intercomparison of snow extent products from earth ...€¦ · modis nsidc (hall et al.) avhrr...
TRANSCRIPT
Intercomparison of Snow Extent Products from Earth Observation Data
Land Product Validation and Evolution Workshop 28 – 31 January 2014
Thomas Nagler, Elisabeth Ripper, Gabriele Bippus, Helmut Rott
Richard Fernandes
Kari Luojus
Sari Metsämäki
Dorothy Hall
David Robinson
Bojan Bojkov
FMI
Contents
• Motivation and Objectives
• Basics for detecting Snow from Optical Satellite Images
• Overview of Snow Extent Products
• Examples of Intercomparison Results
• Summary and Conclusions
• Overview of SnowPEX
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
Motivation and Objectives
3
OBJECTIVES: • Intercompare global / hemispheric (pre) operational snow
products derived from different EO sensors and generated by means of different algorithms, and assess the product quality.
• Prepare for the SNOWPEX study.
(Derksen & Brown 2012)
Time series of Northern Hemisphere June snow cover from NOAA snow chart CDR.
Rate of Snow Extent decrease: 1979-2012: -21.5% / decade
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
Detection of Snow with Optical Satellite Data
4
0.56 0.67 0.87 1.600
5
10
15
20
25
30
35
40
45
snow free< 25 %26-50 %51-75 %76-99 %100 %
Wavelength [mm]
TOA Reflectance [%]
Unforested areas, Baltic Region, 29/3/2011
VIS / NIR together with 1.6 µm Band provides the essential information for detecting snow and for FSC estimation
Thomas Nagler
Spectral Signatures from AATSR Data
Continental to Hemispheric Satellite Snow Extent Products
5
Name Product type
Pixel Spaci
ng
Frequency
Period Main Sensor
Organisation
NOAA IMS
Binary 4 km daily 2004 -
OPT, PMW, + manual
NOAA (Helfrich et al.)
NOAA IMS
Binary 24 km daily 1997 -2004
OPT, PMW
NOAA
GlobSnow
Fractional
1 km daily-
monthly
1996-2012
ATSR2AATSR
SYKE et al.
MOD10 Fractional
0.5 km Daily 2000 -
MODIS NSIDC (Hall et al.)
AVHRR Pathfinder
Binary 5 km daily 1992 -2004
AVHRR CCRS (Zhao, et al)
CryoLand
Fractional
(Europe)
0.5 km daily 2000 - MODIS ENVEO / SYKE et al.
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
Processing Line for SE Product Intercomparison
6
Fractional Snow Map at 0.01 deg Pixel size
Exclude areas with:Snow free in both products, no-data, clouds, open water
Pixel 2 Pixel Comparison
Aggregate binary or fractional SE to FSC at 0.01 deg
• Number of snow pixels• RMSD• Bias• Correlation Coefficient• Scatterplot• Standard Deviation
Fractional SE Map(Binary SE Map)
GlobCover surface mask
Cloud mask
Reference Product
Aggregate to FSC at 0.01 deg
VHR + HR EO
data
MS-Unmixing methodTopographic corrections
Separate FSC for forest/non-forest/plain areas/mountains
Product
Map of Mean Absolute FSC Difference
Aggregate to FSC at 0.01 deg
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
Maximum Snow Extent 1-7 March 2010
7
MOD10
GlobSnow
IMS
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
MOD10 (~500m)
CryoLand (~500m)
Maximum Snow Extent 1-7 March 2010
GlobSnow (~1 km)
IMS (~4 km)
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
MOD10 (~500m)
CryoLand (~500m)
Maximum Snow Extent 1-7 March 2010
GlobSnow (~1 km)
IMS (~4 km)
IMS (~4km)
Mean Absolute FSC Difference MOD10 versus GlobSnow - 1.3.-31.5.2010
number of pixels (N)
unforestedforest
(Σ|FSCMOD10 – FSCGlobSnow|)/N
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
Mean Absolute FSC DifferenceMOD10 versus CryoLand / MODIS 1.3.-31.5.2010
no forestforest (Σ|FSCMOD10 – FSCCryoLand|)/N
Number of pixels
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
Pixel by Pixel Intercomparison of MOD10 versus GlobSNOW FSC product
1228-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
Synergistic use of Snow Products from different sources
13
28-30 Jan 2014 LPVE WS / ESRIN Thomas Nagler
1
IMS merged with SSM/I and MODIS snow extent
Products Indicating Snow
1: All2: MODIS, SSM/I3: MODIS, IMS4: MODIS5: SSM/I, IMS6: SSM/I7: IMS
10: Permanent ice11: Snow free land
2 3 4 5 6 7 10 11
February 6, 2006
Time series of weekly Max. Snow Extent March – May 2010 PanEuropean Domain
14
16/ October 2012 Sentinel-3 OLCI SLSTR Preparatory WS
0
5
10
1520
25
30
35
4045
50
55
60
12 13 14 15 16 17 18 19 20 21 22 23 24
%
week of year 2010
0
5
10
1520
25
30
3540
45
50
5560
12 13 14 15 16 17 18 19 20 21 22 23 24
%
week of year 2010
0
5
10
1520
25
30
35
4045
50
55
60
12 13 14 15 16 17 18 19 20 21 22 23 24
%
week of year 2010
MOD10
GlobSnow
CryoLand
IMS
0
5
10
1520
25
30
3540
45
50
5560
12 13 14 15 16 17 18 19 20 21 22 23 24
%
week of year 2010
IMS
% o
f to
tal a
rea
% o
f to
tal a
rea
% o
f to
tal a
rea
% o
f tot
al a
rea
Summary and Conclusions
15
• A limited data set of different snow extent products from various optical satellite data and applying different algorithms from has been analyzed and reveals systematic differences in the snow extent between the products.
• FSC differences of the various products show a typical pattern with significant differences in forested areas, and less pronounced differences in high mountain regions, but more detailed analysis is needed.
• A short time series of weekly snow extent maps from different sensors has been generated revealing differences in the snow coverage and trend of snow extent, but longer time series need to be analyzed. On the hand it indicates the need for intercalibration of snow extent products from different sensors and algorithms for studying long term trends in the snow extent.
• The synergistic use of snow information from different sensors (optical, microwave) has a high potential to improve the quality of snow cover monitoring. Thomas Nagler
SNOWPEXSatellite Snow Products Intercomparison & Evaluation Excercise An ESA initiative contributing to WMO Global Cryosphere Watch and WCRP CLiC
carried out under the lead of ENVEO in cooperation with FMI, SYKE, EC and CCRS.
US National Ice Center / NOAA/ NESDIS, Cryospheric Processes and RS Laboratory, City University of NY
Global Snow Lab, Rudgers University Cryospheric Sciences Laboratory, NASA
and associated international contributors
SnowPEX Objectives
The primary objectives are
• Intercompare and evaluate global / hemispheric (pre) operational snow products derived from different EO sensors and generated by means of different algorithms, assessing the product quality by objective means.
• Evaluate and intercompare temporal trends of seasonal snow parameters from various EO based products in order to achieve well-founded uncertainty estimates for climate change monitoring.
• Elaborate recommendations and needs for further improvements in monitoring seasonal snow parameters from EO data.
SnowPEX aims to bring together scientists and institutions of seasonal snow pack monitoring for assessing the quality of current satellite-based snow products derived from EO data, and working out guidelines for improvement.
The project will support the setup of a consolidated operational satellite snow observation system for the Global Cryosphere Watch Initiative of the WMO and help to improve the snow cover data base for climate monitoring, as addressed by the WCRP-CliC programme
Main Tasks within SnowPEX
• Review of Algorithms and products focusing on Snow Extent (SE) and Snow Water Equivalent (SWE)
• Definition of Protocols and methods for intercomparison of Snow Extent and Snow Water Equivalent products
• Compilation of global reference data sets for quality assessment of SE and SWE products
• Intercomparison of SE / SWE products from various institutions and quality assessment against reference data base
• Hemispheric/Global SE and Snow Mass Trend Analysis and potential synergy of SE and SWE products
• Conclusion and recommendations for satellite snow monitoring
SNOWPEX will organize 2 International Workshops for Snow Product Intercomparison (IWSPI) with the aim to bring together all major providers of continental to global snow products. The 1st IWSPI is planned to be held in June/July 2014.
19
Pixel by Pixel Intercomparison of MOD10 versus CryoLand/Modis FSC product
20
SNOWPEXSatellite Snow Products Intercomparison &Evaluation Excercise An ESA initiative in response to the recommendations of the WMO Global Cryosphere
Watch WS Toronto, January 2013
carried out under the lead of ENVEO in cooperation with FMI, SYKE, EC and CCRS.
OBJECTIVESThe project aims to bring together scientists and institutions monitoring
seasonal snow for improving the quality of snow products from EO data.
The project will intercompare and evaluate global to continental (pre-) operational snow products (snow extent; SWE) from different sensors
and algorithms for improving the observation of the spatial and temporal variability of the seasonal snow pack.
Planned Project Start: Q1 / 2014
Us National Ice Center, Cryospheric Processes and RS Laboratory, NYC
Interactive Multisensor Snow and Ice Mapping System (IMS)
22
Thomas Nagler
Interactive Multisensor Snow and Ice Mapping System (IMS)
Satellites
GOES (E & W)
MeteoSat (MSG & 7)
MTSAT
NOAA Automated Snow & Ice
AVHRR (Channels 1 & 3)
MODIS (Channel 8)
ASCAT
AMSU (Derived snow, ice, rain)
Other Sources
Radar
Models
Surface Observations
Webcams
Buoys
Charts
Pre-Processing
Indirect Sources
Analyst Derived Output
February 6, 2006
4 km & 24 km Northern Hemisphere Analysis
Snow & Ice Cover Produced daily at U.S. National
Ice Center
GlobSnow AATSR Snow Extent Product
23
• SE product (1 km pixel) is based on: ATSR-2 (1995) & AATSR (2002)
• Hierarchical Cloud Detection Alg. using (A)ATSR Bands only (SCDA developed at SYKE)
• SE Mapping Procedure• Masking of snow free areas (0.67, 1.6 µm) • Fractional Snow Extent:• SCAmod Alg. (0.67µm). (Metsämäki et al, 2005).
Monthly Snow Extent Product (May 2003)
24/5/200824-25/5/2008
24-26/5/200824-29/5/2008
MODIS Fractional Snow Product
24
Thomas Nagler
Fractional Snow Cover (FSC) is estimated by a regression relationship using the NDSI:
FSC = -0.01 + 1.45 * NDSI
where NDSI = (band 4* – band 6*) / (band 4 + band 6)
FSC is calculated for the Normalized Difference Snow Index (range 0 – 1.0):
Quality Maps:Several screens are the applied to a FSC pixel; if a screen is failed then the pixel is set to a “not snow” value and/or a quality assessment (QA) flag is set. Results of the screens are stored in a QA data layer in the product. The purpose of these screens is to alleviate snow commission errors.
Maximum Snow Extent 1-7 April 2010
MOD10_ (~500m)
CryoLand (~500m)
GlobSnow (~1 km)
IMS (~4 km)
Maximum Snow Extent 15-22 May 2010
MOD10_ (~500m)
CryoLand (~500m)
GlobSnow (~1 km)
IMS (~4 km)
CryoLand MODIS Fractional Snow Extent Product
27
Thomas Nagler
Processing:
• Input: MODIS data as input
• Binary snow classification using NDSI with spatially and seasonal variable threshold map
• FSC calculation applies SCAmod algorithm using transmissivity map and predefined reflectance thresholds for forest canopy, bare ground and wet snow
• Covers Pan-European Domain