P and evapotranspiration (ET) are important terms in the water cycle, and even minor shifts in their relative magnitudes can have critical impacts on surface water availability, streamflow, and groundwater storage. Recent changes in P over western Canada have shown regional and seasonal variations, with annual and winter increases in volume in the north and more significant winter decreases in the southern interior (Vincent et al., 2015; DeBeer et al., 2016). Pervasive warming has led to notable declines in the fraction of winter P falling as snow (Vincent et al., 2015; Dumanski et al., 2015). Historical variations and patterns of ET in western Canada have shown mixed trends, in part, due to the challenges with measurement, data availability, and modelling of ET (Mortsch et al., 2015). ET is affected by many variables, including precipitation, air temperature, surface and soil moisture availability, net radiation, wind speed, humidity, and vegetation characteristics. Thus, it is spatially highly variable over heterogeneous landscapes and is sensitive to changing climate and to land cover change (Zha et al., 2010). Changes in P amount and character are controlled to a considerable degree by global- and continental-scale conditions and their influence on regional circulation, air mass characteristics, and smaller-scale variability. These changes are presented and discussed by Stewart et al. (2019). Some further consideration of interactions at the surface and land–atmosphere feedbacks that affect local P processes is presented here.

Regional moisture recycling between P and ET is prevalent and provides a significant portion of the warm-season P across much of the Saskatchewan and Mackenzie River Basins (Szeto, 2007; Szeto et al., 2008). It represents an important mechanism of moisture transport, in some instances leading to intense rainfall and flooding (Li et al., 2017), and may have an important role in sustaining wet (or dry) conditions on seasonal to inter-annual timescales. For example, there are important feedbacks between ET and P, where an increase in P is likely to increase ET, but the increase in P itself could also be a result of increasing land ET and stronger moisture recycling (Trenberth, 1999; Dirmeyer et al., 2009). In future, under a warming climate, earlier disappearance of the seasonal snow cover will act to increase regional ET in spring as a result of the reduction in surface albedo, increase in net radiation to the ground surface, increase in overall surface temperature, thaw of frozen ground, and increase in exposure of wet soils. Shorter ice cover duration, especially in more northern lakes, will lead to increased lake evaporation and will therefore also play an important role in providing local moisture sources to downwind regions. These effects, together with earlier onset of ET from vegetation as a result of changes in the timing of leaf emergence, will enhance local atmospheric moisture supply in spring, possibly further enhancing the projected increase in March–April–May P (see Fig. 5 of Stewart et al., 2019). Later freeze-up in the fall can have similar effects, producing more lake effect snowfall, for example.

Kurkute et al. (2020) simulated future changes in P and ET over the Saskatchewan and Mackenzie River Basins using a pseudo global warming (PGW) approach with a high-resolution (4 km) Weather Research and Forecasting (WRF) model (Li et al., 2019). Under the Representative Concentration Pathway 8.5 W m-2 (RCP8.5) radiative forcing scenario, their results show increases in P, ET, and moisture recycling in both basins for the late 21st century (2085–2100) relative to their control period (2000–2015) but with considerable seasonal and spatial variations (Fig. 2). In the early spring (March and April), increases in P are projected to exceed increases in ET, leading to increasing snowpacks and/or soil moisture, but by May, the earlier snowmelt and increased atmospheric evaporative demand lead to greater increases in ET compared to P and drying of soils over much of the Prairies and Boreal Forest of western Canada (see Figs. 9 and 10 of Kurkute et al., 2020). This pattern continues into summer, and by July and August, simulated future P decreases in these parts of the region (most of the Saskatchewan River Basin), due in part to the decrease in soil moisture and surface water availability in the antecedent spring months. Although there is a simulated increase in moisture recycling in the warm season, the excess of ET over P is associated with an increase in atmospheric moisture divergence (i.e., transport out of the region).

Changes in ET also occur as a consequence of land cover and vegetation changes. Vegetation cover in turn is influenced by soil moisture (which is controlled by topographic position and surficial geology) but also by disturbance and succession dynamics (Ireson et al., 2015). The main vegetative controls on ET include leaf and canopy characteristics (vegetation height, LAI, leaf shape, stomatal behavior) and rooting depth and dynamics (Zha et al., 2010; Black and Jassal, 2016; Nazarbakhsh et al., 2020). Margolis and Ryan (1997) showed that, due to physiological limitations to transpiration in Boreal needleleaf trees, they have much lower ET rates than deciduous species, even when soil water is abundant. This is consistent with observations at the Boreal Ecosystem Research and Monitoring Sites (BERMS) flux towers (Fig. 1, site 7), showing a mature aspen stand with higher ET than a mature black spruce, which had higher ET than a jack pine stand (Fig. 3). Kljun et al. (2006) attributed these differences to a combination of type of tree species, topography, and soil type. Very young forest stands have also been shown to have much lower rates of ET than older stands (Granger and Pomeroy, 1997). Thus, shifts in Boreal Forest composition and structure, from coniferous to deciduous or mixed wood, or from black spruce to jack pine (discussed in Sect. 3.4 below), will have potentially large but species-specific effects on regional ET.

In the northern parts of the CCRN region, thaw-induced landscape change (Sect. 3.5) and expansion of shrubs (Sect. 3.4) are among the key drivers of changes in ET. Increasing thaw depth and shrinkage of permafrost-underlain areas impact growth and physiological processes of the trees through drying of the rooting zone, driving decreases in the productivity of black spruce-dominated sub-Arctic forests and reduction of sap flow and ET (Patankar et al., 2015; Sniderhan and Baltzer, 2016). At the same time, however, the conversion from forest to wetland associated with permafrost thaw acts to expand areas of open, freely evaporating water surfaces, counteracting this effect (e.g., Carpino et al., 2018). Warren et al. (2018) demonstrated that at Scotty Creek (Fig. 1, site 13), ET attributable to black spruce accounted for less than 1 % of landscape ET, suggesting areas of open water are of much greater importance to the water balance regionally. The expansion of shrubs in northern tree line and Tundra environments will likely increase regional ET in the snow-free period. For example, Zwieback et al. (2019b) found that rainfall interception losses from birch shrubs at Trail Valley Creek in the southern Arctic (Fig. 1, site 11) reduced below-canopy rainfall by 15 %–30 % but that losses depend on shrub species and density. Shrubs can efficiently reduce stomatal conductance under conditions of high vapor pressure deficit and their shading effect can act to limit surface evaporation under dense shrubs (Lund, 2018), further complicating the responses to shrub expansion. Shrub–snow interactions (Sect. 3.2) essentially act to retain winter snowfall and increase post-melt water availability, resulting in greater ET (Pomeroy et al., 2006; Ménard et al., 2014).

Over western Canada during the past several decades there has been a widespread reduction in snow depth, snow cover extent, and seasonal duration, with a shorter snow cover period of between 1 and 2 months, mostly due to earlier melt in spring (Brown et al., 2010, 2020; Mudryk et al., 2018; Marsh et al., 2019). Projected climate warming over the coming decades will continue to cause ubiquitous changes in snow regime, including (i) a greater fraction of P in the form of rain as opposed to snowfall, especially during shoulder seasons, at lower elevations, and in more southerly locations, (ii) more frequent rain-on-snow events, (iii) warmer and wetter snowfall, (iv) more mid-winter melt events as air temperature crosses the freezing point more frequently, and (v) earlier spring melt and snow cover depletion (Fig. 4). This will also cause distinct changes in runoff, with further transition from snowmelt to rainfall-dominated regimes. The transitions from snowfall to rain and from snow-dominated to rain-dominated hydrological systems are particularly sensitive where and when conditions are relatively warm and large amounts of P occur near 0 ∘C (Mekis et al., 2020). For example, analysis by Harder and Pomeroy (2014) in the Rocky Mountain Front Ranges at Marmot Creek (Fig. 1; site 2) showed that a significant proportion of the observed P events, recorded as either snowfall or rain, occurred within just a few degrees plus or minus of 0 ∘C as air temperature or hydrometeor temperature. Even slight warming could lead to rain becoming dominant at such locations. Shi et al. (2015) described the effects of increased rainfall during the snowmelt runoff period at Trail Valley Creek.

Hillslope-scale snowmelt runoff is potentially highly vulnerable to warming temperatures and associated changes in the amount and phase of precipitation. For instance, at three small (5 ha) hillslopes in the Saskatchewan Prairie, Coles et al. (2017) found that increases in summer rains were buffered by the unfrozen, deep, high-infiltrability soils. In contrast, winter and spring melt onto frozen ground with limited soil infiltrability resulted in runoff responses that more closely mirrored the snowfall and snowmelt trends. Increasing occurrence of mid-winter melt events can also alter the timing and magnitude of depression-focused groundwater recharge (Pavlovskii et al., 2019) and may lead to more basal ice formation, producing complex runoff responses in spring. Follow-on hillslope-scale analysis by Coles and McDonnell (2018) found evidence for filling of micro- and meso-depressions on the slope, followed by macro-scale, whole-slope spilling. While surface topography is relatively unimportant under unfrozen conditions on low relief and high infiltrability Prairie sites, surface topography was of critical importance for connectivity and runoff generation when the ground was frozen during the brief, annual snowmelt pulse. Under climate warming, losing this brief period of surface topographic control on runoff generation could have large implications for hillslope runoff, depending on basal ice formation, among other factors.

Warming can also lead to other important, and sometimes unanticipated, responses in snow accumulation, redistribution, and ablation processes (Fig. 4). Earlier onset of spring melt of the seasonal snow cover shifts snowmelt timing to conditions of lower incoming solar radiation (Pavlovskii et al., 2019). Paradoxically, this can lead in some cases to a reduction in daily and seasonal average ablation rates and a longer overall period of melt (Pomeroy et al., 2015; Musselman et al., 2017), but not in the Arctic, where both earlier and faster overall melts are predicted (Krogh and Pomeroy, 2019). This is counterintuitive and would not be captured by simple temperature-index melt models (Pomeroy et al., 2015). Warmer and wetter snow has lower susceptibility to wind transport (Li and Pomeroy, 1997), leading to a potential reduction in blowing snow transport and sublimation losses, which can partially offset reductions in snow water equivalent (SWE) due to direct effects of climate warming. Model results by Pomeroy et al. (2015) at Marmot Creek indicate a reduction of blowing snow transport by up to 50 % and decrease in sublimation losses by up to roughly 30 % with warming of up to 5 ∘C. This would also have important but at present poorly understood consequences for the redistribution of snow, the variability and patterns of SWE over the landscape, and the timing and rate of snow cover depletion (e.g., DeBeer and Pomeroy, 2017). Suppression of blowing snow would lead to a more uniform spatial distribution and thus more rapid decline of snow-covered area that could not be compensated for by the variability in melt energy (Schirmer and Pomeroy, 2020).

Snow–vegetation interactions further affect hydrological responses, and the impacts of vegetation change can equal or exceed those due to climate alone (Rasouli et al., 2019). A conceptual summary is shown in Fig. 4. With rising temperatures, warmer and wetter intercepted snow is more likely to fall to the ground instead of remaining in the forest canopy, where it would otherwise mostly sublimate. Snowfall interception efficiency is relatively insensitive to air temperature (Hedstrom and Pomeroy, 1998) and thus warming is unlikely to lead to large changes in initial interception amounts. But retention of the intercepted snow load is highly temperature dependent (Ellis et al., 2010) and so warming promotes faster unloading and a lower sublimation loss. This acts in combination with reduced wind transport of snow on the ground to offset reductions in SWE due to direct warming effects (Pomeroy et al., 2015). Rising temperatures can shift sub-canopy snowmelt energetics such that the reduction in melt rates due to shading is overcome by the increase in melt due to enhanced long-wave radiation from the canopy (Lundquist et al., 2013). Forest canopy structure, density, and species composition also significantly influence interception loss. Thinning of existing forest cover, reduction in leaf area index (LAI), and transition from coniferous to deciduous species, which are expected as a result of increasing human and natural disturbance and wildfire (Sect. 3.4), will lead to greater surface snow accumulation due to the reduction in canopy interception and sublimation but at the same time will expose more of the snow surface to increasing net radiation and an accompanying increase in ablation rates.

In open, windswept environments dominated by short vegetation such as grasses, crops, and shrubs, expansion across the landscape and/or increasing height and density of vegetation influences surface water availability and land–atmosphere energy and moisture exchanges. Shrub expansion acts to enhance local snow accumulation through more trapping of wind-blown snow and suppression of blowing snow redistribution and sublimation (Pomeroy et al., 2006; Ménard et al., 2014; Wallace and Baltzer, 2019). Shrubs reduce albedo in the spring but are buried in winter and have little effect on albedo in summer. Their canopy reduces latent heat fluxes from snow in the spring and initially accelerates melt when partly exposed and then retards snowmelt when the shrub canopy is fully exposed (Pomeroy et al., 2006; Wilcox et al., 2019). In the Prairies, increasing crop stubble height acts to retain more snow and to increase melt rates, infiltration, and meltwater runoff (Harder et al., 2019).

In western Canada and globally, glaciers have been predominantly losing mass and retreating in extent, with an apparent acceleration in their wastage in recent decades (Demuth et al., 2008; Demuth, 2018; Menounos et al., 2019; DeBeer et al., 2020). Even in the absence of further warming, many of these glaciers are out of balance with the current climate, given their present configuration (Zemp et al., 2015; Marzeion et al., 2018). This indicates that they will further recede to adjust their geometry to the current climate, with a typical response time of several decades for glaciers in western Canada (Marshall et al., 2011; Marzeion et al., 2018). Ongoing climate change is expected to further exacerbate the current imbalance and lead to additional retreat (Clarke et al., 2015).

Mass balance (the net gain or loss of snow and ice averaged over the glacier surface) responds directly to climate perturbations, whereas glacier extent, form, and flow patterns exhibit delayed and modified responses to mass balance changes (Cuffey and Paterson, 2010). Glacier responses are also influenced by secondary factors such as temperature effects on ice flow and meltwater availability at the glacier bed, which affects glacier sliding. In general, warmer air temperatures lead to greater specific ablation rates and a longer melt season, and may reduce accumulation depending on the area–elevation distribution of individual glaciers and the nature of P changes. Many glaciers and icefields in the CCRN region receive snowfall year round at high elevations and some rainfall in the summer. With climate warming, the proportion of rainfall events increases and the late summer snowline moves to higher reaches of the glaciers, exposing firn and bare ice, which melt faster than snow due to their lower albedo. Dust, impurities, and algae in the snow and ice become more concentrated on glacier surfaces as a consequence of high melt rates, in turn reducing the albedo and further enhancing melt (Williamson et al., 2019; DeBeer et al., 2020). There may also be an interaction with wildfire in western Canada, with deposition of black carbon and forest-fire fallout further reducing glacier albedo and providing nutrients to microbial communities (e.g., Marshall and Miller, 2020). High thinning rates in the upper accumulation area of many glaciers in western Canada indicate that these processes are well under way (Anderson, 2017; Pelto et al., 2019; Ednie and Demuth, 2018), while reductions in accumulation zone extent can lead to rapid glacier disintegration, and even complete disappearance. Glacier fragmentation and detachment of tributary ice streams leads to loss of ice supply to lower reaches, which can then become stagnant and melt out.

There are other important glacier–climate feedbacks. Energy balance conditions shift in response to glacier retreat; for example, ice-free marginal areas and valley walls contribute turbulent energy supply and long-wave radiation fluxes to the glacier, and these fluxes can be enhanced as glaciers thin and retreat, increasing ablation rates. The presence of glacial ice helps to regulate local climates and preserve cold conditions. As reduced snow accumulation leads to a reduction in glacier mass balance, so a reduction in glacier extent leads to a reduction in snow accumulation, given that the glacier surface, which is ≤0 ∘C, helps retain snow cover (Marshall et al., 2011).

Projections of future glacier change indicate that glaciers in the Rocky Mountains will lose roughly half their total area and volume by mid century, regardless of RCP scenario, and from about 75 % to 95 % or more by the end of the 21st century under the RCP2.6 to RCP8.5 scenarios, respectively (Clarke et al., 2015). By mid century, many valley glaciers will have retreated substantially up-valley, and by late in the century even high elevation glaciers and icefield plateaus will be greatly reduced or will have disappeared entirely (Fig. 5). Even the Columbia Icefield (see Demuth and Horne, 2017, and Tennant and Menounos, 2013), the largest and among the highest elevation ice masses in the Rocky Mountains, is projected to disintegrate into several small vestigial patches of ice near the tops of the highest peaks by the late 21st century. There are no comparable studies for the glaciated regions of the Mackenzie Mountains, Northwest Territories, but the observed patterns of recent change are similar to glaciers in the Rockies and the future change is expected to be similar (Demuth et al., 2014; Ednie and Demuth, 2018).

From a hydrological perspective, as glacier loss progresses, glacier wastage contributions and enhanced ablation will increase glacial contributions to discharge towards “peak water” (Huss and Hock, 2018), followed by a decline in glacier runoff due to loss of ice-covered area, even with further warming and increasing specific ablation rates. It remains uncertain where, when, and over what scales this will occur, although some studies have indicated that peak water has already passed in parts of southwestern Canada (Moore et al., 2020). Clarke et al. (2015) projected that peak runoff of glacier meltwater will occur between 2020 and 2040. Projections of glacier decline on the eastern slopes of the Rocky Mountains by Marshall et al. (2011) indicate a substantial decline in glacier contributions to discharge from about 1.1 km3 yr-1 early this century to 0.1 km3 yr-1 by late in the 21st century. With the loss of glaciers, the buffering effect that glacial storage can provide for discharge variations (e.g., during drought years) in the mountain headwaters will become increasingly diminished (Demuth and Ednie, 2016).

Ecosystem change can have profound effects on hydrological response and land–atmosphere feedbacks, yet the complexity of expected change and the associated uncertainty are often overlooked in hydrological projections. Across the CCRN region, contemporary climate change is already having direct impacts on northern ecosystems, defined here as including the southern Boreal Forest and its transition with the Prairies and the Cordillera. The interior of western Canada has been identified as a region of maximum ecological sensitivity (Bergengren et al., 2011). Forests in the southern Boreal region of western Canada have shown signs of declining productivity and increasing mortality associated with drought stress or insect disturbances, including widespread dieback and mortality of aspen (Hogg et al., 2008), stand fragmentation, and increases in tree mortality of up to 2.5 % yr-1 (Peng et al., 2011). Farther north, remote sensing indices of vegetation greenness indicate that substantial areas of Tundra and northern Boreal Forest have been increasing in vegetation productivity (Ju and Masek, 2016; Keenan and Riley, 2018; Sulla-Menashe et al., 2018). This is largely due to expansion of woody shrubs, such as alders and tall willows (Myers-Smith et al., 2011, 2019; Lantz et al., 2013), infilling of forests near the northern tree line (Lantz et al., 2019), and increases in tree growth rates (Sniderhan et al., 2020). Advancement of the Taiga–Tundra tree line in response to recent trends of climate warming has been more variable (Harsch et al., 2009; Dearborn and Danby, 2018). Lantz et al. (2019) showed infilling of forests below the tree line in the Northwest Territories but no increase in tree density above the tree line in the Tundra. To the south in the Rocky Mountains, Trant et al. (2020) observed widespread upward advance in alpine tree lines and increases in tree density, with changes in growth form from krummholz to erect tree form.

Climate change alters terrestrial ecosystems broadly through changes to (1) composition (vegetation, soils, and wildlife), (2) configuration and disturbance patterns, and (3) function. This includes structural changes to the current vegetation (above- and below-ground biomass, plant density, canopy height, LAI, and rooting depth); changes to land cover distribution patterns (resulting from changes in the disturbance regime and changes in competition, colonization, ecosystem resilience and vegetation succession following disturbance); and functional changes (surface albedo, snow accumulation and melt, soil freeze and thaw, ET, ecosystem productivity, decomposition, biogeochemical cycling, and wildlife habitat). The direct climatic drivers of vegetation change include rising atmospheric CO2 concentrations and temperature- and moisture-induced shifts in plant community function and vegetation distributions. However, over the 21st century the greatest impacts of climate change on vegetation dynamics are expected to be indirect, via increased frequency and intensity of disturbance (wildfire, insect outbreaks, and other landscape-scale disturbances; Turetsky et al., 2017) leading to losses of ecosystem resilience. These intensified disturbance processes can cause ecosystems to reach critical tipping points, triggering ecological state change (reviewed by Johnstone et al., 2016). Imposed on the climate-induced changes in vegetation will be the potential for changing human activities (e.g., logging, land clearing for agriculture and mining; Landhausser et al., 2010; Hannah et al., 2020), some of which will interact with climate change to accelerate vegetation change.

Northern ecosystems are expected to be most resilient to disturbances and environmental conditions that are within the historic range of variability and previous adaptation (Keane et al., 2009; Johnstone et al., 2016; Seidl et al., 2016). Many northern ecosystems may be initially resistant to change, because feedbacks associated with long-lived vegetation help to maintain environmental conditions and ecological functions that support ecological stability, even during directional environmental change (Chapin et al., 2004). While fire has been a foundational process in the functioning and ecology of the Boreal Forest for more than 5000 years, an increase in the frequency of high-intensity fires, coupled with a warming climate, may weaken ecosystem resilience and disrupt the historically stable cycles of forest succession. The result may be a regime shift from one plant community to another and from one stability domain to another (Johnstone et al., 2010c, 2016). Wildfire activity has increased in recent decades across the Boreal Forest (Hanes et al., 2019) and there are indications that fires are burning more severely (Turetsky et al., 2011) and deeper into stored legacy carbon (Walker et al., 2019), creating novel conditions for forest regeneration (Johnstone et al., 2010a; Pinno et al., 2013). For example, stands may burn at young ages before trees are old enough to generate seeds; these events, especially when they occur in combination with unusually dry or warm years, can trigger regeneration failures and cause shifts to non-forested states (Brown and Johnstone, 2012; Whitman et al., 2018). Stand-replacing wildfires initiate new phases of forest regeneration where seedlings may be much more sensitive to climate conditions than in an established stand where canopy trees substantially alter the local microclimate (Johnstone et al., 2010b; Davis et al., 2019; Hart et al., 2019). There is consensus that in northern forests, fire frequency and severity will continue to increase (Rogers et al., 2020).

Projections of future wildfire-induced ecosystem change in the Boreal Forest are challenging and highly uncertain. Increasing fire will result in a younger forest, widespread replacement of black spruce stands, and higher proportions of deciduous broadleaf species or jack pine (e.g., Johnstone et al., 2010a), with greater change in the south than the north. CCRN developed a plausible scenario of post-fire replacement of evergreen needleleaf forest (ENF) with deciduous broadleaf forest (DBF) across the Boreal Forest, as described in the Appendix, for the purpose of use in hydrological model future projections (Fig. 6). Although this is simply a scenario, and not a projection with an associated confidence level, the resulting forest change due to increasing wildfire is potentially great. For both the mid- and late-century periods, there is a considerable reduction in DBF across the southern parts of the Boreal Plain, as a result of increasing fire and the conversion of forest to grassland. Farther north and west, in the Taiga Plain, the Shield, and the Western Cordillera, there is extensive and progressive replacement of ENF with DBF as a result of both climate and fire-driven changes in forest succession. In reality, DBF and jack pine stands tend to be more resilient to fire (Hart et al., 2019), and less flammable in the case of DBF, and so their expansion may partially counter the increase in fire occurrence expected under a warmer climate.

Insects represent another form of disturbance with high potential for disrupting forest successional patterns, and may also lead to the replacement of black spruce stands by mixed-wood and deciduous species (Pureswaran et al., 2015). Forest insects may expand northwards if warmer winter temperatures increase potential rates of population growth (Post et al., 2009; Bentz et al., 2010). For the first time, pest populations of mountain pine beetle have been found in the Northwest Territories (GNWT, 2013). Likewise, unusual outbreaks of spruce bark beetle in the Yukon and Alaska have been associated with warm winter temperatures that allow increased insect survival through the winter (Berg et al., 2006). In some cases, forests have exhibited high levels of resilience to new disturbance conditions, as in the rapid recovery to bark beetle outbreaks in the southwestern Yukon (Campbell et al., 2019).

Across the northern and alpine tree line and tundra areas, displacement of shrubs by ENF and larch forest will occur in areas where sparse forest cover exists (e.g., Mamet et al., 2019), while above the tree lines, shrub expansion into tundra environments will likely continue with warmer temperatures and increasing water availability. Large shifts in tree line position are not expected over the 21st century due to both biological and geological constraints. At the northern tree line, the limited reproductive capacity of the tree species results in low seed availability, which restricts the rate of tree expansion into tundra ecosystems (Brown et al., 2019; Harsch et al., 2009), although this is dependent on the nature of the tree line, as expanded upon in Harsch et al. (2009). Similarly, the advance of the alpine tree line is restricted by geological and geomorphological controls such as avalanching, soil limitations, slope configurations that generate harsh winds, and other seed establishment and growth-limiting factors (Macias-Fauria and Johnson, 2013; Davis and Gedalof, 2018). Northern and montane shrub tundra areas will expand and continue the greening trend, with conversion of dwarf-shrub and graminoid-dominated tundra to tall-shrub tundra, resulting in more and taller shrubs, and an increase in LAI for existing patches. At fine scales, the rate and location of shrub expansion are very heterogeneous due to combined moisture and nutrient-driven responses (Wallace and Baltzer, 2019). For instance, although most infilling and recruitment is expected to occur in valley bottoms, low-lying areas, and other locations with sufficient water availability, excess moisture can carry nutrients downslope. Shrub Tundra is also susceptible to disturbance-induced changes. Large fires can occur in Tundra environments (Mack et al., 2011), and increased fire activity may occur if temperatures cross climate thresholds that have regulated fire activity in the past (Young et al., 2017) or as fuel accumulates due to shrub expansion. Permafrost thaw also affects shrub colonization (see Sect. 3.5). Shrub expansion can have multi-directional hydrological impacts (Grünberg et al., 2020), including shrub–snow interactions (Sect. 3.2) and increasing ET (Sect. 3.1), warmer soils, greater thaw depth, and thermokarst and subsidence, altering supra-permafrost layer storage, flow paths, and lake development (Sect. 3.5).

In addition to the forest cover change scenario, CCRN developed a plausible scenario of 21st century shrub expansion into Tundra, Grassland, and Barren areas, described in the Appendix and shown in Fig. 7. While there is uncertainty and this does not represent a confident projection, prolific shrub growth over the Boreal and Taiga Cordillera, the southern Arctic, and the Taiga Shield ecological regions is expected. The gradual expansion northward is evident through the increase in shrub cover along the northern part of the Mackenzie River Basin and the movement of this growth zone to higher latitudes later in the century.

Climate warming has led to warming and increased thaw depth of permafrost across northern Canada (Smith, 2011), with associated changes in characteristics of seasonally frozen soils (e.g., timing of freezing and thawing, frequency of freeze–thaw cycles, depth of frost). In southerly locations where permafrost is discontinuous, shallow, and relatively warm (i.e., at or near the freezing point depression), there has been widespread thawing and degradation of permafrost, with increasing supra-permafrost layer thickness – including both the active layer (seasonally frozen) and the talik (perennially thawed) (Connon et al., 2018). As a result of warming and shallower re-freeze depths during winter, active layer thickness has been decreasing. Where ice-rich soils occur, there has been active thermokarst development, slumping, and ground surface subsidence (Olefeldt et al., 2016; Turetsky et al., 2019). In permafrost lowlands of the Taiga Plain, soil thawing has led to subsidence and inundation of ground surfaces resulting in extensive forest loss, fragmentation, and concomitant wetland expansion and conversion mostly to sphagnum-dominated bogs (Baltzer et al., 2014; Helbig et al., 2016). In the southern Arctic, increased permafrost thawing is leading to changes in channel permafrost conditions, increasing winter groundwater flow in the channel, and increasing occurrence of aufeis formation (Ensom et al., 2020).

Many northern ecosystems are underlain by ice-rich permafrost that is highly sensitive to thawing during warm summers (Segal et al., 2016; Lewkowicz and Way, 2019) or following other disturbances (Williams et al., 2013). Wildfire and combustion of insulating moss and peat layers affects permafrost temperatures and can trigger thaw (Holloway et al., 2020). The lateral expansion of thermokarst features increases following wildfire activity; for example, Gibson et al. (2018) found that wildfire was estimated to be responsible for 30 % of permafrost thaw expansion in the southern Northwest Territories. Some of the energy driving the thaw of permafrost enters the permafrost bodies laterally from adjacent permafrost-free terrains (Kurylyk et al., 2016). As such, the rate of permafrost thaw and forest loss is accelerating as patches of permafrost plateau become more fragmented, leading to greater proportional plateau edge to total plateau area (Quinton and Baltzer, 2013; Baltzer et al., 2014; Carpino et al., 2018). Reduced soil stability during thaw events can cause substantial mass wasting through thermokarst and retrogressive thaw slumps, with impacts on local vegetation and downstream drainage (Schuur and Mack, 2018). Once the vegetation is disturbed, colonization by tall shrubs can cause a persistent change in vegetation state due to altered patterns of snow accumulation and soil temperatures (Lantz et al., 2009; Schuur and Mack, 2018).

Quinton et al. (2009) proposed a conceptual model of canopy thinning and permafrost thaw in which canopy thinning due to fire, disease, or other disturbance allows for an increase in local solar energy input and leads to preferential ground thaw (Fig. 8). A local depression forms in the relatively impermeable frost table and underlying permafrost table. Such thaw depressions introduce a hydraulic gradient that directs subsurface flow towards them so that thaw depressions soon become local areas of elevated soil moisture content. Since the thermal conductivity of wet soil is far more than that of dry soil, the vertical conduction of energy to the thaw depressions increases due to the increased moisture content, and as a result, a positive feedback is initiated, which accelerates the thaw of the disturbed areas. Wet conditions prevent trees from re-establishing and a new, isolated flat bog is formed. Many areas within the Taiga Plain are highly susceptible to thaw through this process (e.g., Gibson et al., 2020) and widespread replacement of forest-covered peat plateaus by wetlands is expected over the coming decades. A caveat is that these ecosystems represent some of the strongest ecosystem-protected permafrost, so undoubtedly a portion of permafrost peatland will linger, but this will depend on the degree of warming and also fire (Stralberg et al., 2020).

The loss of permafrost is impacting water cycling across the northern parts of the CCRN region. Land-surface subsidence and the collapse of peat plateaus to wetlands in the Taiga Plain alters drainage networks, surface and groundwater storage distribution, and the transit of water across the landscape (Fig. 8; Connon et al., 2014, 2018; Haynes et al., 2018; Quinton et al., 2019). This incorporates individual wetlands into the runoff contributing area, which expands deep into the interior of extensive plateau–wetland complexes as hydrological connections form between wetlands. The process results in both transient increases to basin discharge through the dewatering of incorporated wetlands, and longer-term increases in discharge arising from an expanded contributing area (Quinton et al., 2019). Another mechanism by which thaw influences runoff processes is by opening previously inaccessible subsurface flow pathways. Talik expansion provides an additional drainage path for wetland dewatering – one that conducts water throughout the year (Connon et al., 2018; Devoie et al., 2019). While this may give rise to transient increases in basin discharge due to the increased connectivity and dewatering of wetlands (Quinton et al., 2019), the process is not sustainable and may result in eventual drying of the landscape with increasing ET (Stone et al., 2019). Regeneration of black spruce forest may ultimately occur in the absence of permafrost, as has been observed further south near the Northwest Territories–British Columbia border (Carpino et al., 2018). In the Taiga Shield landscape, lake storage state can rapidly change the contributing area for runoff downstream and the landscape has a distinct threshold–response runoff regime (Ali et al., 2013). Wetlands are important “switches” in controlling the state of hydrological connectivity in the watersheds (Spence and Phillips, 2015). Permafrost thaw (and ultimately disappearance) may significantly affect this functioning, but it is unclear at what fraction of thaw progression major hydrological changes will occur.

Over much of the Prairies and the Boreal Plain, groundwater discharge from shallow sand and gravel aquifers sustains year-round base flow in some small streams and can be an important component of the water balance of wetlands and of some lakes. Groundwater is thus important with respect to local water resources and in maintaining surface hydrological connectivity and ecosystem function. Groundwater provides rural water supplies and in some cases municipal supplies (Peach and Wheater, 2014), and whilst it is not used as a major source for irrigation water outside of the southern–central parts of Manitoba, an issue facing some parts of the Prairies is the increasing reliance on groundwater as water demand rises and surface water becomes over-allocated (Council of Canadian Academies, 2009). Regional-scale groundwater depletion is not common in Canada, unlike other parts of North America (Rodell et al., 2018), but there have been numerous examples of isolated, human-induced local-scale depletion in Alberta (e.g., Munroe, 2015). The water-table records in shallow (<20 m) observation wells in the Prairie region show regular seasonal variations, with rises in spring and declines through the rest of the year. There were no large long-term changes during 1960–2000, a noticeable drop during the 2000–2004 drought, followed by a rise in the following decade (Hanesiak et al., 2011). There are very few long-term observation wells in the Boreal Plain, but the detailed records of water-table variations at the BERMS (Site 7, Fig. 1), together with hydrometeorological records, demonstrate the responses of the water table to changes in net water input to the subsurface throughout dry and wet periods and in various typical settings including peatlands and dry uplands (Anochikwa et al., 2012; Barr et al., 2012).

In the Prairies and Boreal Plain, lateral groundwater flow is slow due to the relatively flat terrain and the low permeability of the clay-rich glacial sediments underlying most of the landscape. As a result, subsurface water movement is mostly vertical – downward with infiltration, upward by root uptake – and the soil water and groundwater form a hydrological continuum. Rises of the water table are primarily driven by snowmelt infiltration and by focused recharge beneath ephemeral ponds in small wetlands and depressions that dry out within days or weeks after filling with snowmelt runoff (Bam et al., 2020). Recharge processes are sensitive to changes in snow accumulation, redistribution, and ablation processes (Sect. 3.2) and to land-use conversion (e.g., native grassland to cultivated fields, change in tillage practice), which influences soil hydraulic properties and snowmelt infiltration and runoff (van der Kamp et al., 2003). Most summer P infiltrates only to the root zone and is taken up by vegetation, driving a seasonal decline of the water table (Hayashi et al., 2016). However, summer infiltration can lead to rises of the water table where it is near the ground surface, as in wetlands. As a result, the dynamics of the shallow groundwater table are strongly controlled by the balance between infiltration and ET in response to weather, vegetation, and seasons. It is also sensitive to inter-annual and inter-decadal fluctuations in P (Hayashi and Farrow, 2014). The water table in the Prairie and Boreal regions can fluctuate quickly but is generally limited in range. When the water table rises near the ground surface, ET is increased and lateral groundwater flow to surface waters becomes important within the highly fractured near-surface materials (Hayashi et al., 2016; Brannen et al., 2015). This causes the water table to decline and provides the negative feedbacks to limit the range of water-table fluctuations to a few meters.

Groundwater processes are closely linked to the water regime (i.e., hydroperiod) of wetlands. Prairie wetlands occur in the form of shallow marshes (“sloughs” or “potholes”) with little accumulation of organic matter, whereas Boreal wetlands primarily occur as peatlands. The spatial transition from Prairie marshes to Boreal peatlands is coincident with the transitional ecotone between the Prairie and Boreal Plain regions, described in Sect. 2 (see Ireson et al., 2015). The hydroperiod of prairie wetlands is essentially controlled by a balance between water inputs from snowmelt runoff and P, versus ET losses and sporadic overflow in wet periods (Hayashi et al., 2016). Groundwater outflow from these wetlands due to ET in the riparian zone also has a strong influence on the hydroperiod. Long-term (50+ years) data collected at the St. Denis WECC observatory (Fig. 1, site 8) have demonstrated the dominance of precipitation amounts in controlling the multi-decadal-scale variability in hydroperiod (Hanesiak et al., 2011; Hayashi et al., 2016).

The hydrology of Boreal peatlands has not been studied as extensively as that of Prairie wetlands, but studies of a fen in the BERMS (Barr et al., 2012) have shown that it has a large water storage capacity and supplies base flow to streams and to support the shallow water table in surrounding uplands during dry periods. In contrast, the fen sheds water quickly to streams during wet periods when the water table rises above the peat surface. Long-term studies in northern Alberta have shown that the type of glacial sediments has a large influence on the groundwater exchange and runoff generation from the peatlands (e.g., Devito et al., 2017).

Groundwater replenishment to deeper aquifers is restricted by the low permeability of overlying layers of clay, clay-rich glacial till, and shale, and by the position of the aquifers within larger regional groundwater flow systems (Cummings et al., 2012). In the Prairies, replenishment rates to confined aquifers generally range from a few mm to a few tens of millimeters per year (van der Kamp and Hayashi, 1998). Recharge to the water table represents a residual in the water balance and is highly sensitive to changes in the balance between P and ET; however, replenishment to deep aquifers is not sensitive to variations of the water table and therefore responds slowly to climate change.

In the Western Cordillera the interaction of groundwater with surface waters is in many ways different from the groundwater dynamics in the Boreal Plain and the Prairies. Groundwater plays an essential role in sustaining base flow in the mountain headwaters of large river systems (Paznekas and Hayashi, 2016), and may be of growing importance under climate change. Above the tree line in the Rocky Mountains, primary aquifers are sedimentary landforms such as talus, moraine, and rock glacier (Hood and Hayashi, 2015; Harrington et al., 2018; Hayashi, 2020; Christensen et al., 2020), except in areas with substantial karst systems. Groundwater storage in these landforms is relatively small compared to the SWE contained in the seasonal snow cover (Hood and Hayashi, 2015), and groundwater discharge exhibits a fast recession after snowmelt or rainfall events. However, this is generally followed by a slower recession and the remaining storage allows these aquifers to sustain stable base flow during the rest of the year when there is little recharge (Hayashi, 2020). The high topographic relief, together with significant heterogeneity in bedrock and surficial deposits, influences patterns of vertical and lateral groundwater flow and recharge and discharge processes. At lower elevations, aquifers include glacial and alluvial deposits of highly permeable sands and gravels that drape mountainsides and underlie valley bottoms, usually 10 s to 100 m thick, but in some instances up to several hundred meters in thickness (Toop and de la Cruz, 2002). These store larger quantities of water and provide a reliable supply for municipal and industrial uses. In floodplain areas, the water table is usually near the ground surface and fluctuates with river levels. Although mountain aquifers are able to buffer base flow against climate warming and associated changes in surface water availability (e.g., Paznekas and Hayashi, 2016), anecdotal evidence has indicated that they cannot sustain high flows in drought years, such as in 2015, when the spring–summer discharge of the Bow River fell to about half its median rate at Banff and to less than 10 % at its mouth.

Based on field studies and understanding from the WECC observatories, efforts were directed primarily at improving functionality and expanding the capability of handling complex cold-region processes within the Cold Regions Hydrological Modelling (CRHM) platform (Pomeroy et al., 2007; https://research-groups.usask.ca/hydrology/modelling/crhm.php, last access: 6 April 2021). CRHM is a flexible modelling system that can be used to generate a process hydrology model, specific to the needs of the user and to the availability of driving meteorological data and of basin biophysical information to select parameters. A functioning model is built by selecting various process modules from a library; the modules incorporate algorithms or sub-models that are based on several decades of hydrological research. Process algorithms cover a wide range of phenomena specific to cold-region hydrology, which are then linked together to represent specific elements of the hydrological system and cycling over distinct landscape units termed “hydrological response units” (HRUs). Process studies and model developments focused on blowing snow transport and sublimation over complex terrain (Aksamit and Pomeroy, 2018, 2020); snowmelt in disturbed forests and on slopes; water flow through snowpacks (Leroux and Pomeroy, 2017, 2019); glacier snow, firn, and ice melt (Samimi and Marshall, 2017; Marshall and Miller, 2020; Anderson, 2017; Pradhananga, 2020); snow avalanching; soil moisture and hydraulic conductivity (Zwieback et al., 2019a); and freezing and thawing of soils (Krogh et al., 2017; Williamson et al., 2018; Rowlandson et al., 2018; Lara et al., 2020).

Improving process representation ultimately translates into increased realism and predictive capability. Marmot Creek (Fig. 1; site 2) provides a prime example of how the WECC observatories are valuable testbeds for model development. Pomeroy et al. (2012) and then Fang et al. (2013) developed a model for this basin using the CRHM platform, with parameterizations based on decades of intensive field science here and elsewhere, and showed very good model performance for snow accumulation and melt, soil moisture, groundwater, and stream discharge at the basin scale and for smaller sub-basins. They conducted “falsification” tests by neglecting parameterizations for forest canopy snow mass and energy, blowing snow, intercepted rain evaporation, and sublimation, which showed that without these, the model prediction of basin and sub-basin streamflow was substantially degraded. This approach, verified by successful internal simulation of sub-basin hydrological variability and model falsification, lends more confidence in the face of non-stationarity and for ungauged basins, where parameter equifinality and conceptual uncertainty are otherwise major challenges.

CRHM was applied at a number of the WECC observatories as well as other sites in western North America and run for historical periods using local meteorological observations, ERA-Interim (Dee et al., 2011), and/or bias-corrected WATCH (Weedon et al., 2011, 2014) forcing data. It was verified using field observations and then used to diagnose hydrological function of these basins, and predict and diagnose historical change, such as the impact of changing climate, wetland drainage, glacier shrinkage and ice exposure, permafrost thaw, and shrub growth/expansion on hydrological processes, cycling, and streamflow hydrographs. It has also been run for late 21st century climates, downscaled using statistical and dynamical methods. Future sensitivity and change was examined by perturbing climate forcing using high-resolution WRF modelled pseudo global warming under RCP8.5 (see Krogh and Pomeroy, 2019) or using results from the North American Regional Climate Change Assessment Program (NARCCAP) consisting of 11 regional climate models driven by outputs from multiple global climate models (GCMs) for the SRES A2 emission scenario (see Rasouli et al., 2019). Hydrological responses to changing vegetation, soils, and land cover were examined using current and expected future states of the basins.

CCRN worked with partners in Environment and Climate Change Canada (ECCC) to advance the Modélisation Environmentale Communautaire (MEC) – Surface and Hydrology (MESH) model. MESH is a stand-alone land-surface–hydrology scheme designed for both forecasting and open loop simulations (Pietroniro et al., 2007). It uses a “grouped response unit” (GRU) approach to represent spatial heterogeneity for parameter identification, with CLASS as the surface water and energy budget simulation model for open loop simulations. As a hydrological modelling system, MESH captures many of the important land-surface processes necessary for cold-region simulation, provides a flexible modelling framework that facilitates inter-comparison of alternative algorithms and models (e.g., land-surface schemes and routing schemes), and can be applied over large river basins.

Over the course of CCRN, major advancements in the MESH system were made in terms of basic operability, scalability, and parallelization, as well as in its ability to handle sloping and complex terrain, permafrost (Sapriza-Azuri et al., 2018; Elshamy et al., 2020), lakes and wetlands, snow processes and glacier representation, vegetation processes including snow–canopy interactions (Bartlett and Verseghy, 2015; Asaadi et al., 2018), frozen soils and Prairie hydrology including variable hydrological connectivity (Mekonnen et al., 2014), and water management impacts including reservoirs, diversions, and irrigation (Yassin et al., 2019). The work has progressed to a point at which functioning MESH models for the Mackenzie and Saskatchewan River systems have been developed, calibrated, and tested (Yassin et al., 2017, 2019; Bahrami et al., 2020).

MESH has been run for historical (1980–2010) and future (2025–2055; 2070–2100) climates at a 10 km resolution, incorporating these advancements in process and water management representation, to examine changes in regional hydrology and river flows. Forcing data included WATCH and ERA-Interim products (Weedon et al., 2011, 2014) with bias correction using regional datasets such as the combined Global Environmental Multiscale (GEM) atmospheric model forecasts and the Canadian Precipitation Analysis (CaPA) (Fortin et al., 2018). Regional climate projections for future MESH simulations to the end of the 21st century were derived from 15 ensemble members from the CORDEX-NA CanRCM4 under the RCP8.5 emissions scenario. Climate fields were spatially downscaled and bias-corrected against the WATCH ERA-Interim reanalysis – GEM – CaPA product (Asong et al., 2020). Evaluation of the historical simulations has shown that the model performed quite well, with Nash–Sutcliffe Efficiencies for river discharge greater than 0.5 at 60 % of observation stations in the Saskatchewan River Basin and greater than 0.6 for most stations along the main stem of the Mackenzie River and its major tributaries (Peace, Athabasca, Slave, Liard, and Great Bear). Major efforts have been needed to develop robust algorithms for simulation of permafrost, glacier, and vegetation change, and the development of scenarios of future land cover change. These have now been prepared and the next phase of the work is to run the models for full (i.e., climate and land cover) future assessment. Scenario results are currently pending, but some preliminary insights are discussed below.