Figure 1 . Differences between the MANIPULATION runs (S1-S4) and the HISTORICAL runs (S0) for winter GT (°C; left column), non-winter GT (°C; middle column), and winter snow depth (%; right column), at each site.
3.3.2 Impacts on biogeochemical variables
Our results further showed large WWE-induced impacts on all biogeochemical variables, with magnitudes often comparable to impacts deriving from altered winter climatologies (Figures 2 and G3). GT and water availability are key drivers of the ecosystem C cycle, as they influence the start of the growing season, nutrient availability, vegetation dynamics, and soil Rh.
The modelled impacts of the manipulation experiments on biogeochemical variables generally followed the same direction as those observed for GT and GWC. We noted substantial reductions in GPP (Figure 2i-l) under the WWE experiments S1 and S3 of up to 25% at the fen, 20% at the peat plateau, and 10% at the birch site, and smaller reductions under ROS events alone (S2). The tundra site showed weaker reductions in GPP than the other sites, ranging from a few percent to up to 5% in the milder WWE experiments (driven by SSP119), and increases of up to 10% under CanESM5 SSP585. In contrast, for S4, the model simulated weak changes in GPP in the mildest CMIP6 scenarios, but sizable increases in GPP in the CanESM5 SSP585: up to 75% in the peat plateau, c. 50% in the birch forest, c. 25% in the tundra, and c. 10% in the fen site. The impacts on Ra mirrored those described for GPP (Figure 2e-h). Noticeably, the relationship between GT, and both GPP and Ra, is not linear because the processes involved in plant dynamics are multiple and complex. For example, observational and manipulation studies have reported considerable vegetation damage, delays in bud phenology, and reductions in vegetation greenness due to phenological and frost stress following WWEs (Bokhorst et al., 2009, 2010). These impacts on vegetation are not explicitly represented in the model, as it does not yet take into account these winter/spring stress-related processes.
Increases in GT can stimulate microbial activity and accelerate litter and soil decomposition rates, resulting in increased Rh(Natali et al., 2019). Accordingly, winter Rh increased or decreased by <5% at the fen, and up to 25% at the tundra and peat plateau, following the GT responses (Figure G1). Contrastingly, in response to the CanESM5 SSP585 scenario at the birch forest site, despite the overall lower winter GT, winter Rh increased by up to >200%. This may be explained because the modeled winter Rh in the birch forest is very low (6 g C m-2, compared to >30 g C m-2 at the other sites), and even small increases in Rh during WWEs can result in substantial relative increases in winter emissions. A recent synthesis of in situ observations across the Arctic indicates that winter Rhaccounts for a substantial portion of the arctic’s annual C budget (Natali et al., 2019). Winter Rh alone currently offsets ~40% of the measured annual vegetation C uptake at our sites. Applying the winter Rh response curve to GT by Natali et al., (2019) (Q10 = 2.9), WWE-induced impacts on winter GT of the magnitudes reported here could change total winter Rh by up to 25%. Hence, realistically simulating the effects of WWEs on GT is important for improving estimates of GHG emissions in high latitudes.
The modelled CH4 emissions at the peat plateau and fen sites decreased under all WWE experiments due to WWE-induced decreases in GT and GWC (Figure G2). At the fen site, emissions halved mostly due to lower water tables in the non-winter season.
Overall, the year-round ecosystem C exchange (i.e. NEE) of the birch forest, tundra, and peat plateau sites decreased (i.e., became a smaller C sink) considerably under most WWE experiments (Figure G3), generally between 20% and 50%, and up to 90% occasionally in the tundra site, due to larger reductions in vegetation C assimilation compared to the C losses from Rh. Contrastingly, NEE at the fen site increased weakly (i.e., became a stronger C sink), mostly due to large reductions in Rh. Noticeably, the changes in NEE caused by WWEs (S1-S3) are substantial and of similar magnitude to those caused by shifts in future winter climatologies (S4). This further indicates that WWEs, despite their short duration, may have the potential to induce changes in high-latitude ecosystem C cycling of magnitudes comparable to those induced by long-term climatic trends.
Given the observed discrepancies in modeled vs measured GT responses to WWEs (section 3.2.1), the modeled impacts on ecosystems C fluxes reported here should not be interpreted as a direction prediction of future impacts of WWEs, but rather as a sensitivity test of the current model’s responses to altered levels of WWEs.