thomas holzer-popp (dlr), stefan kinne (mpi-m) & the aerosol_cci team
DESCRIPTION
Thomas Holzer-Popp (DLR), Stefan Kinne (MPI-M) & the Aerosol_cci team. aerosol_cci. URD - sources. GCOS as baseline CMUG as model-oriented update Applications: model development, assimiation, decadal forecasting, trend monitoring AEROCOM as aerosol_cci CRG - PowerPoint PPT PresentationTRANSCRIPT
Thomas Holzer-Popp (DLR), Stefan Kinne (MPI-M)& the Aerosol_cci team
aerosol_cci
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 2
URD - sources
• GCOS as baseline
• CMUG as model-oriented update
Applications: model development, assimiation, decadal forecasting, trend monitoring
• AEROCOM as aerosol_cci CRG
Applications: process studies, trend monitoring
• MACC (currently added)
Application: assimilation (re-analysis)
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 3
GCOS requirements
• Aerosol optical depthgoal threshold
• accuracy 0.01 0.02• stability 0.005 / decade N/A• resolution 1 km / daily 10 km / weekly
• Other aerosol properties
• to supplement AOD
• e.g. single scattering albedo• accuracy 0.02• stability 0.015 / decade
• Comprehensive ground-based independent validation
-> can not be met (per pixel) by any satellite product
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 4
URD - CMUG
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 5
URD - requirements
• Overall user needs:
Easy availability (netCDF)
proven and documented quality
aerosol properties should identify aerosol species linked to source categories
observation of long-term trends (many years) for regions and globally
analysis of specific issues (absorption, aerosol above clouds, vertical profile, …)
prepare for easy and complete re-processing with new versions
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 6
URD - requirements
• Level2 products for data assimilation:
AOD at 4 wavelengths (440, 550, 670, 870 nm) and several layers
500 (1000) observations per hour
1 - 3 years covered
20 km ( 5 km) horizontal
3 km vertical
accuracy and precision 0.05 (0.02)
with pixel level uncertainty (random + systematic)
consistent with clouds and fire
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 7
URD - requirements
• Level3 products for process studies and trend monitoring:
AOD at 4 wavelengths
Angstrom coefficient (440-870), fine mode fraction (D<1μm), dust fraction
Absorption aerosol optical depth (or single scattering albedo)
aerosol vertical extinction / AOD profile (any information is valuable)
accuracy: combined absolute (low AOD) and relative (high AOD)
error characteristics
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 8
URD - requirements
derived stability need
regional AOD range (0.1 – 0.5) * 5% trend detection -> 0.005
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 9
URD - requirements
Satellite variable (reference dataset)
required RMS at superpixel level of 10x10 km2
required RMS at climate model grid level of 1°x1°
required RMS at regional level of 1000x1000 km2
Aerosol optical depth at 550nm and other wavelengths (Aeronet global dataset daily mean)
20% or 0.05 10% or 0.02 0.02
Fine mode fraction (Aeronet global dataset daily mean)
20% or 0.1 20% or 0.1 0.1
Dust fraction (Coarse fraction from Aeronet global dataset in known dusty conditions daily mean)
30% or 0.2 30% or 0.2 30% or 0.2
Absorption optical depth (Absorption optical depth computed from SSA and size of Aeronet daily mean)
20% or 0.05 20% or 0.02 0.02
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 10
URD - requirementsSatellite variable (reference dataset)
required monthly RMS month-to-month stability at regional level of 1000x1000 km2
required monthly RMS year-to-year stability at regional level of 1000x1000 km2
required monthly RMS decadal stability at regional level of 1000x1000 km2
Aerosol optical depth at 550nm and other wavelengths (Aeronet regional mean)
0.01 0.005 0.01
Fine mode fraction (Aeronet global dataset daily mean)
0.05 0.05 0.05
Dust fraction (Coarse fraction from Aeronet global dataset in known dusty conditions daily mean)
0.05 0.05 0.05
Absorption optical depth (Absorption optical depth computed from SSA and size of Aeronet daily mean)
0.002 0.005 0.005
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 11
URD - consistency
• With CMUG requirements
Same basic parameters
CMUG gives higher priority and detail to vertical information
CMUG adds depolarisation ratio
Aerosol_cci adds aerosol type properties
• horizontal / temporal resolution:
CMUG one dimension more demanding
aerosol_cci adds regional level
• accuracy
aerosol_cci links to horizontal resolution and adds relative criterium
aerosol_cci has lower goal for aerosol type variables
agreement on regional level for AOD, difference for AAOD
• stability
aerosol_cci defines target per temporal analysis grid
agreement on decadal level for AOD, difference for AAOD
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 12
URD - colocation
Done in URD
Done through discussions within AEROCOM
Iteration with CMUG, WMO-GAW SAG and other CCIs initiated, more iteration with AEROCOM to do
Is taken into account
Done through WMO-GAW SAG (membership of several aerosol_cci partners)
Template from ESA used and published on project website
Iteration with CMUG ongoing (March meeting)MACC input upgraded
Done in PSD
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 13
URD - evolution
Aerosol_cci URD_v1.3 accepted
Continuing iterations adding MACC data assimilation requirements (v1.4)
Further iteration with other ECVs to complete their requirements
Feeding aerosol_cci URD into WMO-GAW SAG aerosol input to GCOS
aim to harmonize between different requirements –> iteration with contributors
SAG chair: “The aerosol_cci document is far superior to anything else that I have seen
concerning requirements.” – including CMUG table
Apreaciates tying to applications (CMUG) and accuracy resolution-dependance (aerosol_cci)
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 14
URD - harmonization
SAG chair questions and recommendations
• overall process
• use the WMO RRR excel sheet to avoid missinterpretations and gaps
• add 1 line for each application / resolution
• questions
• can we harmonize application domains?
• can we harmonize / justify vertical layers (3km – 5km)?
• CMUG: Why 2 values for accuracy and precision?
• CCI: 1° x 1° : not equal area / shall we restrict requirements to level2 products?
• recommendations
• add duration, coverage (for process studies)
• add requirements for sub-orbital data (model/sat validation, assimilation)
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 15
PSD - overviewProduct name Parameter(s) sensors level comment
Tropospheric / total column products
Single-sensorAOD / type
Multi-spectral AODAerosol type probability
ATSR-2 / AATSRMERISPOLDER
Level2,3 Multi-spectral AOD depending on instrument capabilities.
Ångstrom coefficient can be derived from multi-spectral AOD.Aerosol type may include information on fine / coarse mode fraction and chemical components, which together best describe the observations
SynergeticAOD / type
Multi-spectral AODAerosol type probability
AATSR/SCIAMACHYATSR-2/GOMEAVHRR/GOME-2
Level2,3
AAI Absorbing aerosol indexaveraging kernel
OMISCIAMACHYGOME
Level2,3
Merged AOD / type Multi-spectral AODAerosol type probability
Combining several level2 with appropriate wighting
Level3
Aerosol type “climatology”
Aerosol type probability / dominant aerosol type
All AOD products Level3 Based on one year of data
Stratospheric products
Extinction Gridded extinction profile
GOMOS(SCIAMACHY)
Level3
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 16
PSD vs. URD (1)No. User requirement Product specification response
Product parameters
AOD at 4 wavelengths 4 AOD, depending on instrument capabilities
Ångström coefficient(based on AOD at 440 and 870 nm)
Implicit from 4 AODs where feasable
Fine mode fraction(based on 4 AOD)
Implicit from 4 AODs where feasable
(dominant) aerosol speciation or dust fraction
Implicit in aerosol type probability where feasable; focus of synergetic product
Absorption: AAOD or SSA Implicit in aerosol type probability where feasable; focus of AAI product
Extinction profile Not feasible with the sensor suite used
Numerical specifications
Accuracy To be established by validation
Precision To be established by validation
Horizontal resolution 10 x 10 km2 / 1° x 1°
Temporal resolution Daily (finer not possible with sensor suite used)
Stability To be established by validation
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 17
PSD vs. URD (2)
Data access
Easy availability To be taken into account in system design
Long-term record Only partial (2 years – options to extend further)
Error documentation(indicating source of error)
Will be available via web portal, with metadata of each product pointing to it
System specifications
Easy reprocessing To be taken into account in system design
Technical specifications
Format netCDF as required
Metadata CF1-4 convention as required
Grid lat-lon for level3 as required
Product levels Level3 as requiredLevel2 based on MACC need / assimilation
Anciliary parameters
Cloud fraction(also for pixels without AOD)
Missing value for pixels without AOD result
Local error Pixel error contained in products
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 18
Planned product use
Model development and inter-comparison -> AEROCOM
Data assimilation -> MACC (as far as possible)
Trend monitoring: only later (longer dataset needed)
options proposed for larger dataset processing
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 19
ECV consistency
Other ECV participants in Aerosol_cci algorithm workshops
cloud masking (cloud, fire, landcover, ozone)
surface treatment (oceancolour, landcover, fire)
aerosol model -> interest of other ECVs to get our recommendation / tech note)
Consistent 3-level cloud masks with cloud_cci
Need recommendations from oceancolour_cci and landcover_cci
for chlorophyll / sediments and BRDF auxiliary datasets + reference data
exchange of URDs between ECV
aerosol_cci URD distributed to CMUG and ECVs -> collect further requirements
aerosol_cci contributed to cloud_cci URD
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 20
Product uncertainties
Known uncertainty contributors
cloud screening
surface treatment
(assumed) aerosol optical properties
radiometric calibration
pixel size /atmospheric noise
radiative transfer / look-up tables
Characterization of uncertainty contributions
production of test datasets (1 month) with different algorithm versions
harmonization of critical modules for round robin – where possible
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 21
Product uncertainties
Error source Absolute error contribution comment
calibration TBD Can be critical
Radiative transfer negligible Error due to Mie calculations for dust TBD
Noise due to pixel size 0.05 Varies for different continents
Cloud fraction TBD Validation showed that pixels up to cloud fraction of 35% can be exploited for the retrieval
Surface reflectance Land: 0.05 – 0.15 (for visible surface reflectance from 0 – 0.25)
Ocean: 0.03 – 0.1 (for visible surface reflectance from 0 – 0.05)
Choice of wrong aerosol type (spectral extinction, absorption, phase function)
0 – 0.2 for AOD550 0. – 1. Is estimated for each pixel by identifying ambiguous aerosol types through comparing quality of fit with difference between aerosol types with similar spectra)
An example: synergetic retrieval algorithm (first estimation based on validation results)
Aerosol_cci > Thomas Holzer-Popp > ESA Living Planet Symposium, Bergen, 1 July 2010
slide 22
ECMWF data needs
These have already iterated with D. Tann -> aerosol_cci DARD
ECMWF CodeOutput field
Units Analysis or Forecast Model levels Model Grid Needed for
165 10m east wind component m s-1 Analysis Surface GG (N128) EO retrieval
16610m north wind
componentm s-1 Analysis Surface GG (N128) EO retrieval
Parameters from ERA Interim, Atmospheric model, Analysiso Requested analysis times: 0000, 0600, 1200, 1800 UTCo Dates: 01/01/1997 to 31/12/1997; 01/10/2008 to 31/12/2008o Requested representation: Lat/long grido Requested representation: 0.7 degreeo Requested area: Global
Grib number
Grib Abbreviation
Units Name
165 10U m s-1 10m east wind component
166 10V m s-1 10m north wind component