Subnivean Arctic and sub-Arctic net ecosystem exchange (NEE)

Abstract

In the Arctic and sub-Arctic, up to half of annual net ecosystem exchange (NEE) occurs during the snow season. Subnivean soil respiration can persist at a greater rate when the overlying snowpack has a lower thermal conductivity, and the rate of photosynthetic uptake at the start and end of the snow season can be diminished by fractional snow cover. Although recent studies have indicated that uncertainty in model estimates of NEE can be reduced by representing the influence of a modeled snowpack on soil respiration, models of NEE have not represented the influence of snowpack dynamics on processes such as subnivean photosynthesis or CO₂ diffusivity, and have not used remote sensing observations to characterize snow season processes. We therefore: (1) review snow season processes and their effects on NEE; (2) assess the suitability of cryospheric remote sensing approaches for models of NEE; and (3) suggest strategies for representing snow season processes in models of NEE. Strategies include: using observations of fractional snow cover in spring and fall to restrict estimates of photosynthetic uptake; combining observations of snow accumulation and soil freeze/thaw with observations of air temperature to generate more realistic estimates of soil temperature and soil respiration; and using observations of depth to estimate the influence of snow accumulation and tree wells on soil respiration. Including remote sensing observations of snow properties in models of NEE could reduce uncertainty in snow season estimates of NEE, resulting in a better understanding of the northern carbon cycle and how it is responding to climate-driven changes in the interconnected biospheric, atmospheric and cryospheric systems.

Keywords

net ecosystem exchange permafrost photosynthesis polar remote sensing respiration snow soil carbon

I Introduction

Anthropogenic emissions have the potential to increase global atmospheric concentrations of CO₂ to twice the pre-industrial level by approximately 2100, which is predicted to cause warming of high-latitude regions (Christensen et al., 2007). Concern exists that climate change is likely to be amplified through positive climate-carbon-cycle feedbacks (Cox et al., 2000; Friedlingstein et al., 2001), such as the biospheric release of CO₂ and CH₄ from thawing permafrost (Tarnocai, 2006). Northern cryospheric soils currently contain 1400–1850 Pg stores of organic carbon (1 Pg=10¹⁵g) (McGuire et al., 2010) which, if all released into the atmosphere as CO₂, would theoretically equate to an increase of 666–880 ppm in global atmospheric concentrations. Simulations using the Intergovernmental Panel on Climate Change A1B emission scenario for 2200 indicate that permafrost thaw could increase atmospheric concentrations of CO₂ by 87 ± 29 ppm (Schaefer et al., 2011). As permafrost thaws, a portion of the increase in CO₂ efflux is predicted to be offset by greater uptake of carbon by vegetation, although uncertainty exists regarding the magnitude of the biospheric response to warming and the time period over which it will persist (Cramer et al., 2001; Euskirchen et al., 2006), partially due to the complexity of the predicted response of high-latitude net ecosystem exchange to rising atmospheric concentrations of CO₂ (Figure 1).

Figure 1.

Flowchart representing how northern high-latitude NEE is predicted to respond to rising levels of atmospheric CO₂, and subsequent warming. Warming-induced increases are indicated in red (+), and decreases are indicated in blue (–).

Models provide local to global scale estimates of the land-atmosphere exchange of carbon based on locally observed associations between carbon cycling, meteorological conditions and land surface characteristics. The modeling approach used to generate estimates of carbon cycling can be described as either process-based or remote sensing based, where process-based models predominantly represent interacting physical processes and remote sensing based models rely on spatial observations of the land surface (Cramer et al., 1999). Several components of the land-atmosphere exchange of carbon can be estimated using these models, including net primary productivity (NPP), net ecosystem exchange (NEE) and net ecosystem productivity (NEP) (Cramer et al., 1999). NPP refers to the net carbon uptake by vegetation in g/m²/year, and is equal to the total carbon fixed by photosynthesis (gross primary productivity, GPP) minus the carbon returned as CO₂ to the atmosphere during vegetation growth and maintenance respiration (autotrophic respiration, R_a) (Bonan, 2002):

N P P = G P P - R_{a}

In terrestrial ecosystems, NEE refers to the instantaneous vertical flux of CO₂ into and out of an ecosystem in μmol/m²/s which occurs primarily through processes such as photosynthesis and respiration. Terrestrial NEE is equal to total carbon fixed by photosynthesis minus autotrophic respiration and carbon loss through soil organic matter decomposition by microorganisms (heterotrophic respiration, R_h). Calculations of NEE do not include inorganic sources and sinks of CO₂ (e.g. precipitation and weathering) which can be substantial in aquatic ecosystems but are minimal in terrestrial ecosystems. Therefore, when gaseous fluxes of CH₄, lateral fluxes of dissolved organic carbon and dissolved inorganic carbon are minimal, terrestrial NEE is equal in magnitude to NEP, but with a sign reversal to reflect the fact that NEE is negative when CO₂ is removed from the atmosphere (Lovett et al., 2006):

- N E E = N E P = G P P - R_{a} - R_{h} = N P P - R_{h}

Recent Arctic and sub-Arctic field studies of NEE have indicated that the snow season accounts for up to 50% of annual CO₂ efflux at sub-Arctic (Aurela et al., 2004; Zimov et al., 1996), low Arctic (Mikan et al., 2002; Sullivan et al., 2008) and high Arctic (Elberling and Brandt, 2003) sites. Studies have also indicated that greater snow accumulation can encourage greater releases of CO₂ at sub-Arctic (Larsen et al., 2007a), low Arctic (Walker et al., 1999), and high Arctic (Morgner et al., 2010) sites. The focus of this paper is on exploring the possibility that incorporating remote sensing observations of snow properties and their influence on NEE could reduce uncertainty in model estimates of Arctic and sub-Arctic NEE. The locations at which key in situ studies of Arctic and sub-Arctic snow season NEE were conducted are summarized in Table 1, and mapped in Figure 2.

Table 1.

Locations of selected key in situ studies of snow-CO₂ interactions, classified as sub-Arctic, low Arctic and high Arctic according to Walker et al. (2005). Sub-Arctic sites indicated with an asterisk are considered to be south of the treeline by Walker et al. (2005), but are characterized by environmental conditions typical of low Arctic regions.

Site	Type	Latitude	Longitude	Snow-CO₂ research conducted at this site
S1	Sub-Arctic	61°08–10’N	149°38–49’W	Sullivan et al. (2010)
S2	Sub-Arctic	68°20–21’N	18–19°E	Björkman et al. (2010); Bokhorst et al. (2010); Grogan and Jonasson (2005, 2006); Grogan et al. (2004); Larsen et al. (2007a, 2007b)
S3	Sub-Arctic	67°29’N	162°12’W	Sullivan (2010)
S4	Sub-Arctic	69°08’N	27°17’E	Aurela et al. (2004)
T1	Sub-Arctic*/Low Arctic	58°45’N	94°04’W	Lafleur et al. (2001)
T2	Sub-Arctic*/Low Arctic	63°52’N	149°15’W	Schuur et al. (2009)
T3	Sub-Arctic*/Low Arctic	64°75’N	148°25’W	Grogan and Chapin (1999)
T4	Sub-Arctic*/Low Arctic	68°N	134°W	Jones et al. (1999)
T5	Sub-Arctic*/Low Arctic	68°18’N	18°51’E	Fox et al. (2008)
T6	Sub-Arctic*/Low Arctic	68°30’N	161°24’E	Zimov et al. (1993, 1996)
L1	Low Arctic	64°50–52’N	111°34–38’W	Buckeridge and Grogan (2010); Buckeridge et al. (2010); Humphreys and Lafleur (2011); Nobrega and Grogan (2007); Oelbermann et al. (2008)
L2	Low Arctic	68°37–38’N	149°18–38’W	Euskirchen et al. (2012); Mikan et al. (2002); Nowinski et al. (2010); Oberbauer et al. (1998); Olsson et al. (2003); Starr and Oberbauer (2003); Sullivan et al. (2008); Walker et al. (1999)
L3	Low Arctic	68°28’N	155°44’W	Olsson et al. (2003)
L4	Low Arctic	71°19’N	156°36’W	Panikov et al. (2006); Tieszen (1974)
H1	High Arctic	74°30’N	20°30–21’W	Christensen et al. (2012); Elberling and Brandt (2003); Elberling et al. (2008)
H2	High Arctic	78°10’N	16°04’E	Björkman et al. (2010); Morgner et al. (2010)
H3	High Arctic	78°N	19°E	Elberling (2007)

Figure 2.

Pan-Arctic locations at which field studies of snow season NEE were conducted (red points) and locations of FLUXNET stations (black triangles). Regions are defined as low Arctic (yellow), high Arctic (red) and below the treeline (green) according to the Circumpolar Arctic Vegetation Map (Walker et al., 2005). Mean AVHRR NDVI values from the CAVM are shown for regions south of the treeline, where darker shades of green correspond to larger values of NDVI.

II Modeling snow season NEE

In situ observations of NEE can be upscaled to generate estimates over larger areas of the Arctic using: (1) meteorological inputs and physiological relationships between temperature, radiation, hydrology, phenology and respiration (Vourlitis et al., 2000); (2) direct observations of NEE (McGuire et al., 2012); or (3) a combination of observed NEE and remote sensing estimates of leaf area index and land cover (Marushchak et al., 2012). However, due to the size and spatial heterogeneity of landscapes in the Arctic and sub-Arctic (Lantz et al., 2010; Nobrega and Grogan, 2008), errors can arise when upscaling is used to generate circumpolar estimates of NEE (La Puma et al., 2007). Challenges in upscaling NEE are worsened by the sparse and uneven distribution of eddy covariance towers at high latitudes (Baldocchi, 2008).

Estimates of biospheric carbon cycle variables such as GPP, respiration, NEE and NPP are typically modeled from a variety of remotely sensed, meteorological and eddy covariance-derived inputs, in addition to measured or optimized region-specific parameters (Baker et al., 2008; Bondeau et al., 2007; Cramer et al., 1999; Krinner et al., 2005; Randerson et al., 2009; Sitch et al., 2007). Biospheric models vary greatly in terms of the approach used (remote sensing based versus process-based), inherent simplifying assumptions, initial conditions, complexity with which various processes are represented, representation of land-use change and whether or not they can be applied prognostically. Generally, however, models must be capable of describing how land surface characteristics (e.g. vegetation, soil) and meteorological conditions (e.g. temperature, photosynthetically active radiation) result in given levels of NEE. These model relationships must be adequately generalized to yield estimates over regions where in situ observations are unavailable (Cramer et al., 1999; Dietze et al., 2011; Huntzinger et al., 2012; Sitch et al., 2007).

The Simple Biosphere Model (SiB) is a process-based model that calculates the regional land-atmosphere exchange of energy, mass and momentum by representing small-scale physical processes (e.g. transpiration, runoff, respiration, photosynthesis) using equations with a physical or biological basis (Sellers et al., 1986). Models such as SiB are highly useful for simulating the land surface energy balance in a biophysically accurate manner, and can be used alone or in combination with general circulation models for predictive and simulation purposes. In a full description of the SiB3 numerical scheme, Baker (2005) indicates that although growing season processes are comprehensively described, and snow properties are calculated according to the Common Land Model (CLM) formulation by Dai et al. (2003), grid cells could only be described as either snow-covered or snow-free. Including a representation of fractional snow cover may therefore improve estimates of snow season photosynthesis and respiration by SiB3.

The Vegetation Photosynthesis and Respiration Model (VPRM) is a remote sensing based model that estimates NEE according to a simple mathematical structure. Respiration is described as a linear function of air temperature, and photosynthetic uptake is calculated as a function of air temperature, incoming shortwave radiation, and estimates of vegetation biomass and land surface moisture from visible and optical remote sensing derived indices. Whereas process-based models contain complex representations of processes, VPRM contains only four parameters per vegetation class (Mahadevan et al., 2008). Therefore, although VPRM cannot be used for biophysical simulation of processes, the resulting estimates of NEE have shown good agreement against observational data, and the errors and uncertainties can be traced to specific inputs (Lin et al., 2011), and the utility of these models is therefore in generating regional estimates of NEE. Remote sensing based models of NEE such as VPRM have not explicitly represented the influence of snowpack parameters on NEE.

Recently, several process-based models have represented the influence of a model snowpack on soil temperature and respiration. For example, process-based simulations of soil CO₂ efflux by Kucharik et al. (2000) and Pumpanen et al. (2003) made use of land surface temperature, including snow surface temperature when available; however, no mention is made of specific adaptations made to the model for describing subnivean production or diffusion of CO₂. McGuire et al. (2000) generated estimates of heterotrophic respiration both with and without a representation of the insulation provided by snow using the following three process-based models: Century, the Terrestrial Ecosystem Model (TEM) and the Carnegie-Ames-Stratford Approach (CASA). When the model deemed snow to be present, soil temperature was set to 0ºC. Evaluations of the resulting models indicated that models showed better agreement against observational data when the influence of a snowpack was explicitly represented, and that this prevents heterotrophic respiration from being underestimated during the snow season.

Modifications have been made to several process-based models to allow them to better simulate the physical processes driving Arctic NEE such as snow and vegetation. For example, Wania et al. (2009) modified the Lund-Potsdam-Jena Dynamic Global Vegetation Model (LPJ-DGVM, by Sitch et al., 2003) so that it would include snow densification over time, permafrost, northern hydrology and peatland vegetation. Snow was simulated as a homogeneous layer that increased in density from 150 to 500 kg/m³ during the snow season, with snow season length assumed to be the same length as the previous snow season and thermal conductivity calculated as a function of density according to parameters developed by Sturm et al. (1997). Similarly, snow is modeled as a homogenous layer in ORCHIDEE, which changes over time through melt, sublimation and snowfall, all of which are calculated according to the surface energy balance. The thermal conductivity of the snowpack is typically fixed, but more realistic estimates of soil carbon have been generated by incorporating field-based observations of snow properties (Gouttevin et al., 2012).

Therefore, although several models have benefited from including a simple representation of snow processes or field-based observations of snow characteristics, the influence of snow on soil respiration is not usually represented in models of NEE, and processes such as subnivean photosynthesis are not represented in existing models. Furthermore, when models of NEE do simulate snow accumulation, they do so in a relatively simple manner, such that snow metamorphism and the formation of ice lenses or depth hoar do not occur, although these influence snowpack thermal conductivity and diffusivity. Site characteristics, such as vegetation and topography, also may not have a realistic influence on snowpack development in existing models. Due to the challenges in accurately simulating snowpack dynamics, and due to the spatial dependence of snow and NEE, we suggest that remote sensing observations of snow may provide an optimal approach for characterizing snow season influences on NEE throughout the Arctic and sub-Arctic. Remote sensing observations could either be directly included as inputs into remote sensing based or process-based models of NEE, or assimilated into process-based simulations in order to constrain estimates of snow accumulation. Remote sensing observations of snow season characteristics could thereby enable models of NEE to simulate the complex responses of subnivean respiration and photosynthesis to snowpack properties, and may allow uncertainty to be reduced in Arctic and sub-Arctic estimates of snow season NEE.

III Uncertainties in Arctic and sub-Arctic NEE

The following section explores uncertainties in local and regional scale observations of NEE in order to provide a context for later discussion of strategies which may reduce uncertainty in snow season estimates of high-latitude NEE. Uncertainties in observations of Arctic site-scale NEE (<10⁵ m) exist due to a number of factors involved in acquiring accurate observations (Björkman et al., 2010) and quantifying the portion of subnivean NEE that was produced during the snow season (Panikov, 2009). Regional scale (>10⁵ m) estimates of the carbon budget can be generated using a variety of approaches such as atmospheric inversion, process-based modeling, spatial upscaling of local observations, and remote sensing based modeling (Cramer et al., 1999; McGuire et al., 2012; Sitch et al., 2007). Recent syntheses by Huntzinger et al. (2012) and McGuire et al. (2012) have indicated that, although convergence exists regarding decadal trends in the Arctic carbon cycle, substantial uncertainties exist in estimates of circumpolar Arctic NEE.

1 Uncertainties in measurements of NEE

Regional surface CO₂ fluxes can be estimated using observations from Lagrangian, ‘airmass-following’ aircraft observations (Lin et al., 2004) and satellites such as the Greenhouse gases Observing Satellite (GOSAT) (Chevallier et al., 2010). However, satellites are limited in yielding accurate observations of CO₂ over Arctic and sub-Arctic regions due to the uncertainties introduced by snow/ice-covered surfaces, limited availability of cloud-free observations and a large solar zenith angle at high latitudes (Boesch et al., 2011). Observations from aircrafts are, by necessity, limited in their temporal resolution. Year-round observations of CO₂ fluxes and net carbon uptake can therefore be accurately and directly measured only at local scales (<10⁵ m), especially at high-latitude sites (Bréon and Ciais, 2010).

Local measurements of NEE can be acquired at the plot scale (10–100 m) using chambers, or at the patch scale (10⁴–10⁵ m) using eddy covariance measurements (Fox et al., 2008), which measure the land-atmosphere turbulent exchange of CO₂ (Baldocchi et al., 1988). Although the accuracy of eddy covariance observations can be compromised by cold season heat transfer in open path eddy covariance systems (Amiro, 2010), uncertainty can arise in point measurements due to landscape heterogeneity and site selection bias. Previously, these factors have resulted in 60% overestimates of chamber measured NEE compared to footprint modeled eddy covariance NEE, as observed at site S2 by Fox et al. (2008). Financial and logistical challenges associated with gathering field measurements limit the collection process itself, the types of measurements that can be taken, and the locations at which they can be acquired (Bäckstrand et al., 2010; Campbell et al., 2005), resulting in a bias towards data collection during the growing season. Recently, however, the eddy covariance technique has been used to characterize high-latitude NEE year-round at site L2 by Euskirchen et al. (2012).

A majority of snow season observations of NEE have been collected using point measurement techniques. Point observations of snow season NEE are collected using chambers connected to infrared gas analyzers either at the top of the snow surface or over a patch of ground surface from which snow has very recently been cleared. Alternatively, a trace gas technique can be used to create a time series of CO₂ efflux based on multiple sampling within the snowpack. NEE can also be measured using air sampling of CO₂ concentrations at two points above and below the snowpack then estimating CO₂ efflux based on an assumption of Fickian diffusion. However, the accuracy of these estimates may be compromised by deviations from simple Fickian diffusion approaches due to non-steady-state convection (Jones et al., 1999).

A comparison of the aforementioned CO₂ sampling techniques by Björkman et al. (2010) at a sub-Arctic (S2) and a High Arctic (H2) site showed that the trace gas and diffusion based methods found little difference in CO₂ between high and low snow accumulation conditions, whereas chamber sampling found greater effluxes of CO₂ arising from the sites with greater accumulation of snow. The application of chamber-based techniques may be especially prone to errors since chambers at the top of the snowpack measure only fluxes from the snow surface (Morgner et al., 2010), and chamber measurements taken at the soil surface immediately after snow removal can initially show very large fluxes of CO₂ before leveling off (Grogan and Jonasson, 2005). As a result, different techniques have yielded observations of NEE that vary by up to two orders of magnitude at a single site (Björkman et al., 2010). The discrepancies found between observations collected using different techniques limits the potential for comparisons to be made between sites.

The aforementioned in situ approaches are also prone to errors because they sample CO₂ at a very small scale and are therefore influenced by small-scale heterogeneities. Furthermore, snowpack stratigraphy and diffusion are altered in the process of acquiring non-automated observations of NEE, such that the manual collection of NEE observations during the snow season may be biased. As a result of discrepancies in the spatial scale and magnitude of observed NEE from different techniques, as well as the overall shortage in snow season observations at high-latitude sites, it is not possible at this point to conduct a thorough inter-site comparison of relationships between snow and NEE. However, it is still possible to rely on findings from individual sites regarding NEE and its association with snow accumulation. Future improvements in automated measurement techniques and greater generalization of sampling strategies would also allow an improved understanding of how associations between snow and NEE vary across study sites. Furthermore, the greater application of snow sampling strategies to characterize the density, grain characteristics and total accumulation of snow relative to its snow water equivalent (SWE) could further assist researchers in understanding the relationship between snow dynamics and NEE and how it varies between study sites.

2 Uncertainties in the partitioning of snow season NEE

Uncertainty exists regarding how much of the NEE released during the snow season is being produced through subnivean respiration and what portion of CO₂ released throughout the snow season was produced during the antecedent growing season. Although the exact ratio varies by season and site, it is important to evaluate the relative contributions of production and release as these define the main mechanisms and snowpack characteristics that affect snow season NEE, as snow influences both the rate of production and diffusion of CO₂.

Using laboratory testing of soil from L4, Panikov et al. (2006) found that a majority of snow season effluxes of CO₂ occurs due to microbial activity, and that variation in CO₂ release (v) can be estimated in a laboratory setting based on incubation temperature (T), the unfrozen soil water content (W), and three constants (A, λ and k): $v = A^{λ T + k W}$ (r²>0.98). This study indicated that a smaller portion of CO₂ arises from abiotic releases of accumulated CO₂. However, in regions such as T2 that are undergoing permafrost thaw, abiotic releases of CO₂ can form an important portion of net carbon exchange (Schuur et al., 2009). The exact portion of apparent CO₂ flux that arises from abiotic releases and microbial activity therefore varies according to site conditions. Elberling and Brandt (2003) found that 40% of total annual soil respiration in a Greenland heath (site H1) arose from subnivean respiration, and that 80% of subnivean respiration between 0 to –9°C became trapped in soil. Although abiotic releases of CO₂ account for a more substantial portion of efflux during the snow season than during the growing season, most of the CO₂ annually released by soil tends to arise from soil respiration (Panikov et al., 2006; Pumpanen et al., 2003; Zimov et al., 1993).

The main limitation on snow season respiration relates to the influence of sub-zero soil temperatures on reducing availability of unfrozen water (Panikov, 2009). In loam soils at site T5, free water has been observed at temperatures of –10°C (Zimov et al., 1993). Unfrozen water has even been detected in clayey permafrost soil at temperatures of –60°C (Ananyan 1970, cited by Wagner, 2008). In a laboratory setting, subsoil microbial activity has been observed to occur at temperatures of –39°C in an Arctic soil sample acquired from site L4 (Panikov et al., 2006). Subnivean temperatures are typically much warmer than –39°C, and the amount of unfrozen water in soil has been observed to decline only marginally at temperatures below –5°C at a high Arctic site (H1) (Elberling and Brandt, 2003), thereby allowing substantial subnivean production of CO₂. Furthermore, as respiration is an exothermic reaction, greater rates of respiration during the snow season also induce elevated soil temperatures and rates of decomposition (Khvorostyanov et al., 2008; Koven et al., 2011; Zimov et al., 1993).

Snow accumulation influences both the rate of diffusion and rate of production of CO₂ through respiration, and so it is difficult to quantify the exact contribution of snow accumulation to the release and production of CO₂ while uncertainty remains regarding the partitioning of snow season NEE. Field studies have indicated greater effluxes of CO₂ and warmer soil temperatures in regions with greater snow accumulation using measurement approaches that observe both instantaneous diffusion through a snowpack (e.g. Sullivan et al., 2010) as well as net accumulated effluxes of accumulated CO₂ throughout the snow season (e.g. Nobrega and Grogan, 2007). Likewise, field studies employing experimental winter warming (e.g. Bokhorst et al., 2010) and experimental nutrient enrichment (e.g. Zimov et al., 1996) have found increases in snow season CO₂ effluxes relative to control plots. Since snow season changes in temperature, nutrient enrichment and snow accumulation alter the total release of CO₂, it is clear that a substantial portion of snow season NEE arises from subnivean respiration. Furthermore, as more CO₂ is released from sites with greater snow accumulation, snow has a more important role in encouraging greater rates of respiration by insulating the soil from cold temperatures than in limiting diffusion of CO₂. Therefore, despite uncertainties regarding the exact portion of NEE that arises through biotic production and abiotic releases, it is clear that snow season production of CO₂ accounts for a substantial portion of snow season NEE (Elberling and Brandt, 2003; Panikov et al., 2006) and that subnivean NEE is sensitive to soil temperature and snow accumulation (Bokhorst et al., 2010).

3 Uncertainties in regional estimates of NEE

Recently, the North American Carbon Program conducted an intercomparison of 19 terrestrial biospheric models (Huntzinger et al., 2012), including Can-IBIS (Wang et al., 2011), CLM-CASA (Randerson et al., 2009), LPJ-wsl (Bondeau et al., 2007), ORCHIDEE (Krinner et al., 2005), SiB3 (Baker et al., 2008) and MODIS (Running et al., 2004). These models vary in terms of how photosynthesis is calculated, which assumptions are made, which driver data sets are used, as well as which processes and land surface characteristics are represented, and at what level of complexity, as described in detail by Huntzinger et al. (2012). Overall, model estimates of North American NEP ranged from –0.7 to 2.2 PgC/yr, indicating uncertainty regarding whether North America was a net carbon source or sink (Huntzinger et al., 2012). Estimates of NEE are complicated by the non-linearity of processes influencing CO₂ exchange at various scales (Levy et al., 1999), inherent challenges in optimizing numerous parameters (Prihodko et al., 2008), and innate challenges in complex representations of processes (Abramowitz et al., 2007).

Physiological models are used to model NEE over a variety of landscapes, including temperate rainforests (Coops et al., 2007), evergreen needleleaf forests (Xiao et al., 2004a) and temperate grasslands (Wu et al., 2008). However, a process-based approach can be more challenging to implement in Arctic and sub-Arctic regions due to challenges in representing processes that drive northern NEE such as fire regimes and soil characteristics, as well as challenges in accurately representing a highly heterogeneous landscape (La Puma et al., 2007). Model calibration and validation are further complicated in high-latitude regions due to the limited number of Arctic and sub-Arctic sites at which eddy flux observations are acquired (ORNL DAAC, 2011), and the clustering of northern field studies in Alaska and Scandinavia (Figure 2). Furthermore, as eddy covariance observations are collected preferentially during the growing season in high-latitude regions, it can be difficult to calibrate and evaluate the accuracy of snow season estimates.

In light of the many challenges faced in the process of collecting and analyzing observations of snow season NEE, and in generating regional estimates of Arctic NEE, we suggest that uncertainty in model estimates of snow season NEE may be reduced by incorporating objective, regional scale, remote sensing derived observations of snow properties. These snow season processes and the suggestions of how they may be represented in models of NEE also appear in Table 2.

Table 2.

Influences of fractional snow cover area (SCA), snow water equivalent (SWE), soil state as frozen/unfrozen, snow state as wet/dry, ice lenses and snow depth (SD) on NEE, and how remote sensing observations (R.S. Obs.) of the aforementioned snow characteristics can be incorporated into models of NEE in order to represent snow season processes, and thereby reduce uncertainty in estimates of snow season NEE. The Normalized Difference Vegetation Index (NDVI) is suggested as a manner to estimate vegetation biomass from visible and infrared remote sensing observations in order to represent snow-vegetation interactions, where appropriate. DGVM=dynamic global vegetation model).

Processes	Influences on NEE (relevant section)	R.S. Obs.	How to represent processes in models using R.S. Obs.
Soil cooling by snowfall	Diminished respiration and photosynthetic uptake (5.1).	SCA or soil freeze	Soil temperatures could be reduced upon initial appearance of snow or initial soil freezing.
Snow limits photosynthesis	Less photosynthetic uptake, especially by trees (4.2).	SCA	Photosynthesis could be reduced in accordance with the fractional snow cover observed.
Litter loss at start of snow season	Respiration in mid to late winter limited by low levels of soil nutrients (5.1).	SCA	Model soil nutrient levels could be initially elevated for the first ∼30 days following initial snowfall, then diminished for the rest of the snow season.
Snow limits soil heat loss	Subnivean respiration persists in soils, as these remain warmer than the air (4.1).	SWE or SD	Estimate soil respiration as a function of snow accumulation (SWE or SD) and air temperature.
Snowpack metamorphism	Metamorphism (5.2) and the formation of faceted crystals (4.2) alter subnivean temperature.	SWE and SD	Incorporate observations of snowpack properties (using SWE, SD, vegetation maps, and fieldwork) in estimates of soil respiration.
Formation of tree wells	Tree wells increase the rate of heat loss from soil, which can diminish rates of soil respiration (4.2).	NDVI or LiDAR	Soil respiration could be diminished in regions with dense forest cover (according to growing season NDVI) or where LiDAR observations indicate tree wells.
Extreme winter warming events	Warming first increases rates of soil respiration. Loss of snow thermal gradient. Resulting frost damage to vegetation can limit their photosynthetic capacity (5.2).	Snow state, SCA or NDVI	Mid-winter declines in fractional snow cover area can be indicative of winter warming events. Declines in photosynthetic capacity can be monitored with NDVI, and changes in NPP can be simulated in DGVMs.
Wind/ice layers form	Ice layers limit snow diffusivity, and allow CO₂ to be stored within the snowpack (5.2).	Ice lenses	Ice lenses can be observed using passive microwave observations.
Onset of snowmelt	Respiration increased by warmer temperatures and nutrient availability (5.3). Release of CO₂ from snowpack.	Snow wetness or SCA	Releases of CO₂ and increased rates of respiration following snowmelt can be simulated according to the timing of snow wetness and diminished snow cover.
End of snowmelt	Diminished snow cover and warmer air temperatures allow photosynthesis (4.3).	SCA	Photosynthetic uptake at the end of the snow season can be reduced according to fractional snow cover area.

IV Spatial associations between snow and NEE

Snow season NEE has been observed to vary according to the thickness of snow, soil temperature and vegetation composition at site H3 (Elberling, 2007). The strength of the associations between air temperature, soil temperature and NEE depends in part on the duration of snow cover and the physiological activity of vegetation during winter thaws (Starr and Oberbauer, 2003; Sullivan et al., 2008). Furthermore, the timing and magnitude of winter CO₂ transport depend on snow conditions, CO₂ production, site characteristics and weather conditions (Jones et al., 1999).

The acquisition of regional estimates of NEE is complicated by the scale dependence of NEE as well as the spatial variability of the various controlling influences on NEE, such as snow, soil, vegetation and microclimate status. Arctic transect studies have indicated that macro-scale (>10⁴ m) vegetation spatial patterns are influenced by latitudinal climatic gradients, topography, precipitation and active layer depth (Gould et al., 2003; McGuire et al., 2002). At the site scale (10–10⁴m), vegetation varies according to micro-topography and soil type, which both influence soil moisture (Svensson and Callaghan, 1988). Similarly, spatiotemporal variability of snow is scale dependent. At the micro (10–10² m) scale, transport and surface roughness influence distributions of snow. At the macro scale, variations in snow are controlled by elevation, orography and latitude, as well as distance from barriers and water; at the meso (10²–10³ m) scale, variations tend to be influenced by slope, elevation, aspect and vegetation (height, extent, density, etc.) (Bonan, 2002). Therefore, both vegetation and snow vary across the landscape and both are scale dependent. Likewise, a portion of the spatial variability in NEE is likely due to the influence of snow and vegetation.

Due to the spatial variability of the factors controlling snow season NEE, it is important that attempts to reduce uncertainty in sub-Arctic and Arctic NEE estimates during the snow season be founded upon the current state of knowledge regarding the influence of snow on the uptake and efflux of CO₂ at Arctic and sub-Arctic study sites. The following section therefore provides a synthesis of recent in situ findings regarding the site-scale associations between snow, vegetation and snow season NEE. Implications of these findings for representing the influence of snow on winter NEE are then discussed based on the physical processes driving these interactions (Table 2).

1 Greater efflux of CO₂ from sites with greater snow accumulation

Snow acts as an insulator because it has a small volumetric heat capacity, undergoes little heat loss by convection, and has a low thermal conductivity. The thermal conductivity of snow has been observed to range from 0.025 to 1.61 W/(m°C) in snowpacks with densities of 100–800 kg/m³ (Gray and Male, 1981). Generally, low values of thermal conductivity (e.g. 0.06 W/(m°C)) are observed in the Arctic and sub-Arctic (Sturm, 1992). The thermal conductivity of a snowpack is influenced by factors such as snowpack depth, bonding, temperature, porosity, ventilation, tortuosity and grain characteristics, but it can be can be estimated as a simple, positive function of snow density (Sturm et al., 1997). A greater accumulation of snow is generally associated with diminished thermal conductivity of the land-atmosphere interface (Gray and Male, 1981). Snowpacks can therefore decouple air and soil temperatures (Olsson et al., 2003) since less heat is transferred from the soil when an overlying layer of snow is present (Bonan, 2002). As a result, the frozen soil at 40–120 cm depth from which CO₂ is emitted has been observed to be 10–40°C warmer than the surface air temperature at a Siberian sub-Arctic site T6 (Zimov et al., 1993). For example, at two low Arctic sites located at L3, minimum 16 cm soil temperatures in the –10 to –5°C range were observed with minimum air temperatures of –30°C (Olsson et al., 2003).

Microbial fluxes have been observed to increase as a function of temperature over both organic and mineral soils at site L1 (Oelbermann et al., 2008). The association between soil respiration and temperature is driven by intracellular desiccation or extracellular barriers to diffusion in frozen soils, and by soil organic matter in thawed soils (Mikan et al., 2002). The influence of snow accumulation on CO₂ flux has been detected experimentally at a variety of sites using snow fences, which cause a local increase in snow accumulation through wind deposition. Across a sub-Arctic gradient in Sweden (S2), comparisons of treatment and control plots indicated that snow accumulation had an important influence on soil temperature. Between 41 and 75% of the variation in respiration could be explained as a function of soil temperature at site S2 (Grogan and Jonasson, 2006). Near the Arctic treeline in Alaska (S3), CO₂ flux varied significantly (p-value<0.01) between sites and years, and according to both snow depth and soil surface temperature. Soil surface temperature and snow depth were closely correlated (r²=0.73, p-value<0.01). Soil temperature and snow depth were greater in forested regions than at the treeline, and both increased in forests with greater stand density (Sullivan, 2010).

At a low Arctic site in northern Alaska (L2), warmer soil temperatures and greater rates of respiration were observed at plots that had received a greater accumulation of snow due to the installation of snow fences (Walker et al., 1999). At low Arctic site L1, observations of net seasonal CO₂ flux were collected using soda lime traps, which accumulate mass as CO₂ is released throughout the season. These measurements indicated a 60% increase in total snow season CO₂ efflux in snow accumulation plots (1 m of snow) relative to control plots (0.3 m of snow) (Nobrega and Grogan, 2007). At an Alaskan tussock tundra site (L2), snow accumulation acted as a stronger control on CO₂ distributions than air temperature (Sullivan et al., 2008). Experimental snow accumulation (30–150 cm) at a high Arctic site in Svalbard (H2) led to significant increases in both soil temperature and ecosystem respiration, leading to a doubling of CO₂ efflux at both heath and meadow sites (Morgner et al., 2010). Greater snow accumulation over several varied Arctic and sub-Arctic ecosystems therefore induced increased CO₂ efflux.

2 Influence of vegetation on snow and snow season NEE

Interactions between snow and vegetation have been observed to alter snow season photosynthesis (Larsen et al., 2007b) and respiration (Grogan and Jonasson, 2006) at site S2. In regions where tree wells form, the size of tree wells and the resulting soil heat loss depend on micro-scale (10–10² m) vegetation distributions (Sturm, 1992). The species distributions of vegetation at a site can therefore influence the soil temperature and respiration. At the meso scale (10²–10³ m), groupings of trees act as windbreaks, causing an increase in snow deposition nearby (Gray and Male, 1981), which could potentially increase the rate of soil respiration.

Vegetation species composition also has an influence on subnivean photosynthesis and respiration. Shrub stems encourage the growth of large faceted crystals via metamorphism under strong temperature gradients. Faceted crystals have one fifth to one twentieth of the effective thermal conductivity of a high-density wind slab (Zhang et al., 1996), resulting in the release of greater effluxes of CO₂ from areas containing shrubs, as observed at site S2 (Sullivan, 2010). Also, although cold temperatures and diminished light availability reduce photosynthetic assimilation of CO₂ (Billings and Mooney, 1968), the species present at a site influence the resulting rate of photosynthesis.

While in a dormant state, evergreen vegetation protects itself from light stress through non-photochemical dissipation of absorbed light (Öquist and Huner, 2003). Following frost hardening, evergreen trees have been observed to undergo no growth and ‘no measurable net photosynthesis’ as long as sub-zero air temperatures persist (Öquist and Huner, 2003). Although vascular plants cannot conduct photosynthesis when their tissues are frozen, evidence exists that lichens can perform photosynthesis at temperatures below –10ºC (Kappen, 1993). Furthermore, indications exist that low Arctic tundra vegetation, especially mosses and evergreens, may be capable of performing photosynthesis at greatly diminished rates under thin snow cover, as long as conditions are suitably warm (Tieszen, 1974).

Larsen et al. (2007b) reported that cold season (October to May) photosynthesis accounted for up to 19% of annual photosynthesis in a mesic, moss-dominated sub-Arctic heath with little (30–40 cm) snow (site S2). Similarly, Starr and Oberbauer (2003) found evidence of subnivean photosynthesis by Arctic evergreens on the north slope of Alaska (site L2) during a 2–4-week period at the end of the snow season in the presence of air temperatures above 0°C and encouraged by elevated subnivean concentrations of CO₂. At a High Arctic site in Zackenberg (H1), Christensen et al. (2012) observed low levels of photosynthesis to continue to occur at the start of the snow season. Vegetation therefore shows greatly reduced rates of photosynthesis in the presence of snow, even in Arctic species which are adapted to perform photosynthesis in sub-zero conditions.

3 Implications for models of NEE

Remote sensing observations of snow water equivalent (SWE) or snow depth could be used in combination with observations of air temperature to gain more accurate estimates of subnivean temperatures. The influence of greater snow accumulation on increased rates of soil respiration observed at many high, low and sub-Arctic sites (e.g. S2 – Grogan and Jonasson, 2006; S1 – Sullivan et al., 2010; L1 – Nobrega and Grogan, 2007; H2 – Morgner et al., 2010) could therefore be represented in models of NEE. This influence could be represented using remote sensing observations of SWE or snow depth to determine more accurate estimates of soil respiration, or by assimilating remote sensing observations of SWE into model estimates of snowpack accumulation to better constrain estimates of soil respiration.

The initial appearance of snow and concurrent drops in air and soil temperatures greatly limit the rate of photosynthetic uptake in Arctic vegetation (Billings and Mooney, 1968; Carstairs and Oechel, 1978; Öquist and Huner, 2003). The influence of snow on limiting the rate of snow season photosynthesis could be represented in models of NEE using observations of fractional snow cover area, and then diminishing the rate of photosynthesis in regions where snow is accumulating by an appropriate quantity given the vegetation distribution. The differing abilities of vascular and non-vascular plants to conduct photosynthesis in sub-zero conditions could therefore be represented. Greater benefits may be accrued by representing fractional snow cover in autumn than spring as photosynthesis is more heavily influenced by air temperature and photoperiod at the start of the snow season than at the end of the snow season, as described by Euskirchen et al. (2012).

The influence of vegetation on snowpack properties such as the formation of tree wells (Sturm, 1992) and faceted crystals in shrub-dominated regions (Zhang et al., 1996), and their influences on increasing and diminishing local snow thermal conductivities, respectively, could be represented using vegetation species-specific expressions to determine soil respiration as a function of snow accumulation. Another option may be to incorporate field observations of grain characteristics to represent the growth of faceted crystals, or high-resolution LiDAR observations to detect the formation of tree wells.

V Seasonal responses of NEE to snowfall, metamorphism and melt

Observations of seasonal changes in the relationships between snow cover/accumulation and CO₂ are summarized below according to three time periods: initial snowfall, midwinter and final snowmelt. The implications of seasonal transitions in snow-CO₂ associations for incorporating remote sensing estimates of snow into models NEE are then described. These associations and their implications are also summarized in Table 2.

1 Initial snowfall

The timing of initial snowfall represents a transition to the snow season, which is accompanied by greatly diminished rates of photosynthesis and cooler soil temperatures (Olsson et al., 2003). The initial accumulation of snow at the start of the snow season cools the ground surface due to the high emissivity and high surface albedo of snow, which results in diminished absorption of solar radiation (Zhang et al., 1996). Prior to initial snowfall, the soil active layer depth is at an annual maximum but starts to diminish as soil cooling occurs, thereby diminishing the rates of microbial activity (Elberling, 2007) due to indirect effects such as intercellular desiccation and extracellular barriers to diffusion (Mikan et al., 2002). Near the end of the growing season, plant productivity becomes limited by the availability of daylight. Initial snowfall in autumn limits light penetration to vegetation, thereby reducing photosynthetic uptake (Euskirchen et al., 2012).

Litter loss in autumn occurs mostly due to leaching of organic compounds (Bokhorst et al., 2010), and is influenced by both temperature and nutrient availability. Nutrient enrichment with a labile carbon source was found to quadruple September to November CO₂ effluxes at a Siberian sub-Arctic site (T5) (Zimov et al., 1996). Bokhorst et al. (2010) found that 90% of winter litter mass decomposition at sub-Arctic site S2 occurred within the first month of initial snowfall in autumn (Bokhorst et al., 2010). Therefore, the nutrient, soil temperature and snow conditions present during the short time period immediately following initial snowfall can have an important effect on net snow season NEE.

2 Snow metamorphism

Due to soil nutrient and temperature limitations, the rate of CO₂ production is greater during snowfall and snowmelt than during the intermediate portion of the snow season when metamorphism is the dominant process, as observed by Elberling (2007) at H3. However, the middle of the snow season has a long duration and therefore accounts for a large portion of the annual CO₂ budget. In the middle of the snow season, the snowpack is transformed over time through destructive and constructive metamorphism.

Destructive metamorphism refers to the process through which dendritic crystals are broken into rounded ice grains that become joined through sintering as snow crystals move to an equilibrium state, resulting in a minimum ratio of surface area to volume. Destructive metamorphism is controlled by vapor transfer within the accumulating pack, and is the dominant process following snowfall (Colbeck, 1980). Constructive metamorphism refers to the formation of temperature and vapor pressure gradients within snowpacks resulting in the diffusion of water vapor upwards, from warm to cold regions of the snowpack. Once the vapor has risen to a <0°C area of the snowpack, it refreezes in the form of faceted crystals. Cycles of freeze and melt within snowpacks cause the formation of large, irregular grains that grow upwards from the bottom of the snowpack (Gray and Male, 1981).

In terrestrial Arctic or sub-Arctic snowpacks unaffected by melt or strong wind activity, the top layer will remain less dense, weak and contain relatively unchanged snow crystals that have had their dendrites broken off. The middle layer will be dense and strong with small ice crystals due to destructive metamorphism, where air space and crystal size are diminished under densification processes. The bottom layer is generally weak, with low thermal conductivity. It will typically contain faceted crystals, or depth hoar (Gray and Male, 1981). The structure and low density of depth hoar cause it to have a very low thermal conductivity relative to other layers of snow (Zhang et al., 1996). Arctic and sub-Arctic snowpacks often contain wind slabs or ice lenses; when present, multiple layers of depth hoar can be formed (Sturm et al., 1995).

Steady state soil heat flux is influenced by the thermal conductivity, temperature and thickness of the snowpack (Gray and Male, 1981). In other words, the influence of mid-season snow on CO₂ flux depends on the thermal conductivity of snow (in W/(m°C)) as greater water availability with warmer soil temperatures leads to greater rates of CO₂ production (Panikov et al., 2006). In general, the thermal conductivity of a snowpack increases as the quantity of air relative to ice/water decreases. Packed, dense snowpacks therefore tend to transfer heat from soil at the greatest rate (Slaymaker and Kelly, 2007). Thermal conductivity (k_t) can be estimated according to snowpack density (ρ, in g/cm³) using a quadratic expression or logarithmic expression (Sturm et al., 1997). For example, Jeffries et al. (1999) estimated thermal conductivity of snow on the Alaskan North Slope as $k_{t} = 0.138 - 1.01 ρ + 3.233 ρ^{2}$ . The development of new sampling strategies has also allowed direct measurement of thermal conductivity, enabling analysis into the influence of snowpack stratigraphy on thermal conductivity (Sturm et al., 2002).

Surface melt, freezing rain and wet snow refreezing at night cause the formation of ice layers and crusts. Rain can fall in temperatures as low as –10°C when ice nucleation is not present, resulting in ice crusts (Gray and Male, 1981). When ice lenses and wind crusts form, the porosity of the snowpack decreases while the tortuosity increases, resulting in diminished permeability and air flow. Under these conditions, larger CO₂ concentration gradients form between the bottom of the snowpack and ambient air (Jones et al., 1999). Ice cover also limits exchange of gas with the atmosphere, but is not thought to limit seasonal total gas exchange (Gray and Male, 1981). Ice lenses within the snowpack can also influence the accuracy of diffusion-based measurements, since they defy the assumption of a linear CO₂ gradient within the snowpack (Sullivan, 2010). The use of snow pits to examine the presence of ice lenses can, therefore, improve the reliability of measurements (Sullivan, 2010). Erosion and windpacking in tundra snowpacks cause slabs to form, and blizzards cause massive hard drifts.

Air temperature also influences snow-CO₂ relationships, even though this influence is limited by the insulating properties of the snowpack. At a low Arctic tundra site (L1), a cold front was observed to cause a temporary drop in soil temperature at 2 cm below the surface and produced diminished fluxes of CO₂ (Buckeridge et al., 2010). Extreme winter warming events initially encourage microbial activity and soil respiration. Following warming, however, soil refreezes more deeply as the insulation provided by snow is lost (Bokhorst et al., 2010), which can cause damage to vegetation that will substantially diminish net primary productivity over the following growing season (Bokhorst et al., 2009). Meteorological influences during the snow season can therefore alter snowpack properties as well as NEE.

3 End of season snowmelt

Snowmelt occurs due to a combination of rainfall, absorption of solar radiation and sensible heat exchange from the air to ground (Gray and Male, 1981). The process of snowmelt is accelerated by albedo effects as the snowpack water content increases, snow depth decreases and low albedo vegetation is uncovered (Bonan, 2002). Due to the higher thermal conductivity of water relative to air, wet snow cannot maintain temperature gradients or insulate soil as well as dry snow (Gray and Male, 1981). The high thermal conductivity of the snowpack at the end of the snow season therefore limits the influence of snow on soil temperature (Zimov et al., 1996). As a result, changes in snowpack stratigraphy in a wet snowpack are therefore likely to have little influence on CO₂ production during snowmelt. However, in regions characterized by low Arctic vegetation, the diffusivity of the snowpack may continue to influence the rate of soil CO₂ release during snowmelt (S2 – Björkman et al., 2010) and warmer temperatures can increase the rate of soil respiration at the end of the snow season (T5 – Zimov et al., 1996).

Effluxes of CO₂ at the end of the snow season occur due to an increase in CO₂ production as nutrients and moisture are released through soil thaw, and due to snowmelt, which allows CO₂ trapped beneath the snowpack to be released (Elberling et al., 2008). An increase in snow accumulation has been associated with diminished growing season photosynthesis due to delayed onset of the growing season at site H2 (Morgner et al., 2010). Thicker snow cover and later timing of snowmelt have been found to increase the magnitude and delay the timing of CO₂ released in late winter and early spring in mesic low Arctic tundra at site L1 (Buckeridge and Grogan, 2010). Conversely, in upland tundra and sedge fen locations at site L1, variations in the timing of snowmelt did not substantially alter early season or total NEP (Humphreys and Lafleur, 2011). Springtime freeze-thaw cycles have likewise been observed not to significantly influence total springtime CO₂ efflux rates in a sub-Arctic heath (S2) (Grogan et al., 2004). Springtime variations in air temperature at a low Arctic mesic birch hummock site (L1) did not bring about freeze-thaw cycles in soil, and also did not cause pulses of CO₂ (Buckeridge et al., 2010). The timing of snowmelt onset and end therefore appears to have a more important influence on NEE than concurrent fluctuations in air temperature.

4 Implications for modeling snow season NEE

Following the first snowfall of the snow season, soil cooling limits the rate of respiration, and the rate of photosynthesis declines due to cold air temperatures and diminished light availability under snow (Olsson et al., 2003). The timing of these transitions could be represented in models of NEE using remote sensing observations of fractional snow cover (Table 2). Similarly, as nutrient availability declines following initial snowfall (Bokhorst et al., 2010), the limiting influence of nutrients following litter loss could be described according to a time period following initial snowfall.

In midwinter, snow metamorphic processes act as a dominant influence on the thermal conductivity of the snowpack (Gray and Male, 1981). However, little NEE occurs during the middle of the snow season (Elberling, 2007). Midwinter fluxes of CO₂ have been observed to vary little between sites (Zimov et al., 1993) and are minimal due to nutrient limitation (Mikan et al., 2002). Existing process-based model and inversion approaches simulate diminished midwinter NEE, even when the influence of snow on NEE is not explicitly represented (McGuire et al., 2012). Models of NEE are therefore more likely to benefit from improved characterizations of early and late snow season NEE than of midwinter NEE, especially due to the lack of operational remote sensing estimates of snow thermal conductivity or microphysical structure (Table 3), and existing challenges in accurately modeling snowpack thermal conductivity and density, as described by Saito et al. (2012). However, if snowpack metamorphism was to be included in models of NEE, it may be best to use a process-based approach to simulating snowpack dynamics, and assimilating remote sensing observations of SWE to ensure accuracy in snowpack accumulation.

Table 3.

Summary of which snow characteristics can be acquired using each remote sensing approach (Jensen, 2007). Products which make use of two or more categories of remote sensing observations include IMS (snow cover extent), ANSA (fractional snow cover, SWE & snowmelt) and SMAP (soil freeze/thaw). The main advantages and limitations of each approach are briefly described in terms of their potential for being applied in models of NEE.

Image mode	Active/ passive	Resolution	Wavelength	Satellites & products	Snow/soil characteristics	Advantages	Limitations
Imaging	passive	30m - 1 km	Visible (0.4-0.7 µm) and infrared (0.7-15 µm)	Satellites: AVHRR MODIS SPOT Product: MOD10	Fractional snow cover area, or the timing of snow on/off at the start and end of the snow season	Moderate spatial and temporal (daily-weekly) resolution estimates available of fractional snow cover.	Arctic observations can only be collected during the start and end of the snow season due to polar darkness. Observations of fractional snow cover in spring and fall can be limited by cloud cover, which is especially prevalent at high latitude sites.
	passive	5 - 60 km	Passive microwave (1-200 GHz or 0.15-30cm)	Satellites: AMSR-E SSM/I Product: GlobSnow SWE	Snow water equivalent (SWE), snow state (wet/dry) and soil freeze/thaw state	Daily observations of SWE are available across the pan-Arctic, and are unhindered by cloud cover or polar darkness.	Coarse spatial resolution (10s of km) inadequate for local scale applications. Accurate estimates of SWE are complicated by wet snow, deep snowpacks and vegetation.
	active	10m -1 km	Active microwave (L, C, X & Ku bands, or 1-2, 4-8, 8-12.5 and 12.5-18 GHz)	Satellites: QuickSCAT ENVISAT/ASAR	Snowmelt onset timing and soil freeze-thaw status	Daily to weekly coverage. Ku scatterometers can estimate regional SWE and snowmelt onset at a coarse resolution. Active and passive microwave observations can be combined to generate estimates of soil freeze/thaw, which affects NEE.	Existing SAR systems (C- and X-band) are not generally applicable for SWE estimation.
		<30 cm	Airborne LiDAR (1.064 & 0.532 µm)	Satellites: ICESat/GLAS	Snow depth	Very fine spatial resolution and high spatial sampling frequency can characterize snow and land microtopography.	Prohibitively expensive. Observations are therefore limited to individual study sites, with few observations per snow season.
Non-imaging		70 m (laser) – 10 km (radar)	Satellite altimetry (laser: 1.064 & 0.532 µm, radar: Ku (12-18GHz) & Ka (26.5-40 GHz) bands)	Satellites: Topex-Poseidon	Snow depth	Moderate to low spatial and temporal (weekly) resolution. Field studies have described the influence of snow depth on NEE.	Sampling along nadir tracks only. Satellite altimetry is typically used for oceans and ice sheet science applications, rather than terrestrial operational applications.

Extreme winter warming events (Bokhorst et al., 2010) could be identified by tracking fractional snow cover and snow wetness throughout the cold season. At sites where extreme winter warming events occur, the resulting damage to the photosynthetic capacity of vegetation could be examined using satellite observations of mid-growing season vegetation health via the Normalized Difference Vegetation Index (NDVI). In dynamic global vegetation models (DGVMs), the influence of extreme winter warming events on hindering NPP could therefore be simulated using remote sensing observations of both the snow and growing seasons. Representing these processes in models could allow insights into the complex responses of the carbon cycle to warmer snow and growing season air temperatures.

During snowmelt, CO₂ stored within the snowpack is released, and CO₂ production also increases due to warmer temperatures and the release of nutrients as soil thaws (Buckeridge and Grogan, 2010; Elberling et al., 2008). Delayed timing of snowmelt can also lead to a later start of the growing season at certain sites (Morgner et al., 2010). Incorporating remote sensing estimates of fractional snow cover at the end of the snow season, or the snow state as wet or dry, may therefore prove beneficial for representing snow season processes that drive NEE (Table 2). The following section provides a summary of recent approaches used to collect remote sensing observations of snow, and provides an assessment of which strategies may be most useful for reducing uncertainty in snow season estimates of Arctic and sub-Arctic NEE.

VI Remote sensing of influences on snow season NEE

Ideally, remote sensing observations incorporated into models of NEE would be at an appropriate resolution, with reasonable accuracy, and would have the ability to characterize aspects of the land surface with the greatest influence on NEE, as described in Table 2. The advantages and limitations of visible/infrared, altimeter, passive microwave and active microwave approaches for observing snow characteristics are evaluated below in light of their potential contribution to models of NEE. A brief summary of these remote sensing products and their relevance to models of NEE can be found in Table 3.

1 Visible and infrared observations of snow cover area

Fractional snow cover area can be estimated from visible and infrared remote sensing observations using the Normalized Difference Snow Index (NDSI) to differentiate snow from clouds. Although both clouds and snow strongly reflect visible radiation, snow has a much lower reflectance in the mid-infrared range than clouds (Crane and Anderson, 1984). On this basis, estimates of snow cover area from Earth Observing System (EOS) Moderate Resolution Imaging Spectroradiometer (MODIS) observations can be generated using the ratio of green (r_GREEN, 0.545–0.565 μm) to mid-infrared (r_MIDIR, 1.628–1.652 μm) reflectance (Hall et al., 1995):

N D S I = \frac{r_{G R E E N} - r_{M I D I R}}{r_{G R E E N} + r_{M I D I R}}

In regions where low illumination, surface water or vegetation complicate retrievals, thresholds can be set to further differentiate snow-covered surfaces (Hall and Riggs, 2007; Riggs and Hall, 2004). For example, snow and water can have similar values of NDSI but water has a lower r_NIR (0.841–0.876 μm) reflectance than snow. Therefore, non-forested regions with NDSI ≥ 0.4 and where r_GREEN ≥ 10% and r_NIR ≥ 10% can be designated as snow-covered (Foster et al., 2011). Similarly, in regions with dense forest cover, observations of NDVI can be used to set lower NDSI thresholds in order to detect subcanopy snow cover area (Hall et al., 1998).

MOD10 daily, eight-day and monthly composites of snow cover are generated at 500 m to 0.25° grid resolutions using MODIS surface reflectance and masks for cloud cover and land/water (Hall and Riggs, 2007). Surface temperature is also incorporated in order to mask out regions with temperatures of >283°K (Riggs and Hall, 2011). Validation of MOD10 snow cover has shown an accuracy of ∼93% relative to in situ observations and other operational snow cover products, with the main source of error arising from snow/cloud discrimination (Hall and Riggs, 2007). The NDSI has also been used to estimate fractional snow cover from a variety of other platforms ranging from 1 to 25 km in resolution (Xiao et al., 2004b). For example, the NSDI in SPOT VEGETATION images (1 km) has shown good agreement with ground observations of snow cover (Dankers and De Jong, 2004).

A major limitation of visible and infrared observations is that they cannot be used to estimate the quantity of snow because of the short penetration depth of light due to scattering (Dozier and Painter, 2004). Furthermore, the acquisition of visible and infrared observations is limited at high latitudes due to cloud cover and polar darkness. The most suitable applications for optical remotely sensed images of snow are therefore in characterizing the snow cover area at the start and end of the snow season (e.g. Paudel and Andersen, 2011; Vikhamar and Solberg, 2003).

Year-round estimates of snow cover extent are therefore often generated using a combination of visible and infrared derived observations with remote sensing observations from a variety of other sensors. For example, the US National Oceanic and Atmospheric Administration (NOAA) has been generating weekly estimates of Northern Hemisphere snow cover extent since 1966, and have been producing interactive multisensor snow and ice mapping system (IMS) estimates of snow and ice cover since 1999 at 24 km from a variety of inputs (Ramsay, 1998). Since 2004, IMS estimates of snow cover extent have additionally been generated at a daily, 4 km resolution (Helfrich et al., 2007) using visible and infrared observations from MODIS and the Advanced Very High Resolution Radiometer (AVHRR) in additional to several non-optical sources of imagery.

The main difference between the IMS and MOD10 products is that MOD10 is produced in a fully automated manner, whereas IMS requires human intervention (Frei and Lee, 2010). Relative to higher-resolution optical data and passive microwave observations of snow, NOAA snow charts have had a late season bias by up to four weeks (Wang et al., 2005). Conversely, the MODIS snow cover area product tends not to identify snow in autumn over Eurasian deciduous needleleaf forest when this is observed by IMS (Frei et al., 2011). Since autumn snow cover influences the photoperiod and photosynthetic uptake by vegetation more than spring snow cover, it may be best to use IMS products for modeling NEE over Eurasia. However, if the influence of fractional snow cover on NEE is represented in either process-based or remote sensing based models of NEE, and estimates of pan-Arctic fractional snow cover are required after 2000, it may be beneficial to use the MODIS snow cover area product.

2 Altimeter observations of snow depth

Light detection and ranging (LiDAR), or laser altimetry, measures the time between emission and reception of a laser pulse in order to map the three-dimensional aspects of topography. LiDAR has been used for a variety of applications ranging from biophysical canopy mapping (Lefsky et al., 2002) to sea level height (Lemoine et al., 2010). Estimates of snow depth can be made by subtracting measurements of the topography at a time when snow is not present from measurements of topography during the snow season (Deems et al., 2006; Fassnacht and Deems, 2006; Hopkinson et al., 2004).

However, LiDAR based techniques for accurately estimating snow depth (<10 cm) may also be compromised by snowfall, fog or large grain size within the snowpack (Prokop, 2008), and the acquisition of high-resolution LiDAR observations by aircraft is expensive. An alternative may therefore be to use observations from NASA’s Geoscience Laser Altimeter System (GLAS) carried on the Ice, Cloud and land Elevation Satellite (ICESat). ICESat/GLAS collects topographic measurements every eight days over regions with a 70 m diameter every 175 m, and these observations have previously been used to generate daily estimates of snow depth over sea ice (Kwok and Cunningham, 2008), Arctic land cover (Ranson et al., 2004) and topography of land (Atwood et al., 2007). One of the stated potential applications of ICESat was to estimate terrestrial snow depth (Zwally et al., 2002); however, no studies to date have applied ICESat for characterizing terrestrial snow depth. Radar altimeter measurements from Topex-Poseidon have been used to characterize terrestrial snow depth (Papa et al., 2002). Terrestrial snow depth could therefore potentially be characterized using ICESat/GLAS or Topex-Poseidon altimeter observations, and to apply these estimates of snow depth to reduce uncertainty in models of NEE. The low spatial resolution of satellite altimeter observations or the high cost of airborne observations may limit the utility of LiDAR estimates in models of NEE. With the development of ICESat-2, this exploration may be possible.

Although field studies focusing on snow-NEE associations have tended to quantify snow in terms of its depth (e.g. Grogan and Jonasson, 2006; Sullivan, 2010), the regional scale thermodynamic properties of snow cannot be well estimated as a direct function of snowpack thickness alone (Slaymaker and Kelly, 2007). Altimeter observations are likely to be most useful in situations where estimates of NEE are required for a small homogeneous region where the site-scale association between snow depth and NEE has already been characterized from in situ observations. Airborne LiDAR estimates of local snowpack depth and vegetation may also be of use in characterizing local snow microtopography and features such as tree wells, which can alter soil temperature and respiration.

3 Passive microwave observations of SWE

Passive microwave sensors measure the Earth’s brightness temperature (T_b) at frequencies ranging from 1 to 18 GHz. T_b represents the amount of radiation emitted by an object at a given wavelength, as expressed by the radiation emitted by a hypothetical blackbody at the same physical temperature (T). Since real objects emit less energy radiation than perfect (blackbody) emitters, brightness temperature at a given wavelength is a function of the physical temperature and emissivity (∊<1) of a material (Chang et al., 1976):

T_{b} = ϵ T

Passive microwave observation can be used to estimate the timing of freeze/thaw cycles (Smith et al., 2004; Zhang and Armstrong, 2001) and to detect ice lenses (Rees et al., 2010). The state of the soil as frozen or thawed influences soil respiration, and ice lenses alter the diffusion of CO₂ through the snowpack (Elberling et al., 2007; Sullivan, 2010). The quantity of snow accumulation in terms of its snow water equivalent (SWE) can also be estimated from passive microwave observations. SWE is more useful than snow cover area for many hydrological and climatological applications as it relates directly to the amount of water accumulated in the snow cover, rather than the simple presence of snow (Foster et al., 2005).

SWE can be estimated from passive microwave observations using site-specific regression values or inversion modeling approaches. Typically, SWE is estimated according to the amount of volume scatter of microwave radiation through the snowpack, which increases with the depth and density of the snowpack (Ulaby and Stiles, 1980). Estimates of SWE are usually generated according to the difference between brightness temperature at two frequencies, such as 19 and 37 GHz observations from AMSR-E (e.g. Pulliainen and Hallikainen, 2001), since 37 GHz observations display greater scatter than 19 GHz observations with greater snow accumulation. However, several aspects of the Arctic and sub-Arctic environment such as microwave emission from dense snowpacks and high lake fractional coverage necessitate the use of tundra-specific algorithms for SWE using observations at 37 GHz (Derksen et al., 2010). Currently, a leading product containing estimates of SWE across the Northern Hemisphere is GlobSnow, which shows strong agreement against ground-based measurements (Takala et. al., 2011).

A drawback of using passive microwave observations of SWE in models of NEE is that estimates of the SWE from deep (>0.5 m) snowpacks may not be accurate since the scattering signal at 37 GHz loses sensitivity at these snow depths (Shi, 2008). Another drawback is that SWE can only be estimated from dry snowpacks using passive microwave observations. Wet snow mainly emits passive microwave radiation, and the influence of volume scattering on the passive microwave signal is therefore difficult to discern (Stiles and Ulaby, 1980). However, this may not be a great concern in representing the influence of snow accumulation on NEE since only dry snowpacks maintain air-soil temperature differentials that affect the production of CO₂ (Zimov et al., 1996). The main drawback of incorporating passive microwave observations into models of NEE is that the resolution at which observations are collected (25 km) may not be fine enough for characterizing variability in cryospheric influences on NEE over heterogeneous Arctic landscapes. The main advantage of passive microwave observations is that they can be collected throughout the snow season over high latitudes, and are not hindered by polar darkness or non-precipitating cloud cover (Foster et al., 2011). Furthermore, the thermal conductivity of the snowpack, which influences soil respiration, is determined through a combination of snowpack density, tortuosity, grain characteristics and accumulation (Gray and Male, 1981). Although thermal conductivity is not a direct function of snow water equivalent (SWE), strategies do exist for estimating snow temperature using passive microwave derived observations of SWE, when combined with a land surface model and atmospheric forcing (e.g. Sun et al., 2004). The influence of snow accumulation on soil respiration could therefore be estimated by including observations of SWE. Furthermore, in regions where moisture resulting from snowmelt influences soil moisture, estimates of water availability upon snowmelt could be improved by incorporating estimates of SWE at the end of the snow season into hydrological estimates by process-based models.

4 Active microwave observations of snow characteristics

Active microwave observations of backscatter can be used to estimate snow wetness, thermal resistance and snow water equivalent (SWE) (Rott et al., 2010; Shi, 2008). The timing of snowmelt can also be estimated from active microwave observations (Royer et al., 2010; Wang et al., 2008) since backscatter diminishes as the snowpack liquid moisture increases (Chang et al., 1985). Presently, active microwave observations are being collected at the C-band, and these observations can be used to characterize snow liquid water content (Niang et al., 2007).

Measuring the amplitude and phase of polarization state and interferometry from repeat pass observations could also be used to characterize other snow characteristics such as grain size, snow depth, structure and density (Shi, 2008). However, to date, current approaches to estimate SWE and snow characteristics from existing space-borne synthetic aperture radar (SAR) observations have been complicated by the dependence of these algorithms on the study site location and the year. The active microwave signal is influenced by a number of different characteristics of the snowpack that alter the snowpack geometry, composition and volume. Efforts to estimate specific characteristics of the snowpack are therefore complicated by the need to decompose this signal. Ground-based and airborne active microwave observations are currently being collected to assess the potential for snowpack properties to be estimated from twin frequency observations at 9.6 and 17.2 GHz (Rott et al., 2010).

Several products are being developed to make use of active microwave observations in combination with other sensors. For example, daily estimates of the state of the land surface as frozen or thawed will be available at a 3 km resolution from the Soil Moisture Active-Passive (SMAP) mission (Entekhabi et al., 2008; Kim et al., 2010). Similarly, the Air Force Weather Agency/NASA Snow Algorithm (ANSA) combines MODIS visible and infrared observations with active and passive microwave observations in order to generate estimates of SWE, fractional snow cover, snowmelt onset and regions of active melt (Foster et al., 2011). Unfortunately, the active microwave observations were acquired by NASA’s QuickSCAT, which was in operation from 1999 to 2009. Presently, no established products of SWE or snowmelt onset/end exist that use active microwave observations alone. In future, it may be advantageous to use a combined product in models of NEE if estimates of a variety of snow season characteristics are desired.

5 Suitability of different remote sensing approaches for models of NEE

The suitability of various remote sensing approaches in estimating snowpack characteristics that alter snow season NEE are summarized in Table 3. Visible and infrared remote sensing data are most useful in situations where high-resolution estimates (weekly, >30 m) of snow cover area are desired at the start or end of the snow season. Laser altimeter estimates of snow depth may be useful where the association between snowpack depth and NEE is known, and where snow cover is adequately consistent and homogeneous so that point observations could adequately characterize regional snow depth. High-resolution observations (<20 cm) by airborne LiDAR could likewise assist in characterizing the snowpack depth, but are unlikely to be collected routinely or regionally for scientific studies due to the high financial cost. Although a majority of field studies have investigated the influence of snow accumulation on NEE in terms of snow depth, the stratigraphy, SWE and snow grain characteristics are more important for the thermal conductivity and gas diffusivity of a snowpack. Therefore, estimates of SWE are likely to be of greater use in models of NEE than altimeter observations of snow depth, as SWE has a more important influence on soil respiration and soil moisture availability than snow depth.

The inclusion of passive microwave observations in models of NEE would be most likely to be beneficial in regions that are sufficiently homogeneous so as to not be hindered by the coarse (≈25 km) resolution of these products. Presently, active microwave observations can be used to assess snowmelt timing, but existing systems cannot be used to generate regional estimates of SWE or other snow characteristics due to challenges in decomposing the active microwave signal, as the relationships between SWE, snowpack stratigraphy and the active microwave signal vary across study sites and years. Established products exist for passive microwave observations of SWE from GlobSnow, and visible and infrared observations of fractional snow cover area from MODIS 10. Synergistic products indicating soil freeze/thaw such as SMAP and ANSA estimates of SWE and fractional snow cover which are currently in development may prove helpful in future studies involving NEE.

6 Incorporating remote sensing of snow into models of NEE

Remote sensing observations could be incorporated into models of Arctic and sub-Arctic NEE in several ways. First, these observations could be brought in as model inputs in remote sensing or process-based models of NEE in order to represent snowpack properties relevant to NEE (Table 2). Another option is that model calibration or evaluation could consist of comparing remote sensing observations of factors such as the dates of initial snowfall and snowmelt against estimates of NEE for portions of the year when observations of NEE are unavailable in order to better understand model performance. This approach could be used both for models that explicitly represent snow processes, and for those that do not. A final option is for models of NEE to explicitly represent physical processes during the snow season, and to assimilate remote sensing observations of snowpack properties into model representations of snowpacks in order to better constrain estimates of NEE. The optimal strategy depends largely on the model class (prognostic or diagnostic), its formulation (remote sensing or process-based), and the intended purpose of the model.

Incorporating remote sensing observations directly into models of NEE is advantageous when model output is required over large regions or many years, as a comparison of 33 snow models by Rutter et al. (2009) indicated that model performance in estimates of SWE and snow depth varied by site and year. It may therefore be more difficult to examine the sources of uncertainty in models of NEE if these models make use of process-based representations of snowpack properties which are unconstrained by in situ or remote sensing observations.

Certainly, representing the influence of fractional snow cover area on photosynthesis and respiration could be conducted by directly incorporating remote sensing based estimates of snow cover area (Tables 2 and 3). Due to the scale dependence of NEE, snow and vegetation, it is important that model representations of interactions between snow, vegetation and NEE consider whether discrepancies exist in the spatial scales at which these interactions are observed and modeled, and how uncertainties due to this effect can be mitigated. Remote sensing observations are acquired at a coarse resolution that is more similar to typical model resolutions than the scale at which field observations are acquired, and this may prove beneficial in characterizing snow at an appropriate resolution.

The combination of a process-based model of snow and NEE, and remote sensing observations of snow, would be optimal for representing the response of NEE to changes in snowpack characteristics (e.g. timing of snow on/off, ice layers, metamorphism, changes due to warming events). A process-based approach may also be beneficial in estimating light penetration and its influence on subnivean photosynthesis, and in representing the influence of ice lenses on altering the portion of CO₂ which results from biotic and abiotic releases. The effects of snow metamorphism, snow cover and snow-vegetation interactions on albedo could likewise be simulated using a process-based approach. Representing these snow season processes in models of NEE, as well as interactions between vegetation, meteorological conditions, permafrost, soil carbon content, snow albedo, soil respiration, and photosynthesis, could contribute to an improved understanding of the drivers of Arctic and sub-Arctic NEE.

VII Conclusions

In situ studies have indicated that a substantial portion of NEE in the Arctic and sub-Arctic occurs during the snow season (Aurela et al., 2004; Sullivan et al., 2008; Zimov et al., 1996), and that the spatiotemporal dynamics of NEE are influenced by the timing of initial snowfall in autumn (Euskirchen et al., 2012), the timing of final snowmelt in spring (Buckeridge and Grogan, 2010; Morgner et al., 2010), and the quantity of snow accumulated (Nobrega and Grogan, 2007; Sullivan et al., 2008). The main contribution of this paper has been to provide a summary of the physical processes driving NEE, and in providing strategies for representing these processes in models of NEE (Table 2) and selecting an optimal remote sensing approach for characterizing snow season properties (Table 3). Incorporating remote sensing observations of snow season properties of the land surface may assist both in reducing uncertainty in model estimates of Arctic and sub-Arctic NEE and in providing insights into the complex physical interactions that occur between snow and NEE throughout the snow season.

Understanding the interactions between snow, vegetation and the northern carbon cycle is especially important due to the complex and interconnected reactions of biological, cryospheric, hydrological and atmospheric systems to climate change and rising levels of atmospheric CO₂ (Figure 1). Uncertainty regarding the response of Arctic systems to predicted changes in climate remains ‘one of the most critical issues in Arctic [carbon] balance research today’ (Lafleur and Humphreys, 2008). Climate change is predicted to cause a rise in temperature and an increase in precipitation (Christensen et al., 2007), which is predicted to result in diminished snow cover extent and greater snow accumulation across the Arctic (AMAP, 2011). Deeper snowpacks could accelerate the rate of winter respiration, but a longer growing season may result in greater photosynthetic uptake by vegetation. Changes in vegetation composition are likely to increase the uptake of carbon in short-lived tissues as temperature constraints on productivity are relaxed (Schlesinger and Lichter, 2001), but species change may also lead to larger effluxes of CO₂ at Arctic sites since greater shrub prevalence is associated with diminished thermal conductivity of snow (Sturm et al., 2001). The net influence of climate change on the northern carbon balance therefore depends on the relative magnitude of changes in CO₂ uptake and efflux by soil microbial communities resulting from changes in snow, temperature and vegetation (Figure 1). Model estimates of NEE are thus important for quantifying and understanding the net effect of these cumulative changes.

Predictions for the year 2050 indicate that climate change may bring about an increase in maximum snow depth and a 20% decline in average pan-Arctic snow cover duration, which could increase the rate of both subnivean respiration and photosynthetic uptake in spring and fall (AMAP, 2011). The timing of initial snowfall alters the photosynthetic uptake of carbon by vegetation at the end of the growing season (Euskirchen et al., 2012) and changes in the timing of snowmelt can alter the carbon balance of a site (Morgner et al., 2010). Furthermore, permafrost thaw may be accelerated by increased accumulation of snow or warming temperatures, which can result in effluxes of CO₂ (Nowinski et al., 2010; Schuur et al., 2009). Over the next century, thawing permafrost is predicted to release 16–20% of soil organic carbon, resulting in large effluxes of CO₂ and CH₄ through aerobic and anaerobic decomposition (Zhang et al., 2008). An increase in snow accumulation would therefore be likely to bring about greater effluxes of CO₂ through both more extensive permafrost thaw and greater rates of soil respiration. Including observations of fractional snow cover at the start and end of season, as well as snow accumulation, could therefore assist in determining how climate-driven shifts in precipitation, temperature, snow accumulation and snow duration alter the timing and magnitude of the northern carbon budget.

Extreme winter warming events have already begun to be observed regionally in the Mackenzie Basin (Cao et al., 2007), and these events are predicted to increase in prevalence over time due to climate change (AMAP, 2011), which may result in damaged vegetation and diminished photosynthetic uptake during the following growing seasons (Bokhorst et al., 2009). Conversely, diminished snow duration has been observed to increase the uptake of CO₂ at the start and end of the growing season as well as the rate of respiration at an Alaskan tundra site (Oberbauer et al., 1998). Identifying midwinter melt events, and quantifying the subsequent change in soil temperature, could therefore prove crucial in understanding the response of the carbon cycle to warming winter temperatures.

Previously, studies have assessed the response of the biosphere-atmosphere system to changes in snow cover using a combination of longitudinal field studies (e.g. Aurela et al., 2004) and biosphere-atmosphere transfer models (e.g. Yang et al., 1997). Process-based ecosystem models have been applied in order to estimate past (McGuire et al., 2001) and future (Cramer et al., 2001) changes in the global carbon balance resulting from climate change and shifts in atmospheric levels of CO₂. As climate change continues to alter northern ecosystems and the relationships between cryospheric and biospheric components of the Earth system, the ability to accurately quantify the Arctic carbon budget and its response to climate change will be crucial. Including remote sensing observations of the cryosphere in models of NEE will allow these models to simulate snow season NEE, and its response to regional cryospheric changes.

Footnotes

Funding

Research funding from the Vanier Canada Graduate Scholarship program (K. Luus) is gratefully acknowledged.

References

Abramowitz

Pitman

Gupta

. (2007) Systematic bias in land surface models. Journal of Hydrometeorology 8(5): 989–1001.

Amiro

(2010) Estimating annual carbon dioxide eddy fluxes using open-path analyzers for cold forest sites. Agricultural and Forest Meteorology 150: 1366–1372.

Ananyan

(1970) Unfrozen water content in frozen clay at a temperature from –0.6°C to –40°C – –60°C. Merzlotnye Issledovaniya 10: 267–270 (in Russian).

Arctic Monitoring and Assessment Program (AMAP) (2011) SWIPA 2011 executive summary: Snow, water ice and permafrost in the Arctic. Available at: http://amap.no/swipa/SWIPA2011ExecutiveSummaryV2.pdf.

Atwood

Guritz

Muskett

. (2007) DEM control in Arctic Alaska with ICESat laser altimetry. IEEE Transactions on Geoscience and Remote Sensing 45(11): 3710–3720.

Aurela

Laurila

Tuovinen

J-P

(2004) The timing of snow melt controls the annual CO₂ balance in a subarctic fen. Geophysical Research Letters 31: L16119.

Bäckstrand

Crill

Jackowicz-Korczyñski

. (2010) Annual carbon gas budget for a subarctic peatland, Northern Sweden. Biogeosciences 7(1): 95–108.

Baker

(2005) SiB3 prognostic variables and equation set. Available at: http://biodav.atmos.colostate.edu/kraus/Papers/SiB/SiB3_eqns_master.pdf.

Baker

Prihodko

Denning

. (2008) Seasonal drought stress in the Amazon: Reconciling models and observations. Journal of Geophysical Research – Biogeosciences 113: G00B01.

10.

Baldocchi

(2008) ‘Breathing’ of the terrestrial biosphere: Lessons learned from a global network of carbon dioxide flux measurements. Australian Journal of Botany 56(1): 1–26.

11.

Baldocchi

Hicks

Meyers

(1988) Measuring biosphere-atmosphere exchanges of biologically related gases with micrometeorological methods. Ecology 69: 1331–1340.

12.

Billings

Mooney

(1968) The ecology of arctic and alpine plants. Biological Reviews 43(4): 481–529.

13.

Björkman

Morgner

Cooper

. (2010) Winter carbon dioxide effluxes from Arctic ecosystems: An overview and comparison of methodologies. Global Biogeochemical Cycles 24: GB3010.

14.

Boesch

Baker

Connor

. (2011) Global characterization of CO₂ column retrievals from shortwave-infrared satellite observations of the Orbiting Carbon Observatory-2 mission. Remote Sensing 3: 270–304.

15.

Bokhorst

Bjerke

Bowles

. (2010) Impacts of extreme winter warming events on litter decomposition in a sub-Arctic heathland. Soil Biology and Biochemistry 42(4): 611–617.

16.

Bokhorst

Bjerke

Tømmervik

. (2009) Winter warming events damage sub-Arctic vegetation: Consistent evidence from an experimental manipulation and a natural event. Journal of Ecology 97(6): 1408–1415.

17.

Bonan

(2002) Ecological Climatology: Concepts and Applications. New York: Cambridge University Press.

18.

Bondeau

Smith

Zaehle

. (2007) Modelling the role of agriculture for the 20th century global terrestrial carbon balance. Global Change Biology 13(3): 679–706.

19.

Bréon

Ciais

(2010) Spaceborne remote sensing of greenhouse gas concentrations. Comptes Rendus – Geoscience 342 (4–5): 412–424.

20.

Buckeridge

Grogan

(2010) Deepened snow increases late thaw biogeochemical pulses in mesic low Arctic tundra. Biogeochemistry 101 (1–3): 105–121.

21.

Buckeridge

Cen

Layzell

. (2010) Soil biogeochemistry during the early spring in low Arctic mesic tundra and the impacts of deepened snow and enhanced nitrogen availability. Biogeochemistry 99(1): 127–141.

22.

Campbell

Mitchell

Groffman

. (2005) Winter in northeastern North America: A critical period for ecological processes. Frontiers in Ecology and the Environment 3(6): 314–322.

23.

Cao

Stewart

Hogg

(2007) Extreme winter warming over the Mackenzie Basin: Observations and causes. In: Woo

M-K

(ed.) Cold Region Atmospheric and Hydrologic Studies: The Mackenzie GEWEX Experience. Volume 1: Atmospheric dynamics. Berlin: Springer, 83–108.

24.

Carstairs

Oechel

(1978) Effects of several microclimatic factors and nutrients on net carbon dioxide exchange in Cladonia Alpestris (L.) Rabh in the sub-Arctic. Arctic and Alpine Research 10(1): 81–94.

25.

Chang

Foster

Owe

. (1985) Passive and active microwave studies of wet snowpack properties. Nordic Hydrology 16(2): 57–66.

26.

Chang

Gloersen

Schmugge

. (1976) Microwave emission from snow and glacier ice. Journal of Glaciology 16(74): 23–39.

27.

Chevallier

Feng

Bösch

. (2010) On the impact of transport model errors for the estimation of CO₂ surface fluxes from GOSAT observations. Geophysical Research Letters 37(21): L21803.

28.

Christensen

Schmidt

Michelsen

(2012) High Arctic dry heath CO₂ exchange during the early cold season. Ecosystems 15(7): 1083–1092.

29.

Christensen

Hewitson

Busuioc

. (2007) Regional climate projections. In: Solomon

Qin

Manning

, et al. (eds) Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press, 849–940.

30.

Colbeck

(1980) Thermodynamics of snow metamorphism due to variations in curvature. Journal of Glaciology 26(94): 291–301.

31.

Coops

Black

Jassal

. (2007) Comparison of MODIS, eddy covariance determined and physiologically modelled gross primary production (GPP) in a Douglas-fir forest stand. Remote Sensing of Environment 107(3): 385–401.

32.

Cox

Betts

Jones

. (2000) Acceleration of global warming due to carbon-cycle feedbacks in a coupled climate model. Nature 408: 184–187.

33.

Cramer

Bondeau

Woodward

. (2001) Global response of terrestrial ecosystem structure and function to CO and climate change: Results from six dynamic global vegetation models. Global Change Biology 7(4): 357–373.

34.

Cramer

Kicklighter

Bondeau

. (1999) Comparing global models of terrestrial net primary productivity (NPP): Overview and key results. Global Change Biology 5(S1): 1–15.

35.

Crane

Anderson

(1984) Satellite discrimination of snow/cloud surfaces. International Journal of Remote Sensing 5(1): 213–223.

36.

Dai

Zeng

Dickinson

. (2003) The common land model. Bulletin of the American Meteorological Society 84: 1013–1023.

37.

Dankers

De Jong

(2004) Monitoring snow-cover dynamics in Northern Fennoscandia with SPOT VEGETATION images. International Journal of Remote Sensing 25(15): 2933–2949.

38.

Deems

Fassnacht

Elder

(2006) Fractal distribution of snow depth from LiDAR data. Journal of Hydrometeorology 7(2): 285–297.

39.

Derksen

Toose

Rees

. (2010) Development of a tundra-specific snow water equivalent retrieval algorithm for satellite passive microwave data. Remote Sensing of Environment 114(8): 1699–1709.

40.

Dietze

Vargas

Richardson

. (2011) Characterizing the performance of ecosystem models across time scales: A spectral analysis of the North American Carbon Program site-level synthesis. Journal of Geophysical Research 116(G4): G04029.

41.

Dozier

Painter

(2004) Multispectral and hyperspectral remote sensing of alpine snow properties. Annual Review of Earth and Planetary Sciences 32: 465–494.

42.

Elberling

(2007) Annual soil CO₂ effluxes in the High Arctic: The role of snow thickness and vegetation type. Soil Biology and Biogeochemistry 39: 646–654.

43.

Elberling

Brandt

(2003) Uncoupling of microbial CO₂ production and release in frozen soil and its implications for field studies of Arctic C cycling. Soil Biology and Biochemistry 35(2): 263–272.

44.

Elberling

Nordstrøm

Grøndahl

. (2008) High-Arctic soil CO₂ and CH₄ production controlled by temperature, water, freezing and snow. Advances in Ecological Research 40: 441–472.

45.

Entekhabi

Njoku

O’Neill

. (2008) The Soil Moisture Active/Passive mission (SMAP). In: IEEE Geoscience and Remote Sensing Symposium, Boston, USA, 6–11 July, volume 3. Piscataway: IEEE, 1–4.

46.

Euskirchen

Bret-Harte

Scott

. (2012) Seasonal patterns of carbon dioxide and water fluxes in three representative tundra ecosystems in northern Alaska. Ecosphere 3(1): 1–19.

47.

Euskirchen

McGuire

Kicklighter

. (2006) Importance of recent shifts in soil thermal dynamics on growing season length, productivity, and carbon sequestration in terrestrial high-latitude ecosystems. Global Change Biology 12(4): 731–750.

48.

Fassnacht

Deems

(2006) Measurement sampling and scaling for deep montane snow depth data. Hydrological Processes 20(4): 829–838.

49.

Foster

Hall

Eylander

. (2011) A blended global snow product using visible, passive microwave and scatterometer satellite data. International Journal of Remote Sensing 32(5): 1371–1395.

50.

Foster

Sun

Walker

. (2005) Quantifying the uncertainty in passive microwave snow water equivalent observations. Remote Sensing of Environment 94(2): 187–203.

51.

Fox

Huntley

Lloyd

. (2008) Net ecosystem exchange over heterogeneous Arctic tundra: Scaling between chamber and eddy covariance measurements. Global Biogeochemical Cycles 22(2): GB2027.

52.

Frei

Lee

(2010) A comparison of optical-band based snow extent products during spring over North America. Remote Sensing of Environment 114(9): 1940–1948.

53.

Frei

Tedesco

Lee

. (2011) A review of global satellite-derived snow products. Advances in Space Research 50: 1007–1029.

54.

Friedlingstein

Bopp

Ciais

. (2001) Positive feedbacks between future climate change and the carbon cycle. Geophysical Research Letters 28(8): 1543–1546.

55.

Gould

Walker

Biesboer

(2003) Combining research and education: Bioclimatic zonation along a Canadian Arctic transect. Arctic 56(1): 45–54.

56.

Gouttevin

Menegoz

Dominé

. (2012) How the insulating properties of snow affect soil carbon distribution in the continental pan-Arctic area. Journal of Geophysical Research Biogeosciences 117: G02020.

57.

Gray

Male

(1981) Handbook of Snow: Principles, Processes, Management and Use. Frankfurt: Pergamon Press.

58.

Grogan

Chapin

FS III

(1999) Arctic soil respiration: Effects of climate and vegetation depend on season. Ecosystems 2(5): 451–459.

59.

Grogan

Jonasson

(2005) Temperature and substrate controls on intra-annual variation in ecosystem respiration in two subarctic vegetation types. Global Change Biology 11(3): 465–475.

60.

Grogan

Jonasson

(2006) Ecosystem CO₂ production during winter in a Swedish subartic region: The relative importance of climate and vegetation type. Global Change Biology 12(8): 1479–1495.

61.

Grogan

Michelsen

Ambus

. (2004) Freeze-thaw regime effects on carbon and nitrogen dynamics in sub-Arctic heath tundra mesocosms. Soil Biology and Biochemistry 36(4): 641–654.

62.

Hall

Riggs

(2007) Accuracy assessment of MODIS snow products. Hydrological Processes 21: 1534–1547.

63.

Hall

Foster

Verbyla

. (1998) Assessment of snow-cover mapping accuracy in a variety of vegetation-cover densities in central Alaska. Remote Sensing of Environment 66(2): 129–137.

64.

Hall

Riggs

Salomonson

(1995) Development of methods for mapping global snow cover using moderate resolution imaging spectroradiometer data. Remote Sensing of Environment 54(2): 127–140.

65.

Helfrich

McNamara

Ramsay

. (2007) Enhancements to, and forthcoming developments in the Interactive Multisensor Snow and Ice Mapping System (IMS). Hydrological Processes 21: 1576–1586.

66.

Hopkinson

Sitar

Chasmer

. (2004) Mapping snowpack depth beneath forest canopies using airborne LiDAR Photogrammetric Engineering and Remote Sensing 70(3): 323–330.

67.

Humphreys

LaFleur

(2011) Does earlier snowmelt lead to greater CO₂ sequestration in two low Arctic tundra ecosystems? Geophysical Research Letters 38: L09703.

68.

Huntzinger

Post

Wei

. (2012) North American Carbon Program (NACP) regional interim synthesis: Terrestrial biospheric model intercomparison. Ecological Modelling 232: 144–157.

69.

Jeffries

Zhang

Frey

. (1999) Estimating late-winter heat flow to the atmosphere from the lake-dominated Alaskan North Slope. Journal of Glaciology 45(150): 315–324.

70.

Jensen

(2007) Remote Sensing of the Environment: An Earth Resource Perspective, second edition. Upper Saddle River, NJ: Pearson Prentice Hall.

71.

Jones

Pomeroy

Davies

. (1999) CO₂ in Arctic snow cover: Landscape form, in-pack gas concentration gradients, and the implications for the estimation of gaseous fluxes. Hydrological Processes 13(18): 2977–2989.

72.

Kappen

(1993) Plant activity under snow and ice, with particular reference to lichens. Arctic 46(4): 297–302.

73.

Khvorostyanov

Krinner

Ciais

. (2008) Vulnerability of permafrost carbon to global warming. Part I: Model description and role of heat generated by organic matter decomposition. Tellus Series B 60: 250–264.

74.

Kim

Van Zyl

McDonald

. (2010) Monitoring surface soil moisture and freeze-thaw state with the high resolution radar of the Soil Moisture Active/Passive (SMAP) mission. In: Proceedings of the 2010 IEEE Radar Conference, Washington, DC, 10–14 May, 735–739.

75.

Koven

Ringeval

Friedlingstein

. (2011) Permafrost carbon-climate feedbacks accelerate global warming. Proceedings of the National Academy of Sciences 108(36): 14769–14774.

76.

Krinner

Viovy

De Noblet-Ducoudre

. (2005) A dynamic global vegetation model for studies of the coupled atmosphere-biosphere system. Global Biogeochemical Cycles 19(1): GB1015.

77.

Kucharik

Foley

Delire

. (2000) Testing the performance of a Dynamic Global Ecosystem Model: Water balance, carbon balance, and vegetation structure. Global Biogeochemical Cycles 14(3): 795–825.

78.

Kwok

Cunningham

(2008) ICESat over Arctic sea ice: Estimation of snow depth and ice thickness. Journal of Geophysical Research 113: C08010.

79.

Lafleur

Humphreys

(2008) Spring warming and carbon dioxide exchange over low Arctic tundra in central Canada. Global Change Biology 14(4): 740–756.

80.

Lafleur

Griffis

Rouse

(2001) Interannual variability in net ecosystem CO₂ exchange at the Arctic treeline. Arctic, Antarctic, and Alpine Research 33(2): 149–157.

81.

Lantz

Gergel

Kokelj

(2010) Spatial heterogeneity in the shrub tundra ecotone in the Mackenzie Delta region, Northwest Territories: Implications for Arctic environmental change. Ecosystems 13(2): 194–204.

82.

La Puma

Philippi

Oberbauer

(2007) Relating NDVI to ecosystem CO₂ exchange patterns in response to season length and soil warming manipulations in Arctic Alaska. Remote Sensing of Environment 109(2): 225–236.

83.

Larsen

Grogan

Jonasson

. (2007a) Respiration and microbial dynamics in two subarctic ecosystems during winter and spring thaw: Effects of increased snow depth. Arctic, Antarctic, and Alpine Research 39(2): 268–276.

84.

Larsen

Ibrom

Jonasson

. (2007b) Significance of cold-season respiration and photosynthesis in a subarctic heath ecosystem in Northern Sweden. Global Change Biology 13(7): 1498–1508.

85.

Lefsky

Cohen

Parker

. (2002) LiDAR remote sensing for ecosystem studies. BioScience 52(1): 19–30.

86.

Lemoine

Zelensky

Chinn

. (2010) Towards development of a consistent orbit series for TOPEX, Jason-1, and Jason-2. Advances in Space Research 46(12): 1513–1540.

87.

Levy

Grelle

Lindroth

. (1999) Regional-scale CO₂ fluxes over central Sweden by a boundary layer budget method. Agricultural and Forest Meteorology 98: 169–180.

88.

Lin

Gerbig

Wofsy

. (2004) Measuring fluxes of trace gases at regional scales by Lagrangian observations: Application of the CO₂ Budget and Rectification Airborne (COBRA) study. Journal of Geophysical Research 109: D15304.

89.

Lin

Pejam

Chan

. (2011) Attributing uncertainties in simulated biospheric carbon fluxes to different error sources. Global Biogeochemical Cycles 25: GB2018.

90.

Lovett

Cole

Pace

(2006) Is net ecosystem production equal to ecosystem carbon accumulation? Ecosystems 9(1): 152–155.

91.

McGuire

Christensen

Hayes

. (2012) An assessment of the carbon balance of arctic tundra: Comparisons among observations, process models, and atmospheric inversions. Biogeosciences 9: 3185–3204.

92.

McGuire

Macdonald

Schuur

. (2010) The carbon budget of the northern cryosphere region. Current Opinion in Environmental Sustainability 2(4): 231–236.

93.

McGuire

Melillo

Randerson

. (2000) Modeling the effects of snowpack on heterotrophic respiration across northern temperate and high latitude regions: Comparison with measurements of atmospheric carbon dioxide in high latitudes. Biogeochemistry 48: 91–114.

94.

McGuire

Sitch

Clein

. (2001) Carbon balance of the terrestrial biosphere in the twentieth century: Analyses of CO₂, climate and land use effects with four process-based models. Global Biogeochemical Cycles 15(1): 183–206.

95.

McGuire

Wirth

Apps

. (2002) Environmental variation, vegetation distribution, carbon dynamics and water/energy exchange at high latitudes. Journal of Vegetation Science 13(3): 301–314.

96.

Mahadevan

Wofsy

Matross

. (2008) A satellite-based biosphere parametrization for net ecosystem CO₂ exchange: Vegetation photosynthesis and respiration model (VPRM). Global Biogeochemical Cycles 22: GB2005.

97.

Marushchak

Kiepe

Biasi

. (2012) Carbon dioxide balance of subarctic tundra from plot to regional scales. Biogeosciences Discussions 9: 9945–9991.

98.

Mikan

Schimel

Doyle

(2002) Temperature controls of microbial respiration in Arctic tundra soils above and below freezing. Soil Biology and Biochemistry 34(11): 1785–1795.

99.

Morgner

Elberling

Strebel

. (2010) The importance of winter in annual ecosystem respiration in the High Arctic: Effects of snow depth in two vegetation types. Polar Research 29(1): 58–74.

100.

Niang

Dedie

Durand

. (2007) New inversion method for snow density and snow liquid water content retrieval using C-band data from ENVISAT/ASAR alternating polarization in alpine environment. In: Proceedings of the ENVISAT Symposium 2007, Montreux, Switzerland, 23–27 April.

101.

Nobrega

Grogan

(2007) Deeper snow enhances winter respiration from both plant-associated and bulk soil carbon pools in birch hummock tundra. Ecosystems 10(3): 419–431.

102.

Nobrega

Grogan

(2008) Landscape and ecosystem-level controls on net carbon dioxide exchange along a natural moisture gradient in Canadian low Arctic tundra. Ecosystems 11(3): 377–396.

103.

Nowinski

Taneva

Trumbore

. (2010) Decomposition of old organic matter as a result of deeper active layers in a snow depth manipulation experiment. Oecologia 163(3): 785–792.

104.

Oak Ridge National Laboratory Distributed Active Archive Center (ORNL DAAC) (2011) FLUXNET Web Page. Available at: http://fluxnet.ornl.gov.

105.

Oberbauer

Starr

Pop

(1998) Effects of extended growing season and soil warming on carbon dioxide and methane exchange of tussock tundra in Alaska. Journal of Geophysical Research D: Atmospheres 103(D22): 29075–29082.

106.

Oelbermann

English

Schiff

(2008) Evaluating carbon dynamics and microbial activity in Arctic soils under warmer temperatures. Canadian Journal of Soil Science 88(1): 31–44.

107.

Olsson

Sturm

Racine

. (2003) Five stages of the Alaskan Arctic cold season with ecosystem implications. Arctic, Antarctic and Alpine Research 35(1): 74–81.

108.

Öquist

Huner

NPA

(2003) Photosynthesis of overwintering evergreen plants. Annual Reviews of Plant Biology 54: 329–355.

109.

Panikov

(2009) Microbial activity in frozen soils. In: Margesin

(ed.) Permafrost Soils. Berlin: Springer, 119–147.

110.

Panikov

Flanagan

Oechel

. (2006) Microbial activity in soils frozen to below –39°C. Soil Biology and Biochemistry 38(4): 785–794.

111.

Papa

Legresy

Mognard

. (2002) Estimating terrestrial snow depth with the TOPEX/Poseidon altimeter and radiometer. IEEE Transactions on Geoscience and Remote Sensing 40(10): 2162–2169.

112.

Paudel

Andersen

(2011) Monitoring snow cover variability in an agropastoral area in the Trans-Himalayan region of Nepal using MODIS data with improved cloud removal methodology. Remote Sensing of Environment 115(5): 1234–1246.

113.

Prihodko

Denning

Hanan

. (2008) Sensitivity, uncertainty and time dependence of parameters in a complex land surface model. Agricultural and Forest Meteorology 148(2): 268–287.

114.

Prokop

(2008) Assessing the applicability of terrestrial laser scanning for spatial snow depth measurements. Cold Regions Science and Technology 54(3): 155–163.

115.

Pulliainen

Hallikainen

(2001) Retrieval of regional snow water equivalent from space-borne passive microwave observations. Remote sensing of environment 75(1): 76–85.

116.

Pumpanen

Ilvesniemi

Hari

(2003) A process-based model for predicting soil carbon dioxide efflux and concentration. Soil Sciences Society of America Journal 67(2): 402–413.

117.

Ramsay

(1998) The interactive multisensor snow and ice mapping system. Hydrological Processes 12(1011): 1537–1546.

118.

Randerson

Hoffman

Thornton

. (2009) Systematic assessment of terrestrial biogeochemistry in coupled climate-carbon models. Global Change Biology 15(10): 2462–2484.

119.

Ranson

Sun

Kovacs

. (2004) Landcover attributes from ICESat GLAS data in Central Siberia. In: IEEE Geoscience and Remote Sensing Symposium, Anchorage, 20–24 September, volume 2, 753–756.

120.

Rees

Lemmetyninen

Derksen

. (2010) Observed and modelled effects of ice lens formation on passive microwave brightness temperatures over snow covered tundra. Remote Sensing of Environment 114: 116–126.

121.

Riggs

Hall

(2004) Snow mapping with the MODIS Aqua instrument. In: Proceedings of the 61st Eastern Snow Conference, Portland, 9–11 June, 81–84.

122.

Riggs

Hall

(2011) MODIS snow and ice products, and their assessment and applications. In: Ramachandran

Justice

Abrams

(eds) Land Remote Sensing and Global Environmental Change. Remote Sensing and Digital Image Processing volume 11. New York: Springer, 681–707.

123.

Rott

Yueh

Cline

. (2010) Cold regions hydrology high-resolution observatory for snow and cold land processes. Proceedings of the IEEE 98(5): 752–765.

124.

Royer

Goïta

Kohn

. (2010) Monitoring dry, wet, and no-snow conditions from microwave satellite observations. IEEE Geoscience and Remote Sensing Letters 7(4): 670–674.

125.

Running

Nemani

Heinsch

. (2004) A continuous satellite-derived measure of global terrestrial primary production. Bioscience 54(6): 547–560.

126.

Rutter

Essery

Pomeroy

. (2009) Evaluation of forest snow processes models (SnowMIP2). Journal of Geophysical Research 114: D06111.

127.

Saito

Yamaguchi

Iwata

. (2012) Climatic physical snowpack properties for large-scale modeling examined by observations and a physical model. Polar Science 6: 79–95.

128.

Schaefer

Zhang

Bruhwiler

. (2011) Amount and timing of permafrost carbon release in response to climate warming. Tellus B 63(2): 165–180.

129.

Schlesinger

Lichter

(2001) Limited carbon storage in soil and litter of experimental forest plots under increased atmospheric CO₂ . Nature 411(6836): 466–469.

130.

Schuur

Vogel

Crummer

. (2009) The effect of permafrost thaw on old carbon release and net carbon exchange from tundra. Nature 459(7246): 556–559.

131.

Sellers

Mintz

Sud

. (1986) A simple biosphere model (SiB) for use within general circulation models. Journal of the Atmospheric Sciences 43: 505–531.

132.

Shi

(2008) Active microwave remote sensing systems and applications to snow monitoring. In: Liang

(ed.) Advances in Land Remote Sensing: System, Modelling, Inversion and Application. Berlin: Springer, 19–49.

133.

Sitch

McGuire

Kimball

. (2007) Assessing the carbon balance of circumpolar Arctic tundra using remote sensing and process modeling. Ecological Applications 17(1): 213–234.

134.

Sitch

Smith

Prentice

. (2003) Evaluation of ecosystem dynamics, plant geography and terrestrial carbon cycling in the LPJ dynamic global vegetation model. Global Change Biology 9(2): 161–185.

135.

Slaymaker

Kelly

(2007) The Cryosphere and Global Environmental Change. Malden, MA: Blackwell.

136.

Smith

Saatchi

Randerson

(2004) Trends in high northern latitude soil freeze and thaw cycles from 1988 to 2002. Journal of Geophysical Research 109: D12101.

137.

Starr

Oberbauer

(2003) Photosynthesis of Arctic evergreens under snow: Implications for tundra ecosystem carbon balance. Ecology 84(6): 1415–1420.

138.

Stiles

Ulaby

(1980) The active and passive microwave response to snow parameters 1. Wetness. Journal of Geophysical Research 85(C2): 1037–1044.

139.

Sturm

(1992) Snow distribution and heat flow in the Taiga. Arctic and Alpine Research 24(2): 145–152.

140.

Sturm

Holmgren

König

. (1997) The thermal conductivity of seasonal snow. Journal of Glaciology 43(143): 26–41.

141.

Sturm

Holmgren

Liston

(1995) A seasonal snow cover classification system for local to global applications. Journal of Climate 8(2): 1261–1283.

142.

Sturm

Holmgren

McFadden

. (2001) Snow-shrub interactions in Arctic tundra: A hypothesis with climatic implications. Journal of Climate 14(3): 336–344.

143.

Sturm

Perovich

Holmgren

(2002) Thermal conductivity and heat transfer through the snow on the ice of the Beaufort Sea. Journal of Geophysical Research 107: C108043.

144.

Sullivan

(2010) Snow distribution, soil temperature and late winter CO₂ efflux from soils near the Arctic treeline in northwest Alaska. Biogeochemistry 99 (1–3): 65–77.

145.

Sullivan

Arens

Sveinbjörnsson

. (2010) Modeling the seasonality of belowground respiration along an elevation gradient in the western Chugach Mountains, Alaska. Biogeochemistry 101 (1–3): 61–75.

146.

Sullivan

Welker

Arens

. (2008) Continuous estimates of CO₂ efflux from Arctic and boreal soils during the snow-covered season in Alaska. Journal of Geophysical Research G: Biogeosciences 113: G04009.

147.

Sun

Walker

Houser

(2004) A methodology for snow data assimilation in a land surface model. Journal of Geophysical Research 109: D08108.

148.

Svensson

Callaghan

(1988) Small-scale vegetation pattern related to the growth of Lycopodium annotinum and variations in its micro-environment. Plant Ecology 76(3): 167–177.

149.

Takala

Luojus

Pulliainen

. (2011) Remote Sensing of Environment 115(12): 3517–3529.

150.

Tarnocai

(2006) The effect of climate change on carbon in Canadian peatlands. Global and Planetary Change 53(4): 222–232.

151.

Tieszen

(1974) Photosynthetic competence of the subnivean vegetation of an arctic tundra. Arctic and Alpine Research 6(3): 253–256.

152.

Ulaby

Stiles

(1980) The active and passive microwave response to snow parameters 2. Water equivalent of dry snow. Journal of Geophysical Research 85 (C2): 1045–1049.

153.

Vikhamar

Solberg

(2003) Subpixel mapping of snow cover in forests by optical remote sensing. Remote Sensing of Environment 84(1): 69–82.

154.

Vourlitis

Oechel

Hope

. (2000) Physiological models for scaling plot measurements of CO₂ flux across an Arctic tundra landscape. Ecological Applications 10(1): 60–72.

155.

Wagner

(2008) Microbial communities and processes in Arctic permafrost environments. In: Dion

Nautiyal

(eds) Microbiology of Extreme Soils. Soil Biology 13. Potsdam: Springer, 133–154.

156.

Walker

Raynolds

Daniëls

. (2005) The circumpolar Arctic vegetation map. Journal of Vegetation Science 16(3): 267–282.

157.

Walker

Welker

. (1999) Long-term experimental manipulation of winter snow regime and summer temperature in Arctic and alpine tundra. Hydrological Processes 13 (14–15): 2315–2330.

158.

Wang

Sharp

Brown

. (2005) Evaluation of spring snow covered area depletion in the Canadian Arctic from NOAA snow charts. Remote Sensing of Environment 95(4): 453–463.

159.

Wang

(2008) Spatio-temporal patterns of forest carbon dioxide exchange based on global eddy covariance measurements. Science in China, Series D: Earth Sciences 51(8): 1129–1143.

160.

Wang

Grant

Arain

. (2011) Evaluating weather effects on interannual variation in net ecosystem productivity of a coastal temperate forest landscape: A model intercomparison. Ecological Modelling 222: 3236–3249.

161.

Wania

Ross

Prentice

(2009) Integrating peatlands and permafrost into a dynamic global vegetation model: 1. Evaluation and sensitivity of physical land surface processes. Global Biogeochemical Cycles 23: GB3014.

162.

Wang

Xiao

. (2008) Modeling gross primary production of a temperate grassland ecosystem in Inner Mongolia, China, using MODIS imagery and climate data. Science in China, Series D: Earth Sciences 51(10): 1501–1512.

163.

Xiao

Hollinger

Aber

. (2004a) Satellite-based modeling of gross primary production in an evergreen needleleaf forest. Remote Sensing of Environment 89(29): 519–534.

164.

Xiao

Zhang

Boles

. (2004b) Mapping snow cover in the pan-Arctic zone, using multi-year (1998–2001) images from optical VEGETATION sensor. International Journal of Remote Sensing 25(24): 5731–5744.

165.

Yang

Z-L

Dickinson

Robock

. (1997) Validation of the snow submodel of the biosphere-atmosphere transfer scheme with Russian snow cover and meteorological observational data. Journal of Climate 10(2): 353–373.

166.

Zhang

Armstrong

(2001) Soil freeze/thaw cycles over snow-free land detected by passive microwave remote sensing. Geophysical Research Letters 28(5): 763–766.

167.

Zhang

Osterkamp

Stamnes

(1996) Influence of the depth hoar layer of the seasonal snow cover on the ground thermal regime. Water Resources Research 32(7): 2075–2086.

168.

Zhang

Chen

Riseborough

(2008) Transient projections of permafrost distribution in Canada during the 21st century under scenarios of climate change. Global and Planetary Change 60 (3–4): 443–456.

169.

Zimov

Davidov

Voropaev

. (1996) Siberian CO₂ efflux in winter as a CO₂ source and cause of seasonality in atmospheric CO₂ . Climatic Change 35: 111–120.

170.

Zimov

Semiletov

Daviodov

. (1993) Wintertime CO₂ emission from soils of Northeastern Siberia. Arctic 46(3): 197–204.

171.

Zwally

Shutz

Abdalati

. (2002) ICESat’s laser measurements of polar ice, atmosphere, ocean, and land. Journal of Geodynamics 34 (3–4): 405–445.

Subnivean Arctic and sub-Arctic net ecosystem exchange (NEE)

Abstract

Keywords

I Introduction

II Modeling snow season NEE

III Uncertainties in Arctic and sub-Arctic NEE

1 Uncertainties in measurements of NEE

2 Uncertainties in the partitioning of snow season NEE

3 Uncertainties in regional estimates of NEE

IV Spatial associations between snow and NEE

1 Greater efflux of CO2 from sites with greater snow accumulation

2 Influence of vegetation on snow and snow season NEE

3 Implications for models of NEE

V Seasonal responses of NEE to snowfall, metamorphism and melt

1 Initial snowfall

2 Snow metamorphism

3 End of season snowmelt

4 Implications for modeling snow season NEE

VI Remote sensing of influences on snow season NEE

1 Visible and infrared observations of snow cover area

2 Altimeter observations of snow depth

3 Passive microwave observations of SWE

4 Active microwave observations of snow characteristics

5 Suitability of different remote sensing approaches for models of NEE

6 Incorporating remote sensing of snow into models of NEE

VII Conclusions

Footnotes

Funding

References

1 Greater efflux of CO₂ from sites with greater snow accumulation