Can We Map the Measurable Pools of Soil Organic Carbon At Catchment Scale?.

This study aims to map the measurable pools of soil organic carbon related to the RothC carbon model at the catchment scale and to assess the prediction quality for each carbon pools. The measurable pools of the RothC model namely; resistant organic carbon, humidified organic carbon, and particulate organic carbon can successfully substitute the conceptual pools of the RothC model i.e inert organic matter, humus and resistant plant material respectively. The study was carried out in Cox’s Creek catchment in northern New South Wales, Australia. Samples were collected in 2010 using a design-based sampling scheme. Measurable pools of RothC model were predicted by newly developed MIR/PLSR models under the National Soil Carbon Research project lead by CSIRO (2009 – 2012), Australia. We used linear mixed models to create a prediction model for mapping the measurable pools of soil organic carbon across the landscape.  The cross validation results revealed that highest Lin’s Concordance correlation between measured vs. predicted was recorded for resistant organic carbon, followed by humidified organic carbon and particulate organic carbon.  In addition, mean standardized squared deviation ratio was closer to one (1) indicating that kriging variance is a good predictor of error. With regard to spatial auto-correlation, resistant organic carbon and humidified organic carbon recorded a moderate spatial dependence while for particulate organic carbon there was weak spatial dependence. We conclude that it is possible to map measurable pools of carbon related to RothC model at catchment scale. However, future studies should consider sampling designs to capture the short range variation of soil organic carbon pools specifically particulate organic carbon and assess the uncertainties related to MIR/PLSR prediction of respective pools. The predicted soil organic carbon pool maps can be effectively coupled with the RothC model in initialization and validation of the simulated results in spatial context.