Data supporting "Soil organic carbon sequestration under conservation agriculture: a structure-enabled global meta-analysis for sampling depth constraints"

This dataset supports the global meta-analysis entitled “Soil organic carbon sequestration under conservation agriculture: a structure-enabled global meta-analysis for sampling depth constraints.” The study evaluates soil organic carbon responses to conservation agriculture relative to conventional tillage, with particular emphasis on sampling-depth dependence, soil structural accessibility and the analytical conditions governing reported carbon-sequestration responses. The final analytical database contains 574 valid soil organic carbon comparisons derived from 127 articles across 36 countries. It includes nested inverse-variance, structural-accessibility and machine-learning analytical subsets. The principal workbook contains the harmonised and quality-controlled comparison-level database, analytical eligibility fields, pooled multilevel meta-analysis estimates, article-clustered CR2 inference, sampling-depth and soil organic carbon metric analyses, fully adjusted moderator models, structural-accessibility index sensitivity analyses, leave-one-article-out diagnostics and retained quality-control information. A second workbook provides repeated article-blocked machine-learning predictions, validation-fold assignments, model-performance statistics and permutation importance. Additional files document the article-level analytical tiers, PRISMA screening totals, database-search strategies and SHA-256 checksums. The primary effect size is the natural-log response ratio, lnRR = ln(SOC_CA/SOC_CT), where positive values indicate greater reported soil organic carbon under conservation agriculture than under conventional tillage. Because individual articles may contribute multiple comparisons, article-level dependence should be retained in any subsequent analysis.