light saturation point of aboveground biomass estimated by Landsat 8 OLI remote sensing in Simao pine forest of Pu'er city was 106.3 t·hm-2. 2) The highest coefficient of determination(R2=0.373)and the lowest Akaike information index(AIC=4 577.8)of GWR model indicated that it has the better fitting performance than OLS, SLM and SEM model. 3) The results of the independent sample test showed that the estimation accuracy of GWR model was higher than those of OLS, SLM and SEM model, especially at the high value stage (above 100 t·hm-2) and the low value stage (below 50 t·hm-2) based on the Jackknife method test. 4) The aboveground biomass by inversion using GWR model was 66.496 t·hm-2, and its estimation error with the measured data was 23.511%, which was less than the estimation values of OLS, SLM and SEM model. Conclusion: GWR model is superior to OLS model and other spatial global regression models in estimating biomass of Simao pine forest in Pu'er city by remote sensing. Moreover, GWR model can solve the problems of high-value under estimation and low-value over estimation to a certain extent, and reduce the influences of the light saturation point on the accuracy of remote sensing estimation.

Key words: aboveground biomass, light saturation point, spatial regression model, forest resource inventory data, Pinus kesiya var. langbianensis forest

中图分类号: