Using topographic attributes to predict the density of vegetation layers in a wet eucalypt forest

Abstract

ABSTRACT Mapping the structure of forest vegetation with field surveys or high-resolution light detection and ranging (LiDAR) data is costly. We tested whether landscape topography and underlying geology could predict the vegetation density of a 19 km2 area of wet eucalypt forest at the Warra Long-Term Ecological Research Supersite, Tasmania, Australia. Using spatial layers for 12 topographic attributes derived from digital terrain models (DTMs) and a geology layer, we predicted the vegetation density of three strata with a high degree of accuracy (validation root mean square error ranged from 9.0% to 13.7%). The DTMs with 30 m resolution provided greater predictive accuracy than DTMs with higher resolution. The importance of different variables depended on spatial resolution and strata. Among the predictor variables, geology generally had the highest predictive importance, followed by solar radiation. Topographic Position Index, aspect, and System for Automated Geoscientific Analyses (SAGA) Wetness Index had moderate importance. This study demonstrates that geological and topographic attributes can provide useful predictions for the density of vegetation layers in a tall wet sclerophyll primary forest. Given the good performance of the model based on 30 m DTM resolution, the predictive power of the models could be tested on a larger geographical area using lower-density LiDAR point clouds combined with medium-resolution satellite data.