1. GlobSnow Snow Extent and SWE datasets
Kari Luojus, Jouni Pulliainen, Matias Takala, Juha Lemmetyinen Tuomo Smolander, Jaakko Ikonen, Juval Cohen
(Finnish Meteorological Institute),
Sari Metsämäki (Finnish Environment Institute),
Chris Derksen (Environment Canada), T. Nagler (ENVEO) R. Solberg (NR), A. Wiesmann (GAMMA),S. Wunderle (Univ. Bern), F. Fontana (MeteoSwiss), E. Malnes (Norut), W. Schoener (ZAMG)
+ Bojan Bojkov (EUMETSAT) & Simon Pinnock (ESA)
13 June 2016 – Columbus, Ohio, USA
WMO GCW “Snow Watch 2” workshop
Finnish Meteorological Institute

2. Outline ESA DUE GlobSnow-2 Snow Water Equivalent (SWE) dataset
Snow Extent (SE) dataset
Future evolution
SWE & Snow Extent
Finnish Meteorological Institute

3. ESA DUE GlobSnow
ESA DUE GlobSnow project: Production of novel hemispherical snow extent (SE) and snow water equivalent (SWE) climate data records.
Generation of long time-series employing FMI supercomputing facilities at Helsinki (daily, weekly and monthly maps of SE and SWE for northern hemisphere) + NRT processing
Consortium members: Finnish Meteorological Institute (FMI) with ENVEO IT GmbH (Austria), GAMMA Remote Sensing (Switzerland), Norwegian Computing Center, Finnish Environment Institute (SYKE), and Environment Canada (EC). + Univ. Bern, MeteoSwiss, ZAMG & Norut
GlobSnow-1 (3.5 years): 11/2008 – 11/2011 (36 months)
GlobSnow-2 (2 years): 05/2012 – 09/2014 (24 months)
Details and products available at

4. 35 year-long CDR time-series on snow conditions of Northern Hemisphere
First time reliable daily spatial information on SWE (snow cover):
Snow Water Equivalent (SWE)
Snow Extent and melt (+grain size)
25 km resolution (EASE-grid)
Time-series for
Passive microwave radiometer data combined with ground-based synoptic snow observations
Variational data-assimilation
Available at open data archive (
Daily NRT production since 2010
=> Methodology moved to EUMETSAT H-SAF
Takala, M., Luojus, K., Pulliainen, J., Derksen, C., Lemmetyinen, J., Kärnä, J.-P, Koskinen, J., Bojkov, B., “Estimating northern hemisphere snow water equivalent for climate research through assimilation of spaceborne radiometer data and ground-based measurements”, Remote Sensing of Environment, Vol. 115, Issue 12, 15 December 2011, doi: /j.rse

5. GlobSnow SWE time series (FCDR)
Northern Hemisphere
1979 to 1987 (SMMR)
1988 to 2013 (SSM/I, SSMIS)
FPS v to 2011 (AMSR-E)
Daily, weekly, monthly products
Includes error estimates (statistical std of the SWE estimate in mm)
Data format HDF4 & NetCDF CF
EASE-Grid projection (~25km resolution)
snow grain data available as well
Glaciers, mountains & Ice Sheets masked out
Versions: 1.0; 1.3 and 2.0 (current)

6. Distributed validation data, e.g. Northern Eurasia & Canada
Snow Survey data (from the former USSR and Russia)
There are 517 snow path stations with data for (1979 – 2009)
Manual ground-based measurements on snow depth/SWE
1 - 2km snow transects, measurements every 100m - 200m
Land Cover
Reference Dataset
Year n
Mean SWE (mm)
Tundra
Intensive Sites; SnowSTAR 2007 28
120
Northern Boreal
EC Snow Surveys
105
135
SWE <150 mm 73
134
Southern Boreal 57 75
BERMS Towers
468 70
Prairie 41 44
Finnish Meteorological Institute

7. SWE retrieval (data assimilation vs. channel diff.)
Density scatterplot (assimilated vs. satellite only SWE)
Russian INTAS SCCONE SWE transect data as reference

8. SWE product intercomparison (ESA SnowPEx): GlobSnow-2, NASA Standard, NASA Prototype SWE

9. GlobSnow-2 SWE vs. RIHMI WDC (2002-2011)
38197 Coinciding samples of GlobSnow, NASA Standard and NASA prototype SWE Evaluations for the samples available in all 3 products!
”Blended product” = combines satellite and ground-based WS-data

10. NASA Standard SWE vs. RIHMI WDC (2002-2011)
38197 Coinciding samples of GlobSnow, NASA Standard and NASA prototype SWE Evaluations for the samples available in all 3 products!
”Pure satellite product” = utilizes only satellite data for retrieval

11. NASA Prototype SWE vs. RIHMI WDC (2002-2011)
38197 Coinciding samples of GlobSnow, NASA Standard and NASA prototype SWE Evaluations for the samples available in all 3 products!
”Pure satellite product” = utilizes only satellite data for retrieval

12. Density scatterplots SWE vs. ref
Density scatterplots SWE vs. ref. data (ESA SnowPEx): GlobSnow-2, NASA Standard, NASA Prototype SWE

13. SWE analysis on a monthly basis, RMSE
Differences increase towards the end of the snow accumulation season

14. Conclusions (satellite-based SWE)
NASA Pro
GlobSnow
NASA Std
Observed under-estimation of NASA SWE products due to high negative bias with deep snow. Total snow estimates of GlobSnow for NH are more accurate.

15. Summary – All SWE products
Assessed for an uniform time period, ranked by retrieval performance (RMSE)
Time period , Russian snow course data as reference
Datasets
Dataset availability
Retrieval performance
(RMSE)
Bias
Retrieval performance (RMSE)
Bias
GlobSnow v2.0
42.6 mm
-3.8 mm
44.9 mm
-4.3 mm
CROCUS-Era-Interim
45.8 mm
+1.1 mm
48.0 mm
+4.7 mm
GLDAS2.0-Noah
-8.4 mm
49.5 mm
-10.8 mm
MERRA (Standard)
54.9 mm
+12.9 mm
57.9 mm
+15.2 mm
ERA-Interim (ERA-Land)
67.3 mm
+35.4 mm
74.7 mm
+42.4 mm
NASA Standard
67.4 mm
-24.3 mm -
NASA Prototype
72.4 mm
-19.9 mm

16. Long term consistency of GlobSnow SWE FPS
RMS error and retrieval bias calculated independently for each year
Reference data: snow transects from Russia (INTAS-SCCONE)
SWE<150 mm

17. GlobSnow Snow Extent (SE) dataset
17 years SE data record has been produced using optical imagery from ESA ATSR-2 (1995-) and AATSR (2002-) on a hemispherical scale. NPP VIIRS from 2012-
SYKE’s SCAmod method for fractional snow cover mapping implemented for Northern hemisphere
Cloud detection algorithm developed by SYKE (+ contributed by ENVEO, FMI & NR)
Methodology developed especially for forested regions – basically a tough challenge for optical SE retrieval
Uncertainty estimate provided for each grid cell, data available as NetCDF CF
Operational data production at the Finnish Meteorological Institute (FMI)
Metsämäki, S., Mattila, O.-P., Pulliainen, J., Niemi, K., Luojus, K., Böttcher, K. "An optical reflectance model-based method for fractional snow cover mapping applicable to continental scale", Remote Sensing of Environment, Vol. 123, August 2012, pp , doi: /j.rse

18. Daily, weekly and monthly products
Optical data ~ 1km spatial resolution
10 April 2003
3 March 2010
1 March 2010
19 April 2003
April 2003
March 2010
Finnish Meteorological Institute

19. SCAmod retrieval algorithm
Based on reflectance model where forest canopy effect into the observed reflectance is
compensated through pre-determined canopy transmissivity
Designed to provide FSC in forested areas, and overall well applicable globally over any terrain
A single band approach - Applicable to optical wavelengths and to a variety of sensors
Surface scattering
Volume scattering
r λ,obs(FSC) observed reflectance from unit area
r λ, snow wet snow reflectance
r λ ,ground snow-free ground reflectance
r λ ,forest forest canopy reflectance
forest canopy transmissivity for unit area
FSC fraction of snow covered area

20. Mean Absolute FSC Difference MOD10 versus GlobSnow - 1.3.-31.5.2010
(S|FSCMOD10 – FSCGlobSnow|)/N
forest
unforested
number of pixels (N)
Thomas Nagler / ENVEO

21. GlobSnow-2, Suomi NPP VIIRS SE product
~1km spatial resolution, daily hemispherical coverage
10 April 2003
3 March 2010
1 March 2010
19 April 2003
April 2003
March 2010
Finnish Meteorological Institute

22. Higher resolution retrieval
Future evolution
SWE
Higher resolution retrieval
Augmenting SWE retrieval using SE information
Snow Extent
Sentinel-3 OLCI & SLSTR
Extending time series back from 1995 (1980?)
Finnish Meteorological Institute

23. Super resolution (5km) Pan-European SWE product
Super resolution 5km SWE
SWE retrieval based on enhanced GlobSnow approach
Fractional snow cover information used in SWE retrieval for:
improved snow detection during winter snow accumulation period
Improved melt detection during spring
Utilization of optical (VIIRS-based) FSC-data in combination with NOAA IMS (4km) product
Processing for Pan-European domain in NRT since 2015
Example of super resolution SWE product

24. Super resolution (5km) Pan-European SWE product
5km resolution with a new processing approach. New Aux data, including an improved forest stem volume map
Improved handling of snow grain size and variance in emission model, artefacts filtered from proc. chain
Masking of snow extent area using IMS/VIIRS data
RMSE similar to GS-2, but output shows added spatial information with the high resolution SWE retrieval
Applying the 5km SWE data for hydrologial modelling, the model output is improved, compared using GS-2 data
Even as radiometer resolution is ~25km, we are able to achieve sub-pixel resolution on the level of 5km without losing SWE retrieval sensitivity
Shows improved results when SWE data applied for hydrological modeling

25. Super resolution (5km) Pan-European SWE product
Future Evolution Potential
Long term ECV with 5km resolution?? (ESA Snow CCI)?
Utilization of NSIDC super-resolution SSMIS Tb data for 5km retrieval?

26. Combination of SE & SWE products for the generation of concise snow cover information
NRT north-hemisphere daily snow monitoring product combining GlobSnow SWE and SE products (based on SSMI-S and VIIRS+Sentinel-3)
SE information from VIIRS & NOAA IMS applied for (NRT) SWE production
Finnish Meteorological Institute

28. Conclusions
ESA GlobSnow produced NH snow extent (SE) and snow water equivalent (SWE) climate data records (17 and 35 years of snow cover information)
The NRT processing system still online! info & data available at SWE NRT production moved under EUMETSAT HSAF initiative!
Future evolution:
Combination of SE and SWE information (for improved SWE time series)
Moving to 5km resolution with SWE
Future plans (ESA Snow CCI+ ? or a similar initiative)
Improved snow emission model for (SWE)
Improved handling of forests and lake rich regions (SWE)
Dynamic snow grain size (SWE)
Sentinel-3 (SE)
Improved cloud masking & FSC retrieval (SE)
Extended time series (from 1980s to present day?)

29. Impact of Radiometer Derived Information
Difference between final assimilated SWE and background SWE from interpolated synoptic weather station data.

30. SWE Retrieval ‘Saturation’ (PMW signal) “what if we would not apply the synop data?”
Chang et al. EC
GlobSnow