Introduction

How combinations of temperature increases and precipitation changes affect the magnitude and timing of peak snowpack and annual runoff in mountain basins is investigated in this sensitivity study. High elevation mountain headwater basins are hydrologically important, as they store water in the form of snowpack during winter and release it in spring and summer (Barry, 1992; Bales et al., 2006), and, ecologically important as they are key zones for biodiversity due to steep gradients of air temperature, precipitation, and topography (Beniston, 2003). Mountain snowpacks are sensitive to warming (Minder, 2010). Air temperature changes exert important controls on the hydrology of basins where snowmelt is the dominant hydrological process (Marks et al., 1998; Pederson et al., 2011; Sospedra-Alfonso et al., 2015). The contribution of mountain headwaters to the downstream discharge of rivers ranges from 35% in cold and humid river basins to 90% in hot and arid basins (Viviroli and Weingartner, 2004). Mountains cover 25% of the Earth’s land surface (Diaz et al., 2003) and 26% of the world population live in high-elevation areas (Meybeck et al., 2001). The origin of discharges from 50% of the world’s rivers are mountain headwaters (Beniston, 2003). Snowmelt volume and timing play a key role in freshwater availability, flood control, and ecological sustainability of cold region mountain environments (Stewart et al., 2004; Semmens and Ramage, 2013).
The higher sensitivity of snow and frozen soils to warming makes cold region mountain basins, those with mean annual air temperatures near 0oC, suitable study areas for investigating climate change impacts on the hydrological cycle (Barry, 1992; Bunbury and Gajewski, 2012). Climate warming effects have been studied in some mountain headwater basins (e.g., Cayan, 1996; Stewart et al., 2004; Bales et al., 2006), and warming is expected to continue to threaten the ecological and hydrological integrity of these regions (Malmqvist and Rundle, 2002). Fyfe and Flato (1999) showed that elevation becomes important to the pattern of climate change over western North America only when a significant continental-scale warming dominates, and it is not detectable in the early stages of climate change. Late winter and spring temperatures have a key role in the responsiveness of mountain basins to a warming climate and snowmelt runoff timing in regions with near-freezing air temperatures (Stewart et al., 2004; McCabe and Clark, 2005; Rasouli et al., 2019a, 2020). Mote et al. (2005) reported that climatic trends and not changes in land use and forest canopy affect snowpack in western North America. A significant increasing temperature trend, especially in minimum temperature, led to a reduction in number of soil freeze days, earlier occurrence of plant-water stress, and a strong seasonal shift in streamflow throughout 1962-2006 in the high elevation Reynolds Creek Experimental Watershed, USA (Nayak, 2008).
A common approach for investigating the hydrological response to climate change is to apply climate model projections under different greenhouse gas emission scenarios and to downscale regional atmospheric circulations obtained from the climate models to variables at local scales using statistical or dynamical methods (e.g., Jasper et al., 2004; Fowler et al., 2007; McDonald et al., 2010). Mountain hydrometeorology, however, poses challenges to statistical and dynamical downscaling methods. The assumption in statistical downscaling that the predictor–predictand relationship is stationary and future relationships will be the same as past ones (Wilby and Wigley, 1997) does not guarantee that statistical downscaling approaches would perform better than the delta change method (Hay et al., 2000; Fowler et al., 2007; Kay et al., 2009; Sunyer et al., 2012). Dynamical models driven by an ensemble of multiple boundary conditions have high computational cost at the high resolutions needed in mountains and so usually neglect uncertainty and also require bias correction to provide reasonable forcings (Fang and Pomeroy, 2020). A realistic downscaling of atmospheric variables shows a high sensitivity to the choice of downscaling methods (Wilby et al., 2000). These limitations make consideration of an alternative solution necessary for mountainous regions.
As biases due to scale and parametrization issues have not yet resolved by statistical and dynamical downscaling methods, the alternative perturbation method also known as delta change factor method (e.g., Stockton and Boggess, 1979; Semadeni-Davies et al., 2008; Kawase et al., 2009), can produce plausible hydroclimatological changes for the future. The perturbation method represents the changes in climatology between current and future climates for variables such as precipitation and air temperature (Stockton and Boggess, 1979; Pomeroy et al., 2015; Rasouli et al., 2019a, 2019b). The method retains the main hydrometeorological processes present in historical measurements, whilst minimizing computational resources. The perturbation method has been widely used; however, its application has been limited to air temperature changes factors (e.g., ∆T = ±2°C in Nayak, 2008; Pomeroy et al., 2015) and precipitation change factors (e.g., ∆P = ±25% in López-Moreno et al., 2016).
The sensitivity of snow processes to warming were studied in the Canadian Rockies (Pomeroy et al., 2015) and Reynolds Creek, Idaho (Rasouli et al., 2015). The snow and runoff sensitivities to precipitation change and warming were studied in Wolf Creek, Yukon Territory (Rasouli et al., 2014). This sensitivity study applies similar sensitivity analyses to both snow and runoff regimes and in a comparable way for physically based cold regions hydrological models representing the three headwater basins that span much of the northern North American Cordillera. More specifically, this research investigates how the magnitude and timing of peak snowpack and annual runoff respond to combinations of temperature increases (0 to 5°C) and precipitation changes (-20, -10, 0, +10, and +20%). By considering 30 combinations, applied to the three mountain headwater basins, the sensitivity of hydrological responses to changes in forcings can be compared. This increases the understanding of the relationships between changes in forcing and model response in these basins. The main question addressed is whether the impact of warming on mountain snow and runoff hydrology can be offset by precipitation increases. This has not been resolved in the literature (e.g., Arnell, 1999; Prowse et al., 2006, Luo et al., 2008). The specific objective for this sensitivity analysis is to quantify the response of simulated mountain hydrological processes to changes in air temperature and precipitation associated with future climate change.