Sources of spatial variation in methane emission from mires in northern Sweden: A mechanistic approach in statistical modeling

Abstract

Abstract. Methane emissions from six mires in northern Sweden were measured using a closed chamber technique during the frost free season in 1992. The average methane flux over the measurement period, calculated either for each mire or for different plant communities within one mire, ranged from 9 to 83 mg CH4 m '2 d '•. The emission rate on each occasion was related to physical and chemical environmental variables, both in a general data set for all mires (n=836) and in subdata sets for individual mires, using multiple linear regression. The variables with significant contributions to the models were water table, standing water above the vegetation surface, peat temperatures, and principal components of the near infrared reflectance spectra of peat samples reflecting variations in organic chemical composition. To account for the actual contribution of methane production and methane oxidation, variables describing the active parts of the vertically distributed potentials of methane production or oxidation were constructed. The interaction terms between these variables, respectively, describing the active proportion of methanogens and methanotrophs, and the temperature values representing the anoxic and oxic parts of the profile were significantly correlated to the methane emission rate; positively for the production zone and negatively for the consumption zone. By using this mechanistic approach, a significant temperature effect in both the methane production and consumption zone was detected. These constructed temperature variables explain 21% of the variance in the logarithmically transformed methane fluxes using the entire data set (n=836) but only 5% of the variance using peat temperatures from fixed depths. Adding variables describing the organic chemical composition of the peat to the models improved the predictability in 10 of the 11 model sets tested, decreasing the unexplained variance by maximally 50% for a poor fen community model and increasing R: from 0.40 to 0.68. In the general lnodel, R: increased from 0.42 to 0.49 with the inclusion of organic chemical composition. The explained variance in the final models, including the substrate variables, ranged from Re=0.49 to R2=0.75, with one exception, Torsmyran (R2=0.36) which is a mire that most closely relates to an eccentric ombrotrophic bog among the mires studied.

CO2 Deficit in Temperate Forest Soils Receiving High Atmospheric N-Deposition

“…Forest sites with anthropogenic disturbances other than high N-load, such as clearcutting, or even thinning-leading to higher groundwater level as a result of decreased evapotranspiration, and manifesting methane production in the summer (15) (Fig. 1), as well as drained and limed sites-were treated separately.…”

Methane emission from Swedish mires: National and regional budgets and dependence on mire vegetation

“…Four of these site types (types 4 and 6-8) were selected for monitoring methane emissions in this study. According to our previous experience [Granberg et al, 1997;Mikkeli•, 1998] the other vegetation types (types 1-3 and 5) do not emit methane, which is why they were excluded. The distributions of our four target peatland site types are presented in Table 2 …”

“…Standardization of the two laboratories was ensured by using the same standard gas (CH4) for calibration. Samples from the two northern sampling regions were analyzed on a Perkin Elmer Sigma 2B gas chromatograph according to Granberg et al [1997]. Samples from the southern areas were analyzed on a Packard 428 equipped with a Porapak T column and a flame ionization detector.…”

Estimating wetland methane emissions from the northern high latitudes from 1990 to 2009 using artificial neural networks

“…[4] Estimates of wetland CH 4 emissions are often obtained using "bottom-up" approaches, ranging from simple empirical or statistical models [e.g., Andronova and Karol, 1993;Granberg et al, 1997;Levy et al, 2011] to detailed process-based models [e.g., Cao et al, 1996;Walter et al, 2001;Zhuang et al, 2004]. Previous process-based model simulations presented a large uncertainty in the estimates of [5] The uncertainty in these estimates could result from many sources including model structures, assumptions, parameterization, and choice of forcing data.…”