Bridging the gap between models and measurements of peat hydraulic conductivity

Abstract

Peat saturated hydraulic conductivity, K sat, declines strongly with increasing degree of decomposition, providing a potentially important negative ecohydrological feedback that may buffer peatlands from climate-induced drying. However, the quantitative nature of this relationship is poorly understood. We measured downcore changes in K sat and carbon-to-nitrogen concentration quotients (C/N) in 14 shallow (5 m deep, 0.1 m diameter) peat cores from a Swedish raised bog. We used the C/N measurements to approximate the fraction of original peat mass remaining. A linear mixed effects (LME) model predicts log 10 (K sat) from (i) our C/N-derived estimate of fractional remaining mass; (ii) depth; (iii) microhabitat (hummock, hollow); and (iv) location (treeless bog center, treed bog margin). The LME model indicated no significant random effects or interactions between predictors, so we derived a nonlinear multiple regression (NLMR) model to predict K sat on its original scale. Both LME and NLMR models predict that K sat decreases exponentially with depth and that K sat is lower beneath hollows than beneath hummocks for equivalent depths below the surface. Fractional remaining mass was an important predictor in the LME model, but not in the NLMR model. The distinction between central and marginal areas of the bog was not an important predictor. We demonstrate for the first time that the relationship between fractional remaining mass and K sat is log-linear, and suggest revisions that should be made to peatland development models. In particular, depth - usually ignored in modeling studies - exerted a strong control over K sat independently of decomposition and should be included explicitly in model algorithms. Key Points: Depth is the strongest predictor of hydraulic conductivity in shallow peat Peat decomposition is important to order of magnitude of hydraulic conductivity Our data indicate alterations that should be made to peatland development models