1. NOAA AMSR2 SNOW AND ICE PRODUCTS
Jeff Key
NOAA/NESDIS
Madison, Wisconsin USA

2. AMSR-2 Snow and Ice Products
Snow Cover (SC) – Presence/absence of snow
Snow Depth (SD) – The depth of snow on land
Snow Water Equivalent (SWE) – The amount of water in the snowpack
Sea Ice Characterization (SIC) –
Ice concentration (area fraction in a pixel)
Ice type or Age class (first-year or multiyear ice)
Snow and ice algorithms are built around heritage products with a few low-risk improvements.
All products are now operational (September 2016 for snow; March 2017 for ice).
SWE = depth (m) x density (kg/m3) (units: kg/m2)
SWE = depth (m) x density (kg/m3) / density of water (kg/m3) (units: m)

3. NOAA AMSR2 SNOW PRODUCTS
Yong-Keun Lee1, Cezar Kongoli2, Jeff Key3
1Cooperative Institute for Meteorological Satellite Studies (CIMSS), University of Wisconsin-Madison
2Cooperative Institute for Climate Studies (CICS), University of Maryland
3NOAA/NESDIS

4. Snow Algorithms Snow cover: Grody (1991) SSM/I algorithm
Lee, K.-L., C. Kongoli, and J. Key, 2015, An in-depth evaluation of heritage algorithms for snow cover and snow depth using AMSR-E and AMSR2 measurements, J. Atmos. Oceanic Tech., 32, , doi: /JTECH-D
Snow cover: Grody (1991) SSM/I algorithm
Snow depth and SWE: NASA AMSR-E algorithm (Kelly, 2009; Tedesco and Narvekar, 2010)

5. Product Examples: Snow Cover
The Snow Cover product provides the presence/absence of snow cover for every pixel.
Snow cover on January 15, 2015

6. Product Examples: Snow Depth
The Snow Depth product provides the depth of the snow cover (cm).
Snow depth (cm) on January 15, 2015

7. Snow water equivalent (kg/m2) on January 15, 2015
Product Examples: SWE
The Snow Water Equivalent (SWE) product provides the water equivalent (mm) of the snow cover.
Snow water equivalent (kg/m2) on January 15, 2015
SWE = depth (m) x density (kg/m3) (units: kg/m2)
SWE = depth (m) x density (kg/m3) / density of water (kg/m3) (units: m)

8. Snow Cover Validation
With wet snow
Wet snow excluded
The overall accuracy is calculated as the number of pixels where both AMSR2 and IMS detect snow or no-snow divided by the whole number. The whole number is the sum of the number of pixels that are reported as snow or no snow by both AMSR2 and IMS. Snow detection rate is calculated as the number of pixels where both AMSR2 and IMS detect snow divided by the number of pixels where IMS detects snow. Omission error is calculated as the number of pixels where AMSR2 misses snow divided by the whole number. Commission error is calculated as the number of pixels where AMSR2 incorrectly detects snow divided by the whole number. In this study, the term, ‘‘false alarm’’ (commission error) is used to indicate a situation where IMS reports no snow and AMSR2 reports snow.
If wet snow is not included, detection accuracy is higher.
Snow classes are from Sturm et al. (1995)

9. Statistics are computed daily, automatically
November
December
overall accuracy (overall): the number of pixels where both AMSR2 and IMS detect snow or no-snow divided by the whole number => The whole number is the sum of numbers of pixels that are reported as snow or no snow by both AMSR2 and IMS
snow detection rate (snow detect): the number of pixels where both AMSR2 and IMS detect snow divided by the number of pixels where IMS detects snow
omission error (omission): the number of pixels where AMSR2 misses snow (compared to IMS) divided by the whole number
commission error (commission): the number of pixels where AMSR2 incorrectly detects snow divided by the whole number
January
February

10. Snow Depth Validation By elevation By forest fraction Tundra Taiga
Maritime
Ephemeral
Prairie
Alpine
Bias (cm)
4.51
3.77
-5.34
6.05
2.75
-4.45
RMSE (cm)
18.77
20.96
19.37
14.95
18.93
21.97
Mean (cm) of in-situ obs
25.10
19.18
20.20
8.40
18.49
25.14

11. Snow Water Equivalent (SWE) Validation
SWE comparison between AMSR2 retrievals and GHCN:
When 10 < AMSR2 SWE < 100 and 10 < GHCN SWE < 100 and the altitude < 3000m:
bias std rmse mean mean number of pixels
When AMSR2 SWE > 100 and GHCN SWE > 100 and the altitude < 3000m:
Units are mm
mean1: average of AMSR2 SWE
mean2: average of GHCN SWE
bias: mean of AMSR2 SWE - GHCN SWE
GHCN: Global Historical Climatology Network

12. Error Budget Summary Attribute Analyzed Threshold Requirement
Validation Result
Error Summary
Snow cover
80% prob of correct snow/no-snow classification
72-97% correct classification
If wet snow is excluded,
> 90% correct
Snow depth
20 cm snow depth uncertainty
15-22 cm depth uncertainty
If alpine excluded, depth uncertainty < 20 cm
SWE
50-70% uncertainty (shallow to thick snowpacks)
20-60%
Larger validation dataset would improve reliability of results. More thin snowpack cases needed.
NOAA’s requirements are in the extra slides.

13. AMSR2 SEA ICE CHARACTERIZATION
Walt Meier1, Scott Stewart2, and Ludovic Brucker3
1National Snow and Ice Data Center (NSIDC; formerly NASA GSFC)
Cooperative Institute for Research in the Environmental Sciences
University of Colorado, Boulder
2NSIDC contractor
3NASA Goddard Space Flight Center

14. Sea Ice Algorithms
Meier, W.N., J.S. Stewart, Y. Liu, J. Key, and J. Miller, 2017, Operational implementation of sea ice concentration estimates from the AMSR2 sensor, IEEE J. Selected Topics Appl. Earth Obs. Remote Sens. (J-STARS), 10(9), , doi: /JSTARS
Ice concentration is essentially the NASA Team 2 algorithm with the Bootstrap algorithm as secondary

15. AMSR2 Sea Ice Concentration Examples
Examples of AMSR2 sea ice concentration over the Arctic (left) and Antarctic (right) on 9 November 2017.

16. Validation
Comparison of AMSR2 (left) and VIIRS (below) sea ice concentration over the Arctic on 31 January 2015.
Numerous validation studies have been done on BT and NT2 algorithms via comparisons with aircraft and other satellite (vis/IR, SAR) imagery
e.g., Cavalieri et al., 2006; Meier, 2005; Comiso et al., 1997; Comiso and Nishio, 2008; Andersen et al., 2007; Ivanova et al., 2014
Concentration errors for the central ice pack during cold, winter periods are <5%
Errors for melting ice, thin ice, and near the ice edge may be higher
Precision of the ice edge limited by spatial resolution of the channel with the largest footprint (IFOV), ~25 km for AMSR2
AIT test cases were validated through a comparison of the AMSR2 ice concentration results to S-NPP VIIRS sea ice concentration. Differences are expected due to:
Cloud cover – VIIRS retrievals are clear-sky only
Spatial resolution – VIIRS is < 1 km
Additional information on validation is in the notes section of this slide 16 16

17. Sea Ice Concentration Validation
Comparison of AMSR2 and VIIRS sea ice concentration over the Arctic on 31 January 2015.
(animation)

18. Sea Ice Concentration Validation
Comparison of AMSR2 and VIIRS sea ice concentration over the Antarctic on 31 January 2015.
(animation)

19. Sea Ice Concentration Validation: Arctic
Comparison of AMSR2 minus VIIRS ice concentrations for different AMSR2 ice concentration ranges/bins in the Arctic. Note that the y-axis range is different for "All", "90-100%", and the other plots. Data are from January to October 2016.

20. Sea Ice Concentration Validation: Antarctic
Same as previous slide except for the Antarctic.

21. Multiyear Ice Validation
Initial comparison with independent ice age fields (Lagrangian tracking of ice parcels) indicates good agreement in terms of spatial distribution of multi-year ice cover.

22. Ice Type Validation: Ice Charts
Comparison of NOAA vs. Canadian Ice Service (CIS) charts in high Arctic
Performance drops in May (melt onset)
Precision = (TruePositives) / (TruePositives + FalsePositives)
NOTE: Summer months are not included in plot.

23. Ice Type Validation: ASCAT
Comparison of NOAA vs. ASCAT scatterometer
Lower performance expected from ASCAT as well
Precision = (TruePositives) / (TruePositives + FalsePositives)
Performance drops in May
NOTE: Summer months are not included in plot.

24. Ice Type Validation: OSI-SAF Confusion Matrix results, 2012-2015
Average over all 3.5 years (Oct – Dec. 2015)
Mid-October through mid-April each year
OSI-SAF MYI
OSI-SAF no-MYI
NOAA MYI
28.1%
2.1%
NOAA no-MYI
4.8%
65.1%
Accuracy: 93.2 ± 2.3%
Precision: 84.5 ± 8.5%
NOAA agrees with OSISAF
(i.e., “correct” retrieval)

25. Error Budget Summary Attribute Analyzed Threshold Requirement
Validation Result
Error Summary
Concentration
10% uncertainty (see note)
1-4% accuracy
9-15% precision
Most errors well below 10% threshold, higher errors near ice edge
Ice type (MYI)
70% correct typing
80-90% (preliminary) during Arctic winter
Multiyear ice (MYI) detection only

26. Experimental Products
AMSR2/in situ blended snow depth and SWE
Blended Ice Concentration
Blended Ice motion

27. Blended AMSR2 + In Situ Snow Depth
A snow depth bias-correction method has been developed based on optimal interpolation of in situ measurements. The method is being refined and tested with the NOAA AMSR2 snow depth product using in situ data from the Global Historical Climatology Network (GNCN).
AMSR2 Snow Depth
Improved
SD areas
Blended AMSR2 + in situ

28. VIIRS Operational Cryosphere Products
Ice Concentration
Ice Surface Temperature
Ice Thickness/Age
Snow Cover (binary)
Snow Fraction

29. Blended Sea Ice Concentration
Blended sea ice concentration from passive Microwave and infrared/visible
Visible/IR ice concentration:
Pro: high spatial resolution
Con: clear-sky only
Passive microwave ice concentration:
Pro: all-weather
Con: low spatial resolution
Blended ice concentration:
high spatial resolution under all-weather conditions
Blended sea ice concentration at 1 km resolution on June 24, 2015 using AMSR-2 and the S-NPP VIIRS ice concentration products

30. AMSR2 and VIIRS Ice Motion
Sea ice motion is retrieved individually from AMSR2, the S-NPP VIIRS 11 micron band, and from the VIIRS Day-Night Band (DNB)

31. VIIRS+AMSR2 Ice Motion
Left: Combining AMSR2+VIIRS ice motion vectors creates output with high spatial resolution, full Arctic coverage
Right: Ice motion from AMSR2 alone.

32. Applications

33. National Ice Center Interactive Multisensor Snow and Ice Mapping System (IMS)

34. National Ice Center Interactive Multisensor Snow and Ice Mapping System (IMS)

35. NAVOCEANO Operational Sea Ice data for ACNFS/GOFS Assimilation
NAVO AMSR2 Ice Concentration
FNMOC SSMIS Ice Concentration
NIC Ice concentration
NAVO J-1 VIIRS Ice Concentration
FY19
Ice Thickness
NIC 4km IMS
NAVO NPP VIIRS Ice Concentration
FY18
NAVO AMSR2 Ice Concentration
FNMOC SSMIS Ice Concentration
NIC 4km IMS
NAVO NPP VIIRS Ice Concentration
JAXA AMSR2 Ice Concentration
FNMOC SSMIS Ice Concentration
NIC 4km IMS
Current
INTERACTIVE MULTISENSOR SNOW AND ICE MAPPING SYSTEM
test/evaluate the assimilation of ice thickness into GOFS by the end of FY19
Approved for Public Release; Distribution Unlimited

36. Sea ice concentration observations
Assimilation of the NPP VIIRS Sea Ice Concentration (SIC) Data for Arctic forecasts
VIIRS/AMSR2
AMSR2 only
AMSR2 SIC
VIIRS/AMSR2 SIC
7 July 2016
Sea ice concentration observations
Mean ice edge errors (km) between the observed and forecasts
Region
Assimilation without VIIRS data
Assimilation including VIIRS data
Pan-Arctic
45.8
33.4
Greenland
43.8
34.6
Barents
37.7
25.3
Laptev
64.8
51.6
Sea of Okhotsk
40.1
35.8
Bering/Beaufort
43.0
35.6
Canadian Arch
57.6
33.3
Distance Along the Cut (km)
Mean ice edge errors were using NAVO VIIRS ice
Average errors for the time period of Jan – Dec 2016.
Adding VIIRS SIC products into the operational sea ice forecast reduces ice edge error by an average of 25%
Approved for Public Release; Distribution Unlimited