Wildfire and degradation accelerate northern peatland carbon release

The northern peatland carbon sink plays a vital role in climate regulation; however, the future of the carbon sink is uncertain, in part, due to the changing interactions of peatlands and wildfire. Here, we use empirical datasets from natural, degraded and restored peatlands in non-permafrost boreal and temperate regions to model net ecosystem exchange and methane fluxes, integrating peatland degradation status, wildfire combustion and post-fire dynamics. We find that wildfire processes reduced carbon uptake in pristine peatlands by 35% and further enhanced emissions from degraded peatlands by 10%. The current small net sink is vulnerable to the interactions of peatland degraded area, burn rate and peat burn severity. Climate change impacts accelerated carbon losses, where increased burn severity and burn rate reduced the carbon sink by 38% and 65%, respectively, by 2100. However, our study demonstrates the potential for active peatland restoration to buffer these impacts. Northern peatland carbon sink plays a vital role in climate regulation. Here, the authors show that wildfire reduced peatland carbon uptake and enhanced emissions from degraded peatlands; climate change impacts accelerated carbon losses where increased burn rate and severity reduced carbon sink.

Peatlands store approximately one-third of the global soil carbon stock in 3% of the land area, making them the most carbon dense ecosystem on Earth 1 . Northern peatlands, in boreal and temperate regions, account for ~90% of global peatland area 2 and have sequestered ~500 GtC since the last glacial maximum 1,3 , regulating the global climate throughout the Holocene epoch 4 . Yet, the future of this peatland carbon stock is uncertain [5][6][7] , in part, due to the changing interactions of peatlands and wildfire [8][9][10] . Despite the critical role of peatlands in the global carbon cycle, recent reports and literature that may influence policy do not explicitly account for the impacts of fire on peatland emissions estimates (for example, ref. 11). While estimates of the contribution of peatland drainage to global greenhouse gas emissions have been made 12,13 , no such evaluation has been conducted for, or includes, the interacting effects of peatland degradation and wildfire. The absence of this assessment results in additional uncertainty regarding the impact of climate change on the peatland carbon sink 12 .
These relatively small peat carbon losses from combustion can be re-accumulated within 10 to 30 years post-fire 16 , enabling peatlands to remain a net carbon sink over typical fire-free intervals 17,18 . Conversely, peatland degradation, such as peatland drainage, not only increases ignition potential 19 but can also inflate carbon emissions from peatland wildfires by one or more orders of magnitude to 10-25 kgC m −2 equating to 500 to >1,000 years of carbon sequestration 10,15,19,20 . Given that >25 million hectares (Mha) (7%) of boreal and temperate peatlands have been drained for anthropogenic use 21 , with some regional or national estimates of ~50% (ref. 11), these degraded peatlands represent high-risk areas where wildfire could lead to large carbon emissions.
The differences in net carbon fluxes between pristine and drained peatland wildfires are exacerbated when examining post-fire dynamics. Alterations to CO 2 and methane (CH 4 ) fluxes immediately after fire affect the short-term carbon balance [22][23][24] while post-fire vegetation recovery controls the long-term carbon balance 8,16 . While most pristine peatlands return to a net carbon sink post-fire, evidence suggests that the greater burn severity in degraded peatlands increases the potential Article https://doi.org/10.1038/s41558-023-01657-w modelled NEE + CH 4 flux in these scenarios. To further constrain peatland NEE + CH 4 distributions and accurately include peatlands in Earth system models, plot-to ecosystem-scale carbon flux data at varying times post-fire, especially in degraded and restored ecosystems, are a critical research need.

Climate-mediated changes to burn rate and fire emissions
In addition to the impact of degradation, peatland NEE + CH 4 fluxes are also sensitive to the increasing pressures of climate-mediated drying and associated increases in peat carbon loss from combustion 10 . By aggregating a global dataset of fire perimeters from 2001 to 2021 (FIRED 33 ) ( Supplementary Fig. 1), we calculated the average burn rate (percentage of land area burned per year) for boreal and temperate non-permafrost regions over the last two decades (Methods). Average burn rate varied between 0.0001 and 1.48% yr −1 amongst boreal and temperate ecoregions (Supplementary Table 1), with a spatially weighted average of 0.35% yr −1 , equivalent to an FRI of 290 years. Assessment of the relationship between peatland (histosol) areal coverage 3 and burn rate found no notable trends (Extended Data Fig. 2 and Supplementary Figs. 2 and 3), suggesting that peatland cover does not exert a strong control over regional area burned. Compilation of national inventories found that degradation due to drainage for agriculture, horticulture and forestry varies between <1% and 54% of peatland area per country 34 and the proportion of total drained northern peatland area is ~7% (26.1 Mha;ref. 21). These data were used to evaluate the current state of the boreal and temperate non-permafrost peatland system.
At the broadest scale, without accounting for future climate change impacts to peatlands or wildfire regimes, we estimate that the total NEE + CH 4 flux for boreal and temperate non-permafrost peatlands is a small net carbon sink (filled dots Fig. 2); however, the system becomes a net carbon source given an annual average peatland burn rate of >0.77% based on the current estimates of drained peatland area (Fig. 2a). Accordingly, and important for regional carbon balances, a greater percentage of degraded peatlands reduces the burn rate required to switch the system from a net carbon sink to a net carbon source by 0.05% yr −1 per additional 1% degraded peatland area.
Similarly, increased peat carbon loss from combustion reduces the carbon sink strength and may contribute to switching the system to a for ecosystem regime shifts 8 , a change from a carbon-accumulating peatland to a carbon-releasing ecosystem with non-peatland vegetation, further increasing the impact of peatland wildfires on long-term carbon balance. As such, the inclusion of peatland drainage and post-fire net carbon fluxes is paramount for the accurate evaluation of peatland wildfire carbon emissions.
Rapid changes to regional wildfire regimes are compounding the impacts of drainage on peatland wildfire. In the boreal zone, annual area burned 25 and the frequency of extreme fire weather conditions 26 are increasing as enhanced evapotranspiration is leading to drier wildfire fuels, particularly in peatland ecosystems 27 . Similarly, in the temperate zone increased wildfire activity has been associated with severe droughts 28 and long-term drying has been observed in peatlands 29 . Increased lightning occurrence, reduced snowpacks and multiyear droughts are predicted to further increase annual area burned 30 . Such combinations of climate change-mediated stressors in northern peatlands, along with pervasive peatland degradation, are likely to increase peatland burn rate (percentage of peatland area burned per year), peat burn severity and associated carbon losses 9,10 . Despite evidence that individual northern peat fires can produce teragrams of carbon emissions 20,31 , the fire return interval (FRI) in northern peatlands is often only assessed on a regional or per-site basis (for example, ref. 32). The lack of consistent methodology for assessing peatland burn rate across northern regions has hindered the evaluation of the current and future contribution of northern peatland fires to global carbon emissions. Hence, here we provide estimates of spatially explicit northern peatland burn rates and the contribution of peatland wildfire and post-fire dynamics to global carbon emissions. We then illustrate the impact of peatland degradation and climate change on the future of the northern peatland carbon sink.

Peatland net ecosystem exchange and methane emissions
To address this challenge, we undertook a synthesis of empirical datasets from natural, degraded (currently drained or previously drained and unrestored) and restored peatlands in non-permafrost boreal and temperate regions. We then used these data to model the net ecosystem exchange (NEE) (CO 2 ) and CH 4 fluxes of peatlands over time, integrating post-fire dynamics (recovery rate and final NEE) (Extended Data Table 1) and averaging over a distribution of FRIs (Extended Data Table 2) (Methods). The inclusion of peat carbon loss from combustion and post-fire net carbon fluxes reduced the mean (s.d.) NEE + CH 4 flux sink strength from −50.7 (61.8) gC m −2 yr −1 (no burn-natural) to −32.9 (63.2) gC m −2 yr −1 in natural (pristine) peatlands experiencing fire. The moderate (~35%) reduction in the sink strength demonstrates the impact of fire on peatland carbon balance but also the resilience of the natural peatland carbon sink function under a typical wildfire regime (Fig. 1).
Across the variability in burn rate and the impacts of the fire (severity and recovery rate) the NEE + CH 4 of degraded peatlands remained a consistent source of carbon with an average flux of 213 (229) gC m −2 yr −1 to the atmosphere, a 10% increase compared to no burn-degraded (194 (242) gC m −2 yr −1 ). Meanwhile, the restoration of peatlands before fire mitigated extensive carbon release (92% reduction in emissions compared to degraded), yet restored peatlands remained a small source of carbon with average NEE + CH 4 emission of 17.3 (85.5) gC m −2 yr −1 (Fig. 1). As such, our modelling indicates that excluding peatland wildfire from peatland NEE + CH 4 calculations results in a misrepresentation of peatland carbon balance and may impact estimated regional to national emissions budgets, especially in fire-prone areas with a high proportion of degraded peatlands.
Our empirical approach includes uncertainty in the magnitude of peat carbon loss, burn rate, the rate of recovery and the initial and final recovered NEE (Methods; Extended Data Fig. 1). Our synthesis highlighted limited availability of post-fire carbon flux data, especially from degraded and restored sites, resulting in a wider distribution of Categories of peatland states include natural (pristine), degraded, and restored (before fire) and not accounting for wildfire; no burn-natural and no burn-degraded.
Article https://doi.org/10.1038/s41558-023-01657-w net source. Increasing the average peat carbon loss from combustion in pristine peatlands to represent a moderate degree of climate change drying (1.5 kgC m −2 added to the original distribution) (Methods) reduces the annual burn rate required to switch from a carbon sink to source to 0.55% (Fig. 2b). This equates to a required lengthening of the average FRI by ~50 years to maintain active net carbon sequestration at a landscape level. Further, there is a strong interactive effect of percentage degraded and peat carbon loss on NEE + CH 4 (Fig. 2c). Using the spatially weighted average burn rate of 0.35% yr −1 , NEE + CH 4 fluxes are sensitive to changes in percentage degraded, where a relatively small reduction in percentage degraded (for example, by one-third from 15% to 10%) via active restoration counteracts potential increases in average peat carbon loss from combustion caused by climate-mediated drying.

Future of the northern peatland carbon sink
To illustrate the impact of peatland degradation status and climate change on the magnitude of the peatland carbon sink we evaluated cumulative annual net fluxes from our NEE + CH 4 simulations. We developed scenarios that combine different peatland degradation and climate change factors and assessed the impact on total carbon sequestration (or emission) by 2050 and 2100 (Methods). Scenarios include (1) no burn, (2) current state, (3) restoration, (4) increased burn rate, (5) increased (burn) severity and (6) full climate change (increased burn rate and burn severity). Accounting for peatland wildfire emissions reduces the magnitude of the estimated peatland carbon sink by 1.3 GtC, or 57%, by 2050 when comparing the no-burn scenario (−2.2 GtC) to our current state scenario (−0.94 GtC) (Fig. 3). Meanwhile, the restoration of all degraded peatlands (restoration scenario) results in a sink of 1.8 GtC by 2050, an additional 0.82 GtC sequestered compared to the current state scenario, showing the short-term gains to be made from peatland restoration. Restoration increases the carbon sink by almost 90% in 2100, increasing it from a sink of 2.5 GtC (current state scenario) to 4.7 GtC (restoration scenario).
Increasing peatland burn rate (linear increase to 0.7% by 2100) and increasing burn severity (+1.5 kgC m −2 peat carbon loss) decrease the peatland carbon sink by similar amounts by 2050 with a 0.25 and 0.36 GtC reduction relative to the current state scenario, respectively. However, the modelled burn rate increase throughout the remainder of the century results in a large decrease in the carbon sink strength in the increased burn rate scenario by 2100, reducing the total carbon sequestration to a sink of 0.88 GtC, a 65% decrease compared to the current scenario (−2.5 GtC). Concerningly, when the impacts of climate change are combined (full climate change scenario) the system shows a potential switch from a carbon sink to a carbon source, with a mean estimated source of 0.4 GtC to the atmosphere by 2100. The acceleration of carbon release from boreal and temperate non-permafrost peatlands and associated diminishment of the strength of the carbon sink over the coming decades has critical implications for global climate change and emissions targets.

Importance of peatland restoration for carbon emissions
This study highlights the resilience of pristine northern peatland ecosystems to wildfire, with natural peatlands returning to a net carbon sink in most of our simulations across the range of fire severity and   Article https://doi.org/10.1038/s41558-023-01657-w post-fire dynamics. Conversely, we demonstrate unequivocally that degraded peatlands are responsible for the largest peatland carbon emissions 19,20,31 . We show that the restoration of degraded peatlands before fire greatly reduces long-term emissions 35 . Our results add to the growing literature base that suggests climate and land-use change increase the vulnerability of peatland ecosystems and their carbon stocks to fire, with considerable and far-reaching ecological, hydrological and societal consequences 34,36 .
While future anthropogenic fossil fuel emissions can be curbed, the climatic changes already induced by rising atmospheric CO 2 concentrations will probably continue to increase peatland wildfire emissions over the coming century, reducing the strength of the peatland carbon sink. We show that, although the peatland carbon sink is currently resilient, changes in degraded peatland area, average burn rate (FRI) and peat burn severity may lead to climate neutrality or net carbon release. Climate-mediated peatland drying across the spectrum of peatland condition 27,29 could contribute to increases in peatland burn rate 37 and peat carbon loss via enhanced burn severity 10 in line with the increasing availability of critically dry peatland fuels 9 . Forested peatlands (natural or managed) may be more prone to positive (amplifying) ecohydrological feedbacks that promote high-severity smouldering fire 38 , when compared to arable peatlands in northern regions; however, the vulnerability of peatlands to wildfire under different management regimes is currently relatively unstudied.
To maintain the northern peatland carbon sink function, decreases in the area of degraded peatland through active peatland restoration must occur to counteract potential increases in average peat carbon loss due to climate-mediated drying. Our restoration scenario (representing the restoration of all degraded peatlands) resulted in an estimated increase in the carbon sink by almost 90% by 2100 compared to the current scenario. Despite the hypothetical nature of our restoration scenario, it serves to support research highlighting the important role peatlands can play in reducing global emissions if they are protected 39 and restored 40 appropriately. We also strongly advocate for better management of carbon-rich ecosystems alongside behavioural changes to stop accidental and unnecessary ignitions 41 especially areas with a high proportion of degraded peatlands (for example, Europe) 11 . On a regional level we provide evidence of the importance of accurately measuring (degraded) peatland area, as well as burn rate, since these factors will affect the ability of countries/regions to account for emissions and, potentially, to achieve emissions targets. The proportion of peatlands affected by land-use change varies considerably between countries and regions but can be substantial (<1% to ~50% degraded 42 ). While there are probable differences in the ignition potential of different peatland land-uses 19 there is a scarcity of these data in the literature. Peatland type and landscape position have been found to impact burn rate 32 and fire severity 43 , yet peatlands are often misclassified in fire risk, spread and emissions models 44 , highlighting the need to improve peatland mapping for use alongside remotely sensed fire products (for example, ref. 33). Appropriate accounting of carbon emissions from peatlands, accounting for wildfire, may guide national/regional restoration and conservation strategies (for example, in the UK 45 ).
Interdisciplinary collaborations will be crucial to accurately represent the northern peatland carbon balance in Earth system models and ensure that community-to international-level climate policies include important peatland processes, such as fire, in their strategies to maintain the impacts of climate change within liveable bounds. While remote-sensing applications, such as FIRED 33 , have enabled consistent burn rate mapping across large regions, the limited precision and consistency of peatland type and carbon stock maps creates challenges for further reducing the uncertainty surrounding estimates of the strength of the northern peatland carbon sink 39 . Further, our carbon sink estimates do not account for fluvial export of carbon, nor the anthropogenic additions/removals of biomass on agricultural peatlands. Data corresponding to methane emissions in different peatland types immediately post-fire for example 22,23 , would further constrain estimates of the peatland carbon sink.
The direction and magnitude of the peatland-climate feedback will be driven by the combined effects of peatland degradation and restoration and the global emissions pathway that will influence rates of climate-induced drying 27 and changes in burn rate 26 . Northern peatlands have regulated global climate over the Holocene but if the predicted increases in peat burn severity and fire activity outweigh carbon sequestration from peatland expansion in high-latitude regions 46 , the northern peatland system will become a shrinking carbon sink and a potential future carbon source, exacerbating the rapidly closing window of time to avoid the most severe impacts of global climate change. Our scenario results found that increasing burn rate and peat burn severity drastically reduced the amount of carbon sequestered in peatlands but overall the system maintained a carbon sink status in 2050. However, the continued and compounding impacts of climate change resulted in an estimated small net carbon source to the atmosphere by boreal and temperate non-permafrost peatlands by 2100. Notably, this estimate does not account for increased peatland tree growth stimulated by drier conditions 10 . However, given the positive correlation between tree size and burn severity 10 and the dominance of long-term carbon storage in peat rather than above-ground vegetation 47 , it is unlikely that increases in above-ground biomass will translate into appreciable increases in carbon storage over long (>1 FRI) time periods. The probable reduction in peatland carbon sink strength will create further challenges to remaining below critical global climate targets.
Against the global backdrop of increases in burn rate and extreme wildfire weather 26 , integrated regional wildfire management solutions are urgently required to mitigate severe climatic and societal impacts of peatland wildfire 37,38 . In regions with higher proportions of peatland degradation we find that a strong trade-off with burn rate (that is, large investments in direct fire suppression) is required to preserve the critical climate regulation function of peatlands. Where this balance is not maintained, peatland wildfire emissions may represent an underappreciated source component in carbon accounting that could be detrimental to achieving emissions targets. We demonstrate here that, despite notable impacts of peatland burn rate and burn severity, peatland restoration represents a large opportunity to minimize impacts to the boreal and temperate peatland carbon sink over the coming century when accounting for peatland wildfire emissions. Our results suggest an immediate need to start including active restoration of degraded peatlands as a cost-effective tool to support the mitigation of extensive carbon emissions.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41558-023-01657-w. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Methods
Peatland net ecosystem CO 2 exchange (NEE) was estimated for northern peatlands using literature values for burned, pristine, degraded and rewetted/restored sites (Extended Data Table 1). NEE was estimated over an FRI using values for post-fire (1-10 yr post-fire; NEE burned ) and recovered (unburned; NEE recovered ) periods. NEE recovered could represent one of undrained, drained or rewetted/restored NEE. Annual values of NEE were derived using a simple conceptual model of NEE recovery from its post-fire to recovered state (Extended Data Fig. 1). The recovery of NEE was represented by a piecewise linear function with two input parameters (t 1 and t 2 ) which define the time period (time since fire; TSF) over which NEE transitions from its post-fire to recovered state (equation (1)). Annual NEE values were obtained from studies with site estimates of NEE for non-permafrost northern peatlands derived from either chamber or eddy covariance measurements (Supplementary Table 2). In total, 336 site-years of data were collated for pristine (183 site-years), degraded (84 site-years), restored/rewetted (56 site-years) and burned (13 site-years) sites. For studies which did not include year-round measurements, we estimated the non-growing season NEE according to the methods outlined by ref. 48. Non-growing season NEE estimates were used for 42 site-years or for ~12% of the collated site-year dataset. For annual CH 4 flux data for northern hemisphere pristine, drained and restored sites, we used data synthesized by ref. 49. From the collated dataset, we derived four distributions of NEE: (1) burned, (2) pristine, (3) degraded and (4) restored/rewetted (Extended Data Table 1). Two separate distributions of carbon emissions from wildfire were used for degraded and pristine sites (Extended Data Table 1). Peat carbon loss estimates from a total of 22 sites were used to create fire carbon loss distributions (Supplementary Tables 2 and 3). A Monte Carlo simulation (n = 10,000) was used to combine sources of uncertainty and generate an annual average NEE + CH 4 estimate for non-permafrost northern peatlands which includes carbon emissions from wildfire. No peatland eddy covariance flux monitoring has occurred which directly quantifies the recovery rate of NEE post-fire. However, chronosequence chamber measurement data (for example, ref. 16) show that recovery of NEE to a pre-fire state corresponds with vegetation recovery, which includes a multiyear lag before recovery which plateaus around 30 yr post-fire for treed boreal peatlands. Hence, we apply a uniform distribution for recovery initiation (t 1 ) and reaching NEE recovered (t 2 ) (Extended Data Table 1), the distance between the two determining the recovery rate of the simulated fluxes, schematically displayed in Extended Data Fig. 1. (1) CH 4 fluxes were included in the Monte Carlo simulation by randomly sampling from the ref. 49 dataset. For simplicity, and due to the lack of studies examining CH 4 post-fire in peatlands, CH 4 fluxes were assumed to not vary systematically over the FRI. CH 4 fluxes were sampled according to peatland type (pristine, degraded and restored/ rewetted). Annual values of NEE + CH 4 fluxes were converted to gC and summed. Note that fluvial exports and anthropogenic additions/ removal (for example, fertilizer and cutting), as well as above-ground biomass were not considered.
Burn rate was calculated for non-permafrost boreal and temperate biomes using fire polygon (fire perimeter) data from FIRED 33 (Supplementary Fig. 1). FIRED provides easy access to compiled datasets of fire perimeter data created using the MODIS burned area product (MCD 64A1). Fire polygon data cover the period November 2001 to July 2021, providing 19.75 yr of data with consistent methodology used across the time series and all spatial regions (Supplementary Table 1). Area burned was assessed for all boreal and temperate ecoregion polygons (World Wildlife Fund Terrestrial Ecoregions of the World 50 ) in the northern hemisphere which contain peatlands according to the ref. 3 histosol dataset. Ecoregion polygons were clipped to remove permafrost regions using a permafrost extent layer 51 where discontinuous and continuous areas of permafrost were considered permafrost in our study. The exclusion of permafrost areas is due to the interaction of wildfire and permafrost thaw that are not within the scope of this study. Ecoregion polygons were aggregated to boreal and temperate regions by joining according to the biome attribute, then clipped to derive estimates for Europe, Asia and North America. Area burned over the 19.75 yr period was calculated from the FIRED polygons using the area geometry function in QGIS 52 and summed over the intersecting clipped ecoregions. The total area burned was then divided by 19.75 (years of data) and the area of each clipped ecoregion to produce a burn rate (% land yr −1 ) which was then converted to FRI (Extended Data Table 2). Area values include open water area of small inland lakes. All analyses were conducted using QGIS 3.6 (ref. 52).
The distribution of calculated FRIs was in line with FRIs (also called fire-free intervals) recorded in the literature for boreal and temperate peatlands (Supplementary Table 4) using a variety of methods (for example, macroscopic charcoal analysis, tree ring fire scars and GIS analyses). Assessment of the relationship between peatland (histosol) areal coverage and burn rate (proportional to FRI) at an ecoregion (Extended Data Fig. 2) and smaller (one degree hexagon) scale found no notable trends ( Supplementary Figs. 2 and 3), suggesting that peatland cover does not exert a strong control over area burned but that both peatland occurrence and average burn rate are mediated by climatic drivers, such as the climate moisture index ( Supplementary  Fig. 4). Considering this evaluation and the conversion of peatlands to alternate landcover types (for example, forest and agriculture), we assume that peatland burn rate (peatland area burned per yr) is equal to the calculated burn rate (percentage land burned per yr). Moreover, methodologies that use macroscopic charcoal layers to determine FRI have the potential to underestimate fire frequency due to the potential erasure of charcoal layers in successive higher severity fires, the possible loss or relocation of charcoal from a given peatland location before it is incorporated into the peat matrix and via passing through unburned hummocks in the peat profile, biasing towards longer estimates of FRIs than actually occur.
We assessed the interactive effect of fire regime changes (peatland burn rate (% yr −1 ) and peat carbon loss (kgC m −2 )) and land-use change (percentage of peatland area that is degraded (%)) on peatland NEE + CH4 (GtC yr −1 ) by varying two of the three parameters concurrently within realistic bounds. While two parameters were varied between their lower and upper limits, the third parameter was held (relatively) constant. In the case of varying peatland burn rate and percentage degraded (Fig. 2a), peat carbon loss is calculated proportional to percentage degraded by drawing from the respective peat combustion distributions for pristine and drained, resulting in higher peat carbon loss with greater percentage degraded. Assessing peatland burn rate and peat carbon loss from combustion (Fig. 2b), percentage degraded is held constant at 7%. When peat carbon loss and percentage degraded are varied (Fig. 2c) mean annual burn rate is held constant at 0.35% (spatially weighted average burn rate). Repeated simulations (n = 10,000) were run using a Monte Carlo framework where FRI, burned and recovered NEE and recovery time were varied as described in Extended Data Table 1. The black filled dots represent the current boreal and temperate non-permafrost peatland system; that is, 0.35 burn rate (% yr −1 ), 7% degraded land area, 93% pristine, and pristine and drained peat carbon loss from their respective distributions, where restored peatlands are (optimistically) attributed carbon loss from the pristine distribution.
To assess the effect of peatland degradation and climate change factors on total carbon sequestration (or emission) over the twenty-first century, we developed several scenarios and produced Nature Climate Change Article https://doi.org/10.1038/s41558-023-01657-w cumulative values for 2050 and 2100, beginning in 2020. The scenarios are (1) no burn (current state with a burn rate of 0% yr −1 that is, no peat carbon loss from fire and no effect on NEE), (2) current state (93% pristine, 7% degraded, with distribution of burn rates around the 0.35% average), (3) restoration (current state with the conversion of all degraded peatland area to restored peatland), (4) increased burn rate (doubling of peatland burn rate by 2100 via a linear increase from 0.35% to 0.7% yr −1 ), (5) increased severity (additional 1.5 kgC m −2 of peat carbon loss from combustion for all peatland types, no changes to recovery rate or final NEE) and (6) full climate change (additive effect of scenarios 4 and 5).
Across northern regions, annual area burned (burn rate) is predicted to increase drastically by the end of the century for example, doubling in the Canadian boreal by 2100 25,53 . Strong increasing trends in area burned are already found in some countries with a high coverage of peatlands, for example, Canada 25 and Siberia 54 . In line with increasingly extreme fire weather driven by temperature and humidity changes 26 , peatlands are experiencing drying 27 that will leave them vulnerable to increased fire severity. We use the low estimate of probable increase in peat carbon loss, calculated as the difference between undrained and moderately drained plot's mean peat carbon loss (0.9 and 2.4 kgC m −2 , respectively) in a treed northern peatland 10 . Due to a lack of data pertaining to the impact of (further) drying on peat carbon loss from combustion in a range of peatland types (degraded and restored) we apply this estimate to all peatland types. However, we note that changes to burn rate and peat burn severity will be highly variable through space and time, dependent on weather, climate and possible vegetation community changes.

Data availability
Synthesized data are uploaded to a certified repository 55 and are open access.

Code availability
Model simulations code is uploaded to a certified repository 55 and is open access.