Topography-based statistical modelling reveals high spatial variability and seasonal emission patches in forest floor methane flux

Boreal forest soils are globally an important sink for methane (CH4), while these soils are also capable of emitting CH4 under favourable conditions. Soil wetness is a well-known driver of CH4 flux, and the wetness can be estimated with several terrain indices developed for the purpose. The aim of this study was to quantify the spatial variability of the forest floor CH4 flux with a topography-based upscaling method connecting the flux with its driving factors. We conducted spatially extensive forest floor CH4 flux and soil moisture measurements, complemented by ground vegetation classification, in a boreal pine forest. We then modelled the soil moisture with a random forest model using digital-elevation-model-derived topographic indices, based on which we upscaled the forest floor CH4 flux. The modelling was performed for two seasons: May–July and August–October. Additionally, we evaluated the number of flux measurement points needed to get an accurate estimate of the flux at the whole study site merely by averaging. Our results demonstrate high spatial heterogeneity in the forest floor CH4 flux resulting from the soil moisture variability as well as from the related ground vegetation. The mean measured CH4 flux at the sample points was −5.07 µmol m−2 h−1 in May–July and −8.67 µmol m−2 h−1 in August–October, while the modelled flux for the whole area was −7.42 and −9.91 µmol m−2 h−1 for the two seasons, respectively. The spatial variability in the soil moisture and consequently in the CH4 flux was higher in the early summer (modelled range from −12.3 to 6.19 µmol m−2 h−1) compared to the autumn period (range from −14.6 to −2.12 µmol m−2 h−1), and overall the CH4 uptake rate was higher in autumn compared to early summer. In the early summer there were patches emitting high amounts of CH4; however, these wet patches got drier and smaller in size towards the autumn, changing their dynamics to CH4 uptake. The mean values of the measured and modelled CH4 fluxes for the sample point locations were similar, indicating that the model was able to reproduce the results. For the whole site, upscaling predicted stronger CH4 uptake compared to simply averaging over the sample points. The results highlight the small-scale spatial variability of the boreal forest floor CH4 flux and the importance of soil chamber placement in order to obtain spatially representative CH4 flux results. To predict the CH4 fluxes over large areas more reliably, the locations of the sample points should be selected based on the spatial variability of the driving parameters, in addition to linking the measured fluxes with the parameters.Peer reviewe