Characterizing soil spatial variability using bi-dimensional empirical mode decomposition

摘要: Soil spatial variability evolves from the combined action of soil physical, mineralogical, chemical and biological processes that operate with different intensities and at different scales. Soil spatial variability is therefore scale-dependent and variation is often nonstationary and nonlinear. Empirical mode decomposition (EMD) has recently been used in one dimension to separate the overall soil spatial variability into various components, each representing different scales. The technique is known to accommodate nonstationary and nonlinear data series. In this research, we used the two dimensional extension of EMD, known as bi-dimensional EMD (BEMD), to separate multi-scale spatial variability in a two dimensional field. To demonstrate the approach, we used a digital elevation model (DEM) of the Simmons Creek catchment in New South Wales, Australia. The BEMD separated the overall variations in the DEM into 6 different bi-dimensional intrinsic mode functions (BIMFs) according to their characteristic scales. The first function (BIMF1) separated the variations at the scale< 1.67 km, which can be attributed to the terrain/landscape variations at field-to farm-scales. Sub-catchment level variations were separated in BIMF2 (2–5 km) and BIMF3 (8–10 km), and the major topographic trends in the catchment were separated in BIMF4 (12–15 km) and BIMF5 (> 15 km). The slope and wetness index calculated from the BIMFs showed more detailed features in topographic variations than those calculated from the DEM. Separating two-dimensional spatial variability of terrain attributes at multiple scales using BEMD might provide opportunities for multi …