Stepwise multiple regression snow models: GIS applications in the Marmot Creek basin, (Kananaskis Country, Alberta) Canada and the National Park Berchtesgaden, (Bayern) Germany
Title | Stepwise multiple regression snow models: GIS applications in the Marmot Creek basin, (Kananaskis Country, Alberta) Canada and the National Park Berchtesgaden, (Bayern) Germany |
Publication Type | Conference Proceedings |
Year of Conference | 1997 |
Authors | Forsythe, K. K. |
Conference Name | 65th Annual Western Snow Conference |
Series Title | Proceedings of the 65th Annual Western Snow Conference |
Date Published | May 1997 |
Publisher | Western Snow Conference |
Conference Location | Banff, Alberta |
Keywords | GIS, Landsat, Modeling, Principal component analysis, Thematic Mapper |
Abstract | Geographic Information Systems (GIS) and satellite remote sensing are now used extensively in analyses of natural environments. This research examined the usefulness of a GIS and remote sensing based approach for integrating many different and widely varying data sources. The multisource data sets were used in stepwise, multiple linear regression models that analyzed the variation in snow dept and snow-water equivalent (SWE) depth.Models were developed in the Marmot Creek Basin using landsat Thematic Mapper (TM) satellite data, Digital Elevation Model (DEM) data and various mapped data layers in combination with historical snow depth and SWE measurements. The best model for snow depth produced a coefficient of determination or adjusted R2 of 0.6370. The independent variables of elevation, incidence, tree height and principal component four of the TM data could explain approximately 64% of the variation in snow depth. The best model for SWE depth had an adjusted R2 of 0.5814. The independent variables that explained the variation in SWE depth were elevation, incidence, and TM band seven. The National Park Berchtesgaden models build upon the approach used in the Marmot Creek Basin. landsat TM data and a DEM provided the basis for the regression independent variables. Additional analyses of these data were performed within the GIS to calculate variables such as across and down slope curvature and a Normalized Difference Vegetation Index (NDVI). The best model for snow depth produced an adjusted R2 of 0.801 while the SWE result was 0.843. The variables providing the explanation in snow depth variation were elevation and NDVI. SWE depth variation was explained by elevation, across slope curvature, TM band seven and principal component four of the TM data. |
URL | sites/westernsnowconference.org/PDFs/1997Forsythe.pdf |