The Derivation of Snow-Cover "Normals" Over the Canadian Prairies from Passive Microwave Satellite Imagery Joseph M. Piwowar Laura E. Chasmer Waterloo Laboratory for Earth Observations University of Waterloo Waterloo, Ontario, Canada Anne E. Walker Barry E. Goodison Climate Research Branch Atmospheric Environment Service Downsview, Ontario, Canada
Objective to develop a procedure which calculates "normal" snow cover over the southern Canadian Prairies from historical satellite imagery and then uses that normal in comparison with current data to identify snow cover anomalies
Rationale knowledge of the extent of snow pack coverage and its snow water equivalent (SWE) is important to both the local hydrological cycle and to the global climate system most SWE maps produced today are still derived from manually measured snow information retrieved from sparsely distributed snow courses SWE estimates from passive microwave satellite imagery are advantageous over traditional measurement methods because a more spatially and temporally complete picture may be established
Data Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave/Imager (SSM/I) passive microwave imagery in EASE-Grid format from the EOS-DAAC (Earth Observing System - Distributed Active Archive Center) at the National Snow and Ice Data Center, University of Colorado in Boulder the EASE-Grid format is optimal for this work because it is a standard earth reference system that is used not only for multi- temporal analyses but also for multi-sensor comparisons SSM/I data were available in EASE-Grid format for the 8-year period ; this is expected to be extended to 20-years very soon we focussed on the southern Canadian Prairie region where AES has been producing SWE maps for 10 years
Issues is the 8-year period of available SSM/I data in EASE-Grid format too short to derive a meaningful "normal"? l other reviews of short-term climate normals of temperature and precipitation concluded that year periods were better for predicting temperature and precipitation stability in the future than the 30-year WMO (World Meteorological Organization) normal what is the most appropriate statistical approach to defining a short-term normal? l other research has found that although the mean may not be the most comprehensive statistic, it still represents the single most effective descriptor, especially for short-term data periods we concluded that we could derive meaningful climatic analyses using eight-year averages of SSM/I imagery
Analysis Method Êderive the "normal" SWE from all available data Ê select data to represent each year the possibility of incomplete orbital data coverage and/or snow melt events on any specific day necessitates the need for some flexibility in choosing the data that will be used to represent a given year in the normal it is not unreasonable to assume that data from one or two days on either side of a specific date can be used in place of data from that date our procedure allows the user to select any single date, or the average of data from any combination of dates, within the five-day window (also known as a "pentad") centred a specific day to represent that date in the normal Ë calculate the average of the annual data Ëcompare current data with the normal l produce map products: Percent of Normal Difference from Normal Significant Departures from Normal
Percent of Normal SWE conventional map product used by resource managers shows the percentage ratio between data from a given day and the normal for that day l e.g. regions identified with a ratio between 90 and 109% identify areas where the snow cover is close to normal the use of ratio data is a major limitation to this map: there is no indication of the absolute differences in SWE that the percentages represent l e.g. if a current SWE value of 50 mm is compared to a normal value of 100 mm, the map would appropriately show this region as 50%, hopefully alerting resource managers to the potential SWE shortfall for the current year. A 50% value would also be assigned, however, in the situation where the current value of 5 mm is compared with a normal average of 10 mm – hardly an alarming difference
Difference from Normal SWE represents the deviation from normal values in absolute terms rather than as a percentage l e.g. regions where the present SWE differs from the long-term values by less than ±10 mm can be considered normal this type of map gives the resource decision maker quantitative measurements of how present conditions differ from the normal the map is limited by the fact that it does not account for normal SWE variances l e.g. if a current SWE value of 50 mm is compared to a normal value of 100 mm, the map would appropriately show this region as 50 mm below normal. Resource managers may become unduly alarmed if, in reality, the average inter-annual variation of the snow pack across the mapped area is in excess of 80 mm
Significant Departure from Normal SWE accounts for the inter-annual variability of SWE values and shows deviations from the normal with statistical significance l e.g. we can say with statistical confidence that the 1991 SWE values near 52°N 110°W are significantly lower than normal values at the 95% significance level the resource decision maker can confidently focus attention on the areas in the 95% or 99% categories (above and below normal) and not be unintentionally alarmed by large values in the Percent and Difference maps l e.g. review the amount of area coloured red (to indicate very low snow conditions) across all 3 maps
Summary an interactive procedure has been developed to create climatic normal snow water equivalent images from SSM/I brightness temperature files when current SWE values are compared to normal values, three map products are created: l the traditional Percent of Normal SWE map, l a Difference from Normal SWE map, and l a unique Significant Departure from Normal SWE map we show that the Significance map may provide a more realistic view of the locations of important SWE variations