Quantifying crown dimensions using high-resolution aerial imagery to estimate the diametric growth of trees in central African forests

Abstract

Characterising forest dynamics of a forest is essential to its management. Tree crowns are a key factor in these dynamics, but measuring them in tropical forests is not an easy matter. This study tested the use of high-resolution aerial imagery to estimate the tree diameter growth by incorporating detailed measurements of the detected tree crowns. Ortho-images at a resolution of 10 cm/pixel were captured by a fixed-wing drone over a 9 ha plot in the Yoko forest in the Democratic Republic of Congo. Inventories conducted on trees ≥ 10 cm diameter at breast height (DBH) in 2008 and 2016 provided access to a variety of tree dendrometric characteristics, including DBH and species temperament, and allowed the calculation of diameter increments. Mixed linear models were calibrated to predict diameter increment of 163 trees identified both on the ground and on the ortho-images, using variables quantified on the ground only and/ or from variables measured from the ortho-images. From the aerial images, we were able to detect 23.4% of the trees with DBH ≥ 10 cm listed in the ground inventories, representing 75.1% of the stand’s aerial biomass. The probability of detecting the trees varied with their DBH, from 0.09 for trees with DBH < 30 cm to 0.97 for trees with DBH ≥ 60 cm. Predictions of diametric growth improved significantly when the variables quantified by remote sensing were added to the ground variables. The best models for estimating diameter increment include, in particular, a term characterising the size of tree crowns, which can only be measured by remote sensing. Of the variables determined by remote sensing, convex crown area was the most successfull in the models and therefore appears to be the most accurate variable to describe competition between tree crowns. These results open up possibilities to build new tools of data acquisition to support forest planning.