Estimating tree height and biomass of a poplar plantation with image-based UAV technology

Abstract

Poplar is considered one of the forest crops with greatest potential for lignocellulose production, so rapid and non-destructive measurements of tree growth (in terms of height and biomass) is essential to estimate productivity of poplar plantations. As an alternative to tedious and costly manual sampling of poplar trees, this study evaluated the ability of UAV technology to monitor a one-year-old poplar plantation (with trees 4.3 meters high, on average), and specifically, to assess tree height and estimate dry biomass from spectral information (based on the Normalized Difference Vegetation Index, NDVI) and Digital Surface Models (DSM). We used an UAV flying at 100 m altitude over an experimental poplar plantation of 95 × 60 m2 (3350 trees approx.), and collected remote sensing images with a conventional visible-light camera for the generation of the DSM and a multi-spectral camera for the calculation of NDVI. Prior to the DSM generation, several adjustments of image enhancement were tested, which improved DSM accuracy by 19–21%. Next, UAV-based data (i.e., tree height, NDVI, and the result of fusing these variables) were evaluated with a validation set of 48 tree-rows by applying correlation and linear regression analysis. Correlation between actual and DSM-based tree heights was acceptable (R2 = 0.599 and RMSE = 0.21 cm), although DSM did not detect the narrow apexes in the top of the poplar trees (1 m length, on average), which led to notable underestimates. Linear regression equations for tree dry biomass showed the highest correlation with NDVI × Tree-height (R2 = 0.540 and RMSE = 0.23 kg/m2) and the lowest correlation with NDVI (R2 = 0.247 and RMSE = 0.29 kg/m2). The best results were used to determine the distribution of the trees according to their dry biomass, providing information about potential productivity of the entire poplar plantation by applying a fast and non-destructive procedure.