Abstract. Models of carbon cycling in terrestrial ecosystems contain formulations for the dependence of respiration on temperature, but the sensitivity of predicted carbon pools and fluxes to these formulations and their parameterization is not well understood. Thus, we performed an uncertainty analysis of soil organic matter decomposition with respect to its temperature dependency using the ecosystem model LPJ-GUESS. We used five temperature response functions (Exponential, Arrhenius, Lloyd-Taylor, Gaussian, Van't Hoff). We determined the parameter confidence ranges of the formulations by nonlinear regression analysis based on eight experimental datasets from Northern Hemisphere ecosystems. We sampled over the confidence ranges of the parameters and ran simulations for each pair of temperature response function and calibration site. We analyzed both the long-term and the short-term heterotrophic soil carbon dynamics over a virtual elevation gradient in southern Switzerland. The temperature relationship of Lloyd-Taylor fitted the overall data set best as the other functions either resulted in poor fits (Exponential, Arrhenius) or were not applicable for all datasets (Gaussian, Van't Hoff). There were two main sources of uncertainty for model simulations: (1) the lack of confidence in the parameter estimates of the temperature response, which increased with increasing temperature, and (2) the size of the simulated soil carbon pools, which increased with elevation, as slower turn-over times lead to higher carbon stocks and higher associated uncertainties. Our results therefore indicate that such projections are more uncertain for higher elevations and hence also higher latitudes, which are of key importance for the global terrestrial carbon budget.
A novel method for acquiring rigorous temperature response functions for electricity demand at a regional scale
