4.2 Data Fusion Snow Water Estimates for British Columbia, Canada: A Regional Perspective

Tuesday, 24 January 2017: 10:45 AM
310 (Washington State Convention Center )
Andrew M Snauffer, University of British Columbia, Vancouver, BC, Canada; and W. W. Hsieh and A. J. Cannon

Regional-scale estimates of snow water equivalent (SWE) in areas of winter snowfall accumulation are critical for hydrological planning purposes.  Numerous gridded SWE data products based on state-of-the-art remotely sensed data and advanced modeling techniques perform well in simple settings but are less reliable in areas of complex topography, heavy forest cover, and very deep snowpacks, conditions common throughout the province of British Columbia (BC), Canada.  Mean absolute errors (MAEs) for six hemisphere- to global-scale gridded SWE products (ERA-Interim, ERA-Interim/Land, MERRA, MERRA-Land, GLDAS2 and GlobSnow) have been computed based on in-situ manual snow survey measurements over the five physiographic regions of BC.  These products have been averaged and also combined with identified covariates using multiple linear regression (MLR) and artificial neural networks (ANN).  The three best products (ERA-Interim/Land, MERRA & GLDAS2) outperformed the others across the province and in the majority of the regions.  The mean of the best three products was found to be better than both the mean of all six and the individual products themselves, but the top performer was an ANN using the best three as inputs.  Bootstrap MAE differences with the mean of six products revealed that the mean, MLR and ANN of the best three have lower MAE, statistically significant at the 95% level.  The ANN also outperformed MLR at the 95% level for a majority of the five physiographic regions and for the average over the entire province.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner