Modeling Aboveground Forest Biomass Using Airborne Light Detection and Ranging (LiDAR) and NAIP Images in Tennessee, USA
Abstract
Predicting and mapping the spatial distribution of woody biomass is a prerequisite for a continuous supply of feedstock for biofuel production. Recent Remote sensing technologies such as very high-resolution images (NAIP image), Light Detection and Ranging (LiDAR) method of acquiring data has gained popularity among resources managers, researchers, and landowners to estimate forest biomass and carbon stock across the forest landscape. We hypothesized that a data matrix derived from LiDAR point clouds and textural analysis of NAIP image would improve the prediction accuracy of forest stand-level variables such as biomass or carbon stock per unit area. We used the Grey Level Co-occurrence Matrix (GLCM) approach to predict forest characteristics. We paired Forest Inventory and Analysis (FIA) data with LiDAR-derived variables; NAIP-derived variables including tree canopy cover from NLCD, and LIDAR-NAIP variables combined, from selected counties in Tennessee. Both parametric and non-parametric models were fitted and compared for these three data types. We did linear regression as parametric and Random Forest (RF) as a non-parametric approach and compared. Regression analysis outperformed the Random Forest approach. In both methods, LiDAR-NAIP gave better model fit statistics (R2= 0.6614 from regression, r2= 0.3589 from RF). However, we found the LiDAR-only model (r2= 0.6572 and RMSE= 3.56 ton/ha) as the best out of all. Even though LiDAR-NAIP gave better fit statistics, the ANOVA test between LiDAR-only and NAIP- LiDAR model showed no significant improvement by adding NAIP-only variables with the LiDAR-only. It is important to estimate aboveground forest biomass and carbon stock to estimate the role of forest in the regional and global carbon cycle and to develop science-based forest management and climate change mitigation strategy through forest management at local, regional, and national levels.
Subject Area
Plant sciences|Forestry|Remote sensing
Recommended Citation
Durga Prasad Joshi,
"Modeling Aboveground Forest Biomass Using Airborne Light Detection and Ranging (LiDAR) and NAIP Images in Tennessee, USA"
(2021).
ETD Collection for Tennessee State University.
Paper AAI28414729.
https://digitalscholarship.tnstate.edu/dissertations/AAI28414729