Fusing Airborne Laser Scanning and Rapideye Sensor Parameters for Tropical Forest Biomass Estimation of Nepal

This study aimed to integrate and optimise multi-sensor data consisting of airborne laser scanning (ALS) and RapidEye image-derived parameters with field-measured biomass using Random forests (RF) to predict the tropical forest biomass (FB). The multi-sensor parameters which include combined spectral, textural and ALS derived 119 variables provided a better result, R
²
= 0.95 and relative RMSE (relRMSE) = 17.25%, as compared to any single sensor parameter for the FB estimation. The top 20 variables were used to compute FB and its distribution map with the R
²
, RMSE and relRMSE values of 0.93, 35.46 Mg ha
^-1
and 17.40% respectively. Furthermore, its validation was checked with the in-situ biomass to obtain satisfactory accuracy (R
²
= 0.72, RMSE = 47.71 Mg ha
^-1
and relRMSE = 23.41%). The result showed that the integration of multi-sensor data using RF regression algorithm proves to be a reliable algorithm for accurate FB estimation.