Global peatlands under future climate-seamless model projections from the Last Glacial Maximum

Peatlands are diverse wetland ecosystems distributed mostly over the northern latitudes and tropics. Globally they store a large portion of the global soil organic carbon and provide important ecosystem services. The future of these systems under continued anthropogenic warming and direct human disturbance has potentially large impacts on atmospheric CO2 and climate. We performed global long term projections of peatland area and carbon over the next 5000 years using a dynamic global 5 vegetation model forced with climate anomalies from ten models of the Coupled Model Intercomparison Project (CMIP6) and three scenarios. These projections are continued from a transient simulation from the Last Glacial Maximum to the present to account for the full transient history. Our results suggest short to long term net losses of global peatland area and carbon, with higher losses under higher emission scenarios. Large parts of today’s active northern peatlands are at risk. Conditions for peatlands in the tropics and, in case of mitigation, eastern Asia and western north America improve. Factorial simulations 10 reveal committed historical changes and future rising temperature as the main driver of future peatland loss and increasing precipitations as driver for regional peatland expansion. Additional simulations forced with two CMIP6 scenarios extended transiently beyond 2100, show qualitatively similar results to the standard scenarios, but highlight the importance of extended future scenarios for long term carbon cycle projections. The spread between simulations forced with different climate model anomalies suggests a large uncertainty in projected peatland variables due to uncertain climate forcing. Our study highlights 15 the importance of quantifying the future peatland feedback to the climate system and its inclusion into future earth system model projections.

CMIP6 also includes extended versions of the scenarios SSP1-2.6 and SSP5-8.5 that range until 2300. To date, however 10 only a handful climate models have run these extended scenarios. In the climate model sample, three out of ten models provide output for these extended scenarios (see sect. 2.5 and Fig. S1). For these climate models we performed additional simulations with transient climate and CO 2 forcing until 2300.
To disentangle future changes in peatlands that are induced by changes in climate, CO 2 and land-use up to 2014 from those induced by future changes in these drivers, we performed an additional simulation with constant boundary conditions at 2014 Table 1. CMIP6 earth system models used to force the LPX-Bern. Output data was used for monthly precipitation, surface temperature and cloud cover from the 'r1i1p1f1' variant of the respective historical simulations and future scenarios SSP1-2.6, SSP2-4.5, and SSP5-8.5

Model
Model reference Data DOI  Seland et al. (2019a, b) the distances are 0.28°C, -0.31°C, and 0.03°C respectively. For precipitation, anomaly total range, inter quartile range, and median differ between ensemble and sample by 9.7 mm, 0.64 mm, and -0.15 mm over all land area and 7.3 mm, 0.54 mm, and -0.49 mm over peatland area respectively. The optimization procedure thus yielded a sub-sample representative of the larger ensemble, although, given the number of possible combinations, optimization could be improved further with further sampling.

10
Gross changes reveal, however, also regions of local peatland expansion (Fig. 4). North eastern Canada, northern Europe, and east Asia are regions with large losses, whereas north west Canada, north east Asia, and south east Asia see an increase in peatland area up to 2300.
Peatland carbon decreases together with global area in most simulations, with new peatland area showing lower carbon density as lost areas. Total peat carbon, depending on the overall balance of accumulation and decay rather than on peatland 15 area dynamics, is changing only slightly but is declining in eight out of ten simulations. Taken together, the simulations suggest a small to moderate peat carbon loss to the atmosphere over the next 300 years given 2014 conditions. Uncertainties however are large. The spread between the simulations increases significantly after 2014 despite boundary conditions being kept constant. At the year 2300, the simulated global peatland area anomaly relative to 1995-2014 averages ranges from -13 to +4 %, with a median of -4 % and inter quartile range (IQR) from -6 to -2 % ( Table 2). Global carbon stored in active peatlands 20 and global total peat carbon are simulated to change by -9 (total range: -16 to -0; IQR: -10 to -7) % and -1 (-2 to +1; -1 to -0) % respectively. The increasing uncertainty highlights how relatively small differences in forcing can propagate and result in large long term ecosystem and carbon cycle uncertainties.

Committed after 2300
Some simulated peatland responses to historic changes in climate and land-use are delayed even beyond 2300. Between about 25 2700 and 3500 all simulations see a rapid peatland expansion. At 3500 gobal peatland area anomaly compared to 1995-2014 averages is +8 (-1 to +16; +5 to +11) %. With that peatland area is simulated even larger than at present, however, with a dramatically shifted global and regional distribution. The delayed peatland expansion is limited to the northern highest latitudes and the tropics. The resulting peatland distribution at 3500 shows loss of sizable parts of today's northern peatlands with new peatlands partly expanding into permafrost regions and the tropics.

30
Carbon accumulation within old as well as newly formed active peatlands continues over millennia reaching a global peatland carbon stock of +4 (-4 to +16; +1 to +8) % at the year 3500 compared to 1995-2014 averages, illustrating a large long term accumulation potential (Fig. S2). For global total peat carbon, the expansion in peatland area also results in a trend reversal in most simulations. The large accumulation in the newly established peatlands helps to shift the balance from decay dominated to accumulation dominated.
At 3500 total peat carbon is simulated at +1 (-5 to +7; -1 to +2) % compared to 1995-2014 averages and continues to increase with continued accumulation until the end of the simulation. From 2100 to 2300 global peatland area, peatland carbon, and total peat carbon continue to decrease for scenarios SSP2-4.5 and SSP5-8.5 despite constant boundary conditions after 2100 (Table 2). Only under the strong mitigation scenario SSP1-2.6, most simulations show an increase in peatland area and partly in carbon compared to 2100. Medians are similar to the simulations under constant 2014 forcing. However the uncertainty, represented by the spread between the simulations, is larger in the SSP scenarios than in the commitment simulations. For all scenarios this uncertainty increases with time.

5
Spatial anomaly patterns at 2300 for the strong mitigation scenario SSP1-2.6 ( Fig. 7), are similar to the committed changes ( Fig. 4). Regions of peatland area loss are north eastern Canada, northern Europe, central Russia and east Asia and peatland area increases can be found in north west Canada, north east Asia, and south east Asia. Losses and gains are further amplified with respect to the committed changes. Thus, the increase in global area after 2100 in SSP1-2.6 is not due to a recovery of lost peatlands, but rather due to a stronger increase of peat area in the regions of local peatland expansion.
The higher the scenario-based emissions the more extensive the regions of peatland area and carbon loss in the northern high latitudes become and the more reduced are regions of gains. In the high emission scenario SSP5-8.5, losses dominate  Peatland area dynamics translate directly and indirectly into the simulated carbon dynamics. In the strong mitigation scenario, the simulated net loss of northern peatland carbon and total peat carbon is mainly a result of the northern peatland dynamics, rather than of declining carbon accumulation rates. Figure 7 (b) shows that the net ecosystem production (NEP) of active peatlands, which represents the net carbon uptake from the atmosphere per year, changes only slightly until 2300 with decreases throughout the tropics and in parts of the northern latitudes. Regional increases in NEP are simulated in central and and western Asia, with most mid-to high-latitude active peatlands turning from a carbon sink to a carbon source and thus contributing directly to the net carbon loss. Regionally NEP increases are simulated again mostly in east Asia, with larger increases compared to SSP1-2.6. (Fig. 8 (b)). It has to be noted that in case of regional peatland expansion NEP might increase independent of environmental drivers, simply due to the dilution of soil carbon. The results are in broad accordance with a

After 2300
Similar to the committed changes, the future scenarios see a delayed rapid expansion in global peatland area between about 2700 and 3500 CE (Fig. 5, Fig. 6, and Table 2 Taken together, the long term response of global peatlands to future climate change, suggest that under strongly limited future climate change and after negative effects dominating over centuries, potential gobal peatland area and peat carbon could increase compared to today, and even to pre-industrial levels on a millenial timescale. Higher emission scenarios however show a negative effect on global peatland area and a reduced peat carbon storage potential persisting for millennia compared to constant 2014 conditions. Uncertainties towards the end of the simulations, indicated by the sample spread, however become very large.

Extended scenarios
The assumption of stable climate and atmospheric CO 2 levels over millennia after 2100 is a highly idealized one, intended 2100 can change simulated long term peatland responses strongly, depending on the emission pathway. As the 21st century becomes ever shorter, the main focus of future projections and climate policy remains on the next few decades up to 2100. To better understand the long term effects of past and future emissions on global peatlands and to assess their potentially large feedbacks on future climate, the horizon of the future must be expanded beyond 2100. Scenarios extended to 2300 should be elevated to standard practice for future climate projections.

Driver contributions
A factorial analysis was used to attribute the positive and negative changes in peatland variables to individual forcing drivers (see sect. 2.4). Figure 10 shows the calculated mean driver contributions to the global gross positive and negative anomalies in peatland area, peatland carbon, and total peat carbon.
Increases in peatland area up to 2300, both in the high and low latitudes, are driven mostly by committed changes (constant 10 2014 conditions) and an increase in regional precipitation. In northern permafrost regions, this is further strengthened given strong mitigation and moderately rising temperatures resulting in longer growing seasons and larger water retention (Fig. A1).
Simulated peatland area losses are driven mostly by committed changes and increasing temperatures. Higher temperatures lead to an increase in evapotranspiration, especially in boreal peatlands (Helbig et al., 2020b), and thus a decrease in the 5 regional water balance which is not compensated despite potential concurrent increase in annual precipitation. This corresponds to the already observed decade to century long drying trends in northern Europe (Swindles et al., 2019;Zhang et al., 2020) and eastern Canada peatlands (Pellerin and Lavoie, 2003;Pinceloup et al., 2020;Beauregard et al., 2020), regions of large simulated committed area loss, which are found to result in negative effects on carbon accumulation rates and strong trends of woody encroachment. These trends are expected to continue and amplify under future climate change.

10
In the LPX-Bern, a decreasing water balance can lead to a positive feedback on the retreating water table. A long term draw down of the mean gridcell water table leads to a reduction in peatland area, which in turn reduces the mean gridcell water

20
Peatland carbon and total peat carbon dynamics up to 2300 are dominated by committed and temperature driven losses.
The decline in peatland area directly reduces global peatland carbon and indirectly affects the total peat carbon balance by reducing overall accumulation. At the same time, global decay in active as well as former peatlands is increased by the higher temperatures. Peatland NEP in the northern high latitudes is also driven in large parts by committed changes and increasing temperatures which can both increase or decrease NEP given the balance between respiration and productivity and their effect 25 on permafrost ( Fig. A1 and A2). Increasing CO 2 , precipitation, and non linear interactions between the drivers have strong positive effects on NEP in east Asia. In the tropics, the negative trend in NEP is mostly driven by the higher temperatures and non linear effects.
Up to 2300, land-use change and atmospheric CO 2 have a comparatively small impact on the global scale. Increasing CO 2 concentrations have a positive effect on carbon accumulation which is progressively larger with increasing emission scenarios.

30
This also translates into a moderate peatland area gain at 2300. Effects of land-use change are negative for all peatland variables, but remain small. One reason for this small impact, might be that only net land-use area increases are considered to affect peatlands in the model. The global net increase in land-use area, however, is much smaller in the future scenarios, as for the historic period (sect. 2.6). Especially in regions where peatlands are directly targeted for land-use conversion, such as in Indonesia (Dommain et al., 2018;Hoyt et al., 2020), our approach might significantly underestimates the negative effect of land-use change. The late expansion after 2300 is driven by peatlands newly establishing in model gridcells with no previous peatland presence. In some gridcells in the northern high latitude and in south Asia, the historic climate change and atmospheric CO 2 rise lead to the fulfilment of criteria for peatland establishment, targeting the peatland water and carbon balance. The stronger the 5 boundary conditions change under additional future scenarios, the more gridcells, especially in the tropics, become able to support peatlands (see sect. 3.2.3). This change is driven by increases in precipitation, CO 2 , temperature and the non-linear interactions between them. In the model, newly established peatlands start from a small seed and their growth is restricted to 1% of their size per year. They thus reach noticeable size only centuries after their initial establishment. Outside of the model world, the speed of lateral expansion of growing peatlands is dependent on multiple factors including local topography, hydrology and peatland type (Charman, 2002;Ruppel et al., 2013). Topography can constrain lateral expansion velocities.
Depending on terrain slopes, peat accumulation can be limited to a small area or depression for centuries to millennia until the peat column grows tall enough or expand quickly over a flat plain (Bauer et al., 2003;Loisel et al., 2013;Broothaerts et al., 5 2014; Le Stum-Boivin et al., 2019). This heterogeneity and complexity in lateral expansion of newly established peatlands is not represented by the model used here. The magnitude and timing of the simulated late expansion should therefore be taken with care. However, the results suggest that historic and future climate change might create the potential for newly forming peatlands in regions where conditions have been mostly unsuitable before.

Climate forcing uncertainty 10
The spread between simulations forced with climate anomalies from the different CMIP6 climate models indicates a large climate anomaly related uncertainty in simulated peatland variables. This is in line with previous studies that also found a large uncertainty propagation from climate variables to peatland and carbon cycle variables in general Ahlström et al., 2017;Qiu et al., 2020;Müller and Joos, 2020).
The magnitude of uncertainties is regionally different with large uncertainties in the northern high latitudes for peatland area 15 (Fig. S3) and total peat carbon (Fig. S4), and in the mid latitudes for peatland NEP (Fig. S5). The climate variables driving the uncertainty depend on the region and peatland variable in question. Linear regressions for each gridcell were used to investigate how the differences between the model climate anomalies translate to the simulated peatland variables. Differences in northern high latitude peatland area between simulations were found to be dominantly a factor of climate model temperature.
Warmer anomalies resulted in less peatland area in most gridcells, except for north east Asia, where warmer temperatures 20 facilitate peat expansion in some gridcells (Fig. S3). In the tropics, the difference in precipitation is the best predictor for most gridcells, with anomalies from wetter models resulting in larger peatlands. Total peat carbon, determined by the balance between total accumulation and total decay of peat carbon, shows a similar regional pattern (Fig. S4). Northern high and mid latitude peat carbon is reduced with higher temperature anomalies as the area for accumulation declines and heterotrophic respiration increases. In the tropics precipitation remains the dominant predictor, increasing the accumulation area and limiting 25 respiration. For peatland NEP, precipitation minus evapotranspiration, as a measure of the moisture balance, resulted in a larger number of gridcells with significant (regression p-value < 0.05) results compared to temperature or precipitation alone (Fig.   S5). Climate models resulting in a more positive water balance mostly also resulted in a higher peatland NEP due to the controls of peatland water table depth both on productivity and respiration. In permafrost regions also temperature on its own is a strong positive factor for simulated peatland NEP.

30
The results show that the differences in peatland responses to different climate forcings can be explained mostly by the same drivers and mechanisms as the transient changes (discussed in sect. 3.3). They also reveal the large dependence of peatland and carbon cycle projections on key properties of climate models. Further constraining model climate sensitivity is thus essential to reduce uncertainty in carbon cycle projections. Here the climate model sample was selected to best represent the full CMIP6 ensemble and with this the fullest possible range of projections. However model performance, compared to different targets is highly variable (Harrison et al., 2014), and different sample selections or weighted ensemble medians might be preferable in future work, depending on the focus.
The simulations presented here are also subject to other large, but less quantifiable uncertainties. Structural and parameter uncertainties, not only in the peat module, but through all components of the model, are unavoidable in simplified global 5 models such as the LPX-Bern, especially on regional and local scales. Implementation of peatlands in DGVMs is still in its beginning and comprehensive model comparison and structural uncertainty evaluation is still mostly lacking. With the inclusion of peatlands into more and more DGVMs and earth system models, comparative studies might identify the most promising model developments and thus pave the way for more robust peatland and carbon cycle projections. Additional simulations with extended SSP scenario climate forcing from three different climate models showed that con-15 tinuing transient forcing along a scenario trajectory can substantially change the simulated results. Extending the SSP1-2.6 scenario to 2300 with global temperature anomalies decreasing again after 2100 led to a reduction in the response, positive or negative, relative to the standard scenario. The extension of SSP5-8.5 on the other hand lead to a drastically increased loss of peatland area and carbon due to the extreme increases in mean global temperature until 2300. These results highlight the importance of extended emission pathways to project long term effects of anthropogenic climate change not only on peatlands 20 but on the carbon cycle and the climate system as a whole. As the current century grows shorter the next phase of CMIP should aim to extend projections beyond the end of the century as a standard practice.
Driver contributions to future changes were analyzed using factorial simulations. Besides committed changes, increasing temperature was identified as the main driver of peatland area and carbon losses and increasing precipitation as the main driver of gains. After 2300 influences of CO 2 and non-linear interactions on peat initiation become more apparent, when peatland 25 area begins to expand more widely. Cloud cover was found to have only small influences on global peatland variables. Future changes in the net area under land use are small (< 11 %) in the scenarios compared to the historical changes and have a small impact on global peatlands in our simulations. Here a simplified assumption was taken with peatlands being affected by land-use change proportional to their size. However, this might not be the case if peatlands are directly targeted for conversion to land-use areas. Future studies might try to integrate specific peatland -land-use conversion scenarios to better quantify the 30 effect of potential future land-use conversion within a global modeling framework.
The spread between the simulations forced with different climate anomalies from the ten sample climate models, reveals that a large uncertainty is propagated from the climate anomalies to the global peatland and carbon cycle variables. Propagation was found to be mediated different regionally by temperature, precipitation or the combination of both. The uncertainty increases with time even after climate forcing is kept constant due to the long response time scales important for peatlands. Even in 35 the case of the 2014 commitment simulations, which only see 40 years of slightly diverging climate anomalies, uncertainties grow large over time. This shows that small differences in climate forcing can propagate to large long term differences in peatland and carbon cycle variables. In future studies, uncertainties could be reduced by including a skill criterion into the climate model sample selection or the subsequent ensemble analysis. Structural model uncertainties are harder to quantify but could potentially be equally large. A focus of future work must be to quantify these structural uncertainties in peatland model peatlands could be compensated or even superseded by increased emissions from expanding tropical peatlands. In addition, methane emissions increase with temperature (Turetsky et al., 2014). The sign of the methane feedback therefore is dependent on multiple factors. The absence of these potentially important feedbacks between peatlands and the climate system in the stateof-the-art future projections such as produced by the CMIP is a potential limit to formulating adequate climate policy. Future work should focus on the production of fully coupled peatland-climate simulations to assess the magnitude of the potential 20 feedbacks, as well as the integration of peatland modules into the next generation of earth system and integrated assessment models (Loisel et al., 2021).