1. Trying to Stop a Leak in the Operational Global Model Thomas J. Galarneau, Jr. Mesoscale Dynamics Section Mesoscale and Microscale Meteorology Laboratory National Center for Atmospheric Research NCAR Day of Networking and Discovery 17 April 2015

2. Conclusions NCEP Global Forecast System (GFS) model loses atmospheric water vapor over tropical west Pacific during “quiet” regimes – Reduced precipitation/synoptic-scale outflow – Difficulty in predicting TC development Behavior of physical parameterizations is important for medium-range forecasts 2

3. Initial Motivation to Examine GFS Significant degradation of medium-range GFS track forecasts relative to the HFIP “baseline” Track errors motivate examination of synoptic- scale flow in GFS forecasts Are there systematic medium-range forecast errors in synoptic-scale circulation features? Are there linkages between forecast errors in circulation features and the behavior of convection and water vapor in the tropics? GFS North Atlantic Track Error Statistics 2012–2014 GFS mean absolute track error HFIP “baseline” track error 3

4. Synoptic-Scale Flow Errors (2014) Synoptic-scale eddies too weak in GFS forecasts Difficult to maintain amplified (or, blocked) flow 4 ×10 6 m 2 s −1 168-h GFS Forecast Error: 500 hPa ψ and ψ error Composite Mean 8 Jul–7 Nov 2014 anticyclonic cyclonic

5. Pa s −1 descent ascent 250 hPa wind speed (blue contours in m/s), 250 hPa V div (arrows in m/s), and 600–400 hPa mean vertical velocity (shaded in Pa/s) Composite mean 8 July–7 Nov 2014 = WPAC region Decrease in tropical convection over western and central Pacific Reduced divergent outflow from tropics toward NH midlatitude jet/waveguide Reduced divergent outflow from NH tropics toward SH 5 GFS Analysis 168-h GFS Forecast

6. Precipitable water (PW) error (shaded in mm) Forecast minus analysis Composite mean 8 July–7 Nov 2014 6 24-h GFS Forecast 168-h GFS Forecast = WPAC region GFS loses water vapor over tropical west Pacific warm pool PW also decreases over parts of eastern Pacific and southeast Atlantic mm too moist too dry

7. May 26 June 5 June 15 June 25 July 5 July 15 July 25 July 13 July 23 Aug 2 Aug 12 Aug 22 Sep 1 Sep 11 = Active Tropical Cyclone 2014 GFS Precipitable Water: WPAC TC Fengshen Fengshen precursor PW loss is episodic Prevalent during “quiet” regimes 7

8. GFS Misses Development of Fengshen 250 hPa Height (black contours in dam), 700 hPa wind (arrows > 5 m/s), and PW (shaded in mm) Initialized 00Z/27 Aug GFS forecast fails to develop Fengshen disturbance as PW dramatically decreases bad forecast 8

9. GFS Recovers SLP (black contours in hPa), 10-m wind (arrows > 5 m/s), and wind speed (shaded in m/s) Initialized 00Z/3 Sep Later GFS forecast improves when Fengshen disturbance is initialized good forecast 9

10. May 26 June 5 June 15 June 25 July 5 July 15 July 25 July 13 July 23 Aug 2 Aug 12 Aug 22 Sep 1 Sep 11 = Active Tropical Cyclone 2014 GFS Precipitable Water: WPAC TC Fengshen Fengshen precursor PW loss is episodic Prevalent during “quiet” regimes 10

11. 2014 GFS Precipitable Water: EPAC and NATL May 27 June 6 June 16 June 26 July 6 July 16 July 26 Eastern Pacific North Atlantic PW forecasts are much better in other basins 11

12. 2014 GFS Relative Humidity: WPAC 13-Sep3-Sep24-Aug 14-Aug 4-Aug25-Jul15-Jul Fig. 4a in Holloway and Neelin (2009, JAS) “quiet” regime 12

13. Precipitation Quiet regime GFS appears to produce too much precipitation everywhere in 0–24 h forecast Precipitation gradually decreases with increasing forecast lead during quiet regime Reduction of precipitation is consistent with static stabilization 13

14. Precipitation and Vertical Velocity Vertical Velocity (cm/s) Height (km) ascent “Quiet” Regime* over tropical west Pacific Precipitation (mm) Total rain Cumulus rain CMORPH *quiet regime: 1 Aug–5 Sep and 10–31 Oct 2014 (n=58) GFS produces too much cumulus precipitation in 0–24 h forecast and gradually decreases thereafter Cumulus scheme driving too much deep convection early in forecast? 14

15. GFS T v Profile and LHFX: WPAC Virtual Temperature Bias (K) Height (km) warm bias cold bias “Quiet” Regime* *quiet regime: 1 Aug–5 Sep and 10–31 Oct 2014 (n=58) too much heating aloft early in forecast column stabilizes through 96-h deep cold bias by 120–168-h (as convection weakens?) Surface Latent Heat Flux (W m −2 ) Reduced surface fluxes may contribute to inability to offset deep layer stabilization Lower fluxes linked to low wind speed bias of 0.5–1 m/s at surface (0.5 m/s ≈ 10 W m −2 ) 15

16. Summary GFS model loses atmospheric water vapor over tropical west Pacific during “quiet” regimes – Reduces precipitation – Hinders TC development Atmosphere overturns early in forecast (0–24 h) and does not recover – Cumulus scheme too active early – Surface fluxes too low to aid in destabilization Behavior of physical parameterizations is important for medium-range forecasts 16