Designing national forest inventories for accurate estimation of soil carbon change

Detecting changes in forest soil carbon stocks is critical for national carbon budgets, yet remains challenging due to high spatial variability and relatively small temporal changes. We use data from Canada’s National Forest Inventory (NFI), which includes repeated measurements of organic and mineral soil horizons across 532 plots. We quantified within- and between-plot variability, assessed minimum detectable differences (MDD), and explored design improvements through simulations. Variation in soil carbon stocks was substantial: coefficients of variation were ~40% for mineral and ~70% for organic horizons, with within-plot comparable to between-plot variability. Consequently, MDDs were also high, at ~4.1 and 4.6 Mg ha^-1 10 yr^-1 for the mineral and organic horizons. This implies that only large, widespread changes would be detectable with the current data. Simulations showed that ~700 plots with 4 subsamples could reduce MDD to the current estimate of soil carbon change. Grouping plots by ecozone provided inconsistent benefits at the national level because of ecozones with high spatial heterogeneity. The data also had patterns consistent with the statistical phenomenon of regression to the mean, which implies that any change in carbon stock may be a statistical artifact. Indeed, soil carbon stocks appeared to grow by 2.3 Mg C ha^-1 10 yr^-1 during the first remeasurement interval, while a small number of second remeasurement interval data showed a completely unrelated pattern. Our analysis suggests that NFI’s design characteristics of sampling multiple microplots and collecting longitudinal data per microplot are critical to providing robust estimates of soil carbon stock changes.