Digital soil mapping of organic carbon concentration in paddy growing soils of Northern Sri Lanka

In this study we explore the environmental covariates that are useful in mapping SOC contents in tropical paddy growing soils in Northern Sri Lanka. We carried out digital mapping of SOC contents across the study area using a spatial soil prediction function with auto correlated errors via linear mixed models (LMMs). Two separate LMMs were fitted considering two depth intervals, 0 to 0.15 m and 0.15 to 0.30 m respectively. Results revealed that environmental covariates of SOC content for the two depth intervals considered in this study were different despite being situated within a zone typically considered as the top soil (0 to 0.3 m). Landsat 8 bands and its products derived from those bands (namely Landsat 8 band 2, band ratios 4/3 and 4/7) were identified as significant environmental covariates for both depth intervals. In terms of spatially auto correlated residuals, optimum spatial models for both depth intervals were identified as a spherical model. Residuals were found to be spatially auto correlated up to 10,000 m and 5000 m for 0 to 0.15 m and 0.15 to 0.30 m depth intervals respectively. Cross validation results suggest that fitted LMMs are adequate to carry out mapping across the study area with little difference between estimated bias (-0.006% for 0 to 0.15 m and -0.004% for 0.15 to 0.30 m depth intervals) and accuracy (0.441% for 0 to 0.15 m and 0.367% for 0.15-0.30 m depth intervals) for fitted LMMs considering two depth intervals. However, in case of Lin's concordance correlation value, 0 to 0.15 m depth interval reported higher value (0.81) compared to 0.15 to 0.30 m depth interval (0.60) suggesting upper depth interval model is much superior to the lower one. In conclusion, this study provides; a) first hand information on current status of SOC contents in northern paddy growing soils in Sri Lanka, b) generated information that are useful to optimize spatial sampling of SOC in future applications and c) provided firsthand information vital to establishment of a national carbon accounting system in future. (c) 2016 Elsevier B.V. All rights reserved.