Exploration on nested sampling designs to study scale-dependent nature of soil carbon, nitrogen, moisture and their interactions

Abstract

The term ‘Scale’ is extensively used in many environmental-related sciences which refers the spatial, temporal and hierarchical dimension of a phenomenon. Soil variation cannot be avoided and therefore it is necessary to understand the dominant spatial scales of variation and focus our sampling efforts and modelling at these scales. Most published models have been developed using data from one scale, therefore they may not necessarily perform similarly at other scales. Therefore it is a prerequisite to capture the spatial heterogeneity/dominant spatial scales as much as possible through an appropriate sampling design for future land management purposes. This study is focused on a nested sampling design with four hydrological spatial scales to study the scale-dependency of soil nitrogen, carbon-water dynamics and modelling of them in an intermediate size forest catchment over a three-year study period. The study recognized that most carbon and nitrogen variation at surface soil came from ≤ 5 m and 30 m scales while soil moisture variation from all scales. At the C soil horizon carbon and nitrogen variation was from the ≤ 5 m and 566 m scales while at the B horizon it was from all scales. It was found that for surface soil carbon and nitrogen, a 30 × 30 m grid with 4 replicate samples per grid will account for almost all of variation. The relationship between carbon and other properties, as measured by the correlation, did change with scales and it implies that different model inputs are needed for different spatial scales. It provides guidance for size of the grid for soil sampling for soil carbon accounting and soil water monitoring and also offers the evidence of scale-dependency behaviour of carbon, water and related models. The study also details a sampling design (nested, grid and random) comparison for geostatistical parameter estimates at simulated field and found that the nested design can estimate them similar to other designs in medium size data sets.