The global land sector is increasingly expected to contribute towards net zero emission goals. Boreal forests can be managed to enhance forest sinks, but are also at risk from natural disturbances and climate change. Understanding the impacts of forest management, natural disturbances and climate change on future greenhouse gas emissions and removals requires scientifically-credible tools with which to establish baseline emissions as well as the outcomes of alternative management scenarios. The Generic Carbon Budget Model (GCBM) is an open-source, modular, spatially-explicit model of forest carbon dynamics built on the Full Lands Integration Tool (FLINT) platform of moja global. The GCBM incorporates the science modules of the Carbon Budget Model of the Canadian Forest Sector (CBM-CFS3) and implements IPCC Tier 3 methods that link biomass, dead organic matter and soil carbon dynamics to estimate emissions and removals as affected by management, land-use change and natural disturbances. Here we provide an overview of the GCBM, highlight the development of new science modules (moss, peatlands, albedo, fire severity), provide examples of the the application of the model to estimate past and projected future GHG emissions and removals, in managed and unmanaged northern forests of Canada, and discuss future improvement plans. The open-source approach facilitates ongoing collaboration with model developers and scientists in Canada and internationally. Models such as the GCBM are also required to identify and quantify the benefits of forest carbon management investments aimed at enhancing forest carbon sinks, such as tree planting, rehabilitation after natural disturbances, and fuel management actions. Nesting of such activities within a spatially-explicit framework will also enable future reporting of carbon investment outcomes within Canada’s National Forest Carbon Monitoring, Accounting and Reporting System.

Heterotrophic respiration is a major process releasing carbon to the atmosphere and a major source of uncertainties in forest carbon budget projections. We analysed the seasonal dynamics of heterotrophic (microbial) respiration of soil and downed woody debris (DWD) based on repeated measurements of CO2 fluxes with a closed dynamic chamber method using an infrared gas analyzer in a mixed middle boreal old-growth forest of Piceetum myrtillosum-oxalidosum type, located in the Kivach reserve, Russia. The soils are podzolic surfacely gleish sandy-loamy (Alic Stagnosols).

Mean CO2 emission from the soil surface was 89.5 g C m-2 h-1 with the variability of 25-39% due to microsite variation. Total annual CO2 emission from the soil surface averaged 639 g C m-2. The mean share of microbial respiration was 79%. The seasonal dynamics of soil respiration was 44-62% dependent on soil temperature at the depth of 10 cm. The Q10 varied from 2.15 to 3.65. The cold season (November-April) contributed 15-22% to the total annual CO2 emission from soil surface. The differences in the temperature regime among microsite types were the most pronounced in warm period.

The CO2 flux from DWD varied from 4 to 395 mg C m-2 h-1, and averaged 98.3 mg C m-2 h-1. The factors influencing this variability – climatic variables and physicochemical substrate traits – were interrelated. The air temperature, tree species identity and decay class limited the total CO2 flux from the DWD surface. The trends in the dynamics of share of wood and bark in microbial respiration of DWD were tree species specific. The respiration from DWD comprised 15-16% of the total microbial respiration (soil + DWD) during the vegetation period of 180 days.

The study was financially supported by the Russian Foundation for Basic Research (#19-04-01282).

The importance of the boreal biome in the global carbon cycle is well recognized. However, current estimates of its sink-source strength at regional scales and its responses to climate change rely primarily on models and thus remain uncertain. The aim of this study is to quantify the C balance over the north Scandinavian boreal region by integrating observations of land-atmosphere fluxes and atmospheric CO2 concentrations at landscape to regional scales and to better understand the impact of 2018 drought on this region. The -Svartberget site (SVB), an established ICOS (Integrated Carbon Observation System) station in Northern Sweden offers a unique range of observations, from plot scale chamber measurements to EC fluxes and tall-tower concentration measurements. Here in this study, apart from flux and concentration observations, we used several vegetation models and an atmospheric transport model to connect the different spatial scales for the period 2016-2018. We found that northern Sweden region remained as a C sink for the study period with models differed in sink strength between -25 to -275 gC m-2 year-1. All models indicate similar but small reductions up to 50 gC m-2 year-1 in the net CO2 uptake for the drought year 2018 in northern Sweden except LPJ-GUESS that reveal limitations which call for further model improvement. With the LUMIA/FLEXPART forward model runs we could show that atmospheric concentrations are sensitive to ecosystem fluxes within the study domain and can be used further to constrain inverse model approaches. Our work highlights the interest of using combined ecosystem-atmosphere ICOS sites such as SVB in the Scandinavian region and shows that it is a promising way forward to monitor CO2 fluxes and concentrations at the regional scale.

Disturbances in the boreal forest, such as eastern spruce budworm (Choristoneura fumiferana) outbreaks, impact the strength and persistence of the forest as a carbon sink. Salvage harvests are a typical management response to widespread tree mortality, but our prior research has shown that the decision to salvage mortality has large implications for the fate of carbon stocks (including forest carbon and harvested wood products) in the near and long terms. We used a range of economic discounting rates and applied decision tree machine learning algorithms as an interpretable rule-based classification technique to develop a more refined knowledge of which forest stand conditions should salvaging be encouraged by policy-makers to support emissions reduction and carbon sequestration objectives. We derived carbon sequestration outcomes from growth and harvest simulation models that incorporated a life cycle assessment of the fate of harvested material in comparative “salvage” vs. “no salvage” scenarios using US Forest Service Forest Inventory and Analysis plot data for the northeastern United States. We found that long-term carbon sequestration is negatively impacted by salvage logging when the volume of mortality is high. However, interpretable rule-based classification methods can be used to determine specific forest stand characteristics where salvaging is likely to lead to more beneficial sequestration outcomes for specific stand conditions. Applying economic discount rates to future carbon stocks further affects rule-based classification model utility, where higher discount rate scenarios suggest more carbon sequestration can be achieved from using classification model decision-making compared to lower discount rate scenarios. Our findings provide important considerations for forest managers and policy-makers to optimize boreal forest carbon sinks following major disturbances.

Boreal ecosystems store 30% to 40% of terrestrial carbon (C). Thus, changes to the boreal C balance will impact the global C cycle and climate. Decomposition is a critical component of C balance, and decay models are frequently used in global C cycling models. Yet, numerous decay models exist without consensus on models best fit for boreal ecosystems. Here, we compare three decay models (single exponential, asymptotic, and Weibull) fit to mass loss data from litterbags deployed in two boreal forest types in Interior Alaska: black spruce (Picea mariana) and paper birch (Betula neoalaskana). We selected three blocks containing adjacent stands of spruce and birch. Within each block, we randomly located five plots in each stand type (15 plots per type) and installed five sets containing one bag of each litter type (birch, spruce, and moss) in every plot. We collected one set from each plot annually over five years. We also measured annual litter inputs collected in traps. We fit decay models for each environment-litter combination and estimated mean residence time (MRT). We calculated stand-level MRT by weighting the litter MRT to annual litter inputs and used a linear mixed effects model to test differences in MRT between stand types and models. The single exponential model was the best fit for most environment-litter combinations. However, it performed poorly when performance metrics were combined, indicating it did not provide as much flexibility as the other models. Models produced similar estimates of stand-level MRT for birch stands (~8 years) but differed for spruce (p<0.001), ranging from 4 to 27 years. Variability in spruce stands is likely driven by poor estimates of moss decomposition, which was too slow to fully capture here. These results suggest that decay model selection substantially influences interpretation of ecosystem-level turnover time, an important component of C balance.

Boreal forests are facing multiple threats from climate change and increasing wildfires, but models seldom capture all the critical drivers and feedbacks that are important in these ecosystems. We developed a new spatially-explicit model that tightly couples climate, vegetation, soil (including permafrost), hydrology and wildfire. The model, DGS-Succession (pronounced “digs”), combines the DAMM-McNiP model of soil C and N dynamics, the heat balance and hydrology model of SHAW and the permafrost module of GIPL. Using this new model, we determined the relative importance of soil moisture and soil temperature in driving long-term changes in above- and belowground C in interior Alaska under wildfire and climate change. Model projections of aboveground carbon showed good agreement with FIA data; modeled soil moisture and temperature generally reproduced the interannual variability and differences between depths in the empirical data across all three sites used for calibration. Preliminary results showed that tree-available soil moisture at the landscape-scale was similar between the two climate scenarios initially, but then declined under climate change in ~2030 and remained 29.5% lower than historical levels until mid-century. Active layer thickness at the landscape-scale declined dramatically after only ten years and by mid-century, there was a 6-fold decline. Soil moisture was the most influential driver of long-term changes in aboveground biomass, though soil temperature also played an important role, limiting growth by 10-40%. Modeling efforts that integrate climate, wildfire, hydrology, permafrost, and species dynamics are essential for projecting non-analog species assemblages, feedbacks, and above- and belowground carbon dynamics in boreal forests.

The Canadian Model for Peatlands (CaMP) is a site- to national-level peatland carbon model developed to better estimate greenhouse gas (GHG) emissions and removals across a range of open and forested peatland types in Canada. The CaMP is a module for the Generic Carbon Budget Model (GCBM), a new open-source, spatially-explicit carbon dynamics model that builds on the science of the Carbon Budget Model of the Canadian Forest Sector (CBM-CFS3). The CBM-CFS3 is currently used to estimate and report GHG emissions from Canada’s managed forest area. This version of the CaMP quantifies peatland carbon exchanges with the atmosphere (CO2, CH4, CO) by modelling the productivity of different vegetation layers (trees, shrubs, sedges, and mosses), and their turnover and decay. Decomposition is modelled using a Q10 relationship with mean annual air temperature and explicit representation of acrotelm and catotelm dynamics. The CaMP utilizes the maximum annual value of the widely available Canadian Fire Weather Index System’s Drought Code to predict the long-term and annual fluctuations in water table position for different peatland types. Spatial layers with actual area burned across Canada are applied to estimate wildfire impacts on GHG emissions, transfers between C pools due to plant mortality, and change in water table depth. Here, we present estimates of GHG emissions and removals from peatlands across Canada for the period 1990 to 2018. Peatland GHG exchanges will be compared across different ecoregions, peatland types, and severity of wildfire impacts. Future steps towards reducing model uncertainty and increasing the model’s ability to project the impacts of anthropogenic disturbances and climate change will also be discussed.

The soil carbon models may be accurately predicting the change in forest carbon stock in mineral soils but their predictions under water saturated conditions are biased. This could be because under typical conditions the soil surface, for which the models were developed, is the most active zone of carbon dynamics but deeper soil carbon pools, for which the models deviate most, is poorly represented. The empirical temperature and water response functions of decomposition vary greatly between current biogeochemical models and so vary their soil carbon estimates and predictions to global warming.

We evaluated the soil carbon stocks and soil CO2 emissions with Yasso07 soil carbon model coupled with Dual Arrhenius Michaelis-Menten (DAMM) like representation of empirical temperature and moisture functions and with Bayesian data assimilation. We have found that the Yasso07 model with default environmental controls represented reasonably well CO2 emissions along the whole forest – mire gradient and, as expected, failed in estimation of peatland soil carbon stock. However, the DAMM environmental functions accounting for substrate dynamics, when calibrated along the forest - mire ecotone, represented both soil carbon stocks and emissions accurately. For accurate representation of decomposition, it was crucial that the moisture optimum was in the range of conditions of mineral soils and decomposition was exponentially reduced in mires.

Unlike for functions of most soil carbon models derived from a small range of moisture conditions of mineral soil forest, the environmental controls based on the whole range of well drained to water saturated soils reflect the microbial substrate dynamics and can be used for accurate modelling of soil carbon stocks and CO2 emissions of both mineral and organic soil forests.

In a common effort to mitigate climate change, the parties of the United Nations Framework Convention on Climate Change (UNFCCC) have to produce annual reports on their national greenhouse gas (GHG) emissions. These reports are a valuable source of information and they are often used to measure the effectiveness of national climate mitigation strategies over time. However, large parts of GHG inventories rely on estimated quantities. As a consequence, the reported figures are uncertain. Quantifying this uncertainty is crucial as it may affect our ability to distinguish the true trends from the intrinsic variability.

In this study, five statistical techniques for uncertainty quantification, among which two are recommended by the Intergovernmental Panel on Climate Change (IPCC), were evaluated as to their ability to correctly estimate the variance. The standard Monte Carlo estimator, which is one of the two techniques recommended by the IPCC, overestimated the true variance. It was no better than a naı̈ve estimator. The propagation-based estimator, which is the second technique recommended by the IPCC, also overestimated the true variance but to a lesser extent. Goodman’s estimator and a rescaled Monte Carlo estimator were both unbiased and they should be preferred when evaluating GHG emissions.

Estimates of carbon sink of Russia's forests in both peer-reviewed publications and the national communications to international organizations (IPCC, UN FAO) during the last decades are in the range from 150 to 1000 Tg C yr-1. We present the systems requirements to and results of a full and verified carbon account of the forest ecosystems of Russia (FCA) which allows to get unbiased proxy of the carbon (C) sink (NECB, NBP) and corresponding uncertainties. The methodology used is based on understanding the FCA as an underspecified (fuzzy) system that requires integration of different independent methods of the carbon account. The landscape-ecosystem approach (LEA) serves as the semi-empirical and structural basis of the FCA. The results, obtained by other methods (process-based models; inverse modelling; eddy covariance; “multi-concept” of remote sensing (RS)) are used for mutual constraints of the final results and uncertainties of the FCA. The information background is presented in form of the Integrated Land Information System (ILIS), which integrates relevant on-ground and remote sensing (RS) data and contains maximum available information about forest ecosystems and landscapes, spatially referred to the hybrid land cover with spatial resolution 150-200 m. Major features of the reanalysis includes: 1) availability of time series of basic forest parameters by forest enterprises (RS products of the “Space Science Observatory of Carbon of Russia” Project); 2) modified models and reference data for assessing amount and emissions due to decomposition of Coarse Woody Debris; 3) modelling system for assessing net and gross growth; 4) modified models and algorithms for assessing spatially distributed emissions caused by disturbances (fire, harvest, biogenic); 5) modified reference data for assessing dynamics of soil carbon stocks and emissions; and some others. We discuss the strengths and weaknesses of the methods used in the FCB and major results received for 2001-2018.