Abstract Modelling is widely used in ecology and its utility continues to increase as scientists, managers and policy‐makers face pressure to effectively manage ecosystems and meet conservation goals with limited resources. As the urgency to forecast ecosystem responses to global change grows, so do the number and complexity of predictive ecological models and the value of iterative prediction, both of which demand validation and cross‐model comparisons. This challenges ecologists to provide predictive models that are reusable, interoperable, transparent and able to accommodate updates to both data and algorithms. We propose a practical solution to this challenge based on the PERFICT principles (frequent Predictions and Evaluations of Reusable, Freely accessible, Interoperable models, built within Continuous workflows that are routinely Tested), using a modular and integrated framework. We present its general implementation across seven common components of ecological model applications—(i) the modelling toolkit; (ii) data acquisition and treatment; (iii) model parameterisation and calibration; (iv) obtaining predictions; (v) model validation; (vi) analysing and presenting model outputs; and (vii) testing model code—and apply it to two approaches used to predict species distributions: (1) a static statistical model, and (2) a complex spatiotemporally dynamic model. Adopting a continuous workflow enabled us to reuse our models in new study areas, update predictions with new data, and re‐parameterise with different interoperable modules using freely accessible data sources, all with minimal user input. This allowed repeating predictions and automatically evaluating their quality, while centralised inputs, parameters and outputs, facilitated ensemble forecasting and tracking uncertainty. Importantly, the integrated model validation promotes a continuous evaluation of the quality of more‐ or less‐parsimonious models, which is valuable in predictive ecological modelling. By linking all stages of an ecological modelling exercise, it is possible to overcome common challenges faced by ecological modellers, such as changing study areas, choosing between different modelling approaches, and evaluating the appropriateness of the model. This ultimately creates a more equitable and robust playing field for both modellers and end users (e.g. managers), and contributes to position predictive ecology as a central contributor to global change forecasting.

Abstract Modelling is widely used in ecology and its utility continues to increase as scientists, managers and policy‐makers face pressure to effectively manage ecosystems and meet conservation goals with limited resources. As the urgency to forecast ecosystem responses to global change grows, so do the number and complexity of predictive ecological models and the value of iterative prediction, both of which demand validation and cross‐model comparisons. This challenges ecologists to provide predictive models that are reusable, interoperable, transparent and able to accommodate updates to both data and algorithms. We propose a practical solution to this challenge based on the PERFICT principles (frequent Predictions and Evaluations of Reusable, Freely accessible, Interoperable models, built within Continuous workflows that are routinely Tested), using a modular and integrated framework. We present its general implementation across seven common components of ecological model applications—(i) the modelling toolkit; (ii) data acquisition and treatment; (iii) model parameterisation and calibration; (iv) obtaining predictions; (v) model validation; (vi) analysing and presenting model outputs; and (vii) testing model code—and apply it to two approaches used to predict species distributions: (1) a static statistical model, and (2) a complex spatiotemporally dynamic model. Adopting a continuous workflow enabled us to reuse our models in new study areas, update predictions with new data, and re‐parameterise with different interoperable modules using freely accessible data sources, all with minimal user input. This allowed repeating predictions and automatically evaluating their quality, while centralised inputs, parameters and outputs, facilitated ensemble forecasting and tracking uncertainty. Importantly, the integrated model validation promotes a continuous evaluation of the quality of more‐ or less‐parsimonious models, which is valuable in predictive ecological modelling. By linking all stages of an ecological modelling exercise, it is possible to overcome common challenges faced by ecological modellers, such as changing study areas, choosing between different modelling approaches, and evaluating the appropriateness of the model. This ultimately creates a more equitable and robust playing field for both modellers and end users (e.g. managers), and contributes to position predictive ecology as a central contributor to global change forecasting.

Abstract Modelling is widely used in ecology and its utility continues to increase as scientists, managers and policy‐makers face pressure to effectively manage ecosystems and meet conservation goals with limited resources. As the urgency to forecast ecosystem responses to global change grows, so do the number and complexity of predictive ecological models and the value of iterative prediction, both of which demand validation and cross‐model comparisons. This challenges ecologists to provide predictive models that are reusable, interoperable, transparent and able to accommodate updates to both data and algorithms. We propose a practical solution to this challenge based on the PERFICT principles (frequent Predictions and Evaluations of Reusable, Freely accessible, Interoperable models, built within Continuous workflows that are routinely Tested), using a modular and integrated framework. We present its general implementation across seven common components of ecological model applications—(i) the modelling toolkit; (ii) data acquisition and treatment; (iii) model parameterisation and calibration; (iv) obtaining predictions; (v) model validation; (vi) analysing and presenting model outputs; and (vii) testing model code—and apply it to two approaches used to predict species distributions: (1) a static statistical model, and (2) a complex spatiotemporally dynamic model. Adopting a continuous workflow enabled us to reuse our models in new study areas, update predictions with new data, and re‐parameterise with different interoperable modules using freely accessible data sources, all with minimal user input. This allowed repeating predictions and automatically evaluating their quality, while centralised inputs, parameters and outputs, facilitated ensemble forecasting and tracking uncertainty. Importantly, the integrated model validation promotes a continuous evaluation of the quality of more‐ or less‐parsimonious models, which is valuable in predictive ecological modelling. By linking all stages of an ecological modelling exercise, it is possible to overcome common challenges faced by ecological modellers, such as changing study areas, choosing between different modelling approaches, and evaluating the appropriateness of the model. This ultimately creates a more equitable and robust playing field for both modellers and end users (e.g. managers), and contributes to position predictive ecology as a central contributor to global change forecasting.

Abstract Climate change is increasing the frequency and extent of fires in the boreal biome of North America. These changes can alter the recovery of both canopy and understory vegetation. There is uncertainty about plant and lichen recovery patterns following fire, and how they are mediated by environmental conditions. Here, we aim to address these knowledge gaps by studying patterns of postfire vegetation recovery at the community and individual species level over the first 100+ years following fire. Data from vegetation surveys collected from 581 plots in the Northwest Territories, Canada, ranging from 1 to 275 years postfire, were used to assess the influence of time after fire and local environmental conditions on plant community composition and to model trends in the relative abundance of several common plant and lichen species. Time after fire significantly influenced vegetation community composition and interacted with local environmental conditions, particularly soil moisture. Soil moisture individually (in the absence of interactions) was the most commonly significant variable in plant and lichen recovery models. Patterns of postfire recovery varied greatly among species. Our results provide novel information on plant community recovery after fire and highlight the importance of soil moisture to local vegetation patterns. They will aid northern communities and land managers to anticipate the impacts of increased fire activity on both local vegetation and the wildlife that relies on it.

Abstract Climate change is increasing the frequency and extent of fires in the boreal biome of North America. These changes can alter the recovery of both canopy and understory vegetation. There is uncertainty about plant and lichen recovery patterns following fire, and how they are mediated by environmental conditions. Here, we aim to address these knowledge gaps by studying patterns of postfire vegetation recovery at the community and individual species level over the first 100+ years following fire. Data from vegetation surveys collected from 581 plots in the Northwest Territories, Canada, ranging from 1 to 275 years postfire, were used to assess the influence of time after fire and local environmental conditions on plant community composition and to model trends in the relative abundance of several common plant and lichen species. Time after fire significantly influenced vegetation community composition and interacted with local environmental conditions, particularly soil moisture. Soil moisture individually (in the absence of interactions) was the most commonly significant variable in plant and lichen recovery models. Patterns of postfire recovery varied greatly among species. Our results provide novel information on plant community recovery after fire and highlight the importance of soil moisture to local vegetation patterns. They will aid northern communities and land managers to anticipate the impacts of increased fire activity on both local vegetation and the wildlife that relies on it.

Abstract Key message Black spruce ( Picea mariana (Mill.) B.S.P.) has historically self-replaced following wildfire, but recent evidence suggests that this is changing. One factor could be negative impacts of intensifying fire activity on black spruce seed rain. We investigated this by measuring black spruce seed rain and seedling establishment. Our results suggest that increases in fire activity could reduce seed rain meaning reductions in black spruce establishment. Context Black spruce is an important conifer in boreal North America that develops a semi-serotinous, aerial seedbank and releases a pulse of seeds after fire. Variation in postfire seed rain has important consequences for black spruce regeneration and stand composition. Aims We explore the possible effects of changes in fire regime on the abundance and viability of black spruce seeds following a very large wildfire season in the Northwest Territories, Canada (NWT). Methods We measured postfire seed rain over 2 years at 25 black spruce-dominated sites and evaluated drivers of stand characteristics and environmental conditions on total black spruce seed rain and viability. Results We found a positive relationship between black spruce basal area and total seed rain. However, at high basal areas, this increasing rate of seed rain was not maintained. Viable seed rain was greater in stands that were older, closer to unburned edges, and where canopy combustion was less severe. Finally, we demonstrated positive relationships between seed rain and seedling establishment, confirming our measures of seed rain were key drivers of postfire forest regeneration. Conclusion These results indicate that projected increases in fire activity will reduce levels of black spruce recruitment following fire.

Abstract Key message Black spruce ( Picea mariana (Mill.) B.S.P.) has historically self-replaced following wildfire, but recent evidence suggests that this is changing. One factor could be negative impacts of intensifying fire activity on black spruce seed rain. We investigated this by measuring black spruce seed rain and seedling establishment. Our results suggest that increases in fire activity could reduce seed rain meaning reductions in black spruce establishment. Context Black spruce is an important conifer in boreal North America that develops a semi-serotinous, aerial seedbank and releases a pulse of seeds after fire. Variation in postfire seed rain has important consequences for black spruce regeneration and stand composition. Aims We explore the possible effects of changes in fire regime on the abundance and viability of black spruce seeds following a very large wildfire season in the Northwest Territories, Canada (NWT). Methods We measured postfire seed rain over 2 years at 25 black spruce-dominated sites and evaluated drivers of stand characteristics and environmental conditions on total black spruce seed rain and viability. Results We found a positive relationship between black spruce basal area and total seed rain. However, at high basal areas, this increasing rate of seed rain was not maintained. Viable seed rain was greater in stands that were older, closer to unburned edges, and where canopy combustion was less severe. Finally, we demonstrated positive relationships between seed rain and seedling establishment, confirming our measures of seed rain were key drivers of postfire forest regeneration. Conclusion These results indicate that projected increases in fire activity will reduce levels of black spruce recruitment following fire.

Abstract Over the past several decades, various trends in vegetation productivity, from increases to decreases, have been observed throughout Arctic–Boreal ecosystems. While some of this variation can be explained by recent climate warming and increased disturbance, very little is known about the impacts of permafrost thaw on productivity across diverse vegetation communities. Active layer thickness data from 135 permafrost monitoring sites along a 10° latitudinal transect of the Northwest Territories, Canada, paired with a Landsat time series of normalized difference vegetation index from 1984 to 2019, were used to quantify the impacts of changing permafrost conditions on vegetation productivity. We found that active layer thickness contributed to the observed variation in vegetation productivity in recent decades in the northwestern Arctic–Boreal, with the highest rates of greening occurring at sites where the near‐surface permafrost recently had thawed. However, the greening associated with permafrost thaw was not sustained after prolonged periods of thaw and appeared to diminish after the thaw front extended outside the plants' rooting zone. Highest rates of greening were found at the mid‐transect sites, between 62.4° N and 65.2° N, suggesting that more southernly sites may have already surpassed the period of beneficial permafrost thaw, while more northern sites may have yet to reach a level of thaw that supports enhanced vegetation productivity. These results indicate that the response of vegetation productivity to permafrost thaw is highly dependent on the extent of active layer thickening and that increases in productivity may not continue in the coming decades.

Abstract Over the past several decades, various trends in vegetation productivity, from increases to decreases, have been observed throughout Arctic–Boreal ecosystems. While some of this variation can be explained by recent climate warming and increased disturbance, very little is known about the impacts of permafrost thaw on productivity across diverse vegetation communities. Active layer thickness data from 135 permafrost monitoring sites along a 10° latitudinal transect of the Northwest Territories, Canada, paired with a Landsat time series of normalized difference vegetation index from 1984 to 2019, were used to quantify the impacts of changing permafrost conditions on vegetation productivity. We found that active layer thickness contributed to the observed variation in vegetation productivity in recent decades in the northwestern Arctic–Boreal, with the highest rates of greening occurring at sites where the near‐surface permafrost recently had thawed. However, the greening associated with permafrost thaw was not sustained after prolonged periods of thaw and appeared to diminish after the thaw front extended outside the plants' rooting zone. Highest rates of greening were found at the mid‐transect sites, between 62.4° N and 65.2° N, suggesting that more southernly sites may have already surpassed the period of beneficial permafrost thaw, while more northern sites may have yet to reach a level of thaw that supports enhanced vegetation productivity. These results indicate that the response of vegetation productivity to permafrost thaw is highly dependent on the extent of active layer thickening and that increases in productivity may not continue in the coming decades.

Resilience of plant communities to disturbance is supported by multiple mechanisms, including ecological legacies affecting propagule availability, species' environmental tolerances, and biotic interactions. Understanding the relative importance of these mechanisms for plant community resilience supports predictions of where and how resilience will be altered with disturbance. We tested mechanisms underlying resilience of forests dominated by black spruce (Picea mariana) to fire disturbance across a heterogeneous forest landscape in the Northwest Territories, Canada. We combined surveys of naturally regenerating seedlings at 219 burned plots with experimental manipulations of ecological legacies via seed addition of four tree species and vertebrate exclosures to limit granivory and herbivory at 30 plots varying in moisture and fire severity. Black spruce recovery was greatest where it dominated pre-fire, at wet sites with deep residual soil organic layers, and fire conditions of low soil or canopy combustion and longer return intervals. Experimental addition of seed indicated all species were seed-limited, emphasizing the importance of propagule legacies. Black spruce and birch (Betula papyrifera) recruitment were enhanced with vertebrate exclusion. Our combination of observational and experimental studies demonstrates black spruce is vulnerable to effects of increased fire activity that erode ecological legacies. Moreover, black spruce relies on wet areas with deep soil organic layers where other species are less competitive. However, other species can colonize these areas if enough seed is available or soil moisture is altered by climate change. Testing mechanisms underlying species' resilience to disturbance aids predictions of where vegetation will transform with effects of climate change.The online version contains supplementary material available at 10.1007/s10021-022-00772-7.

Abstract Modelling is widely used in ecology and its utility continues to increase as scientists, managers and policy‐makers face pressure to effectively manage ecosystems and meet conservation goals with limited resources. As the urgency to forecast ecosystem responses to global change grows, so do the number and complexity of predictive ecological models and the value of iterative prediction, both of which demand validation and cross‐model comparisons. This challenges ecologists to provide predictive models that are reusable, interoperable, transparent and able to accommodate updates to both data and algorithms. We propose a practical solution to this challenge based on the PERFICT principles (frequent Predictions and Evaluations of Reusable, Freely accessible, Interoperable models, built within Continuous workflows that are routinely Tested), using a modular and integrated framework. We present its general implementation across seven common components of ecological model applications—(i) the modelling toolkit; (ii) data acquisition and treatment; (iii) model parameterisation and calibration; (iv) obtaining predictions; (v) model validation; (vi) analysing and presenting model outputs; and (vii) testing model code—and apply it to two approaches used to predict species distributions: (1) a static statistical model, and (2) a complex spatiotemporally dynamic model. Adopting a continuous workflow enabled us to reuse our models in new study areas, update predictions with new data, and re‐parameterise with different interoperable modules using freely accessible data sources, all with minimal user input. This allowed repeating predictions and automatically evaluating their quality, while centralised inputs, parameters and outputs, facilitated ensemble forecasting and tracking uncertainty. Importantly, the integrated model validation promotes a continuous evaluation of the quality of more‐ or less‐parsimonious models, which is valuable in predictive ecological modelling. By linking all stages of an ecological modelling exercise, it is possible to overcome common challenges faced by ecological modellers, such as changing study areas, choosing between different modelling approaches, and evaluating the appropriateness of the model. This ultimately creates a more equitable and robust playing field for both modellers and end users (e.g. managers), and contributes to position predictive ecology as a central contributor to global change forecasting.

Abstract Modelling is widely used in ecology and its utility continues to increase as scientists, managers and policy‐makers face pressure to effectively manage ecosystems and meet conservation goals with limited resources. As the urgency to forecast ecosystem responses to global change grows, so do the number and complexity of predictive ecological models and the value of iterative prediction, both of which demand validation and cross‐model comparisons. This challenges ecologists to provide predictive models that are reusable, interoperable, transparent and able to accommodate updates to both data and algorithms. We propose a practical solution to this challenge based on the PERFICT principles (frequent Predictions and Evaluations of Reusable, Freely accessible, Interoperable models, built within Continuous workflows that are routinely Tested), using a modular and integrated framework. We present its general implementation across seven common components of ecological model applications—(i) the modelling toolkit; (ii) data acquisition and treatment; (iii) model parameterisation and calibration; (iv) obtaining predictions; (v) model validation; (vi) analysing and presenting model outputs; and (vii) testing model code—and apply it to two approaches used to predict species distributions: (1) a static statistical model, and (2) a complex spatiotemporally dynamic model. Adopting a continuous workflow enabled us to reuse our models in new study areas, update predictions with new data, and re‐parameterise with different interoperable modules using freely accessible data sources, all with minimal user input. This allowed repeating predictions and automatically evaluating their quality, while centralised inputs, parameters and outputs, facilitated ensemble forecasting and tracking uncertainty. Importantly, the integrated model validation promotes a continuous evaluation of the quality of more‐ or less‐parsimonious models, which is valuable in predictive ecological modelling. By linking all stages of an ecological modelling exercise, it is possible to overcome common challenges faced by ecological modellers, such as changing study areas, choosing between different modelling approaches, and evaluating the appropriateness of the model. This ultimately creates a more equitable and robust playing field for both modellers and end users (e.g. managers), and contributes to position predictive ecology as a central contributor to global change forecasting.

Abstract Modelling is widely used in ecology and its utility continues to increase as scientists, managers and policy‐makers face pressure to effectively manage ecosystems and meet conservation goals with limited resources. As the urgency to forecast ecosystem responses to global change grows, so do the number and complexity of predictive ecological models and the value of iterative prediction, both of which demand validation and cross‐model comparisons. This challenges ecologists to provide predictive models that are reusable, interoperable, transparent and able to accommodate updates to both data and algorithms. We propose a practical solution to this challenge based on the PERFICT principles (frequent Predictions and Evaluations of Reusable, Freely accessible, Interoperable models, built within Continuous workflows that are routinely Tested), using a modular and integrated framework. We present its general implementation across seven common components of ecological model applications—(i) the modelling toolkit; (ii) data acquisition and treatment; (iii) model parameterisation and calibration; (iv) obtaining predictions; (v) model validation; (vi) analysing and presenting model outputs; and (vii) testing model code—and apply it to two approaches used to predict species distributions: (1) a static statistical model, and (2) a complex spatiotemporally dynamic model. Adopting a continuous workflow enabled us to reuse our models in new study areas, update predictions with new data, and re‐parameterise with different interoperable modules using freely accessible data sources, all with minimal user input. This allowed repeating predictions and automatically evaluating their quality, while centralised inputs, parameters and outputs, facilitated ensemble forecasting and tracking uncertainty. Importantly, the integrated model validation promotes a continuous evaluation of the quality of more‐ or less‐parsimonious models, which is valuable in predictive ecological modelling. By linking all stages of an ecological modelling exercise, it is possible to overcome common challenges faced by ecological modellers, such as changing study areas, choosing between different modelling approaches, and evaluating the appropriateness of the model. This ultimately creates a more equitable and robust playing field for both modellers and end users (e.g. managers), and contributes to position predictive ecology as a central contributor to global change forecasting.

Increased fire activity due to climate change may impact the successional dynamics of boreal forests, with important consequences for caribou habitat. Early successional forests have been shown to support lower quantities of caribou forage lichens, but geographic variation in, and controls on, the rates of lichen recovery has been largely unexplored. In this study, we sampled across a broad region in northwestern Canada to compare lichen biomass accumulation in ecoprovinces, including the Saskatchewan Boreal Shield, the Northwest Territories Taiga Shield, and Northwest Territories Taiga Plains, divided into North and South. We focused on the most valuable Cladonia species for boreal and barren-ground caribou: Cladonia mitis and C. arbuscula, C. rangiferina and C. stygia, and C. stellaris and C. uncialis. We developed new allometric equations to estimate lichen biomass from field measurements of lichen cover and height; allometries were consistent among ecoprovinces, suggesting generalizability. We then used estimates of lichen biomass to quantify patterns of lichen recovery in different stand types, ecoprovinces, and with time following stand-replacing fire. We used a hurdle model to account both for the heterogeneous nature of lichen presence (zero inflation) and for the range of abundance in stands where lichen was present. The first component of the hurdle model, a generalized linear model, identified stand age, stand type, and ecoprovince as significant predictors of lichen presence. With a logistic growth model, a measure of lichen recovery (time to 50% asymptotic value) varied from 28 to 73 yr, dependent on stand type and ecoprovince. The combined predictions of the hurdle model suggest the most rapid recovery of lichen biomass across our study region occurred in jack pine in the Boreal Shield (30 yr), while stands located in the Taiga Plains (North and South) required a longer recovery period (approximately 75 yr). These results provide a basis for estimating future caribou habitat that encompasses some of the large variation in fire effects on lichen abundance and vegetation types across the range of boreal and barren-ground caribou in North America.

Increased fire activity due to climate change may impact the successional dynamics of boreal forests, with important consequences for caribou habitat. Early successional forests have been shown to support lower quantities of caribou forage lichens, but geographic variation in, and controls on, the rates of lichen recovery has been largely unexplored. In this study, we sampled across a broad region in northwestern Canada to compare lichen biomass accumulation in ecoprovinces, including the Saskatchewan Boreal Shield, the Northwest Territories Taiga Shield, and Northwest Territories Taiga Plains, divided into North and South. We focused on the most valuable Cladonia species for boreal and barren-ground caribou: Cladonia mitis and C. arbuscula, C. rangiferina and C. stygia, and C. stellaris and C. uncialis. We developed new allometric equations to estimate lichen biomass from field measurements of lichen cover and height; allometries were consistent among ecoprovinces, suggesting generalizability. We then used estimates of lichen biomass to quantify patterns of lichen recovery in different stand types, ecoprovinces, and with time following stand-replacing fire. We used a hurdle model to account both for the heterogeneous nature of lichen presence (zero inflation) and for the range of abundance in stands where lichen was present. The first component of the hurdle model, a generalized linear model, identified stand age, stand type, and ecoprovince as significant predictors of lichen presence. With a logistic growth model, a measure of lichen recovery (time to 50% asymptotic value) varied from 28 to 73 yr, dependent on stand type and ecoprovince. The combined predictions of the hurdle model suggest the most rapid recovery of lichen biomass across our study region occurred in jack pine in the Boreal Shield (30 yr), while stands located in the Taiga Plains (North and South) required a longer recovery period (approximately 75 yr). These results provide a basis for estimating future caribou habitat that encompasses some of the large variation in fire effects on lichen abundance and vegetation types across the range of boreal and barren-ground caribou in North America.

Fungi play key roles in carbon (C) dynamics of ecosystems: saprotrophs decompose organic material and return C in the nutrient cycle, and mycorrhizal species support plants that accumulate C through photosynthesis. The identities and functions of extremophile fungi present after fire can influence C dynamics, particularly because plant-fungal relationships are often species-specific. However, little is known about the function and distribution of fungi that survive fires. We aim to assess the distribution of heat-resistant soil fungi across burned stands of boreal forest in the Northwest Territories, Canada, and understand their functions in relation to decomposition and tree seedling growth. We cultured and identified fungi from heat-treated soils and linked sequences from known taxa with high throughput sequencing fungal data (Illumina MiSeq, ITS1) from soils collected in 47 plots. We assessed functions under controlled conditions by inoculating litter and seedlings with heat-resistant fungi to assess decomposition and effects on seedling growth, respectively, for black spruce ( Picea mariana ), birch ( Betula papyrifera ), and jack pine ( Pinus banksiana ). We also measured litter decomposition rates and seedling densities in the field without inoculation. We isolated seven taxa of heat-resistant fungi and found their relative abundances were not associated with environmental or fire characteristics. Under controlled conditions, Fayodia gracilipes and Penicillium arenicola decomposed birch, but no taxa decomposed black spruce litter significantly more than the control treatment. Seedlings showed reduced biomass and/or mortality when inoculated with at least one of the fungal taxa. Penicillium turbatum reduced growth and/or caused mortality of all three species of seedlings. In the field, birch litter decomposed faster in stands with greater pre-fire proportion of black spruce, while black spruce litter decomposed faster in stands experiencing longer fire-free intervals. Densities of seedlings that had germinated since fire were positively associated with ectomycorrhizal richness while there were fewer conifer seedlings with greater heat-resistant fungal abundance. Overall, our study suggests that extremophile fungi present after fires have multiple functions and may have unexpected negative effects on forest functioning and regeneration. In particular, heat-resistant fungi after fires may promote shifts away from conifer dominance that are observed in these boreal forests.

© 2020 The Authors. Ecography published by John Wiley & Sons Ltd on behalf of Nordic Society Oikos Climate change is altering disturbance regimes outside historical norms, which can impact biodiversity by selecting for plants with particular traits. The relative impact of disturbance characteristics on plant traits and community structure may be mediated by environmental gradients. We aimed to understand how wildfire impacted understory plant communities and plant regeneration strategies along gradients of environmental conditions and wildfire characteristics in boreal forests. We established 207 plots (60 m2) in recently burned stands and 133 plots in mature stands with no recent fire history in comparable gradients of stand type, site moisture (drainage) and soil organic layer (SOL) depth in two ecozones in Canada's Northwest Territories. At each plot, we recorded all vascular plant taxa in the understory and measured the regeneration strategy (seeder, resprouter, survivor) in burned plots, along with seedbed conditions (mineral soil and bryophyte cover). Dispersal, longevity and growth form traits were determined for each taxon. Fire characteristics measured included proportion of pre-fire SOL combusted (fire severity), date of burn (fire seasonality) and pre-fire stand age (time following fire). Results showed understory community composition was altered by fire. However, burned and mature stands had similar plant communities in wet sites with deep SOL. In the burned plots, regeneration strategies were determined by fire severity, drainage and pre- and post-fire SOL depth. Resprouters were more common in wet sites with deeper SOL and lower fire severity, while seeders were associated with drier sites with thinner SOL and greater fire severity. This led to drier burned stands being compositionally different from their mature counterparts and seedbed conditions were important. Our study highlights the importance of environment–wildfire interactions in shaping plant regeneration strategies and patterns of understory plant community structure across landscapes, and the overriding importance of SOL depth and site drainage in mediating fire severity, plant regeneration and community structure.

Boreal forest fires emit large amounts of carbon into the atmosphere primarily through the combustion of soil organic matter1,2,3. During each fire, a portion of this soil beneath the burned layer can escape combustion, leading to a net accumulation of carbon in forests over multiple fire events4. Climate warming and drying has led to more severe and frequent forest fires5,6,7, which threaten to shift the carbon balance of the boreal ecosystem from net accumulation to net loss1, resulting in a positive climate feedback8. This feedback will occur if organic-soil carbon that escaped burning in previous fires, termed ‘legacy carbon’, combusts. Here we use soil radiocarbon dating to quantitatively assess legacy carbon loss in the 2014 wildfires in the Northwest Territories of Canada2. We found no evidence for the combustion of legacy carbon in forests that were older than the historic fire-return interval of northwestern boreal forests9. In forests that were in dry landscapes and less than 60 years old at the time of the fire, legacy carbon that had escaped burning in the previous fire cycle was combusted. We estimate that 0.34 million hectares of young forests (<60 years) that burned in the 2014 fires could have experienced legacy carbon combustion. This implies a shift to a domain of carbon cycling in which these forests become a net source—instead of a sink—of carbon to the atmosphere over consecutive fires. As boreal wildfires continue to increase in size, frequency and intensity7, the area of young forests that experience legacy carbon combustion will probably increase and have a key role in shifting the boreal carbon balance.

Climate warming and drying is associated with increased wildfire disturbance and the emergence of megafires in North American boreal forests. Changes to the fire regime are expected to strongly increase combustion emissions of carbon (C) which could alter regional C balance and positively feedback to climate warming. In order to accurately estimate C emissions and thereby better predict future climate feedbacks, there is a need to understand the major sources of heterogeneity that impact C emissions at different scales. Here, we examined 211 field plots in boreal forests dominated by black spruce (Picea mariana) or jack pine (Pinus banksiana) of the Northwest Territories (NWT), Canada after an unprecedentedly large area burned in 2014. We assessed both aboveground and soil organic layer (SOL) combustion, with the goal of determining the major drivers in total C emissions, as well as to develop a high spatial resolution model to scale emissions in a relatively understudied region of the boreal forest. On average, 3.35 kg C m−2 was combusted and almost 90% of this was from SOL combustion. Our results indicate that black spruce stands located at landscape positions with intermediate drainage contribute the most to C emissions. Indices associated with fire weather and date of burn did not impact emissions, which we attribute to the extreme fire weather over a short period of time. Using these results, we estimated a total of 94.3 Tg C emitted from 2.85 Mha of burned area across the entire 2014 NWT fire complex, which offsets almost 50% of mean annual net ecosystem production in terrestrial ecosystems of Canada. Our study also highlights the need for fine-scale estimates of burned area that represent small water bodies and regionally specific calibrations of combustion that account for spatial heterogeneity in order to accurately model emissions at the continental scale.

Increased fire frequency, extent and severity are expected to strongly affect the structure and function of boreal forest ecosystems. In this study, we examined 213 plots in boreal forests dominated by black spruce (Picea mariana) or jack pine (Pinus banksiana) of the Northwest Territories, Canada, after an unprecedentedly large area burned in 2014. Large fire size is associated with high fire intensity and severity, which would manifest as areas with deep burning of the soil organic layer (SOL). Our primary objectives were to estimate burn depth in these fires and then to characterise landscapes vulnerable to deep burning throughout this region. Here we quantify burn depth in black spruce stands using the position of adventitious roots within the soil column, and in jack pine stands using measurements of burned and unburned SOL depths. Using these estimates, we then evaluate how burn depth and the proportion of SOL combusted varies among forest type, ecozone, plot-level moisture and stand density. Our results suggest that most of the SOL was combusted in jack pine stands regardless of plot moisture class, but that black spruce forests experience complete combustion of the SOL only in dry and moderately well-drained landscape positions. The models and calibrations we present in this study should allow future research to more accurately estimate burn depth in Canadian boreal forests.