GRWG Web Meeting 2020-06-18

GSICS Web Meeting on ECMWF/EUMETSAT NWP SAF Workshop on the treatment of errors in satellite data assimilation

The workshop website can be found at https://www.ecmwf.int/en/learning/workshops/ecmwf-eumetsat-nwp-saf-workshop-treatment-random-and-systematic-errors-satellite-data

Agenda

1. Tim Hewison (EUMETSAT) – Introduction
2. Heikki Pohjola (WMO) + Dave Duncan (ECMWF) – Assessment of Value of Provision of Spectral Response Functions from CGMS Agencies
3. Greg Kopp (LASP) + Tim Hewison (EUMETSAT) – Input from CEOS/GSICS Workshop of SI-traceable Space-based Climate Observing System
4. Group Discussion to Consolidate GSICS inputs for workshop

CMA: Scott Hu

CNES: Clemence Pierangelo

ECMWF: Dave Duncan

ESA: Stefano Casadio, Fabrizio Niro

EUM: Dorothée Coppens, Tim Hewison, Rosemary Munro, Viju John

JMA: Kazutaka Yamada, Hideaki Tanaka, Arata Okuyama, Kazuki Kodera

KMA: Dohyeong Kim, Minju Gu, Wonhyeong

NASA; Ben Scarino, Dave Doelling

NOAA: Likun Wang, Fred Wu, Manik Bali, Fangfang Yu, Larry Flynn

UCAR: Greg Kopp

WMO: Heikki Pohjola

1. Tim Hewison (EUMETSAT) – Introduction

Tim introduced the ECMWF/EUMETSAT NWP SAF Workshop on the treatment of errors in satellite data assimilation and described the aim of the web meeting: to coordinate potential contributions on this topic from GSICS.

2. Dave Duncan (ECMWF) – Assessment of Value of Provision of Spectral Response Functions from CGMS Agencies

Dave presented the themes of the CGMS paper on the the provision of accurate SRFs for NWP:

• This has been a CGMS Best Practice for IR and MW instruments since CGMS-44

• Accurate SRFs are increasing vital to support microwave spectral management

• He also showed examples of where NWP statistics benefits from use of detailed SRFs, instead of approximating them as simple Top-hat functions, as has been done

He went on to review the proposed actions to be considered by CGMS:

• CGMS Members shall ensure that accurate channel Spectral Response Functions (SRFs) for all microwave and infrared instruments are measured and made available as described in the CGMS Best Practise.

• CGMS Members shall also make available their validated instrument SRFs together with uncertainty information through their instrument calibration landing pages. In addition, a document summarizing the currently available SRFs and their status (accurate/inaccurate) as well as identifying any missing information shall be provided through the landing pages.

  • it was noted that these information could be made available through OSCAR instrument landing pages

• WMO will establish links to this information through the relevant instrument entries in the OSCAR/Space database. This information will be updated with the help of the OSCAR/Space Support Team though the regular requests for satellite status updates.

Discussion:

• Definition of precise? High-precision pre-launch knowledge

• The difficulty of diagnosing whether the root cause of observed bias patterns are radiometric calibration or spectral characterisation was recognised. It was pointed out that while SRF shifts may be easy to characterise, changing SRF shapes are less so.

• The applicability of NWP bias monitoring statistics to investigate the SRF of infrared window channels was discussed, but felt to be difficult due to the lack of atmospheric signal.

• The question of how to specify and use SRF uncertainty information remains open. It was suggested that offline studies could be run using RTMs, and that a GSICS convention could be established to provide SRF uncertainty information

• The spectral resolution needed for SRFs depends on the channel, but may require tens to hundreds of wavelengths spread across the passband. Again, it was suggested that offline studies could be run using RTMs.

• Furthermore, it was pointed out that SRF may need to be characterised across the full spectrum (especially in the reflected solar band) due to sensitivity to out of band signal.

• Instrument manufacturers would appreciate guidance on what should be avoided - e.g. sensitivity to particular atmospheric lines.

• Greg pointed out CLARREO studies, which showed inter-calibration uncertainty can be dominated by SRF uncertainties.

• WMO plan to start monitoring the availability of SRF info

Action: A.GRWG.20200618.1: Tim Hewison (EUMETSAT) to consider providing FCI SRF uncertainties to ECMWF for analysis

Discussions lead to the following recommendations for the workshop:

1) NWP to define the minimum level of uncertainty on the SRF (e.g. accurate and resolution requirements) from the workshop and provide the feedback to the GSICS community.

2) NWP to define the minimum level of uncertainty on instrument's calibration from the workshop and provide the feedback to the GSICS community - this may provide focus for GSICS future developments.

3) NWP community to communicate with the instrument manufacturers to avoid adverse spectral range observed with the assimilation results.

3. Greg Kopp (LASP) – Input from CEOS/GSICS Workshop of SI-traceable Space-based Climate Observing System

Greg Kopp (UC/LASP) summarized the outputs of the CEOS/GSICS workshop of SI-traceable space-based Climate Observing System, including the recommendations and issues for IR, VNIR and MW instruments calibration as well as the recommendations for ECMWF/NWP error workshop. These focused on the roles of SI traceability in absolute measurements and on-orbit stability & continuity using GSICS inter-calibration/comparison methods.

• SITSATs offer key advantages for NWP & Climate applications:
  • Provision of an irrefutable reference
  • Can demonstrate traceability of on-orbit calibration to SI
  • Can be assimilated without bias correction
  • To diagnose model errors, independent of instrument errors
  • Resilience against gaps/drifts outside overlaps between reference instruments ==> Allow generation of Robust and harmonised Fundamental Climate Data Records
• Additional benefits depend on SITSAT design:
  • Ability to resolve residual dependencies in calibration of other sensors:
  • Scan angle, Non-linearity, Diurnal variations, Spectral (Gap-filling, SBAF)
  • Challenges include extending to other spectral bands
    
    
• Priority is to establish usable and accurate SI-traceable references for LEO and GEO sensors

4. Discussion on potential topics that could be presented by GSICS for discussion at the workshop

• The role of anchor observations that can be assimilated into NWP without bias correction - to help diagnose model errors. These include some temperature sounding channels of microwave sounders as well as GNSSRO, which have led to large improvements in the NWP representation of the upper stratosphere.

• NWP can provide GSICS with feedback on the bias correction applied in the assimilation of GSICS Reference instruments (IASI, CrIS, AIRS, …)

  • And could potentially contribute to the IRRefUTable report

• ...and observations found to have negative impact on NWP (e.g. due to large or unstable error characteristics)

• Seek feedback from the NWP community on the use of FCDRs as inter-calibration references - e.g. could inter-calibration products derived from such algorithms be useful in NWP Re-Analysis campaigns?

• Seek feedback from the NWP community on the minimum acceptable level of bias in observations

• In general, we should aim for consistent treatment of systematic, random and structure errors - including timescale and spatial dependence in NWP and GSICS communities.

• Can GSICS provide information to help determine whether observed biases are due to radiometric calibration or spectral characterisation errors?
  e.g. SRF retrievals based on comparison with hyperspectral references, Diagnose whether observed bias patterns consistent with SRF shifts, etc

• SRF Uncertainties:
  • How to express uncertainties in the SRFs (vectors or covariance matrices)
  • Need to account for correlation of uncertainty across each channel’s SRF
  • How can SRF uncertainties be propagated into radiance space?
  • Related to change in weighting functions for sounding channels
  • e.g. using RTMs + standard atmospheres + cloud + …?
  • GSICS instrument agencies to provide the SRF documents for the SI traceability.
  • GSICS to provide guidance to show the bias at different spectral scenes caused by erroneous SRF.
    • In the GEO-LEO and LEO-LEO inter-calibration, sometimes it is hard to distinguish the bias caused by radiometric calibration error from spectral calibration uncertainty.
    • One example is the detector non-linear response may have the similar bias pattern as inaccurate SRF.
      
      
• It was noted that the outputs of this workshop could provide inputs for the forthcoming GSICS/CEOS workshop of Pre-flight Characterisation.
  • Noting that polarisation sensitivity is becoming increasingly important.
    
    
• Many small satellites/CubeSats may have very limited on-board calibration devices, which will make the inter-calibration activity more important.
  • Their error budgets should be published and available to the public.
    

Action: A.GRWG.20200618.2: Clemence Pierangelo (CNES) to investigate whether it is possible to share the analysis of systematic vs random error performed for greenhouse gases observation missions.