INNOVATION March-April 2015

f ea t u r e s

Aqua

Terra

0% 20% 40% 60% 80%

FIGURE 1 Snow-covered area maps for July 5, 2011 in the Cheakamus Basin, using four different satellite remote sensing methods. Elevation band contours are shown in white.

MODSCAG

Landsat

100% Cloud

event in the fall, when the spatial coverage of snow available for melting would otherwise be unknown. Such practical successes notwithstanding, much remains to be done. Recently, BC Hydro has moved to in-house processing of MODIS data for integration into a hydrologic modelling plat- form and potential direct ingestion into operational hydrologic models. However, more formal quantitative uses of MODIS- based SCA data in the hydrologic modelling chain will likely require rigorous data assimilation schemes. SCA uncertainties also remain an issue. A visual comparison of four SCA estimation methods is provided in Figure 1 for the Cheakamus Basin, a watershed in southwestern BC where BC Hydro operates the Cheakamus hydropower project. These snow cover maps are derived from three MODIS products (Terra binary, Aqua binary, and MODSCAG from the Terra observations) plus Landsat. In these maps, the Aqua, Terra, and Landsat pixels only use a binary snow classification: either 0% or 100% snow cover (or cloud cover). In contrast, the product derived from MODIS data using the MODSCAG spectral unmixing algorithm, developed and provided by the NASA Jet Propulsion Lab, provides fractional SCA values ranging continuously from 0% to 100%, which can be beneficial for accurately tracking spatiotemporal snow ablation patterns in mountain environments using the relatively low spatial resolution MODIS data. We delineated SCA on the Landsat image using the NDSI threshold method on the higher-resolution 30 m pixels, adjusting the threshold downwards for vegetation. Overall, SCA patterns appear similar between all four data products. On closer inspection, however, the binary method from both Aqua and Terra delineated significant cloud cover on this day, whereas neither MODSCAG or Landsat did. Further, when examined on an elevation-band by elevation-band basis (not shown), there are some significant differences between the methods. Much of this variation is again associated with the ways that cloud cover and

fractional snow cover within a pixel are handled by the various methods. Because all these observations have uncertainty, it is dif- ficult to say which is closest to the truth. Finally, though SCA is useful, accurate SWE estimation by sat- ellite remote sensing remains the ultimate prize. This is particular- ly valuable to longer-term seasonal water supply forecasting, which requires the total volume of SWE in the watershed. Such SWE monitoring methods are used elsewhere, but these have enjoyed relatively little success in Pacific Canada due to steep terrain, heavy evergreen forest cover, and annual peak SWE values that can ex- ceed 3,000 mm with total snowpack depth much greater than that (see Table 1). Similarly, snow data modelling applications, akin to the American SNODAS system (which also provides partial cover- age in southern BC), are another avenue to explore. Exciting times are ahead Remote sensing has been used in many ways for water resources applications, and these uses will only expand in the future. As technology marches forward, we are experiencing ever-increasing resolution, greater sensor accuracy, and simplification of data delivery. The difficulties of remote sensing in a complex region such as BC are being chipped away year by year, and a wide range of industries and applications will reap the benefits. About the authors Joel W. Trubilowicz, EIT, is a PhD candidate with the Department of Geography at the University of British Columbia (UBC). Emma Chorlton is a student with the UBC Department of Engineering Physics. Dr. Stephen J. Déry is with the University of Northern British Columbia, Environmental Science and Engineering Program. Dr. Sean W. Fleming, P.Phys., A.C.M., P.Geo., is with the Meteorological Service of Canada, UBC, and Oregon State University. The authors can be contacted via joel.trubilowicz@geog.ubc.ca.

2 0

M A R CH/A P R I L 2 015

i n n o v a t i o n