Transcript The NOAA/NESDIS - ESA Data User Element

The NOAA/NESDIS/STAR Long Term
Strategy of Hyper Spectral Fundamental
Climate Data Records and
Environmental Climate Variables
Antonia Gambacorta, Chris Barnet, Walter Wolf,
Eric Maddy, Thomas King, Murty Divakarla,
Mitch Goldberg
NOAA/NESDIS/STAR
Camp Springs, MD, USA
1
Outline
 Part I. The NOAA/NESDIS/STAR long term
hyper spectral strategy for climate
applications
 Part II. Temperature & water vapor retrieval
products: characteristics and validation results
2
Part I
The NOAA/NESDIS long term hyper
spectral strategy for climate
applications
3
NOAA/NESDIS/STAR long term
hyper spectral sounding strategy
AQUA
1:30pm orbit
(May 4th, 2002)
NPP
METOP
1:30pm orbit
9:30am orbit, (October 19th 2006)
(October 25th, 2011)
2002
•
AIRS/AMSU retrieval algorithm into operations
2006 - 2016
•
Migrate the AIRS/AMSU retrieval algorithm into operations for METOP IASI/AMSU/MHS phase
A, B and C (2006, 2011, 2016)
•
Metop A IASI operations approved on June 18th 2008; operational since August 14th 2008
•
Currently studying differences between instruments (AIRS, IASI and CrIS in simulation)
2010 - 2025
•
Migrate the AIRS/IASI algorithm into operations for NPP ( October 25th, 2011 ) & NPOESS
(~2013,~2018) CrIS/ATMS/VIIRS (NOAA NDE program)
The NOAA/NESDIS/STAR processing system is a modular
design compatible with multiple instruments
AIRS
IASI
CrIS
Retrieval
Code
Retrieval
products
Diagnostic tools
ECMWF
NCEP
SONDES
ATOVS
A long term uniform data record of atmospheric variables - cloud
cleared radiances, temperature, water vapor, trace gases - by employing:
 the same retrieval algorithm
 the same underlying radiative transfer spectroscopy (UMBC SARTA)
 a modular design for re-processing
Sequence of Steps of the
Retrieval Algorithm

1) A microwave retrieval module which derives cloud liquid water flags
and microwave surface emissivity uncertainty;

2) A fast eigenvector regression retrieval for temperature and moisture
that is trained using the ECMWF analysis and observed cloudy radiances;

3) A cloud clearing module that uses a set of microwave and IR channels
to produce the cloud-cleared IR radiance product and reject those cases
violating the cloud-clearing requirements;

4) A fast eigenvector regression retrieval for temperature and moisture
that is trained using the ECMWF analysis and IASI cloud cleared
radiances;

5) The final IR retrieval module, which uses the regression retrieval as an
initial solution and produces the final version of the physical retrieval by
an iterated regularized least squared minimization.

We start with the temperature retrieval, because temperature is the most
linear component of the RTA equation, followed by water vapor, ozone,
6
etc.
Climate quality temperature and water vapor
products require compelling instrument
Specifications

Need High Spectral Resolution & Spectral Purity
» Ability to isolate spectral features → vertical resolution.
» Ability to minimize sensitivity to interference signals.

Need Excellent Instrument Noise & Instrument Stability
» Low NEΔT is required.
» Minimal systematic effects (scan angle polarization,
day/night orbital effects, etc.)

Need Large Spectral Coverage (multiple bands) & High Sampling
» Increases the number of unique pieces of information.
» Ability to remove cloud and aerosol effects.
Dedicated efforts in selecting retrieval channel lists

Need high accuracy in the retrieval of temperature, moisture and surface parameters
» Extracting small signals from noisy data
» Errors in the retrieval of temperature, water vapor and surface parameters are of the same order
of magnitude of climate trend signal
Dedicated efforts in validating temperature & moisture products
7
Instrument comparison (I):
Information content of AIRS, IASI and CrIS
AIRS, IASI and CrIS have comparable information content (first 500
eigenvalues shown for brevity. Global ensembles used for the SVD)
 IASI highest information content
 CrIS in simulation: reason for higher information content wrt AIRS
Instrument comparison (II):
AIRS, IASI and CrIS instrument noise
• Comparable Longwave band noise – IASI shortwave band noise highest
• Channels used in retrieval scheme carefully selected to avoid high NEDT
NOAA/NESDIS channel selection
methodology
Sensitivity to:
Water vapor 10% perturbation
1k temperature perturbation
1K SST perturbation
10% ozone perturbation
2% methane perturbation
Dashed vertical lines indicate selected water vapor channels
• Core water vapor lines are selected based on low NEDT and low interference from other
atmospheric components. From spectroscopy, the interference signals of other species (methane,
ozone, etc.) are well known and are used as off-diagonal terms of the noise covariance matrix
Retrieval error assessment: this kind of error sensitivity analysis allows to
discriminate correlation sources in within the retrieval scheme from natural signals.
10
Instrument comparison (III):
Water vapor channel selection used in the NOAA/NESDIS
retrieval algorithm for AIRS, IASI and CrIS
AIRS version 5
IASI version 5
CrIS version 1
AIRS version 6
candidate water
vapor channel list
improvement
Improvement work
in progress
11
• Low water vapor sensitivity in the lower stratosphere (feature common to all latitudes)
Instrument Line Shape distortion in presence of FOV in-homogeneities:
What’s error introduced in the radiance measurement?
 Nominal Geometric Centroids
obtained from the IASI
Instrument Point Source
Function (IPSF)
 Radiometric Centroids obtained
from the IIS measurement and
the IASI IPSF
 Spatial inhomogeneities
introduce a shift between the
geometric and radiometric
centroids
 The higher the spatial
inhomogeneity, the largest the
radiometric centroid shift
IIS Imager (64x64 pixels) and
IASI FOVs (black contour)
12
ILS dependence on radiometric
centroid shift ( )


~ FOV 

Radiometric
centroid shift
Geometric
centroid angular
position


simplified case of an ideal
monochromatic source
In general a non uniform light
source introduces a distortion
in the pixel ILS. The frequency
shift  of the peak is the
dominant effect.
This frequency shift is a source
of error in the radiance
spectrum that we try to quantify
(next slides).
13
IASI Instrument Line Shift Effects
in presence of non uniform scenes
FOV Radiometric Centroid shift
MEAN = 10-6 rad
SIGMA = 0.1 mrad
NEDT
Band 1
Error becomes
noticeable at 3 sigma
shift
NEDT
Band 2
Error becomes
noticeable at 3 sigma
shift
NEDT
Band 3
Error becomes
noticeable at 2 sigma
shift
Error induced is negligible except for very rare cases (that can be flagged out)
Part II
The NOAA/NESDIS Temperature and
Water Vapor retrieval products:
characteristics and validation results
15
NOAA/NESDIS water vapor and temperature
key features for climate applications

Top of atmosphere to surface vertical retrievals of temperature and water vapor
»

Cloud-clearing algorithm allows for spatial uniformity
»

no clear-sky bias typical of IR hyper spectral sensors.
Vertical resolution
»
»

The full vertical extent of the retrieval profiles allows to perform a complete study of water
vapor’s sensitivity to temperature variations in the upper, middle and lower troposphere.
High vertical resolution is fundamental to capture the vertical structure in the climate trends
of temperature and water vapor
AIRS v5 tropospheric temperature (moisture) retrieval resolution, as determined by the fullwidth-half-maximum of the averaging kernels, ranges between ~2.5km (3km) near the
surface and 6km (4km) near the tropopause (Ref.: Maddy and Barnet, 2007)
High retrieval accuracy
» Extensively validated retrieval algorithm (see ahead)
(Ref: Fetzer et al., 2003; Divakarla et al., 2006; Tobin et al., 2006; Fetzer et al., 2008;
Gambacorta et al., 2008, etc.)
A robust temperature and water vapor data set
for climate applications
16
IASI Temperature and Water Vapor vs ECMWF
Temperature
Water Vapor
RMS
statistics
Global Ensemble – Ocean Night Only – Ocean Night Clear Only
Solid: physical retrieval; Dashed: First guess
• Results consistently improve for both T and q due to better characterized surface
17
emissivity and no cloud clearing errors; physical retrieval improves from first guess
IASI Temperature and Water Vapor vs ECMWF
Temperature
Water Vapor
SDV
statistics
Global Ensemble – Ocean Night Only – Ocean Night Clear Only
Solid: physical retrieval; Dashed: First guess
• Results consistently improve for both T and q due to better characterized surface
18
emissivity and no cloud clearing errors; physical retrieval improves from first guess
IASI Temperature and Water Vapor vs ECMWF
Temperature
Water Vapor
BIAS
statistics
Global Ensemble – Ocean Night Only – Ocean Night Clear Only
Solid: physical retrieval; Dashed: First guess
• Ocean night Clear only water vapor mid-trop bias degradation due to low
statistics (case by case study not shown)
19
IASI validation vs ECMWF
by latitudinal regions
Temperature
Water Vapor
RMS
statistics
Polar region – Mid latitude region – Tropics
Solid: physical retrieval; Dashed: Night Ocean Only
• Results consistently improve from high to low latitudes:
20
 Surface emissivity better characterized; sharper vertical temperature gradient; less uniform cloud formations
IASI validation vs ECMWF
by latitudinal regions
Temperature
Water Vapor
SDV
statistics
Polar region – Mid latitude region – Tropics
Solid: physical retrieval; Dashed: Night Ocean Only
• Results consistently improve from high to low latitudes:
21
 Surface emissivity better characterized; sharper vertical temperature gradient; less uniform cloud formations
IASI validation vs ECMWF
by latitudinal regions
Temperature
Water Vapor
BIAS
statistics
Polar region – Mid latitude region – Tropics
Solid: physical retrieval; Dashed: Night Ocean Only
• Results consistently improve from high to low latitudes:
22
 Surface emissivity better characterized; sharper vertical temperature gradient; less uniform cloud formations
RAOBs Validation
courtesy of Murty Divakarla
[email protected]

Bias and RMS difference Statistics with Reference to RAOBs.
» Statistics for MetOp-IASI (9:30AM/PM)
– Acceptance Criteria: Mid-Troposphere Temp Flag = 0
– Solid Lines for IASI (9:30 AM/PM)
l
l
l
l
l
RAOB
RAOB
RAOB
RAOB
RAOB
vs. IASI Final - Physical Retrievals
vs. IASI Fast Regression (FG)
vs. MetOp-ATOVS Retrievals
vs. ECMWF Forecast
vs. NCEP-GFS Analysis/Forecast Fields
» Similar Statistics for Aqua-AIRS (1:30 PM/AM)
– Dotted Lines
l
Similar Color Convention
– Acceptance Criteria: Mid-Troposphere Temp Flag = 0
– (I do have comparison with V5 QA also)
» These Figures are only a few selected ones. Go to our website to
see RMS and Bias Plots for
– Any region (tropics, midlat, polar, global)
– Any category: land/sea; day/night etc.
http://www.orbit2.nesdis.noaa.gov/smcd/spb/iosspdt/iosspdt.php#1 23
IASI- T(p), q(p) RMS Difference
Global (L+S+Coast) NSAMP=12,666 Yield: 55%
Acceptance Criteria: Mid-Troposphere Temp Flag = 0
100
Pressure (hPa)
Pressure (hPa)
100
1000
1000
0
0.5
1
1.5
2
2.5
3
0
Temperature RMS Difference (K)
IASI_PR
IASI_FG
NCEP_GFS
ECMWF
M02_ATOVS
20
40
60
80
100
Water Vapor RMS Difference (%)
IASI_PR
IASI_FG
NCEP_GFS
ECMWF
M02_ATOVS
RAOB vs. IASI AVN ATOVS ECMWF FG
24
AIRS- T(p), q(p) RMS Difference
Global (L+S+Coast) NSAMP=5,993, Yield: 36%
Acceptance Criteria: Mid-Troposphere Temp Flag = 0
100
Pressure (hPa)
Pressure (hPa)
100
1000
1000
0
0.5
1
1.5
2
2.5
3
Temperature RMS Difference (K)
AIRS_PR
AIRS_FG
NCEP_GFS
ECMWF
N18_ATOVS
0
20
40
60
80
100
Water Vapor RMS Difference (%)
AIRS_PR
AIRS_FG
NCEP_GFS
ECMWF
N18_ATOVS
RAOB vs. AIRS AVN ATOVS ECMWF FG
Dotted Lines : AIRS
25
IASI- T(p), q(p) RMS Difference
Sea Samples NSAMP=2,848
100
Pressure (hPa)
Pressure (hPa)
100
1000
1000
0
0.5
1
1.5
2
2.5
3
0
Temperature RMS Difference (K)
IASI_PR
IASI_FG
NCEP_GFS
ECMWF
M02_ATOVS
20
40
60
80
100
Water Vapor RMS Difference (%)
IASI_PR
IASI_FG
NCEP_GFS
ECMWF
M02_ATOVS
RAOB vs. IASI AVN ATOVS ECMWF FG
26
AIRS - T(p), q(p) RMS Difference
Sea Samples NSAMP=540
100
Pressure (hPa)
Pressure (hPa)
100
1000
1000
0
0.5
1
1.5
2
2.5
3
Temperature RMS Difference (K)
AIRS_PR
AIRS_FG
NCEP_GFS
ECMWF
N18_ATOVS
0
20
40
60
80
100
Water Vapor RMS Difference (%)
AIRS_PR
AIRS_FG
NCEP_GFS
ECMWF
N18_ATOVS
RAOB vs. AIRS AVN ATOVS ECMWF FG
27
IASI Validation Summary
 Globally (except high latitude regions)
» Temperature rms: ~1.5 K (lower troposphere) to ~1k (upper troposphere)
» Temperature bias: <+/-0.5
» Water vapor rms: <10% (lower troposphere) to 25% (upper troposphere)
(vs ECMWF)
» Water vapor bias: < 10% (lower troposphere) to +/- 3% (upper troposphere)
(vs ECMWF)
 AIRS shows comparable results, 10% degraded in the mid-upper
tropospheric water vapor profile (improvements are underway).
 Ultimate target: meet the GCOS climate variables
specification requirements globally
28
Conlcusions
 NOAA/NESDIS temperature and water vapor
characteristics:

Top of atmosphere to surface vertical retrievals of temperature and water vapor

Cloud-clearing algorithm
»

Error Sensitivity analysis
»

High vertical resolution is fundamental to capture the vertical structure in the dynamics and
climate trends of temperature and water vapor
Re-processing capability
»

allows to discriminate between correlation sources in within the retrieval scheme from natural
signals.
Vertical resolution
»

allows for spatial uniformity no clear-sky bias typical of other IR hyper spectral data bases.
Allows to keep data sets updated with latest optimization of the retrieval algorithm
High retrieval accuracy
»
The NOAA/NESDIS temperature and water vapor products show promising results towards
29
fully meeting the GCOS climate variables specification requirements almost globally.
30