Temporal variability is a major source of uncertainty in soil carbon measurements

Authors: Wuest, Stewart; Durfee, Nicole

Submitted to: Soil Science Society of America Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/22/2024
Publication Date: 3/21/2024
Citation: Wuest, S.B., Durfee, N.M. 2024. Temporal variability is a major source of uncertainty in soil carbon measurements. Soil Science Society of America Journal. 88(3):830-845. https://doi.org/10.1002/saj2.20660.
DOI: https://doi.org/10.1002/saj2.20660

Interpretive Summary: Monitoring changes in soil organic carbon over time is important for agricultural and environmental efforts. Despite decades of measurements, temporal variability in soil carbon measurements has not been studied extensively. Prompted by a study of monthly variation in soil properties, several sets of monthly samples extending about three years each were collected at four locations representative of dryland farming in the Pacific Northwest, USA. The variance from month-to-month averaged 20%. This was often greater than the variance between replications (17%). This means that a single sample on a particular date can deviate substantially from the actual long-term site average. We recommend that confidence intervals for soil organic carbon estimates be calculated based on variance from a large population of samples rather than from a single sample set at one timepoint.

Technical Abstract: Monitoring changes in soil organic carbon over time is important for agricultural and environmental efforts. Despite decades of measurements, temporal variability in soil carbon measurements has not been studied extensively. Prompted by a study of monthly variation in soil properties, several sets of monthly samples extending about three years each were collected at four locations representative of dryland farming in the Pacific Northwest, USA. The variance from month-to-month was 15 to 32% of the random error, averaging 20%. This was often greater than spatial heterogeneity, that is, the variance between replicate experimental units (2 to 42%, averaging 17%). At certain sites, there was some autocorrelation between sequential samples, but no consistent trend patterned on annual seasons. This means that a single sample, even with extensive subsampling, or the sampling of multiple experimental units, can deviate substantially from the long-term site average and this would only be discovered by subsequent sampling. We recommend that confidence intervals for soil organic carbon estimates be calculated based on variance from a large population of samples rather than from a single sample set at one timepoint. For example, in this study, coefficients of variation of the temporal component of error averaged 3.7% which could be converted to a standard deviation and added to the standard deviation used for confidence interval calculations.