Download - TH3.TO4.3.ppt
Jul 28 2011
IMPLEMENTING HEMISPHERICAL SNOW WATER EQUIVALENT PRODUCT ASSIMILATING WEATHER STATION OBSERVATIONS AND SPACEBORNE MICROWAVE DATA
M. Takala, K. Luojus, J. Pulliainen, C. Derksen, J. Lemmetyinen, J-P. Kärnä, J. Koskinen, B. Bojkov
Jul 28 2011
Introduction• Properties of snow cover (SCA, SWE, SD, melt) are
important in investigating hydrological, climatological, and greenhouse gas processes (such as CO2 and CH4)
• In this work a time series of SWE for 30 years has been produced
• The algorithm used is based on data assimilation (Pulliainen 2006) and integrates data of snow clearance (Takala et al. 2009) and auxiliary data (forest coverage etc.)
• The results show significant improvement to traditional algorithms which are based on using either spaceborne derived estimates or interpolated values only
Jul 28 2011
Principle of SWE algorithm I• Weather station snow depth data is obtained from
European Centre for Medium-range Weather Forecasts (ECMWF) and kriging interpolated over the area in question -> SWE estimate & SWE Var estimate
• Spaceborne radiometer data is obtained from National Snow and Ice Data Center (NSIDC). Data is either SMMR, SSM/I or AMSR-E.
• Snow grain size (and variance) is estimated using SD data and HUT Snow model for SD station locations. Values are interpolated over area under investigation.
• From spaceborne data estimates of the SWE are obtained using inversion of HUT model.
Jul 28 2011
Principle of SWE algorithm II• If snow is dry: weighing different data sources
applying their respective statistics an assimilated SWE is estimated
• If snow is wet: only kriging interpolated data is used
• To correctly track down new snow a cumulative dry snow mask has been used
• To correctly track down snow melt snow clearance date product has been integrated to SWE system
• The final product is SWE and SWE variance map of whole Northern Hemisphere in EASE Grid
Jul 28 2011
Principle of SWE algorithm III• Example of snow
clearance date product for year 2008
• Time series of 30 years available from author
• For details see Takala et al. 2009
Jul 28 2011
Example of SWE product
Jul 28 2011
SWE algorithm assesment I• Difference
between assimilated SWE estimate and kriging interpolation only fields
• Weather stations marked in yellow
Jul 28 2011
SWE algorithm assesment II
• Histogram of difference between assimilated SWE result and kriging interpolated background field
• Typically increases accuracy in areas with sparse SD data
Jul 28 2011
SWE sensitivity I• Density scatterplot
• Ground truth data is INTAS SCCONE SWE path data
Jul 28 2011
SWE sensitivity II
Jul 28 2011
SWE sensitivity III
0
20
40
60
80
100
120
140
160
180
305 317 329 341 353 365 12 24 36 48 60 72 84 96 108 120
Day of Year 2005/06
SW
E (
mm
)
Old Jack Pine
1975 Harvest
1994 Harvest
2002 Harvest
GlobSnow SWE V0.9.2
0
20
40
60
80
100
120
140
160
180
305 317 329 341 353 365 12 24 36 48 60 72 84 96 108 120
Day of Year 2006/07
SW
E (
mm
)
Old Jack Pine
1975 Harvest
1994 Harvest
2002 Harvest
GlobSnow SWE V0.9.2
0
20
40
60
80
100
120
140
160
180
305 317 329 341 353 365 12 24 36 48 60 72 84 96 108 120
Day of Year 2007/08
SW
E (
mm
)
Old Jack Pine
1975 Harvest
1994 Harvest
2002 Harvest
GlobSnow SWE V0.9.2
Jul 28 2011
SWE Animation
Jul 28 2011
Thanks for your attention!• SWE data freely
available at
www.globsnow.info
• Manuscript has been submitted to a peer reviewed journal