1. Introduction 2. Goals 3. Methodology 4. Results 5. Other Results 6. Conclusions

GODAE OceanView / CLIVAR GSOP Technical Workshop on Observing System Evaluation and Inter-comparisons, 13-17 June 2011, UCSC

A Sensitivity Study of Short-Range Forecasts over the Atlantic METAREA V with HYCOM and the Cooper and Haines Scheme Using

Different Sea Surface Heights

Jean Oliveira1, João Alvarenga1, Raquel Mello1, Ana Freitas1, Lucimara Russo1 , Clemente Tanajura2

1 Brazilian Navy Center of Hydrography (CHM), Brazil2Institute of Physics, Federal University of Bahia (UFBA), Brazil

Oceanographic Modeling and Observation Network (REMO)www.rederemo.org

1. Introduction

2. Goals

3. Methodology

4. Results

5. Other Results

6. Conclusions

Outline

• REMO is a research group in physical oceanography and operational oceanography established in January 2008

• The members are researchers and students from four Brazilian universities, the Brazilian Navy and the PETROBRAS research center CENPES

1. Introduction

1. Introduction (cont.)

- Pre-operational short-range ocean forecasts with HYCOM (7 days) and ROMS (3 days) using NCEP GFS forcing since Oct ober 2009

- Operational forecasts with HYCOM in the Brazilian Navy CHM with the Cooper & Haines scheme (1996) (C&H) since February 2010

HYCOM 1/4o 21L HYCOM 1/12o 21L ROMS 5-13 km 32L

2. Goals - Evaluate HYCOM 1/12o Forecasts

- Investigate the behavior of the model predictability along time

- Investigate the model sensitivity to different initial conditions

- Develop routines to validate model simulations and forecasts according to GODAE OceanView standards

3. Methodology - Construct initial condition with u, v, T and S from the HYCOM+NCODA

system on 00Z 1 Feb 2010 available in z-coordinates at www.hycom.org

- 7-day forecasts from 1 Feb 2010 until 31 Mar 2010 using GFS forcing

- HYCOM 1/4o : initial condition from C&H with the 24 h forecast as background and SSH from HYCOM+NCODA

- HYCOM 1/12o nested in the HYCOM 1/4o : initial condition from C&H with the 24 h forecast as background and SSH from

HYCOM+NCODA (NRL), Mercator-océan (MERCATOR) and SSH analysis (ALONG_TRACK)

3. Methodology (cont.)

- SSH analyses were performed daily with (filtered) along-track SSH anomalies (SSHA) from Jason-1 and Jason-2 obtained from AVISO.

- OI was employed to assimilate SSHA according to Ezer & Mellor (1997)

- The background SSHA was obtained from

- The analysis was calculated from

SSHSSHSSHA bb model mean

RH(HBBH=K ' 1)

222 /exp),( LrjiB iji

)SSHAHK(Y+SSHA=SSHA bba )(

R is diagonal with constant values equal to (0.02 m)2

L = 170 km

is the SSHA model variance at point i

is the distance between point s i and j222 )()( jijiij yyxxr

2i

3. Methodology (cont.)

- 747 Argo profilers with T-S were used to evaluate the forecasts

- Only the 24 h forecasts were investigated

- Temperature and salinity from Argo and model were interpolated to the standard Levitus vertical levels

Argo profilers1 Feb - 31 Mar 2010

4. Results

- NRL = C&H with HYCOM+NCODA

- ALONG_TRACK = C&H with SSHA analysis

- MERCATOR = C&H with Mercator-Océan 1/4º SSH

Control

Argo

Mean Temperature Profile Mean Salinity Profile

Experiments

Control

Argo

Experiments

- Standard deviations (Std) of the difference between the forescast and ARGO data for temperature and salinity were calculated

- Smallest values were obtained for the NRL followed by the CONTROL , ALONG_TRACK and MERCATOR experiments

4. Results (cont.)

Mean Temperature and Std (ΔT) Mean Salinity and Std (ΔS)0.2oC < Std (ΔT) < 2.4oC 0.2oC < Std (ΔT) < 2.4oC

Argo

Argo

NRL (dashed)

NRL (dashed)

4. Results (cont.)

0 – 2000 m

0 – 100 m 125 – 800 m

Taylor Diagrams for Temperature

A – ARGOB – NRL C – ALONG_TRACKD – MERCATORE – CONTROL

4. Results (cont.)

0 – 2000 m

0 – 100 m 125 – 800 m

Taylor Diagrams for Temperature

A – ARGOB – NRL C – ALONG_TRACKD – MERCATORE – CONTROL

4. Results (cont.)

0 – 2000 m

125 – 800 m0 – 100 m

Taylor Diagrams for Salinity

A – ARGOB – NRL C – ALONG_TRACKD – MERCATORE – CONTROL

16/09/10

5. Other results

- Initial condition: “interpolated” HYCOM+NCODA fields available in z-coordinate in www.hycom.org on 00Z 1 Jan 2011

- 7-day (168 h) Forecasts performed daily with Cooper and Haines, SSH from HYCOM+NCODA

- NCEP GFS forcing

Forecast minus PIRATA (8oS. 30oW) Forecast minus PIRATA (19oS. 34oW)

20 m 20 m300 m

300 m

100 m

100 m

Solid lines = 24 h forecastDashed lines = 168 h forecast

Tem

p.

Dif

fere

nce

(o

C)

Tem

p.

Dif

fere

nce

(o

C)

Leonardo Lima; Luiz Gavaza

16/09/10

5. Other results (cont.)

- Background: “interpolated” HYCOM+NCODA fields available in z-coordinate in www.hycom.org on 00Z 1 Jan 2011

- Bergthorsson and Doos analysis with 19 profiles of T and S from Argo T and S data from surface to 2000 m performed in the Levitus z-levels;

- Application of Cooper and Haines with SSH from HYCOM+NCODA

Davi Mingac; Luiz Gavaza

BackgroundBackground

Analysis+C&H

Analysis

RMSE for Temperature (oC) RMSE for Salinity (PSU)

Dep

th (

m)

Analysis+C&H

Analysis

16/09/10

6. Conclusions- The system HYCOM 1/12o – initialized with “interpolated”

HYCOM+NCODA fields on 1 Feb 2010 – restrained daily by Cooper and Haines with HYCOM+NCODA SSH produces good agreement with Argo data;

- When assimilation of AVISO along-track SSHA data was employed, similar (worse) results were obtained;

- The use of SSH from Mercator-Océan produced similar results to the control run without Cooper and Haines;

- In another forecast experiment (from 1 Jan 2011 to 31 Mar 2011) the 24h forecasts for T with respect to PIRATA buoys close to Brazil were very similar to the 168h forecasts

- The analysis of vertical profiles of T and S from Argo performed in z-coordinates were substantially modified by Cooper and Haines;

- An OI scheme to assimilate synthetic T(z) and ρ(z) is under evaluation and a multivariate OI system is being pursued to jointly assimilate SSHA, SST, T(z) and layer thickness.

Thanks!