An alternative approach to reduce algorithm-derived biases in monitoring soil organic carbon changes.

Quantifying soil organic carbon (SOC) changes is a fundamental issue in ecology and sustainable agriculture. However, the algorithm-derived biases in comparing SOC status have not been fully addressed. Although the methods based on equivalent soil mass (ESM) and mineral-matter mass (EMMM) reduced biases of the conventional methods based on equivalent soil volume (ESV), they face challenges in ensuring both data comparability and accuracy of SOC estimation due to unequal basis for comparison and using unconserved reference systems. We introduce the basal mineral-matter reference systems (soils at time zero with natural porosity but no organic matter) and develop an approach based on equivalent mineral-matter volume (EMMV). To show the temporal bias, SOC change rates were recalculated with the ESV method and modified methods that referenced to soils at time t1 (ESM, EMMM, and EMMV-t1) or referenced to soils at time zero (EMMV-t0) using two datasets with contrasting SOC status. To show the spatial bias, the ESV- and EMMV-t0-derived SOC stocks were compared using datasets from six sites across biomes. We found that, in the relatively C-rich forests, SOC accumulation rates derived from the modified methods that referenced to t1 soils and from the EMMV-t0 method were 5.7%-13.6% and 20.6% higher than that calculated by the ESV method, respectively. Nevertheless, in the C-poor lands, no significant algorithmic biases of SOC estimation were observed. Finally, both the SOC stock discrepancies (ESV vs. EMMV-t0) and the proportions of this unaccounted SOC were large and site-dependent. These results suggest that although the modified methods that referenced to t1 soils could reduce the biases derived from soil volume changes, they may not properly quantify SOC changes due to using unconserved reference systems. The EMMV-t0 method provides an approach to address the two problems and is potentially useful since it enables SOC comparability and integrating SOC datasets.