|dc.description.abstract||Snow is a critical component of the earth’s overall energy budget and it contributes significantly to water resources especially in mountainous regions, coining the term the “water towers” for downstream communities (Viviroli et al., 2006). Studies have shown an increase in snow cover variability due in part by climate change. Most evident throughout the research is an earlier freshet period throughout the northern hemisphere, elevation-dependent warming in mountainous regions and regional climate models indicating transitions from snow to rain dominated basins (Pepin et al., 2015; Rangwala & Miller, 2012). Studies throughout British Columbia have shown evidence of earlier peak runoff from river gauges, a decrease in snow duration and increases in temperature by 1.4ᵒ (Shrestha et al., 2012; Kang et al., 2014; Islam et al., 2017). The Thompson Okanagan region is a semi-arid snow dominated region located in the southern portion of British Columbia (Kang et al., 2014). The spring freshet in Thompson Okanagan is affected by large atmospheric systems as well, including the Pacific North American Pattern (PNA), the Pacific Decadal Oscillation (PDO) and the Oceanic Nino Index (ONI).
This research focuses on identifying variations in snow cover during the spring freshet (April 1st-June 30th) in Thompson Okanagan with remote sensing observations from 2003-2019. Snow cover mapping is achieved using visible-infrared observations of snow. High albedo is easily distinguishable in the visible spectrum; however, cloud contamination impedes analysis using visible infrared observations. Steps to mitigate the impact of cloud cover adopted a multi-step methodology. This improved the ability to characterize snow cover extent variability during the spring freshet. The methodology includes: i) a daily combination of Terra/Aqua (from 2003-2012) and VIIRS (from 2012-2019) observations; ii) an adjacent temporal deduction (ATD) technique which replaces cloud pixels with non-cloudy pixels from +/-2 adjacent days; iii) a spatial filter to interpolate snow in cloudy pixels; iv) and the identification of a regional snowline elevation above which cloud-labelled pixels are classified as snow, and cloud pixels below the elevation for no-snow are classified as no-snow. This methodology significantly reduced cloud cover from an average of 71.5% to 1.6% annually.
Using stratified random sampling approach, reference points were gathered for a range of elevation bands for four watersheds within the region to test the snow mapping accuracy. The last day of snow (LDS) was extracted for each point from 2003-2019. Large scale atmospheric patterns (Pacific Decadal Oscillation (PDO), Pacific-North American (PNA) teleconnection pattern and Oceanic Nino Index (ONI)) were analyzed using simple and multiple linear regression to assess the variability within the LDS dataset that could be explained by these patterns. This analysis showed that the PNA did not significantly account the variability, but the PDO did with an R2 value reaching 64%, with a significance level of >95%. The simple linear regression models showed that the ONI explained 78% of the LDS variation during the March-April-May (MAM) months, with p>95%; this was more than any other 3-month interval studied. Also, the ONI R2 value decreased as elevation increased. Overall, El Nino years showed snow disappearance of ~23 days earlier than La Nina years at low elevation, ~18 days sooner at mid elevation and ~13 days sooner at high elevations. Earlier snow melt-out during El Nino phases have implications for water resources in the region, for residential and crop use as well as economic impacts for tourism (Westering, 2016; Winkler et al., 2017). This also contributes to area burned in forest fires and rapid melting snow can cause flooding in surrounding urban areas within Thompson Okanagan. Extending the study period into the future could allow further insights on potential effects of climate change within the region.||en