ozone cci experience on producing and assessing harmonised...

Post on 06-Jun-2020

2 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Ozone_cci experience on producing and assessing harmonised long-

term ECV series

7th CCI Collocation meeting, ESA/ESRIN, Frascati, 4-6 October 2016

M. Van Roozendael, BIRA-IASB on behalf of the Ozone_cci consortium

Outline

•  GCOS requirements for ozone

•  Building consistent and stable multi-sensor data products

Ø  The total ozone case

•  Assessing drift and biases on existing multi-sensor data sets

Ø  The limb ozone profile case

•  Summary and conclusion

GCOS requirements for ozone

Target Requirements

Variable/ Parameter Horizontal Resolution

Vertical Resolution

Temporal Resolution Accuracy Stability

Ozone profile in upper stratosphere and mesosphere

100-200 km 3 km Daily 5-20 % < 1%

Ozone profile in upper troposphere and lower stratosphere

100-200 km 1-2 km 4h 10 % 1%

Total ozone 20-50 km N/A 4h Max (2%; 5 DU) < 1%

Tropospheric ozone 20-50 km 5km 4h 10-15% 1%

GCOS-154, Dec 2011

Total ozone case •  Requirement on stability: < 1%/decade •  Sensors: GOME, SCIAMACHY, GOME-2, OMI, OMPS, …

Nadir UV hyperspectral spectrometers, all belonging to the same family of instruments but with different design and different individual performances

•  Approach adopted for CCI: 1.  Use single retrieval baseline applicable to all sensors à direct-

fitting scheme selected based on Round-Robin conducted ahead of CCI programme

2.  Apply level-1 corrections on individual sensors before level-2 processing (soft-calibrations)

3.  Further adjust (small) inter-sensor residual bias before merging

Level-1 issues

•  Different sensors treated with identical retrieval baseline sometimes (still) show significant discrepancies

•  Problem is sensor-dependent (largest for SCIAMACHY) •  Related to level-1 errors •  Two types of radiance errors

•  Broadband

•  Spectral features •  Erroneous key data (e.g. ISRF) •  All can affect retrievals

à Need for more corrections

Dynamical adjustment of ISRF •  ISRF (slit function) is a critical parameter for trace gas retrieval •  Measured in the lab (pre-launch), but can evolve in-flight due to

instrumental instabilities •  Dynamical adjustment as part of the level-2 based on analysis of

solar lines width, impact ozone product at percent level

Dikty and Richter, 2011

Soft-calibration of level-1 (1) •  Principle: use simulated radiances over well characterised

reference sites to identify systematic radiance errors

Lerot et al., 2014

Soft-calibration of level-1 (2)

•  Examples of level-1 systematic errors identified and corrected in a level-2 pre-processing step

GOME radiance degradation

East Nadir West

SCIAMACHY spectral features

Lerot et al., 2014

GOME, SCIAMACHY, GOME-2 reprocessing

•  Full reprocessing of 3 sensors show excellent consistency with independent NASA SBUV v8.6 data set

•  Still small (~1%) residual biases between SCIAMACHY and other sensors at high latitudes

•  OMI recently added to the picture, shows excellent stability in time without soft-calibration need (à new reference)

•  Phase-2 level-2 reprocessing currently ongoing

Data merging step

•  Additional empirical small corrections applied by reference to one sensor (GOME)

•  Merging to create on single multi-sensor level-3 data sets (GTO-ECV CCI)

Coldewey–Egbers et al., 2015

Koukouli et al., 2015

Chiou et al., 2014

Final assessment (validation)

Limb ozone profile case •  Stability requirements in stratosphere: < 1%/decade •  Sensors: MIPAS, SCIAMACHY, GOMOS, OSIRIS, SMR,

ACE, MLS Large variety of instrument designs and measurement concepts (UV, TIR, MW), limb scattering, limb emission, solar and star occultation

•  Approach adopted for CCI: q  Minimise impact of inter-sensor bias on trend analysis through

use of anomalies in ozone (instead of vmr) – cf. WMO report q  Systematic analysis of drift and biases to characterise the

stability performance of individual sensors Ø  Pair-wise comparison of sensors Ø  Comparison against independent reference data sets (validation)

Strength of ozone anomalies

Remove biases but not drifts…

Bias and drift analysis

Drift analysis limited in significance by length of data series (10 years)

Drifts smallest in 25-40 km altitude range.

altit

ude

(km

)

From (Rahpoe et al., 2015): 30S -30N

bias ( %) drift ( %/dec)

SCIAMACHY

•  Bias in US + LS depends on profile representation

IUP V2.9

ESA/SGP 5.02

Hubert et al., 2016

Overview Level-2 CRDP decadal drift

Unphysical? to be confirmed (SAGE II, ACE-FTS)

Hubert et al., 2016

Overview drift assessment

Hubert et al. (2016), Atmos. Meas. Tech., 9(6), 2497–2534.

Summary and conclusion

•  Tracking drift and bias sources is an essential part of the work for scientists working on CDR generation

•  There is generally not one single and generic approach to drift and bias minimisation, instead a combination of different ‘good practices’:

•  Algorithm harmonisation and FCDR (level-1) understanding should always be the first step where possible

•  Data set customisation can be useful for trend analysis (cf. bias-removal through use of ozone anomalies)

•  Do not stick to one method but think about multiple ways to assess drifts and biases (is there a convergence?)

top related