Hierarchical and time-varying controls on snowmelt runoff generation: a coupled analysis of temporal trends and spatial patterns

Abstract:

The patterns and processes influencing runoff generation are many and complex. In northern, semi-arid, gently-sloping landscapes, threshold-like activation of infiltration-excess overland flow has been observed to be both temporally and spatially variable. In order to quantify and predict hydrological connectivity and runoff generation, it is important to identify the multiple controls on hydrological connectivity and runoff generation, their hierarchy of importance, and their time- and space-varying nature. Here, we capitalize on a hitherto unused 52-year catchment-integrated data set (including runoff, soil characteristics, and meteorological data) from three 5-hectare experimental hillslopes in southern Saskatchewan, Canada. Through data mining and regression tree techniques, we show that the key controls on snowmelt runoff generation over a multi-decadal time period are 1) snow water equivalent, 2) soil moisture in the preceding fall, 3) depth of the surface thawed layer during snowmelt, and 4) topography (Fig. 1). The hierarchy of these controls is time-varying, and acts to offset and supplement the influence of other controls. Building on this temporal analysis, we map the high-resolution spatial patterns of these four key controls on runoff generation during the 2014 snowmelt season to understand how they interact to generate hydrological connectivity. We find that soil moisture and thawed layer depth show little spatial variability during snowmelt, while the spatial patterns of snow water equivalent (SWE) ablation reveal time-varying hot spots of snowmelt (e.g. Fig. 2). Coupled with analysis of the stable isotope composition of both snowmelt water from the base of the snowpack (high spatial variability) and hillslope runoff (high temporal variability), we show that hillslope runoff has time-varying sources, and that the delivery of water is dependent on the surface micro-topography and the filling and spilling of surface depressions over frozen ground.

Fig. 1. Summary regression tree showing the main controls on runoff amount for the 1962-2013 spring snowmelt seasons.

Fig. 2. a) Example of the spatial patterns mapping, showing the spatial variability of snow water equivalent (SWE) loss from the snowpack on one day (12 March 2014), overlain on a digital elevation model of the experimental hillslope. b) SWE ablation on the hillslope (daily data; blue line), and runoff through the hillslope outlet (15-minute data; black line). Grey shading indicates the particular day (12 March 2014) of SWE mapping shown in a).