Simulating Areal Snowcover Depletion and Snowmelt Runoff in Alpine Terrain

An overwhelming proportion of the flow of some of the major rivers in the western Canadian Prairies (e.g., the South Saskatchewan River) is derived from runoff in the headwaters of the Rocky Mountains, where snowmelt represents the greatest single contribution. Increasing concerns over future regional water resource stresses require better understanding and prediction of some alpine snow hydrology components, which are currently limited due to the large spatial heterogeneity of snow accumulation and melt processes, and problems with the scaling of these processes in hydrological models. The work presented in this thesis was aimed at improving the representation and effects of this variability on simulated areal snowcover depletion (SCD) and snowmelt runoff generation at different spatial scales in alpine environments. To accomplish this, a focused field data collection campaign was carried out at a small (1.2 km2) alpine cirque basin within the Marmot Creek Research Basin in the Front Ranges of the Canadian Rockies in the Kananaskis Valley, Alberta. Measurements here included detailed hydro-meteorological observations, snowcover (spatially distributed snow surveys, LiDAR-derived snowcover mapping, and daily acquisition of terrestrial-based photography of the alpine landscape for spatial–temporal snowcover mapping), and streamflow measurement at the alpine basin outlet. A theoretical framework was developed to upscale physically based point-scale snowmelt simulations for the prediction of areal SCD and meltwater generation, and was based on the lognormal probability distribution for values of snow water equivalent (SWE). The framework was applied and tested using a point-scale snowmelt model (Snobal) developed within the Cold Regions Hydrological Model (CRHM) platform. Finally, a conceptual/process-based hydrological model was developed for this basin using CRHM, and the spatial snowmelt framework was used together with this model to simulate the streamflow hydrograph at the outlet of the basin. This work has led to a number of important findings that advance the state of understanding of alpine snow hydrology, and provide useful tools for prediction outside of well-studied research basins. First, it was shown how the spatial and temporal variability in both pre-melt snowcover and snowmelt energetics control the evolution of the alpine snowcovered area (SCA) during the spring, which is an important variable for both hydrological and climatological applications. Daily terrestrial photographs were re-projected orthogonally over the landscape, and comparison of model predictions of areal SCD with observations from this imagery showed that improvements resulted from considering separate SWE distributions and applied energy to the snowcover on different slope-based landscape units in the basin, relative to using a single, basin-wide distribution with uniform applied energy. It was further shown that at certain times, such as early in spring, the effects of differential warming, ripening, and melt of different initial classes of SWE within a single landscape unit cause an “acceleration” of areal SCD due to the earlier and more rapid melt of areas with a relatively shallow snowpack, and that models that do not properly account for this effect may be in error. This is a feature that is common to all “cold” snowcovers, yet currently this can only be represented by fully distributed simulations applied at a fine spatial scale (i.e., 10 – 25 m), and where difficulties arise in establishing initial snowcover patterns outside of well-studied basins. However, the framework developed here provides a useful approach for resolving all major sources of SWE and melt rate variability, while retaining spatial and computational simplicity, and physical integrity. This is done by making explicit snowmelt computations for different initial classes of SWE (with unique mass and energy states) on different slope-based landscape units; the framework only requires values of mean SWE and CV (coefficient of variation) to establish initial snowcover conditions in a model. Thus, it can easily be applied in other basins by using “representative” landscape-based CV values. Lastly, the work provided insight on how the variability in both pre-melt snowcover and meltwater inputs over the basin influence the snowmelt hydrograph at the basin outlet. Through a comparison of different approaches for representing snowcover, snowmelt, and lower basin forest canopy effects, it was shown that the best correspondence with observed hydrographs was achieved when explicitly accounting for the differential timing, location, and extent of source areas for snowmelt runoff. However, in many other cases realistic appearing hydrographs were obtained, but for the wrong reasons due to cancellation of model errors. The approach here maintains internal “correctness” of the alpine snow components, which is beneficial towards development and parameterization of other process components in hydrological models applied elsewhere in alpine landscapes. The results also showed that the effects of differential melt timing and rate over different SWE classes within a single landscape unit (i.e., inhomogeneous melt) did not become manifested in the overall hydrograph response, despite having an important influence on areal SCD. Thus, if the primary goal of model application is to predict the hydrograph only, then this effect can likely be neglected without serious errors