Spatial distribution and scale-specific controls of soil water-stable aggregates in southeastern China

Soil water-stable aggregates (WSA) are a key indicator of soil stability and are critical for maintaining soil quality and controlling erosion. WSA have spatial heterogeneity and their spatial variability is influenced by different environmental factors at different scales. Few studies, however, have focused on the scale-specific controls of the WSA content on a mountain-plain landform. The objective of this study was to investigate the spatial distribution and scale-specific controls of WSA content using methods of Gaussian geostatistical simulation (GGS) and multivariate empirical mode decomposition (MEMD). A total of 101 soil samples were collected along a sinuous 4200-km transect at a mean interval of 42 km in southeastern China. The WSA content and nine environmental factors, i.e., soil bulk density, clay, silt, sand and organic-matter (SOM) contents, pH, elevation, slope and vegetation coverage (VC) were measured at each sampling site. WSA content was strongly spatially dependent, with a nugget/sill ratio of 7.63%. E-type maps from the GGS with different realizations (10, 25, 50, 100, 200 and 500) had similar spatial patterns, with high WSA content mostly in the south and west of the study area and low WSA content concentrated in the north. The spatial probability map indicated that the locations around Taihu Lake in the north of the study area had a high probability of poor aggregate stability. A multi-scale analysis of influencing factors indicated that intrinsic mode function 1 (IMF1) (23.76%) and IMF2 (23.64%) explained most of the spatial variability of WSA content. The correlations between WSA content and the environmental factors varied with the scale. The dominant controls were soil particle composition at small scale, pH and SOM content at moderate scale and elevation and VC at large scale. The prediction of WSA content at the measurement scale based on the scale-specific controls analyzed with MEMD (R²ₐdⱼ = 0.577) outperformed the prediction of traditional multiple regression (R²ₐdⱼ = 0.356). These results will support accurate practices and management for preventing soil erosion and land degradation.