Soil organic carbon (SOC) measurement is critical for result-based payments for carbon sequestration in agriculture. This study examined (i) the consistency of SOC measurements from a portable near-infrared (NIR) spectroscopy soil scanner and laboratory analyses (Walkley-Black method) using samples from 151 plots across Western Kenya, and (ii) explored the implications of measurement error for SOC stock change estimates in group-based carbon farming schemes. Scanner predictions showed a weak correlation with laboratory results (R2 = 0.10–0.11; RMSE = 5.6–7.1 g/kg). Comparing results between the two laboratories showed only moderate agreement (R2 = 0.34) with a systematic bias of 7.5 g/kg. There was low repeatability (R2 = 0.19; CV = 21.9 %) for samples within the same laboratory. Repeated scans covered an average range of 8.8 g/kg between their highest and lowest readings. Simulations showed that an unbiased measurement error (mean 0, SD 4 g/kg) caused a 22 % deviation in stock change estimates for a group of 50 farmers. However, with increased farmer group size (>1000), the average values converged to the true value (<2 % deviation). A systematic bias of 1 g/kg resulted in 18 % error that persisted regardless of group size. These findings demonstrated a significant variability in the output of different techniques, which could severely undermine the viability of result-based payments, especially for smallholders. Reducing the within-approach measurement error to less than 1 g/kg will be key to enabling effective carbon farming group sizes of up to 30 farmers. To achieve this, there is a need for targeted capacity-building efforts to standardize SOC measurements across laboratories, calibrate soil scanners for the local context of each country and integrate error margins into the design of carbon farming interventions.
Published in Journal of Environmental Management, Volume 398, 15 January 2026, 128562.