Identifying the spatial drivers and scale-specific variations of soil organic carbon in tropical ecosystems : a case study from Knuckles Forest Reserve in Sri Lanka

Soil organic carbon (SOC) is a key driver of ecosystem functioning and may also contribute to climate change mitigation through the sequestration of carbon. Therefore, having an understanding of the key drivers of SOC may inform management changes that will improve ecosystem function and climate change mitigation. The selected study area is ranged from montane forests to tropical grasslands. Extensive soil sampling (0–0.15 m and 0.15–0.30 m) was undertaken across this region to inform our knowledge about key drivers of SOC at different spatial scales. Initially spatial modelling was carried out using spatial linear mixed modelling approach using a variety of environmental covariates. The model had a Lin's concordance correlation coefficient value of 56–60%, and indicated that SOC was predominately influenced by vegetation type and elevation, although the sub-surface (0.15–0.30 m) SOC was influenced by slope and wetness index. Further, four spatial transects with 100 m sampling interval were extracted from the digital maps representing the study area and empirical mode decomposition (EMD) analysis was carried out to examine the scale specific variability of SOC stocks. The EMD, a mathematical analysis, separates dominant frequencies within a spatial/temporal series representing variability created by various underlying processes operating at different scales into a finite number of scale components or intrinsic mode functions (IMFs). Decomposition of SOC spatial series for the considered transects resulted up to 7 IMFs. The scale components with lower IMF numbers separated higher frequency oscillations, whereas higher IMF numbers separated lower frequency oscillations, which is the representative of smaller and larger scale processes, respectively. Spectral analysis was performed to identify the scales of IMFs and the correlation analysis was carried out with different environmental covariates to identify the dominant controlling factors at different depths. Majority of the large-scale variations (e.g. 2037–8149 m for IMF's 6 for depth interval 0–0.15 m for transect 1–4) were attributed to the elevation and climatic factors controlling the forest type, while small-scale (e.g. 69–118 m for IMF's 1 for depth interval 0–0.15 m for transect 1–4) variations were more attributed terrain derived attributes. Similar scales were identified for the depth 0.15–0.30 m. The scale-specific controlling factors at different locations and their relative controlling factors may help in selecting environmental covariates that enables us to model SOC more accurately rather than fitting one global model. The study provided firsthand information on baseline SOC stock values from a tropical forest ecosystem with six different vegetation types. The information revealed in this study will be useful in the conservation of tropical forests in the region and towards providing vital firsthand information to establishing a national carbon accounting system for land sector in the future