American Geophysical Union San Francisco, December 5th - 9th 2011 Snowmelt detection using QuikSCAT over Eastern Canada during the 2000-2009 period and snow simulation results Alexandre Langlois1 The team: Nathalie Thériault1, Alicia Gressent2, Alain Royer1, Libo Wang3, Ross Brown4, Chris Derksen5 1 Centre d’Applications et de Recherches en Télédétection, Université de Sherbrooke, Québec, Canada. 2 Laboratoire de Glaciologie et Géophysique de l’Environnement, CNRS-Université de Grenoble, France. 3 Climate Research Branch, Atmospheric Science and Technology Directorate, Environment Canada, Toronto, Ontario, Canada. 4 Science and Techonology Branch, Climate Processes Section, Environment Canada @ Ouranos, Montréal, Québec, Canada. 5 Climate Research Branch, Meteorological Service of Canada, Environment Canada, Toronto, Ontario, Canada. American Geophysical Union San Francisco, December 5th - 9th 2011
INTRODUCTION Importance of snow melt detection In northern hemisphere, snow exists from 6 to 9 months Spring transition has a significant impact on the climatological (albedo), hydrological (runoff) and ecological (habitat) properties The timing of snowmelt and the associated thawing of soil in the spring is coincident with the seasonal switch from a net source to sink for atmospheric carbon Changes in melt timing patterns in northern latitudes may be a good indicator of climate change in those regions
INTRODUCTION Importance of snow melt detection Sparse manual and/or meteorological observations Hindrance to melt studies Uncertainties in climate models and associated feedback (e.g. snow albedo feedback) Use of satellite remote sensing: microwave scatterometers: Broad spatial coverage at reasonable spatial/temporal resolutions Day/night, less sensitive to clouds High sensitivity to water (through dielectric properties)
OBJECTIVES Monitor snow melt using QuikSCAT 1. Compare two existing methods of melt detection using QuikSCAT 1. Wang et al., (2008) 2. Roy et al., (2010) 2. Look at temporal trends in melt onset days (MODs from QSCAT and CRCM4 model) at 19 stations in Eastern Canada 3. Compare MODs with snow model simulations (SNOWPACK and CLASS) and in-situ observations
METHODS - 19 stations, data available: 1. QuikSCAT melt 2000-2009 2. Met. obs., dates vary (on Fig.) 3. SNOWPACK sims. 1979-2009 4. CLASS sims. 1979-2009 5. CRCM4 sims. 1990-2009 SNOWPACK and CLASS simulations use NARR data, where precipitations are forced on the CANGRD adjusted precipitations product (accounting for wind, evaporation and wetting losses): Mekis, É and L.A. Vincent, 2011: An overview of the second generation adjusted daily precipitation dataset for trend analysis in Canada. Atmosphere-Ocean 49(2), 163-177.
METHODS: 1. Wang et al., (2008) Major melt events were identified when daily σ° was 1.7dB lower than the previous 5-day average for three or more consecutive days; Time series of QS backscatter, air temperature and snow depth observations at a forest site (left) and a tundra site (right) from Wang et al., (2008).
METHODS: 2. Roy et al., (2010) The threshold on backscatter values is adjustable (changes for each year, each pixel). The threshold is optimized linearly on Tmax. If Tmax > 3˚C = melt and if Tmax < -1˚C = no melt Yearly range of ˚ values that correspond well to this condition. Threshold = adjusted for best correspondance
RESULTS: Melt onset days - CRCM4 melt = 3 consecutive days or more where melt rate > 0 mms-1 - Roy et al. (2010) and CRCM4= earlier melt compared to Wang et al. (2008)
RESULTS: Melt onset days Better correlation between CRCM4 and Wang et al. (2008) for the period 2001-2009 in general. Earlier melt from Roy since the method is based on punctual ˚ (whereas Wang is based on a 5-day average ˚ threshold), which is variable in early winter (patchy snow), leading to MOD. Considering after March: much better
RESULTS: Melt onset days Wang et al., 2008 CRCM4 Earlier melt in CRCM4 Needs to be analyzed using Roy et al., 2010 method Comparison with snow model simulations to match MOD with decreasing snow thickness
RESULTS: Snow thickness = related to MOD? Snow thickness simulations: SNOWPACK – CLASS (corrected and non-corrected on CANGRD) QSCAT period for MOD
RESULTS: Snow thickness Snow thickness simulations: SNOWPACK – CLASS (corrected and non-corrected on CANGRD)
RESULTS: Snow thickness vs MOD CRCM4 MOD corresponds well to decreasing SNOWPACK thickness and met observ. Systematically longer snow cover in CRCM4, although earlier MOD No spatial trend No temporal trend within stations, approach pixel by pixel needed
RESULTS: Snow thickness Comparison of: SNOWPACK vs meteorological observations SNOWPACK/CANGRD vs meteorological observations CRCM4 vs meteorological observations SNOWPACK improved using CANGRID 11/15 = 73% of cases CLASS improved using CANGRD 10/16 = 63% of cases CLASS more accurate than SNOWPACK 15/16
RESULTS: Snow thickness Large differences in MOD can be found following the method, hindrance to spatial/temporal analysis of melt Snow models simulations forced on CANGRD can be used to validate MOD from QSCAT products and CRCM4 simulations (well correlated to obs) Within the 2000-2009 period, no spatial/temporal trends were observed at our stations, period too short. ERS data between 1992-2000 to be added to the study.
NSERC, CSA, Environment Canada, Ouranos, MDDEP ACKNOWLEDGEMENTS THANK YOU!!! Funding: NSERC, CSA, Environment Canada, Ouranos, MDDEP