Report from DAOS-WG Roger Saunders and Tom Hamill with input from WG members 1

Overview Highlights from Madison meeting Actions from Madison and ICSC9 WG Matters (membership, meetings) Future of DAOS 2

Mission statement To achieve its mission the DAOS WG, in collaboration with the CBS OPAG-IOS: Addresses Data Assimilation issues including the development of improved understanding of the sources and growth of errors in analyses and forecasts Promotes research activities that lead to a better use of observations and the understanding of their value Provides input and guidance for THORPEX regional campaigns for the deployment of observations to achieve scientific objectives.

DAOS-5 WG meeting Madison, 19-20 Sep 2012 Update on targeting Updates on THORPEX field campaigns Review observing systems Review developments in data assimilation WG matters http://www.ssec.wisc.edu/meetings/daos/agenda.html

Targeted Observations Plan to submit a thinned version to BAMS during 2013. 2. Analysis of WSR 2010/11 cases 3. WSR 2011/12 case

ESRL Impact of Targeted Dropsonde Data on Mid-latitude Numerical Weather Forecasts during the 2011 Winter Storms Reconnaissance Program Presented by Tom Hamill Forecasts and assimilations : Carla Cardinali, ECMWF Data analysis : Fanglin Yang, NCEP also: Sharanya Majumdar, RSMAS, U. Miami Question: will assimilation of mid-latitude dropsonde data have as large an impact as previously shown, given denser observation network, better assimilation systems and models?

2010/11 WSR campaign

Scatterplot of impacts For each case where dropsondes were launched, a downstream target location and verification time are identified. Data are plotted here only for these times/target locations. Verification area here is a +/- 10 degree box centered on target. Verification norm is an approximation to the total-energy norm. Cases above line indicate benefit from targeted data. No obvious beneficial impact.

Selecting target locations for tropical cyclones: SV/adjoint vs Selecting target locations for tropical cyclones: SV/adjoint vs. ETKF Brett Hoover (Univ Wisconsin) SV/adjoint guidance = strictly based on dynamics of perturbation growth. Suggests target locations around cyclone. ETKF guidance = estimates reduction of forecast error variance as a result of assimilation at given location based on linear mappings between ensemble forecast perturbations and analysis perturbations. Suggests target locations also along jet stream, commonly in ridges N of TC. These affect cyclone steering.

Summary of targeting Clear benefits of targeted observations for TC forecasts. In study with 4D-Var assimilation & ECMWF model, little impact seen for mid-latitude systems (2011 WSR). Targeting’s future: Global observational network design: use targeting strategies to adaptively select / thin satellite data for assimilation? Potential to adapt targeting concepts to work in concert with rapidly adaptable observational resources, e.g., cloud-drift winds from rapid-scan imagery. But: need quick, efficient algorithms.

Concordiasi = CONCORDIA-IASI A French-US initiative for climate / meteorology over Antarctica Improve the use of space-borne atmospheric sounders over polar regions, in particular IASI on board MetOp Benefit from the continental French-Italian station Concordia 11

Analysis Uncertainty, 45°S to 70°S Analysis differences: observational gap leads to large differences off Antarctica To the north: Geostationary satellite winds, ship surface obs, commercial aircraft routes To the south: Antarctic raobs and land surface data, MODIS and AVHRR winds Winds AMSU/ MHS Conventional

Observation impact diagnostics, Meteo-France, NASA-GMAO, ECMWF, NRL (obs south of 60 S) NRL AMVs includes LEO-GEO winds, and these have a big positive impact in this area.

CONCORDIASI: Concluding remarks Both Concordiasi dropsonde and gondola information have a positive impact on forecast performance Dropsondes have a larger impact at high latitudes (inside Antarctic continent) Large impact of dropsonde wind observations at high levels, of temperature information at lower levels 14

Messages for THORPEX Became TS Joyce TS Isaac The system on the left is Tropical Storm Isaac (viewed ~1530z) and the system on the right (~1400z) has a very high probability of developing into a tropical storm in the next day or so. See http://www.nhc.noaa.gov/ TS Isaac 15

Private industry to supply hyperspectral data Private industry to supply hyperspectral data? GeoMetWatch, a US Company Privately owned commercial data provider offers “STORM” leveraging GIFTS technology development Oct 2010: Licensed by US Dept. of Commerce for hyper- spectral data collection at 6 sites around the world (under the US Remote Sensing Act of 2003) Promises to restore critical data for severe weather forecasting cancelled from GOES-R and much more at a fraction of the cost, in record time! Potential Customers: US, top sovereign governments world-wide, and & commercial enterprises 16

3. New and future AMVs/products: Closing the gap with Leo/Geo winds Composites of GOES, Meteosat, FY-2, MTSAT, AVHRR, MODIS AVHRR: Metop A, NOAA- 15, 16, 18, 19 MODIS: Terra and Aqua Tracking clouds in infrared window channel, accounting for: Variable pixel time Parallax

3. New and future AMVs/products: Closing the gap with Leo/Geo winds Impact of Leo/Geo winds: NRL superobbed winds in NASA GMAO GEOS-5 Courtesy of Dagmar Merkova and Ron Gelaro

Messages for THORPEX Research satellites/instruments can be valuable as operational data providers, but can fail suddenly without backups (e.g. ENVISAT, AMSR-E are recent examples) Satellite data contributions to the GOS by nations increasing (e.g. FY-3, Oceansat-2), to hopefully help fill future gaps New privately funded initiatives to provide satellite data Impact of satellite data in medium range NWP is dominant in many advanced DA systems Research underway to extend/improve use of advanced IR sounders (cloudy rads, use PCs, more data over land etc.) 19

Distribution of gobal ASAP soundings 2010 [input from ASAP report at SOT-VI meeting 2011, author: Rudolf Krockauer, chairperson ASAP Task Team] 6011 Snd. on the GTS 82% E-ASAP fleet 7% RV Polarstern 11% All other (mainly 2 Japanese ships) Note: Soundings over East Europe are due to longitude errors 3. Performance, Slide 14/30

E-SURFMAR Drifting Buoys The number of operating buoys significantly dropped in 2011. This was mainly due to a decrease in lifetime of Iridium buoys deployed in 2010-2011. Since June 2012 the number of operating buoys has been increasing again. Problems with short buoy lifetimes seem having been fixed. The E-SURFMAR design study (2004) recommended 175 drifting buoys.

Observational issues Transition to BUFR for radiosondes provides new opportunities TAMDAR Water Vapour AIREPS increasing Global provision of GPS total zenith delay data Further delay in ESA wind lidar GEO-LEO AMVs proving beneficial should become operational Common format for rain radar data. To improve estimates of solid precipitation and develop guidance on the accuracy and temporal resolution of solid precipitation parameters. New observations needed for convective-scale models 22

Differences between operational analyses 23 Quantify the uncertainty (differences) in current operational analyses of the atmosphere – height, temperature, winds Consider implications of analysis uncertainty for NWP and plans for the future global observing network Analysis differences are a proxy for actual analysis error, which cannot be precisely quantified

300mb wind speed (2010) GFS / ECMWF differences Root-Mean Square of Analysis Differences: 300mb Wind Speed 24 Langland and Maue 2011 2010 Note the very significant effect of in-situ wind observations: Radiosondes and Commercial Aircraft 2011 2009 2008 ms-1

Radiosonde stations on the budget chopping block Example: Eareckson Air Station (Shemya) 70414 25 Langland and Maue 2011 GFS | ECMWF Root mean square difference of analyzed 300mb wind speed: July 2009 – June 2011 Shemya Island March 2012: NRL received a message from NUOPC (National Unified Operational Prediction Capability), that NOAA was trying to convince USAF not to end soundings at station 70414. Previously in 2008, soundings were reduced to one per-day (12UTC). This station is in top 10% world-wide for error reduction (NRL study). 70414 ms-1 Uncertainty in atmospheric upper-tropospheric wind analyses is substantially lower in locations where radiosonde data is provided. The blue-shaded areas are locations where raobs provide soundings twice-daily (00z and 12z). Station 70414 provides data only at 12z, so the associated reduction in analysis error at that location is mitigated, but still significant.

Analysis spread, 250 hPa u component Large differences in estimates of zonal wind along equator. In initializing global ensemble predictions, are we accounting for this sort of analysis uncertainty?

Power spectra of ensemble perturbations, NCEP ensemble Suggests this NCEP perturbation ensemble have insufficient power at planetary scales. This is consistent with the assumption made in the ETR that the analysis is unbiased while analyses between different centers suggest there is bias. (2) NCEP’s underestimate of initial amplitude is the least for the small baroclinic scales. This may be because the breeding method inside the ETR generates perturbations that project onto the (finite amplitude) Lyapunov vectors, dominated by baroclinic scales.

Aerosol assimilation case Study: Severe weather April 27th, 2011 (Huntsville tornado) MODIS aerosol optical depth (AOD) shows significant aerosol loading over Gulf within warm sector of storm RAQMS surface Black and Organic Carbon (BCOC) analysis shows highly elevated aerosols RAQMS used to initialize and provide lateral boundary conditions for 4km WRF-CHEM with explicit (MOSAIC) aerosol/cloud interaction 26th 27th 28th Aeronet AOD Louisiana WRF-Chem OBS

Case Study: Severe weather April 27th, 2011 (Huntsville tornado) Inclusion of aerosol cloud interactions significantly impacts the predicted precipitation distribution and improves forecast over Huntsville, AL SPC 12hr precip and tornado tracks valid 06Z on April 28, 2011 WRF no aerosols WRF-CHEM simulations conducted by Pablo Saide in collaboration with Greg Carmichael and Scott Spak (U-Iowa) WRF-Chem + RAQMS BC

Ensemble-variational with Integrated Lanczos (EVIL) Tom Auligne, NCAR Ensembles as used in data assimilation typically require a separate cycling of an ensemble methods such as an EnKF. To blend with variational, effectively two different assimilation methods must be run? Can the ensemble cycling be performed within the variational system itself?

Comparison with NCAR “DART” EnKF using WRF regional model EVIL can outperform EnKF method. The stopping criteria when applying the iterated Lanczos algorithm controls how much spread reduction is applied (i.e., prevents possible overfitting to noisy data).

Data Assimilation Terminology Suggestions based on usual current usage. “hybrid” applies to covariance, not method. E.g. “hybrid 4D-Var”  blend of stationary, ensemble-estimated covariances used in 4D-Var EnKF, ETKF, etc. acronyms indicate method of updating ensembles 3D-Var, 4D-Var, EnVar, etc. generate a single best estimate, unless specified e.g. “An ensemble of 4D-Vars” 4D-Var always uses a forecast model and adjoint to generate time−covariances 4D-EnVar, 4DEnKF, etc. use the ensemble to generate time−covariances. (The 4D may be omitted) — in 4D-Var, 3D-Var was standardised by Ide et al 1997 (and QJ), but not elsewhere. It may be omitted in new names. © Crown copyright Met Office Andrew Lorenc ECMWF Seminar 2011 35

Actions from DAOS-5 Carried forward from DAOS 4 DAOS 5 New Actions  Action 4-3: All DAOS members to provide recommendations on links to data monitoring sites and to other field campaign sites of interest.   Action 4-4: Consider organising future DAOS “back to back” with the MWFR meeting as a workshop (Co-Chairs/D. Barker). Action 4-8: The DAOS WG should consider taking a leading role in organising the next WMO DA symposium DAOS 5 New Actions Action 5-1: DAOS-WG members to assess if they have any observations or model outputs to contribute to the four high-impact precipitation African RC case studies: E. Africa October Oct 17-28 1997; S. Africa Nov 11-12 2008; N. Africa Nov 29-30 2010; W. Africa Sept 1 2009 36

Actions from DAOS-5 Action 5-2: Submit peer review paper on 2010/11 ECMWF WSR impact study (T. Hamill, S. Majumdar & C. Cardinali)   Action 5-3: Submit BAMS paper on targeting (S. Majumdar) Action 5-4: Look at observation sensitivities and possibly run a short OSE for Jan 2012 period where dropsondes had a sustained high impact in the GMAO system (R. Gelaro and C. Cardinali) Action 5-5. The terminology for adjoint sensitivity impacts should be discussed and agreed (Ron Gelaro and Carla Cardinali) Action 5-6: The USA and Europe are encouraged to liaise concerning the issue of wider exchange of GPS data (R. Saunders with Z. Toth) 37

Actions from DAOS-5  Action 5-7: The working group recommends to NOAA that GEO-LEO AMVs are made available to NWP centres in BUFR.     Action 5-8: GMAO to continue to investigate why the NRL GOES AMVs give more impact than the operational AMVs and report at the next meeting. (R. Gelaro) Action 5-9: A DAOS representative to attend CBS workshop on radar data exchange in Exeter in Nov 12 (R. Saunders) Action 5-10: The working group recommends to NOAA that the successor to CrIS provides the contiguous infrared spectrum to users to maximise the information content for assimilation in NWP and climate. A reduction in the field of view should also be considered for maximising clear sky scenes.  38

Actions from DAOS-5 Action 5-11: Circulate a proposal for a common terminology for hybrid ensemble variational methods and circulate it to DAOS and WGNE for comment. (A. Lorenc)   Action 5-12: The co-chairs to request approval of the changes to DAOS-WG membership as outlined in the report to ICSC-10. (Co-Chairs) .  39

Actions from ICSC-9 Action 8: The USA and Europe are encouraged to liaise concerning the issue of wider exchange of GPS total zenith delay data (with Zoltan Toth) Action 9: Maintain close links with EGOS on evolution of the GOS and definition of optimum network Action 10: The ICSC encouraged DAOS to consider DA for convective scale models. Action 15: All THORPEX WGs to examine how they can further develop activities to assist the African RC 40

Current proposed membership Tom Hamill(D), Co-chair NASA, USA Roger Saunders(O), Co-chair Met Office, UK Stefan Klink(O) DWD, Germany Carla Cardinali(D) ECMWF Chris Velden(O) Univ Wisconsin-CIMSS, USA Ron Gelaro(D) NOAA, USA Tom Keenan(O) CAWCR, Australia Rolf Langland(D) NRL, USA Bertrand Calpini (O) MeteoSwiss, Switzerland Andrew Lorenc(D) MetOffice, UK Florence Rabier(D/O) Météo-France Prof. Bin Wang(D), Chinese Academy of Sciences, China Michael Tsyroulnikov(D) HydroMet Centre, Russia Mark Buehner (D) Environment Canada Sharan Majumdar (D) RSMAS, Univ Miami, USA Darryl Kleist(D), NCEP, USA Needs approval by ICSC-10 O=Observations D=Data Assimilation

DAOS-WG Future Meetings Propose to hold next dedicated DAOS-WG meeting in 2014 Location TBD (probably Europe) Allow more time for discussions (fewer/shorter presentations) Co-ordinate next WMO Data Assimilation Symposium in Late 2013 42

DAOS-WG Future (1) Recognised as leading group for DA in WMO Acknowledged by WGNE which has less focus on DA. Strong links with ET-EGOS, MWFR, …. DAOS remains a global focus not mesoscale 1 day to seasonal timescales Joint meeting with MFWR under discussion Continue mix of Observations and DA 43

DAOS-WG Future (2) Membership to evolve but size of group remains about the same Proposal is for DAOS-WG to become a sub-group in WWRP Strong links would be retained with any THORPEX-Follow-On project ICSC invited to comment 44