Ocean Initialization System for Coupled Hurricane-Ocean Models and its Transition to HWRF

1
Ocean Initialization System for Coupled
Hurricane-Ocean Models and its Transition to HWRF

• Isaac Ginis
• and
• Richard Yablonsky
• University of Rhode Island
• 61st Interdepartmental Hurricane Conference
• 6 March, 2007

2
Why Proper Ocean Initialization in a Coupled
Hurricane-Ocean Model is Needed for Accurate
Hurricane Forecasting?

• Wind-induced mixing of the upper ocean generates
  the cold wake by entraining cooler water from
  below
• The entrainment rate is greatly dependent on the
  depth of the upper ocean warm layer.

3
Ocean Model Initialization Challenge
Depth of 26oC isotherm on September 15,
2005 (courtesy of Michelle Mainelli, TPC)

• Accurate representation of mesoscale oceanic
  features is needed, such as the penetration of
  the Loop Current into the Gulf of Mexico and the
  shedding of Loop Current eddies.

4
Ocean Initialization Methods that can be Utilized
in a Hurricane-Ocean Coupled Model

• A continuously run ocean model with assimilation
  of in situ and/or remotely-sensed observations
  that is capable of predicting the position of the
  LC and LCEs.
• A feature-based modeling procedure that is
  capable to adjust the background LC position and
  structure and directly insert LCEs with real-time
  in situ and/or remotely-sensed observations.

5
Feature-Based Initialization Procedure

• This procedure is based on a feature modeling
  approach that allows a realistic simulation of
  the cross-frontal temperature, salinity and
  velocity of oceanic fronts and eddies.
• We developed an innovative method of
  cross-frontal sharpening of the background
  temperature and salinity fields, which is guided
  by observed cross-sections obtained in
  specialized field experiments.

6
Identifying LC and LCEs in the Gulf of Mexico
from Sea Surface Height

• During the winter months, the LC and LCE
  positions are often identifiable from the SST
  field, however during the hurricane season the
  SST is nearly homogeneous in the Gulf.
• The LC and LCEs can be identified by the
  difference in the sea surface height. The daily
  sea-surface height anomaly (SHA) is produced at
  Stennis Space Center in Mississippi by blending
  multiple ground tracks from satellites.
• The altimetric SHA is used daily by the Tropical
  Prediction Center to calculate the depth of the
  26C isotherm and integrate it into the
  Statistical Hurricane Intensity Prediction Scheme
  (SHIPS)

7
Identifying LC and LCEs in the Gulf of Mexico
from Sea Surface Height
Daily NCEP SST Analysis
SST GDEM September climatology
Depth of 26C isotherm derived at TPC
SHA from satellite altimetry
September 15, 2005
8
AXBT Measurements
In advance of hurricane Rita, on 15 September
2005, HRD scientists deployed AXBTs
AXBTs provide accurate measurements of the ocean
temperature to a depth of 300-400 m.
Temperature profiles for this study were
provided by E. Uhlhorn at HRD.
9
Results of Feature-Based Model Initialization on
Sep. 15, 2005
Model
TPC Analysis
Depth of 26oC isotherm
10
Evaluation of the Feature-based Initialization
Scheme and Comparison with RSMAS HYCOM Prediction
System

• Evaluation is done by comparing the model and
  observed temperature profiles at each AXBT
  location.
• RSMAS HYCOM dataset from 15 September 2005 was
  downloaded from the HYCOM consortium data server
  website. RSMAS HYCOM assimilates SST and
  satellite altimetry.

11
75-m Temp. (SSH/SST-assimilated) with AXBT
locations
1
2
3
5
7
6
4
12
75-m Temp. (SSH/SST-assimilated) with AXBT
locations
8
9
10
13
12
11
13
75-m Temp. (SSH/SST-assimilated) with AXBT
locations
14
15
16
18
17
14
Ocean Heat Content
Feature-based initialization
OHC provides a quantitative measure of the heat
energy available to an approaching tropical
cyclone
RSMAS HYCOM
OHC is the integrated heat content from the
oceans surface to 26oC
15
Evaluation of OHC at AXBT Locations
TCHP rmse Feature-based 9.5 kJ
cm-2 RSMAS HYCOM 26.0 kJ cm-2
16
Operational implementation

• Feature-based initialization of the LC and
  LCEs was implemented operationally in the GFDL
  coupled model in 2006.

17
Examples of LC and LCE parameters input for the
ocean model initialization
TABLE 1. LC parameters.
Parameter(s) Description
2 LCs intrusion into GoM initialized with 2 axes
-87.3, 27.4 Lat/Lon position of northernmost LC intrusion
-86.6, 23.5 Lat/Lon position of LC as it enters the GoM
-84.6, 24.6 Lat/Lon position of LC at merge with Florida Current
-89.5, 25.5 Lat/Lon position of western edge of LC axis bend
-86.6, 25.5 Lat/Lon position of eastern edge of LC axis bend
0 LCPROFILE is defined using GDEM Caribbean profile

TABLE 2. LCE parameters.
Parameter(s) Description
2 Two rings are simulated in the GoM
1 First ring is a WCR (WCR1)
0.8 Not used in this case because using real data
-93.3, 26.7 Lat/Lon position of western edge of WCR1 major axis
-89.7, 26.7 Lat/Lon position of eastern edge of WCR1 major axis
1.2 Radius (deg) of WCR1 minor axis
1 WCR1PROFILE is defined using real data (see below)
18
Summary

• Using a feature-based modeling approach that
  assimilates satellite-derived SSH and SST, a new
  ocean initialization has been developed and
  implemented operationally for the GFDL coupled
  model in 2006.
• Feature-based initialization provides more
  accurate estimates of OHC than a continuously run
  three-dimensional ocean model (i.e. RSMAS HYCOM)
• This initialization technique is being
  transitioned to the HWRF system for the 2007
  hurricane season.