Simulations of spatial patterns and species distributions in sandy land using unmanned aerial vehicle images

Abstract

The spatial distribution of vegetation in sandy lands is closely related to micro-topography. Point pattern analysis of vegetation distribution from ground surveys and satellite images is a commonly used method but does not capture the influence of spatial heterogeneity at small scales. This study examined long-term ecological observation sites of elm (Ulmus pumila) sparse forest in the Otindag Sandy Land, China. Elevation models from unmanned aerial vehicles (UAVs) and ground survey data on vegetation structure from 3768 elms were used to classify the terrain of sampled sites using a decision tree classification. It combined terrain factors, including slope, aspect, and small-scale altitude differences. Plots were divided into five topographic types: sand flat (53%), sand lowland (17%), sunny slope (13%), shady slope (10%), and sandy ridges (7%). Elm densities varied from 141.7 trees hm−2 on shady slopes to 17.0 trees hm−2 on sand lowland. A 2D Poisson fitting method was applied to the diameter at breast height, crown width, and other vegetation growth characteristics to simulate and verify the distribution of elms in the plots. Multivariate analysis was undertaken to confirm the effect of topographic factors on variation of tree characteristics. The integrated terrain approach could better characterize the spatial distribution of sparse forests. This research demonstrated that UAVs were a useful tool to measure spatial heterogeneity of sand micro-topography. Simulations of the distribution of plant characteristics indicated that the terrain classification matched the spatial pattern analysis of elms in semi-arid regions. Simulation of vegetation distribution is a useful technique for analyzing arid regions. This study will assist with further research on ecological restoration and vegetation protection in semi-arid areas.