Assessing canopy structure in Andean (agro)forests using 3D UAV remote sensing

Abstract

Agroforestry systems (AFS) are important for biodiversity conservation outside protected areas. The presence of shade trees in AFS form structurally complex habitats that provide food for many species. Habitat complexity is considered an essential biodiversity variable and its characterization is now possible using remote sensing techniques, including 3D point clouds derived from images obtained with unmanned aerial vehicles (UAVs). However, studies evaluating the relationship between canopy structure and variables derived from 3D clouds are rare for AFS, especially for the tropical Andes. Here, we show how six important variables of canopy structure can be predicted across a canopy structure gradient from AFS with cacao and coffee to a natural forest using characteristics extracted from the 3D point clouds and multiple linear regression. For leaf area index the best model obtained an R² of 0.82 with a relative RMSE = 24%, for canopy cover an R² of 0.81 and relative RMSE = 13%, for above-ground biomass (AGB) an R² of 0.81 and relative RMSE = 10%, the density of shade trees was predicted with an R² of 0.66 and relative RMSE = 34%, the mean height and the standard deviation of height in the canopy obtained an R² of 0.82 and 0.79 respectively, and relative RMSE of 18% for both. The approach presented in this study allows an accurate characterization of the canopy structure of AFS using UAVs, which can be useful for assessing above-ground biomass and biodiversity in tropical agricultural landscapes to monitor sustainable management practices and derive payments for ecosystem services.