EVALUATION OF UPPER OCEAN MIXING PARAMETERIZATIONS

1. EVALUATION OF UPPER OCEAN MIXING PARAMETERIZATIONS 1 GEST, UMBC/ NASA GSFC, Greenbelt, MD 20771 2 MPO, RSMAS, University of Miami, Miami, FL 33149 S. Daniel Jacob1, Lynn K. Shay2 and George R. Halliwell2 Alan Wallcraft (NRL Stennis) Chet Koblinsky (NOAA)

2. Importance of Ocean Mixing • Based on Observational Analysis (Jacob et al., JPO 2000) • Entrainment is the dominant mechanism in controlling mixed layer budgets. • Mixed layer heat and mass budgets strongly depend upon the entrainment scheme used – crucial for intensity prediction. • Numerical Modeling (Jacob and Shay, JPO, 2003): • Measured and simulated quantities based on different hypothesis are used to compute entrainment mixing. • Bulk schemes in MICOM also indicated strong flux variability.

3. Motivation: HYbrid Coordinate Ocean Model (HYCOM) Results (Bleck 2002, Halliwell 2004) • Configuration • Domain: Gulf of Mexico • Resolution: 0.07 • 50 Levels/ Layers • Closed Boundaries • Initial Conditions • Quiescent Conditions with Gilbert forcing • SST Variability • Significant in the directly forced region • Large range in fluxes • Sensitivity to precipitation

4. Hycom Q Movie Gaspar KPP MY2.5 PWP

5. No Precip 2 Rmax Precip

6. Evaluate and Validate Mixing Schemes to identify the most appropriate parameterizations for use in coupled prediction model. Schemes in the Hybrid Coordinate Ocean Model K Profile Parameterization (Large et al. 1994; KPP) Gaspar (1988) Price et al. (1986; PWP) scheme. Mellor and Yamada (1974; MY2.5) Level 2.5 K-ε scheme Canuto et al. (2001, 2002; GISS) Compare Simulations to Observations Gilbert (1988) Isidore (2002) Lili (2002) Objectives

7. Model Configuration, set up, derivation of initial conditions and forcing – February 2004. Gilbert simulations and comparison – May 2004 Isidore and Lili Simulations and comparison to observations – Jan 2005 Recommendation of the most appropriate mixing scheme for use in coupled prediction models. Aug 2005. Project Schedule

8. Gilbert (1988): Prior HYCOM/ MICOM configuration on Mercator grid at 0.07° resolution at the equator. Buffer zones at open boundaries relaxed to climatology. Simulations from 14 Sep 1988 to 20 Sep 1988. Isidore (2002) and Lili (2002): Domain includes Caribbean Sea and the Gulf of Mexico. Treated as a single case with 20 day integration for each mixing parameterization. Configured as a nest of the basin scale HYCOM that provides conditions at the open boundaries. Model Configurations

9. Gilbert Initial Conditions • Pre-storm variability due to a Loop Current Warm Core Eddy (LCWCE) – Eddy F • Sampled extensively by Minerals Management Service - Yearday 200/300 data. • Propagated westward at 3 to 4 km/day (Average translation speed 5 km/day for Gulf eddies: Vukovich and Crissman 1986). • Possess distinct T-S relationship corresponding to the subtropical water mass. Shay et al. (1998) derived the T-S relationship for the Gulf Common Water and the Loop Current Eddy Water using historical and Yearday 200/300 data set. • Realistic initial conditions are derived by combining the objectively analyzed day 200 data with the Levitus climatology.

10. Temperature-Salinity Diagram LCW GCW

11. Pre-Gilbert Sea Surface Height • HYCOM is initialized following the methodology used by Jacob and Shay (2003). • Location of pre-storm eddy (LCWCE F) is accurately reproduced. • Depth of 26° C and 20° C isotherms compare well to observations.

12. Isidore Initial Conditions • Pre-storm variability due to the Loop Current in the Eastern Gulf of Mexico. • As the source of eddies, the Loop Current Water possess the same distinct T-S relationship corresponding to the subtropical water mass. • Realistic initial conditions from the 0.08° North Atlantic HYCOM. Courtesy of the Data Assimilative Ocean Modeling NOPP project of RSMAS, University of Miami and Naval Research Lab, Stennis Space Center. • The location of boundary currents and eddies are reproduced accurately by the basin-scale nowcast/ forecast system. • Fields evaluated with respect to expendable probe data.

13. Hycom Nowcast/Forecast System (Smedstad et al. 2003) • North Atlantic Hycom at 0.08° resolution extending from 28° S to 70° N. 26 levels/ layers in the vertical. High resolution surface forcing from FNMOC with Salinity relaxation at the surface. 3° buffer zones in the north and south relaxed to climatology. • Satellite altimeter height anomalies from the MODAS operational system at NAVOCEANO. • Mean Sea Surface Height from the 0.08° Atlantic MICOM. • Vertical projection of surface height signature using the Cooper-Haines (1996) technique. • Sea Surface Temperature not assimilated at present.

14. Pre-Isidore Sea Surface Height 19 Sept 2002 • Pre-storm mesoscale variability is accurately reproduced in terms of simulated SSH. • Corresponds to 19 Sept 2002 pre-storm experiment.

15. Sea Surface Height Anomaly and Ocean Heat Content 19 Sept 2002

16. Pre-Isidore Sea Surface Temp 19 Sept 2002 • Pre-storm sea surface temperatures in the assimilated fields are biased lower throughout the domain with 0.5° C in the Caribbean Sea and slightly higher elsewhere.

17. Meridional Cross Section along 84° W The depth of 26° and 20° C isotherms compare well to profiler data in the Caribbean Sea.

18. Meridional Cross Section along 86° W Due to the lower SSTs, the depth of 26° and 20° C isotherms are shallower compared to profiler data by 10 m on the average in the central gulf of Mexico.

19. Zonal Cross Section along 24° N The model captures subsurface structure variability due to assimilated altimetric heights well.

20. Gilbert (1988): Flight level reduced winds and buoy data in the core. ECMWF 6 hourly surface data provides large scale field. Combined using NOAA Hurricane Research Division HWIND program to produce wind forcing every 3 hours. Constant air temperature and relative humidity with synthetic rain rates derived from wind field distribution. Isidore (2002) and Lili (2002): Flight level reduced winds and buoy data analyzed using HWIND available on regular grid (courtesy Dr. Mark Powell). FNMOC surface forcing provides the large scale forcing. Blended using the parameter matrix objective analysis technique of Mariano and Brown (1992) with cubic splines every 3 hours. Rainrate from TMI. Wind Forcing

21. Gilbert Wind Field Structure 16 Sep 88 06UTC The double eyewall structure is captured accurately by the analysis

22. Lili Wind Field Structure 02 Oct 02 00UTC The parameter matrix algorithm works well in blending the large scale data with the HWIND analysis.

23. Summary and Future Work • Initial Conditions and forcing are established to investigate the sensitivity of mixing parameterizations in upper ocean heat content simulations. • While the Gilbert initial conditions compare well with pre-storm observations, the SST prior to Isidore is biased low in the model by an RMS difference of 1.1° C in the domain. Updated fields where SSTs are assimilated will fix this problem for the Isidore and Lili cases. With a RMS differences of about about 0.2 m/s, simulated current field compares well with profiler data. Detailed comparisons are in progress to validate the initial fields. • The parameter matrix based objective analysis technique works well to blend the wind analysis data with large scale model fields. This will be further improved during the course of the project. • The upper ocean response during hurricane Gilbert will be performed in the next phase followed by Isidore and Lili simulations.