Using Airborne Lidar, Multispectral Imagery, and Field Inventory Data to Estimate Basal Area, Volume, and Aboveground Biomass in Heterogeneous Mixed Species Forests: A Case Study in Southern Alabama

Abstract

Airborne light detection and ranging (lidar) has proven to be a useful data source for estimating forest inventory metrics such as basal area (BA), volume, and aboveground biomass (AGB) and for producing wall-to-wall maps for validation of satellite-derived estimates of forest measures. However, some studies have shown that in mixed forests, estimates of forest inventory derived from lidar can be less accurate due to the high variability of growth patterns in multispecies forests. The goal of this study is to produce more accurate wall-to-wall reference maps in mixed forest stands by introducing variables from multispectral imagery into lidar models. Both parametric (multiple linear regression) and non-parametric (Random Forests) modeling techniques were used to estimate BA, volume, and AGB in mixed-species forests in Southern Alabama. Models from Random Forests and linear regression were competitive with one another; neither approach produced substantially better models. Of the best models produced from linear regression, all included a variable for multispectral imagery, though models with only lidar variables were nearly as sufficient for estimating BA, volume, and AGB. In Random Forests modeling, the most important variables were those derived from lidar. The following accuracy was achieved for linear regression model estimates: BA R2 = 0.36, %RMSE = 31.26, volume R2 = 0.45, %RMSE = 35.30, and AGB R2 = 0.41, %RMSE = 31.31. The results of this study show that the addition of multispectral imagery is not substantially beneficial for improving estimates of BA, volume, and AGB in mixed forests and suggests that the investigation of other variables to explain forest variability is necessary.

Keywords

• laser
• forestry
• allometry
• parametric
• non-parametric
• modeling 3DEP