Investigating snowpack across scale in the northern Great Lakes–St. Lawrence forest region of Central Ontario, Canada

Abstract

This study investigates scaling issues by evaluating snow processes and quantifying bias in snowpack properties across scale in a northern Great Lakes-St. Lawrence forest. Snow depth and density were measured along transects stratified by land cover over the 2015/2016 and 2016/2017 winters. Daily snow depth was measured using a time-lapse (TL) camera at each transect. Semivariogram analysis of the transect data was conducted, and no autocorrelation was found, indicating little spatial structure along the transects. Pairwise differences in snow depth and snow water equivalent (SWE) between land covers were calculated and compared across scales. Differences in snowpack between forested sites at the TL points corresponded to differences in canopy cover, but this relationship was not evident at the transect scale, indicating a difference in observed process across scale. TL and transect estimates had substantial bias, but consistency in error was observed, which indicates that scaling coefficients may be derived to improve point scale estimates. TL and transect measurements were upscaled to estimate grid scale means. Upscaled estimates were compared and found to be consistent, indicating that appropriately stratified point scale measurements can be used to approximate a grid scale mean when transect data are not available. These findings are important in remote regions such as the study area, where frequent transect data may be difficult to obtain. TL, transect, and upscaled means were compared with modelled depth and SWE. Model comparisons with TL and transect data indicated that bias was dependent on land cover, measurement scale, and seasonality. Modelled means compared well with upscaled estimates, but model SWE was underestimated during spring melt. These findings highlight the importance of understanding the spatial representativeness of in situ measurements and the processes those measurements represent when validating gridded snow products or assimilating data into models.

Assessing the impact of distributed snow water equivalent calibration and assimilation of Copernicus snow water equivalent on modelled snow and streamflow performance

“…The three rulers were spaced approximately 1 to 2 m apart in front of the time‐lapse cameras to capture fine scale variability. This is a cost‐effective method of automatically measuring daily, spatially distributed snow depths in a remote basin, such as the Petawawa (Beaton et al, 2019; Fortin et al, 2015; Hedrick & Marshall, 2014; Kim et al, 2007; Lundberg et al, 2016; Parajka et al, 2012).…”

“…It is distributed by the European Commission for use by government, research institutions and the hydropower industry for runoff modelling and flood forecasting, planning for hydropower generation, agriculture, forestry, water supply and climate change modelling (European Commission Joint Research Centre, 2022; Luojus et al, 2017). The Copernicus product has known issues during melt due to the assimilation of snow depths measured in open areas that may not be representative of the surrounding landscape as well as the limitations of passive microwave‐derived SWE in wet snow conditions (Beaton et al, 2019; Luojus et al, 2017; Meromy et al, 2013).…”

“…Remotely sensed and modelled gridded SWE products provide estimates across larger extents and at higher frequencies than can feasibly be measured. Several modelled (e.g., GLDAS, ERA-Interim/Land, and MERRA), remotely sensed (e.g., NASA AMSR-E, JAXA AMSR-2) and data assimilated (e.g., Copernicus, ESA Globsnow, SNODAS and CMC Analysis) daily gridded SWE datasets generated by government and academia are available at various spatial and temporal resolutions (Balsamo et al, 2015;Brown et al, 2003;Kelly, 2009;Luojus et al, 2017;Rienecker et al, 2011;Rodell et al, 2004;Takala et al, 2011) may not be representative of the surrounding landscape as well as the limitations of passive microwave-derived SWE in wet snow conditions (Beaton et al, 2019;Luojus et al, 2017;Meromy et al, 2013).…”