Method And Application Of Forest Biomass Estimation Based On LiDAR And OLI Multispectral Data

Estimating forest biomass at region scale quickly、quantificationally、accuracily is an important problem among ecosystem functions evaluation and carbon storage studys. Remote sensing technique can obtain forest biomass continuous distribution information on surface in time and accuracily in large area, because of its wide range of observation、large amount of information、quick information access. This study using Landsat 8 OLI multispectral imagery、small-footprint LiDAR data and plot survey data, firstly, estimating subtropical forest biomass by some metrics extracted from Landsat 8 OLI multispectral imagery, then, exploring the improvement of estimation accuracy based on combining OLI metrics and small-footprint diecrete LiDAR metrics, at last, explored new method of upscale(low cost)forest biomass estimation based on OLI imagery covering the whole study area and a little bit(one strip)of LiDAR data. The results showed that:(1)Estimating forest biomass based on OLI multispectral imagery, the result showed that for all plots with no partition analysis, the OLI model’s accuracy of above-ground biomass model was 0.41 and the accuracy of below-ground biomass model was 0.57. For three different forest types, the accuracy of above-ground biomass model and below-ground biomass model for coniferous forest were 0.67 and 0.70; the accuracy of above-ground biomass model and below-ground biomass model for broad-leaf forest were 0.74 and 0.80; the accuracy of above-ground biomass model and below-ground biomass model for mixed forest were 0.77 and 0.84.Textures had higher correlation with biomass, especially CO、DI、HO were all selected in estimating biomass of all plots、coniferous forest and mixed forest, however, initial bands had lower correlation with biomass. But textures’ interpretation decreased significantly when estimating biomass of broad-leaf forest.(2)Estimating forest biomass based on combining LiDAR and OLI multispectral data, the result showed that for all plots with no partition analysis, the general model’s accuracy of above-ground biomass model was 0.65 and the accuracy of below-ground biomass model was 0.69. For three different forest types, the accuracy of above-ground biomass model and below-ground biomass model for coniferous forest were 0.86 and 0.91; the accuracy of above-ground biomass model and below-ground biomass model for broad-leaf forest were 0.93 and 0.92; the accuracy of above-ground biomass model and below-ground biomass model for mixed forest were 0.83 and 0.92. In addition,the LiDAR model’s accuracy of above-ground biomass model was 0.65 and the accuracy of below-ground biomass model was 0.64. For three different forest types, the accuracy of above-ground biomass model and below-ground biomass model for coniferous forest were 0.82 and 0.75; the accuracy of above-ground biomass model and below-ground biomass model for broad-leaf forest were 0.93 and 0.86; the accuracy of above-ground biomass model and below-ground biomass model for mixed forest were 0.80 and 0.83. The accuracy of the general model was higher than OLI model and LiDAR model. And textures and Percentile heights metrics had higher correlation with biomass.(3)Estimating forest biomass based on a little bit(one strip) of LiDAR data and OLI multispectral data covering the whole study area, the result showed that for all plots with no partition analysis, the LiDAR-OLI model’s accuracy of above-ground biomass model and below-ground biomass model were 0.69 and 0.56, and were higher than OLI2 model’s accuracy(the accuracy of above-ground biomass model and below-ground biomass model were 0.49 and 0.55). It showed that using a little bit of LiDAR data and OLI imagery covering the whole study area can improve accuracy of biomass estimation, at same time, it can reduce the cost relatively.This study explored the new method of estimating forest biomass based on OLI and LiDAR metrics using multiple stepwise regression、remote sensing image processing、LiDAR data processing and metrics extraction. This paper sets the subtropical secondary forests in southern hilly area of Jiangsu province as a research subject, based on OLI imagery and LiDAR data pre processing and analysis, extracting metrics and choosing better metrics by correlation analysis, at last, establishing estimation models combining with plots surey stand characteristics parameters, as well as verifying the accuracy of models. The results showed that textures and percentile heights have high sensitivity with forest biomass, the accuracy of general model and LiDAR-OLI model were the highest and higher than OLI and LiDAR model. OLI metrics describe the horizontal structure information of forest, and LiDAR metrics describe the vertical structure information of forest, they have complementary advantages and can improve accuracy of estimation of forest biomass.