Evaluating Skill in Ocean Model Parameterizations: Taylor Diagrams

Lynn (Nick) Shay and G. Halliwell MPO, RSMAS, Univ. of Miami

PhOD, AOML, NOAA http://isotherm.rsmas.miami.edu/~nick

HFIP Observations and Modeling Teams

T’is the Season…….

Ivan’s Track and Intensity Relative to OHC (left) NRL SEED Mooring Locations in Northern Gulf of Mexico Relative

to Bottom Depth (Right) (Teague et al., JPO, 2007).

14 ADCP moorings- Focus here in Array 9.

Ocean Structure Gulf Versus East Pacific

Strong vertical temperature, salinity and density gradients at base of OML in EPAC…

Implications for mixing…and ocean (SST) cooling.

N~20 cph

Differentiated cooling in the LC system (Jaimes and Shay, MWR, 2009)Differentiated cooling in the LC system (Jaimes and Shay, MWR, 2009)

Loop Current Shedding front

Warm core eddy

T ~ -1oC

T ~ -4.5oC

T ~ -0.5oC

ageostrophicvelocity(cm s-1)

ageostrophicKE

(cm2 s-2)

verticalshear(s-2)

Richardson number

Cluster-averaged temperature profiles

R

f

E

Taylor diagram Taylor Diagrams:

Taylor, K., 2001: JGR, 106, D7, 7183-7192

Taylor diagram Taylor Diagram – Skill Score

S [0.6,0.7]

S [0.9,1.0]

S-isolines: 0.1 intervals, grey shading

Taylor, K., 2001: JGR, 106, D7, 7183-7192

MOTIVATION:

Ivan (2004) over the GOM SSH (cm) from HYCOM (from Halliwell et al., MWR, 2008).

SST Analyses

Northern Cyclone

Southern Cyclone

HYbrid Coordinate Ocean Model (HYCOM)Hurricane Ivan Simulations 10 Sept-6 Oct 04

Configuration:• 0.04° Mercator grid, Gulf of Mexico domain• No data assimilation performed

Initial and boundary conditions from U.S. Navy HYCOM ocean nowcast-forecast system:•Data assimilative ocean nowcast•Navy Coupled Ocean Data Assimilation (NCODA) assimilation

Atmospheric Forcing:•Navy 27 km COAMPS atmospheric model•Vector wind blended with higher resolution fields from HWIND •Wind stress for HWIND calculated using Donelan cd

Hurricane Ivan Modeling Experiment Summary

Model Attribute Baseline Experiment GOM1 Alternate Experiments

Horizontal resolution 0.04° Mercator GOM2: 0.08° Mercator

Vertical resolution 26 layers, 4-8m in OML GOM3: 21 layers, 7.5-15m in OMLGOM4: 31 layers, 3-5m in OML

Vertical mixing KPP GOM5: MYGOM6: GISS

CD Donelan GOM7: PowellGOM8: Large and PondGOM9: Large and Pond (capped)GOM10: Shay and Jacob

CEL, CES COARE3.0 algorithm GOM11: Kara et al.

Atmospheric forcing 27-km COAMPS+H*WIND GOM12: 27-km COAMPS only

Outer model NCODA GoM hindcast GOM13: Free GOM simulation

Ocean Dynamics 3-D Ocean Physics GOM14:1-D Column Models

RMS SST (oC) Differences

Pre Ivan SST

Post-Ivan SST

Pre-Post Ivan ΔSST

TMI and KPP SST Comparisons

Observed/Simulated Current Response at M9 (1.5 Rmax)from 7 ExperimentsHalliwell et al., MWR, (2009)

Below cd using ocean response as a tracer in Shay and Jacob (2006)

Current Time Series Comparisons @ 1.5 Rmax

U (east-west) V (north-south)

Taylor Diag., simulations vs. baseline:

SST;

v Current at M9 (1.5Rmax).

(Halliwell et al., MWR, 2009)

Taylor Diag., simulations vs. observations:

SST;

v Current at M9 (1.5Rmax).

(Halliwell et al., MWR, 2009)

Regression XBT Transect: East Pacific Ocean

XBT Transect-East Pacific

Equivalent OHC and Juliette (2001)

Track and Intensity of TC’s Gustav and Ike Versus AXBTs relative to OHC and 26oC Isotherm Depth.

Gustav : 191 AXBTs 111 GPS Drifters Floats

Ike : 216 AXBTs 111 GPS Drifters Floats

(Shay et al. 2009)

NOAA WP-3D Profiling over MMS Moorings

(Collaboration with AOML HRD, AOC, TPC, NCEP)

Goal: To observe and improve our understanding of the LC response to the near-surface wind structure during TC passages. Specific objectives are:

1. Determine the oceanic response of the LC to TC forcing; and, 2. Influence of the LC response on the atmospheric boundary layer and intensity.

Deliverables include:

V, T, S profiles to 1000 m @ 2-m resolution.

Surface winds (SFMR, GPS) provided by HRD.

Atmospheric profiles of V, T and RH @ 5-m resolution.

Progress and Blueprint For Future

Ivan a clear example of negative feedback (wake cooling/mixing induced by strong winds and Cold Core Ring) as opposed to positive feedback over the Loop Current and Warm Core Rings.

Taylor Diagrams collapse Standard Deviation, Correlation Indices and RMS Differences into one representation relative to an observation of current, temperatures, winds, humidities, variance etc to assess model performance.

Estimate skill by combining standard deviations and correlations.

In Ivan case, 14 sets of model simulations were made differing in only one aspect at a time. Approach shows sensitivity and allows us to isolate physics (Even a 1-D versus 3-D Ocean!)

Applying same approach to OHC variability in assessing uncertainties in satellite retrievals using in situ data as truth.

Field programs to acquire 3-D data……