Space-time monitoring of soil organic carbon content across a semi-arid region of Australia

Monitoring and mapping organic carbon throughout the soil profile is an important task, as land management and fluctuations in rainfall patterns have the potential to substantially alter the levels of soil organic carbon (SOC). This study aims to monitor the change in SOC content between 2002 and 2015 in a semi-arid, irrigated cotton-growing region in Australia under various land uses. One multivariate linear mixed model (MLMM) was used to model SOC content with four response variables – at two sampling depths (0–0.1, and 0.3–0.5 m), and two time points (2002 and 2015). This involved a four-dimensional model (horizontal and vertical space, and time) to monitor the spatio-temporal shifts of SOC content. Such a model utilises the correlation between the four measurements of SOC at the different depths and times. When tested with leave-one-site-out cross-validation, SOC content was predicted with a Lin's concordance correlation coefficient (LCCC) of 0.68, and a root mean square error (RMSE) of 0.21%. The MLMMs produced coherent soil maps, and an ability to detect statistically significant change in SOC content over time. While no change in subsoil (0.3–0.5 m) SOC content was observed across the study area, a small section of the study area revealed a statistically significant increase in SOC content in the topsoil (0–0.1 m). It was not completely clear whether this change was a function of land management, or the extended wet period from 2010 to 2012. The ability of MLMMs to take advantage of the relationship between different soil properties, soil depths, and time points, makes them advantageous for spatio-temporally monitoring soil. •Multivariate linear mixed model of SOC at different depths and time points.•Coherent maps and an ability to detect statistically significant change resulted.•Statistically significant increase in SOC over time observed in only a small area.