A new approach with DTM-independent metrics for forest growing stock prediction using UAV photogrammetric data

Abstract

We present a novel approach for the prediction of forest growing stock volume based on explanatory variables from unmanned aerial vehicle (UAV) image photogrammetry without relying on the availability of a digital terrain model. This DTM-independent approach was developed to avoid the need for a detailed DTM, which is instead required in traditional photogrammetry to obtain relative heights above the terrain. The method, following an Area Based Approach (ABA), was tested in a boreal forest on a flat area in Norway and in a temperate mixed forest in a mountain steep terrain in Italy, on the basis of aerial images acquired with a SenseFly eBee Ag fixed-wing UAV.The plot level predictive performance of the models based on the DTM-independent metrics were evaluated against the results based on two more traditional approaches based on: (i) metrics from UAV photogrammetric data normalized using a DTM from airborne laser scanning (ALS), and (ii) metrics from ALS data. Percent root mean square error of predictions against measured values (RMSE%) was used for quantifying the performance of the different tests.Results revealed that the DTM-independent approach produced comparable results with both the traditional photogrammetric and ALS methods (the RMSE% ranged between 15.9% and 16.7% in Italy, and between 16.3% and 17.9% in Norway).Our results demonstrated that UAV photogrammetry can be used effectively for predicting forest growing stock volume even when high-resolution DTMs are not available, hence increasing the potentiality of UAVs in forest monitoring and inventory.