@phdthesis{Forbrich2011,
  author    = {Inke Forbrich},
  title     = {Methane exchange of a boreal peatland - Integrated measurements and modelling on microform and ecosystem scale at the Salmisuo mire complex, Eastern Finland},
  journal    = {Methanaustausch in einem borealen Moor - Integrierte Messungen und Modellierungen auf Plot- und {\"O}kosystemskala im Moorkomplex Salmisuo, Ostfinnland},
  url       = {https://nbn-resolving.org/urn:nbn:de:gbv:9-000922-1},
  year      = {2011},
  abstract  = {Peatlands cover only about 3\% of the terrestrial surface but are significant players in the global carbon (C) cycle and the climate system, since they store roughly one quarter of the global soil carbon (C) and are among the largest natural sources of methane (CH4). Since the resulting feedbacks on the climate system are uncertain, research efforts aim at identifying key processes and quantifying the C exchange from ecosystem to regional and global scales. To identify peatland ecosystem dynamics requires analysis of yet different scales. The key scale for their C dynamics is the microform scale, which is the smallest entity of the system. To estimate ecosystem dynamics, up-scaling from the microform scale is needed. Up-scaling demands (1) a correct estimation of the spatial heterogeneity and (2) the correct aggregation. In this thesis, the traditional spatial weighting of microform fluxes by the microform distribution is evaluated by (1) analyzing the flux calculation procedure, (2) investigating the effect of the resolution of the landcover maps on the up-scaling and by (3) cross-evaluating the up-scaling result with the directly measured ecosystem flux. Eventually, it is evaluated how these dynamics are considered in a mechanistic ecosystem model (LPJ-WHyMe). CH4 fluxes were measured on the microform scale with the closed chamber technique and on the ecosystem scale with the eddy covariance (EC) technique. The quantification of microform fluxes relies on the correct flux calculation. Since only few gas samples are taken during the closure period, traditionally the linear regression is applied when calculating CH4 fluxes from chamber measurements. Still, the chamber itself affects the diffusion gradient between peat and chamber atmosphere resulting in a theoretically non-linear concentration increase in the chamber. Using data with six data points per measurement from different microform types it is tested whether the linear or exponential regression fits the data better. In the majority of cases, the linear regression fits best. However, the exponential concentration change might still not be detectable resulting in an underestimation of the ’real‘ flux and the test of different techniqes to estimate the slope of a non-linear function with small sample amounts is recommended. To define the spatial heterogeneity of the peatland surface, the application of remote sensing techniques offer the advantage of supplying area-wide information with less uncertainty when compared to vegetation mapping along transects. However, the required resolution to resolve the microform distribution is <1m which in this study was derived from near-aerial photography. Besides for up-scaling, the resulting high-resolution landcover map was used in combination with a footprint model to analyze (1) the effect of landcover on the directly measured ecosystem flux and (2) its spatial representativeness. It was shown that fluctuations of the measured ecosystem flux over periods of several days could be explained by changes of the landcover composition in the source area of the EC measurements. The estimated budget was slightly biased towards the higher emissions from lawns which could be corrected. Still, the seasonal ecosystem CH4 budget was higher than the estimate derived from the up-scaling of microform fluxes. This is most likely due to an underestimation of microform fluxes by the chamber technique. Generally, the budget estimate derived from EC measurements was more accurate, i.e., characterized by less uncertainty than the up-scaled estimate. The developed approach depends on (1) identification and accurate measurements of all relevant microform types and (2) on spatial information which should be smaller than the footprint size of the EC measurements and available on the scale relevant for the studied process, i.e., the microform scale. The demonstrated effect of microform dynamics on the ecosystem flux highlights the importance of dealing with spatial heterogeneity of ecosystems in mechanistic modelling. For example, in LPJ-WHyMe, the ecosystem flux is simulated with mean input variables as water table level. To investigate its model performance, flux data from the rather homogeneous peatland margin and the more heterogeneous peatland centre were compared with the model output. At the homogeneous peatland margin, the ecosystem flux was clearly dominated (with a contribution of 91\%) by one microform flux. In this case, one water table level as input variable could be used to estimate the ecosystem flux. However, for a heterogeneous site such as the peatland centre in this study, only one mean water table would simulate a mean microform flux but not the ecosystem flux. Consequently, it is recommended to incorporate at least one high-emitting and one low-emitting microform type in the model to increase the model performance.},
  language  = {en}
}