The impact of PCW satellite winds filling the gap at high latitudes

1. The impact of PCW satellite winds filling the gap at high latitudes L. Garand, J. Feng, Y. Rochon, S. Heilliette, EC, and A. P. Trishchenko,CCRS Satellite Applications for Arctic Weather and SAR Operations (SAAWSO) Workshop St-Johns, NFLD, 22-24 April, 2013

2. Outline • Advantage of Highly Elliptical Orbit for polar regions • Motivation for a specific AMV impact study • OSSE definition • Impact results of real and simulated AMVs • Conclusion

3. Polar Communications and Weather (PCW) mission in a few words • 2-satellite constellation in highly elliptical orbit • Core meteo instrument similar to ABI (GOES-R) • Extends GEO applications to the pole, 15 min imagery 16-h 3-apogee (TAP) ground track spatio-temporal coverage vs latitude Apogee 43,000 km, at 30 N, 30,000 km

4. 16-h three apogee (TAP) orbit • path over two days • SAT-1 would be 8-h ahead • of SAT-2, and follows same path • each satellite would image • +/- 5-h to apogee Positions represent 1-h intervals

5. Motivation here: impact on NWP of filling the AMV gap in the northern polar region Current AMV coverage After quality control and thinning 4 HEO satellites would Be needed to fill both N and S gaps AMVs would be produced from 15 min imagery At latitudes 45o-90o

6. Single images Image pairs Image triplets HEO (2-sats) vs LEO (2, 4, 7 sats) coveragebased on May 2011 average Ref: Trishchenko and Garand, 2012, Canadian J. Remose Sensing

7. Simulation and assimilation setups • Assimilation model and system: • Operational Global Environmental Multi-scale model (GEM) • 801x600 (~35 km), 80 levels, top 0.1 hPa • 3D-VAR assimilation, FGAT (First Guess at Appropriate Time) • Cycle starts from 5-day forecast from NR • ECMWF Nature Run serves as truth (~40 km resolution) interpolated to GEM grid for validation purposes

8. OSSE definition • Period covered in test cycles (2.5 months): • 15 December 2005 to 28 February 2006 • Simulated from NR all data types assimilated. Positions are those at the same dates in 2008-2009, to include recent types (GPSRO, IASI) not available in 2005-2006. • All-sky (cloudy) IR radiances were simulated from NR. Clear radiances were selected as done operationally (residual cloud contamination is possible). • Background check done once for all (same data assimilated in all cycles).

9. Non-radiance observations assimilated

10. Radiance observations assimilated

11. Observation perturbations • Perturbations applied to the simulated observations using Gaussian-distributed random errors • No applied spatial or inter-channel error correlations. • No applied biases • Calibration of OSSE: • Perturbation is simple multiplier of assigned observation error STD for each data type to get (O-A), (O-F) statistics similar to real corresponding statistics Ref: Rochon et al., 2012, Workshop on Impact of observations, Sedona, AZ

12. Wind errors assigned in assimilationin comparison to AMV MVD errors • AMV error inflated in relation to (O-F) • polar MDV lower than extratropics MVD • perturbation is 0.28 AMV obs error

13. Simulated AMV: NR wind at NR cloud top Cloud fraction 1- tau(cloud) 11m BT From NR AMV NR wind At cloud top Cloud top Pressure Where TOA tau(cloud) =0.9 Ref: Garand et al, Atmosphere-Ocean, 2011.

14. PCW AMV used in assimilation • thinning at 180 km • no data where cloud free • 50-90 N/S coverage • allowed range 100-925 hPa • every 6-h • same obs error (by level) for all AMVs Conditions similar to operational AMVs except +-3-h window for OPE and range 100-700 hPa

15. Definition of OSSE cycles (3dvar)

16. Impact of real and simulated AMVs REAL_CNTL REAL_NOAMV SIM_CNTL SIM_NOAMV

17. Comparing AMV impact In real and simulated Data assimilation systems OSSE REAL 500 hPa temperature STD Differerence w.r.t. Own analyses: REAL_CNTL minus REAL_NO_AMV (left) SIM_CNTL minus SIM_NO_AMV (right) Grey: 95% confidence

18. Impact of adding PCW AMVs in region 50-90 Nin comparison to control (vs NR) SIM_CNTL SIM_PCW1 Note: very similar results vs own analyses beyond day 2

19. Adding PCW AMVs in regions 50-90 N/S:impact comparison vs no AMV vs own analyses vs Nature Run 24-h 120-h

20. Comparing impact at 72-h in regions 50-90 N and 50-90 S (vs NR) SIM_PCW2SIM_NOAMV Larger impact in SH

21. 500 hPa GZ anomaly correlation vs NR SIM_PCW2SIM_NOAMV Predictability gain at days 3-5: ~2-h in NH, ~4-h in SH, accounting for 50% too optimistic results from OSSE

22. A comprehensive OSSE setup was developed Real AMVs have a consistent positive impac up to day 4. OSSE AMV impact tends to be higher than real impact, especially in the Tropics Adding 50-90 N PCW AMVs has a significant positive impact in that region up to day 3, but not significant further south Gain of predictability of order 1-3h at day 3 in region 50-90 N. Gain is twice larger in region 50-90 S, and extends to 30 S. Validation vs NR or own analysis consistant after day 2 OSSE set-up could be used to infer other impacts, e.g. assimilation of PCW radiances at high temporal resolution Conclusion This work soon to be published in JCAM