Widespread Greening but Uneven Soil Carbon Change: Twenty-Year Evidence from a Semi-Arid Eurasian Steppe

Accurate assessment of long-term soil organic carbon (SOC) change requires repeated field measurements and explicit treatment of temporal differences in soil mass. We quantified the net 2004–2024 SOC stock difference at 20 permanent monitoring sites in the semi-arid chestnut-soil zone of Central Kazakhstan using fixed-depth (FD) and equivalent soil mass (ESM) calculations, together with Landsat NDVI and hydroclimatic summaries. At 0–50 cm, FD SOC stock increased from 69.70 ± 17.43 to 79.39 ± 25.00 Mg C ha−1, giving a mean paired difference of +9.69 Mg C ha−1 (95% CI: 2.78 to 16.60; p = 0.008). After ESM adjustment to the site-specific 2004 reference soil mass, the 2024 mean was 75.43 ± 25.24 Mg C ha−1 and the paired difference was +5.73 Mg C ha−1 (95% CI: −0.78 to 12.25; p = 0.081). None of the five depth-specific FD comparisons remained significant after Holm adjustment. Positive annual-composite NDVI slopes occupied approximately 94.7% of the mapped area, but the site-level association between ΔNDVI and ESM-corrected ΔSOC was weak and non-significant (Pearson r = 0.259, R2 = 0.067, p = 0.270; Spearman ρ = 0.241, p = 0.307). Mean aridity index declined from 0.345 in 2004–2008 to 0.308 in 2020–2024 and was interpreted as hydroclimatic context rather than a demonstrated driver of SOC change. Management–soil combinations were examined only as an exploratory site-level comparison because most combinations were unreplicated. These results show that ESM correction materially affects inference and that a two-date resampling design supports interpretation of net endpoint differences rather than continuous SOC change.