Sensitivity analysis of a wetland methane emission model based on temperate and Arctic wetland sites

• Published in: Biogeosciences discussions Vol. 6; no. 5; pp. 9083 - 9126
• Main Authors: Huissteden, Jvan, R Petrescu, AM, Heniks, D, Rebel, K T
• Format: Journal Article
• Language: English
• Published: 17.09.2009
• Subjects: Freshwater; PN, Arctic
• ISSN: 1810-6277; 1810-6285

Modelling of wetland CH sub(4) fluxes using wetland soil emission models is used to determine the size of this natural source of CH sub(4) emission on local to global scale. Most process models of CH sub(4) formation and soil-atmosphere CH sub(4) transport processes operate on a plot scale. For large scale emission modelling (regional to global scale) upscaling of this type of model requires thorough analysis of the sensitivity of these models to parameter uncertainty. We applied the GLUE (Generalized Likelihood Uncertainty Analysis) methodology to a well-known CH sub(4) emission model, the Walter-Heimann model, as implemented in the PEATLAND-VU model. The model is tested using data from two temperate wetland sites and one arctic site. The tests include experiments with different objective functions, which quantify the fit of the model results to the data. The results indicate that the model 1) in most cases is capable of estimating CH sub(4) fluxes better than an estimate based on the data avarage, but does not clearly outcompete a regression model based on local data; 2) is capable of reproducing larger scale (seasonal) temporal variability in the data, but not the small-scale (daily) temporal variability; 3) is not strongly sensitive to soil parameters, 4) is sensitive to parameters determining CH sub(4) transport and oxidation in vegetation, and the temperature sensitivity of the microbial population. The GLUE method also allowed testing of several smaller modifications of the original model. We conclude that upscaling of this plot-based wetland CH sub(4) emission model is feasible, but considerable improvements of wetland CH sub(4) modelling will result from improvement of wetland vegetation data.