Scale-dependent predictability of DEM-based landform attributes for soil spatial variability in a coastal dune system

Abstract

Much soil–landform modeling has shown that the predictability of topographic parameters derived from digital elevation models (DEM) for soil spatial variability is influenced by the selection of DEM's grid size. This study investigates soil–terrain relationships in a coastal dune at multiple DEM resolutions to examine if such scale-dependence is a ubiquitous phenomenon even in low-relief systems with relatively homogeneous substrates. We collected 193 soil samples from the Sindu coastal dunefield in western Korea and analyzed 11 physical and chemical soil properties. Based on principal component analysis, four soil attributes (K +, pH, total P, and nitrate) that reflect dominant pedogeomorphological processes on the dune were selected as dependent variables in spatial regression. Predictors included vegetation types, distance from the coastline, and seven terrain indices extracted from various grid sizes of DEM (5, 10, 20, 30, 40 m). Results demonstrated strong scale-dependence in the model prediction: Spatial patterns of the four soil variables were best predicted by landform parameters extracted from different DEM resolutions. Specifically, we found the lowest Akaike's Information Criterion for K +, pH, total P, and nitrate at scales of 5, 20, 30, and 40 m, respectively. Residuals of each soil attribute acquired at these scales also possessed the lowest degree of spatial autocorrelation identified by Moran's I. The scale-dependent predictability in our soil–landscape modeling implies that fine-scale topographic information would not always be optimal for understanding soil spatial variability. Future pedogeomorphological modelers therefore need to take into account the unique spatial extent and behavior of individual soil attributes to determine optimal spatial resolutions of DEM. We propose that scale-dependent soil–terrain relationships may be a widely applicable notion in soil geomorphology, not only to various terrestrial rugged hillslopes where the notion has been dominantly examined and developed, but also to systems with low topographic ruggedness and uniform substrate such as coastal dunes.