Doctor of Philosophy (PhD)
Geography & Environmental Studies
Faculty of Arts
Tundra snow cover is important to monitor as it influences local, regional, and global scale surface water balance, energy fluxes, and ecosystem and permafrost dynamics. Moreover, recent global circulation models (GCM) predict a pronounced shift in high latitude winter precipitation and mean annual air temperature due to the feedback between air temperature and snow extent. At regional and hemispheric scales, the estimation of snow extent, snow depth and, snow water equivalent (SWE) is important because high latitude snow cover both forces and reacts to atmospheric circulation patterns. Moreover, snow cover has implications on soil moisture dynamics, the depth, formation and growth of the permafrost active layer, the vegetation seasonality, and the respiration of C02.
In Canada, daily snow depth observations are available from 1955 to present for most meteorological stations. Moreover, despite the abundance and dominance of a northern snow cover, most, if not all, long term snow monitoring sites are located south of 550N. Stations in high latitudes are extremely sparse and coastally biased. In Arctic regions, it can be logistically difficult and very expensive to acquire both spatially and temporally extensive in-situ snow data. Thus, the possibility of using satellite remote sensing to estimate snow cover properties is appealing for research in remote northern regions.
Remote sensing techniques have been employed to monitor the snow since the 1960s when the visible light channels were used to map snow extent. Since then, satellite remote sensing has expanded to provide information on snow extent, depth, wetness, and SWE. However, the utility of satellite sensors to provide useful, operational tundra snow cover data depends on sensor parameters and data resolution. Passive microwave data are the only currently operational sources for providing estimates of dry snow extent, SWE and snow depth. Currently, no operational passive microwave algorithms exist for the spatially expansive tundra and high Arctic regions. The heterogeneity of sub-satellite grid tundra snow and terrain are the main limiting factors in using conventional SWE retrieval algorithm techniques. Moreover, there is a lack of in-situ data for algorithm development and testing.
The overall objective of this research is to improve operational capabilities for estimating end of winter, pre-melt tundra SWE in a representative tundra study area using satellite passive microwave data. The study area for the project is located in the Daring-Exeter-Yamba portion of the Upper-Coppermine River Basin in the Northwest Territories. The size, orientation and boundaries of the study area were defined based on the satellite EASE grid (25 x 25 km) centroid located closest to the Tundra Ecosystem Research Station operated by the Government of the Northwest Territories. Data were collected during intensive late winter field campaigns in 2004, 2005, 2006, 2007, 2008, and 2009. During each field campaign, snow depth, density and stratigraphy were recorded at sites throughout the study area. During the 2005 and 2008 seasons, multi-scale airborne passive microwave radiometer data were also acquired. During the 2007 season, ground based passive microwave radiometer data were acquired. For each year, temporally coincident AMSR-E satellite Tb were obtained.
The spatial distribution of snow depth, density and SWE in the study area is controlled by the interaction of blowing snow with terrain and land cover. Despite the spatial heterogeneity of snow cover, several inter-annual consistencies were identified. Tundra snow density is consistent when considered on a site-by-site basis and among different terrain types. A regional average density of 0.294 g/cm3 was derived from the six years of measurements. When applied to site snow depths, there is little difference in SWE derived from either the site or the regional average density. SWE is more variable from site to site and year to year than density which requires the use of a terrain based Classification to better quantify regional SWE. The variability in SWE was least on lakes and flat tundra, while greater on slopes and plateaus. Despite the variability, the interannual ratios of SWE among different terrain types does not change that much. The variability (CV) in among terrain categories was quite similar. The overall weighted mean CV for the study area was 0.40, which is a useful regional generalization. The terrain and landscape based classification scheme was used to generalize and extrapolate tundra SWE. Deriving a weighted mean SWE based on the spatial proportion of landscape and terrain features was shown as a method for generalizing the regional distribution of tundra SWE.
The SWE data from each year were compared to AMSR-E satellite Tb. Within each season and among each of the seasons, there was little difference in 19 GHz Tb. However, there was always a large decrease in 37 GHz Tb from early November through April. The change in ΔTb37-19 throughout each season showed that the Tb at 37 GHz is sensitive to parameters which evolve over a winter season. A principal component analysis (PCA) showed that there are differences in ΔTb37-19 among different EASE grids and that land cover may have an influence on regional Tb. However, the PCA showed little relationship between end of season ΔTb37-19 and lake fraction. A good relationship was found between ΔTb37-19 and in-situ SWE. A quadratic function was fitted to explain 89 percent of the variance in SWE from the ΔTb37-19. The quadratic relationship provides a good fit between the data; however, the nature of the relationship is opposite to the expected linear relationship between ΔTb37-19 and SWE.
Airborne Tb data were used to examine how different snow, land cover and terrain properties influence microwave emission. In flat tundra, there was a significant relationship between SWE and high resolution ΔTb37-19. On lakes and slopes, no strong relationships were found between SWE and high resolution ΔTb37-19. Due to the complexity of snow and terrain in high resolution footprints, it was a challenge to isolate a relationship between SWE and Tb. However, as the airborne footprint size increased the amplitude of variability in Tb decrease considerably to the point that Tb in large footprints is not sensitive to local scale variability in SWE. As such, most of the variability evident in the high and mid resolution airborne data will not persist at the EASE grid scale.
Despite the many challenges, algorithm development should be possible at the satellite scale. The AMSR-E ΔTb37-19 changes from year to year in response to differences in snow cover properties. However, the multiple years of in-situ snow data remain the most important contribution in linking Tb with SWE.
Rees, Andrew, "Tundra Snow Cover Properties from In-Situ Observation and Multi-Scale Passive Microwave Remote Sensing" (2010). Theses and Dissertations (Comprehensive). 1105.