p - GlobAlbedo

www.GlobAlbedo.org
The ESA GlobAlbedo p
project:
j
An Overview
Jan-Peter Muller, Tom Kennedy,
Mullard Space Science Laboratory, UCL
P.Lewis, UCL Geography
Jürgen Fischer,
Fischer Freie Universität Berlin
Pete North, Swansea University
Uwe Krämer,, Brockmann Consult
9th February’10
1
www.GlobAlbedo.org
Scientific Context
•
Land
L
d surface-Atmosphere
f
A
h
iinteractions
i
iimportant to climate
li
prediction,
di i
especially related to Land Parameterisation Schemes for hydrology
•
Albedo required for
– Earth’s surface energy balance
– Parameterisation for Numerical Weather Prediction, climate
monitoring and climate impact assessment
– Modelling of soil and vegetation in land ecosystems
•
Surface albedo includes both snow and ice-free as well as snow/ice
surface cover
•
Surface
S
rface albedo pla
plays
s a role in radiati
radiative
e forcing thro
through
gh changes in land
cover providing a cooling effect and in the role of black Carbon (soot) to
provide a new source of warming
9th February’10
2
www.GlobAlbedo.org
Overall Aims
•
Production of a 15 year record (1995-2010)
(1995 2010) of Land Surface Broadband Albedo
(BBA) from European space assets to provide an independent capability to
generate this Essential Climate Variable (ECV) to be continued into the future
from 2013 with SENTINEL-3
•
BBA will be produced at broadband (0.4-0.7µm, 0.7-3µm and 0.4-3µm)
according to “white-sky/BHR” and “black-sky/DHR” formulations for snow-free
(excluding permanent snow/ice) and including
including_snow
snow on
– an 8-day time-step with a rolling 16-day window length at 1km on a SIN
equal area
– A fixed
fi d monthly
thl titime-step
t on a Plate
Pl t Carrée
C é equall angle
l map projection
j ti att
30 arc-seconds, 0.05 degrees and 0.5 degrees
•
Input
p data will consist of level 1b ((radiometrically
y calibrated,, satellite projection)
p j
)
– ATSR2 (6/1995-12/2008), MERIS and AATSR (6/2002-12/2010)
– VEGETATION (24.3.98-31.1.03) and VEGETATION2 (1.2.03-present)
9th February’10
3
www.GlobAlbedo.org
Overall Aims (2)
•
With each BBA map, a set of bit masks and metadata will be supplied which will relate to
the quality, data source and provenance of the data
•
An estimated uncertainty (variance-covariance matrix) will be produced for each and
every pixel which will be derived from end-to-end estimates radiance sensor intercalibration (using CEOS-GEO QA4EO protocols) through aerosol correction through BBA
•
User requirements specify that there is a need for NWP forecasters to have a
– BBA>0.15; 10% and for
– BBA<0.15; 0.015
•
GlobAlbedo data products will be available via anonymous ftp, via http and via an OGCcompliant web-GIS service based on the CEOS-WGISS ICEDS
•
Plan to collaborate with G. Leptoukh (NASA-GSFC) on inter-operability with GIOVANNI
•
Subsetting of products will be built upon the CEOS Cal/Val portal, MERCI server
developed by Brockmann Consult. Both GUI and programmable interface
•
All code will be written in Java for p
platform independence
p
using
g a cloud computing
p
g
paradigm. All code will be made publicly available as part of the open source BEAM
•
Products will be publicly available from 11/2011
•
Crystal Schaaf
Schaaf, BU; Gabriela Schaepmann-Strub
Schaepmann Strub (U of Zurich) and Nigel Fox (NPL) are
consultants
9th February’10
4
www.GlobAlbedo.org
User inputs
•
Users play
U
l a critical
i i l role
l iin the
h d
development
l
off b
baseline
li requirements
i
and the assessment of whether these products are “fit for purpose”
•
These users include:
– Jean-Louis Roujean (Météo France)
– Edouard Davin/Sonia Isabelle Seneviratne ((ETH Zürich, CH))
– Alexander Loew (MPI Hamburg, D)
– Gunnar Myhre (CICERO, Norway)
– Wolfgang Knorr (QUEST, UK)
– Chris Taylor (CEH Wallingford, UK)
– Samantha Pullen (UK Met Office/Hadley Climate Centre)
•
All users have significant experience with MODIS albedo products
9th February’10
5
www.GlobAlbedo.org
U C
Use
Case E
Example:
l APPLICATION FOR WEATHER FORECAST
(J-L. Roujean, KO Meeting, 5/11/09)
Weather forcast model: ALADIN (~9.5km)
Two experiments: with ALADIN albedo and with Land SAF albedo analysis
Run every day at 00h: 20070201->20070731 (54h forcast)
Surface albedo 15022007 00UTC+12
LandSAF-reference
no assimilation
of
snow/ice pixels
2m temperature 15022007 00UTC+12
LandSAF-reference
Land SAF albedo
is lower than
ALADIN albedo
lower Land SAF albedo
induced warmer
atmospheric T2m
=> ALADIN model indicates a significant cold bias in winter, reduced thanks to sat. obs.
[ Cedilnik,
Cedilnik Carrer,
Carrer Roujean and Mahfouf
Mahfouf, “Analysis
Analysis of satellite derived surface albedo for numerical weather prediction”
prediction ,
to be submitted ]
9th February’10
6
www.GlobAlbedo.org
How will these objectives be achieved?
•
UCL-MSSL will provide overall project leadership (management and scientific)
•
Swansea University (SU)
(SU), Pete North
North, to provide ATBDs and processing for
SDR/SBRDFs from level 1-b input data for ATSR2, AATSR, VEGETATION
(SPOT-4) and VEGETATION2 (SPOT-5)
•
Freie Universität Berlin (FUB)
(FUB), Jürgen Fischer
Fischer, to provide ATBDs and
processing for SDR/SBRDFs from level 1-b input data for MERIS
•
UCL-Geog, P. Lewis to provide ATBDs for SDR-to-BBA retrieval and gap-filling
•
A common optimal estimation system will be developed by SU for atmospheric
correction so that error estimates can be provided on a per SDR pixel basis
•
Code for a distributed processing facility will be provided by Brockmann consult
including production quality code (Java) to run on linux clusters based on
ATBDs provided by UCL, SU and FUB
•
UCL MSSL to process from spectral SDRs to Broadband SDRs to BBA as well
UCL-MSSL
as provide archiving, visualisation, ordering, web-GIS and distribution facility
•
A validation system will be established by Brockmann Consult at UCL-MSSL
based on the MERCI system to allow initially the GlobAlbedo project team and
their consultants and users access to both SDR/SBRDFs and BBA
9th February’10
7
www.GlobAlbedo.org
Legacy - ESA MERIS AlbedoMap
2002-2007 16-day albedo products for MERIS spectral
bands based on MODIS collection 4 BRDFs
9th February’10
8
www.GlobAlbedo.org
Example MERIS broadband albedo* product: DoY 193
(16-day time period : 12-27.7.03): Shortwave
* White-sky
N.B. Greyy areas have missing
g data due to cloud cover
9th February’10
9
www.GlobAlbedo.org
E
Example
l MERIS albedo
lb d bit
bit-maskk product:
d t DoY
D Y 193 (16
(16-day
d titime period
i d : 12
1227.7.03): Source MOD43C2 BRDF_QA
Key.
y Yellow=Full Inversion, Red=Magnitude
g
inversion, White=NO retrieval, Black=NO sun
N.B. Large areas of missing data due to cloud cover
9th February’10
10
www.GlobAlbedo.org
MERIS AlbedoMap Processing Overview: heritage for GlobAlbedo
MERIS SDR
BRDF
Correction
MERIS L1
Smile
Correction
Cloud and Snow
Screening
Cloud Shadows
& Cloud Edges
g
Aerosol
Correction
Land / Water
Reclassification
Albedo
Calculation
Map Projection
MERIS Albedo
Level 3
9th February’10
11
www.GlobAlbedo.org
GlobAlbedo
L1 → SDR Processor
Peter North, Andreas Heckel
SU
Luis Guanter, Rene Preusker, Jürgen Fischer
FUB
9th February’10
12
www.GlobAlbedo.org
Background – GlobAlbedo L1 → SDR Processors
• {MERIS/AATSR/VGT} L1  SDR Processors -- Modular approach in
GlobAlbedo:
Pixel classification (cloud screening)  AOT retrieval  SDR retrieval
• GlobAlbedo requirements on SDR retrieval:
• AOD mustt be
b estimated
ti t d from
f
the
th data
d t to
t be
b processed
d for
f SDR retrieval.
ti
l
• Avoid use of a priori estimates of SDR since this may bias subsequent
albedo retrieval.
• Provide per-pixel error estimates of SDR after propagation of
AOD/instrument/RT errors.
products must be validated.
• AOD and SDR intermediate p
• Single atmospheric RT model (MOMO) to be used in the processing of the
three instruments.
9th February’10
13
www.GlobAlbedo.org
Atmospheric RT simulations: MOMO Radiative Transfer Code
•
Based on doubling
doubling-adding
adding
•
Atmosphere and ocean fully coupled
•
Isotropic and non-isotropic surface reflection
•
Gas absorption through k-distribution technique
•
any spectral and vertical resolution
•
Computationally efficient
•
Thoroughly tested and validated
9th February’10
14
www.GlobAlbedo.org
Schematic of SDR algorithm
Auxiliary data:
O3, H2O, Hbar,
Initial gguess:
AOD at 550nm: a
Aerosol model: Ma
AATSR LTOA
(8 channels)
Iterative
inversion
Atmos correct using LUT:
Options:
F(LTOA, Ma , a , O3, H2O, Hbar,S, V, R, )
-> RSURF(,S, V, R, )
Test fit with surface model:
Optimise a and Ma to
minimise EMOD
8
*
E MOD   w i (Ri, SURF  Ri,MOD
)2
(ii) Spectral model
(VGT, MERIS)
i1
N
EMOD  ETHR ?
Y
(iii) Angular model
((AATSR)
S )
Output: RSURF, a aerosol model
Ma & associated uncertainties
9th February’10
15
www.GlobAlbedo.org
V lid ti
Validation:
MERIS AOD Retrieval
R ti
l - Comparison
C
i
with
ith AERONET data
d t
Guanter et al.,
RSE, 2008
9th February’10
16
www.GlobAlbedo.org
Validation – MERIS SDR Retrieval
Comparison with Inland Water targets
Guanter et al., RSE, 2010
9th February’10
17
www.GlobAlbedo.org
GlobAlbedo
Narrowband-to-Broadband Coefficients
Luis Guanter, Rene Preusker, Jürgen Fischer
FUB
9th February’10
18
www.GlobAlbedo.org
Calculation of Narrowband-to-Broadband conversion coefficients
3 spectral ranges:
4 types of spectral integral (SDR/albedo):
• PAR: ∆λ = [0.4, 0.7] µm
• Apparent: Ed=Ed_dir+Ed_dif
• NIR:
µm
∆λ = [0.7, 3.0]
• Inherent: Ed=Esc
• SW:
µm
∆λ = [0.4, 3.0]
y Ed=Ed_dir
• Black-sky:
• White-sky: Ed=Ed_dif
9th February’10
19
www.GlobAlbedo.org
BroadBand Albedo (BBA) retrieval and gap-filling
P. Lewis, UCL Geography
9th February’10
20
www.GlobAlbedo.org
BBDR-BB model parameters
– Fit ‘observations’ of broad band directional reflectance to BRF
model
• Include constraints
– Linear kernel driven model
R v ,s  fiso,   fvol,  K vol v ,s  fgeo,  K geo v ,s
– To obtain estimates of parameters
f
iso

(), f vol (), f geo ()
– For  = VIS, NIR, SW
9th February’10
21
www.GlobAlbedo.org
Model parameter estimation
•
Minimise discrepancy
1
1
1
e 2    i  
̂i  Cobs,i
̂i   n pi  Csmooth,i
i  
 n pi  pi  pp̂i  C prior,i
pi  pp̂i 
T
T
i
T
i
i
1
ˆ i  Cobs,i
Observations :   i  
i  ˆ i 
T
i
1
smoothness :   n pi  Csmooth,i
 n pi 
T
i
11
priors
i :  pi  pˆ i  C prior,i
pi  pˆ i 
T
i
9th February’10
22
www.GlobAlbedo.org
Gap filling
•
Use off prior
U
i obviates
b i
need
d for
f gap filling
filli
– Provided have ‘full’ specification of prior
• i.e. no gaps
– And good estimate of uncertainty of prior
– Possible that some sites never seen by MODIS?
• Certainly some … e.g. Arctic in winter
– … but can’t produce albedo for this …
•
Temporal smoothness also reduces need for gap filling
– Uncertainty simply increases when there are gaps
– Use double exponential
p
((Laplace
p
fn)) in time
• After Roujean et al.
• Instead of ‘full’ smoothness constraint/Kalman filter
– Issue:
I
definition
d fi iti off decay
d
constant?
t t?
9th February’10
23
www.GlobAlbedo.org
Background albedo
•
Most reliable
M
li bl current estimates
i
for
f priors
i
– MODIS albedo: 2000-2010
– Characteristics:
• Global 500m, gridded
• Every 8 days (16 day window)
• Snow flags
• ‘full’ retrievals
– RTLiSpR inversion
in ersion
• ‘backup’ retrievals
– ‘magnitude inversion’
•
Issues
– How to combine full and backup retrievals?
– How to define temporal weighting?
9th February’10
24
www.GlobAlbedo.org
Number of samples
•
Over 10 year period
period, no
no. of samples will vary
– May be zero for full inversion in places
Full only
Scaled
0-10
Full and Backup
9th February’10
25
www.GlobAlbedo.org
Only Snow: all samples
9th February’10
26
www.GlobAlbedo.org
No Snow: all samples
9th February’10
27
www.GlobAlbedo.org
With Snow: all samples
9th February’10
28
www.GlobAlbedo.org
GlobAlbedo Work Logic
Start: 11/5/09
End: 11/5/12
9th February’10
29
www.GlobAlbedo.org
Sensor inter-calibration using QA4EO
•
•
The production of BroadBand Albedo
(BBA) requires calibration of individual
spectral reflectance values and its
transformation from narrowband to
broadband
Usually this is performed using radiative
transfer simulations
QuickTime™ and a
decompressor
are needed to see this picture.
QuickTime™ and a
decompressor
are needed to see this picture.
Cao, CEOS-30
•
Dome-C
Dome
C (available from Dec
Dec-Jan
Jan each
Comparison of Hyperion EO-1 spectra with
year) can be used for Simultaneous
SeaWiFS.MODIS & AVHRR spectral responsivities
Nadir Observations (SNO) between
satellites with similar spectral bands (e.g.
ATSR2 cff AATSR,
AATSR VGT1 cff VGT2)
•
Hyperspectral ToA radiances can also
be used to transform narrow-tobroadband
•
QuickTime™ and a
decompressor
are needed to see this picture.
Link via NPL consultant to ESA/NOAA
sponsored CEOS intercalibration
g DOME-C
exercises and QA4EO using
SeaWiFS SDR and Nadir-corrected BRDF
9th February’10
30
www.GlobAlbedo.org
Validation approach
Jan-Peter Muller, UCL-MSSL
9th February’10
31
www.GlobAlbedo.org
Validation approach
•
SDRs will be compared for a few validation sites to ensure that inter-calibration has been performed
correctly in collaboration with end users and “friends of GlobAlbedo”.
GlobAlbedo” Will employ CEOS QA4EO process
which will be documented with advice from CEOS-IVOS at NPL
•
Three-point difference statistics will be produced for 16-day products using
– GlobAlbedo
– MODIS
– MISR (to-be-computed from L2AS) 16-day products
– METEOSAT daily products
•
Monthly composites to be used for intercomparisons with
– MISR 0.5º “true monthly” level-3 product
– PARASOL monthly products
•
Intercomparison with in situ data albedometer measurements from SURFRAD and BSRN (collaboration
with Boston University, BU) and from field measurements (University of Zurich, UZ)
•
Inter-comparisons with airborne measurements, where available in collaboration with BU
9th February’10
32
SEVIRI Spectral albedo products
0 6 µm
0.6
0 8 µm
0.8
August
30, 2005
1.6 µm
§ Spatial Resolution: 3km at the Sub
Sub--Satellite Point
§ Projection: native MSG/SEVIRI Projection
§ Products: Spectral & BB for DHR (noon) & BHR
§ Production Frequency: Daily
§ Effective Temporal Resolution: 5 Days
§ Format: HDF5
§ Timeliness: 3 hours
§ Dissemination: EUMETCast, project website
Courtesy of J-L Roujean, Météo-France
www.GlobAlbedo.org
COMPARISON WITH MODIS ALBEDO (1/2)
[0.4µm,0.7µm]
16-day average MSG Albedo
–
MODIS albedo projected
on SEVIRI grid
[0.7µm,4µm]
[0 3µm 4µm]
[0.3µm,4µm]
10-25 of June 2006
Courtesy of J-L Roujean, Météo-France
9th February’10
34
www.GlobAlbedo.org
COMPARISON WITH MODIS ALBEDO (2/2)
12-27 of July 2006
C t
Courtesy
off J-L
JLR
Roujean,
j
Mété
Météo-France
F
9th February’10
35
www.GlobAlbedo.org
ALBEDO TIME SERIES (snowfall episodes)
Modis SW-BH
SW BH albedo
MSG albedo
(boreal forest)
ground measurements
(mixed shrub/tree)
Courtesy of J-L
J L Roujean
Roujean, Météo-France
Météo France
9th February’10
36
www.GlobAlbedo.org
Scientific Exploitation
J-P. Muller, UCL-MSSL
9th February’10
37
www.GlobAlbedo.org
Scientific Exploitation: Overview
•
Four different
F
diff
collaborators
ll b
iinvited
i d to participate
i i
on the
h ffollowing
ll i
general topics:
– assessment of the effects of surface albedo change
g ((including
g land
cover change, biomass burning, variations in snow cover, etc.) on
radiative forcing of climate, either between natural or pre-industrial
versus current conditions
conditions, or over the time pan of the GlobAlbedo
data set,
– DGVM or GCM model assimilation and/or verification, for the
purposes off investigating
i
ti ti th
the effects
ff t off improved
i
d description
d
i ti off the
th
surface energy balance, or of the vegetation phenology,
– Development
p
and testing
g of albedo p
parameterisations
9th February’10
38
www.GlobAlbedo.org
Scientific Exploitation
•
CICERO (Gunnar Myhre, N) for an assessment of the effects of surface albedo
change on radiative forcing of climate (in association with J-P Muller, UCL-MSSL)
•
QUEST (Wolfgang Knorr, UK) for an investigation of the effects of an improved
description of the surface energy balance (in association with P Lewis, UCL-Geo)
•
g ((Alexander Loeuw,, D)) in association with J. Fischer ((FUB)) for
MPI Hamburg
– Assessment of the impact of the new albedo data set on the radiative forcing and
energy budget within ECHAM/JSBACH at the global scale
– Identify regions with large albedo changes and assess the feedback of land surface
albedo dynamics on climate model simulations in those regions
•
CEH (Chris Taylor, UK) in association with P. North (SU) for
– Comparison of existing GCM albedo with new GlobAlbedo values
– Estimation of PFT and background albedo for GCM parameterisation
– Evaluation of impacts of revised albedos on GCM forecasts
9th February’10
39