OPERATIONAL OCEAN FORECAST FOR THE NORTHERN IBERIA PENINSULA

(OOF-NIP)

Investigación e Predición Numérica - MeteoGalicia - Consellería de Medio Ambiente e Ordenación do Territorio Xunta de Galicia

Introduction

The Regional Ocean Modelling System (ROMS) developed by Rutgers University has beenused to set-up the Operational Ocean Forecast for the Northern Iberia Peninsula (OOF-NIP) whichis being used as an additional operational ocean forecast system at MeteoGalicia. In ROMS modelthe primitive equations governing ocean dynamics and thermodynamics are discretized onto anArakawa C-grid to obtain numerical solutions (Haidvogel et al., 2000; Shchepetkin andMcWilliams, 2005). ROMS uses orthogonal curvilinear coordinates in the horizontal and stretchedterrain following sigma coordinates in the vertical (Haidvogel et al., 2000; Song et al., 1994).

For the purpose of computational efficiency the code utilises the natural time scaleseparation of barotropic and baroclinic processes by employing a mode-splitting algorithm whichsolves the vertically-integrated barotropic momentum equations using a much smaller time step(Mason et al., 2010). A specially designed fast-time-averaging procedure prevents aliasing ofprocesses unresolved by the longer baroclinic time step and, at the same time, maintains allnecessary conservation properties (Shchepetkin and McWilliams, 2005).

Model setup

The domain of the Northern Iberia ocean model based on ROMS extends from 14ºW to4.5ºW in the west-east direction and from 38ºN to 46ºN in the south-north direction (Figure 1). Thehorizontal resolution of the Northern Iberia ocean model is 0.02° (approximately 2 km) and it has41 sigma levels in the vertical. The vertical stretching parameters are chosen in such a way that thevertical resolution is highest in the upper part of the ocean.

The lateral boundaries are treated as open where the tracer and momentum fields are relaxedto Mercator global ocean analysis and forecast system fields with radiation condition with nudging(Marchesiello et al., 2001). For the barotropic mode the normal velocity component uses a Flather-type condition (Flather 1976) based on radiation and the prescription of characteristic variables(Riemann invariants: Blayo and Debreu 2005). A Chapman boundary condition is applied to thefree surface.

In order to ensure a more realistic bathymetry near the coast the final topography (Figure 1)of the model was merged with the general bathymetric chart of the oceans (GEBCO) with 30 arc-second resolution and cartography from two governmental institutions: Instituto Hidrográfico de laMarina, Cádiz (IHC) and from Instituto Hidrográfico Português (IH).

Tides are included in the OOF-NIP with eleven tidal constituents (M2, S2, N2, K2, K1, O1,P1, Q1, mf, mm) from the OSU TOPEX/Poseidon Global Inverse Solution version 7.2.

Atmospheric forcing

Meteorological fields are extracted from the 12km horizontal resolution operationalmeteorological forecast system (covering the whole Iberian Peninsula) implemented inMeteoGalicia through the use of the bulk flux algorithm (Fairall et al. 2004). MeteoGalicia runs theWRF model twice a day (00 and 12 UTC) on three different resolution domains (36, 12 and 4km)using as a boundary condition the solution of the Global Forecast System (GFS) model.

Figure 1. Map of model domain used to set-up Northern Iberia ocean model ROMS. In the rightpanel the four sources of the ocean topography: GEBCO, Instituto Hidrográfico de la Marina,Cádiz (IHC) and Instituto Hidrográfico Português (IH) with 0.5 and 1 mile of resolution.

River runoffs

Flow measurements at the mouth of the rivers are very difficult to acquire so the alternativeis to use a hydrological tool which will give flow forecasts for the coming days. The Soil waterAssessment Tool (SWAT model) developed by Agricultural Research Service and Texas A&MUniversity is the most commonly used. Along shoreline the daily average flow and temperaturefrom the main rivers for the Cantabria, Galicia and Portuguese regions are introduced in theNorthern Iberia Ocean model. From north to south: Sella, Nallon, Navia, Eo, Eume, Mandeo, Mero,Xallas, Tambre, Ulla, Umia, Lerez, Verdugo, Minho, Lima, Douro, Mondego and Tagus, (Figure 2).River Lima, Douro, Mondego, and Tagus daily flow measures are obtained from the SistemaNacional de Informação de Recursos Hídricos (SNIRH - Portuguese Information System for WaterResources).

Downscalling of Mercator global ocean analysis and forecast

For acquiring satisfactory initial and boundaries conditions the interpolation process wasdivided into two stages:

1. For a given time a horizontal stage where the parent (Mercator global ocean analysisand forecast) variables 2D and 3D are interpolated to the horizontal coordinates(longitude, latitude) of the new child domain (Northern Iberia Ocean model with

0.02° of resolution).

2. The vertical stage for each 3D variable a vertical interpolation transforms the datafrom the parent (Mercator global ocean analysis and forecast) z coordinates to thechild (Northern Iberia Ocean model with 0.02° of resolution) sigma coordinates.

Depth mismatches between child and parent grids may be significant in these regions andbecome a problem when they lie along open boundaries, because volume conservation is difficult toenforce (Mason et al., 2010). In order to prevent this problem a consistency check is made betweenthe sea surface height and the ocean current velocities.

Figure 2. Rivers implemented in Northern Iberia computational domain (North to South: Sella,Nallon, Navia, Eo, Eume, Mandeo, Mero, Xallas, Tambre, Ulla, Umia, Lerez, Verdugo, Minho,Lima, Douro, Mondego and Tagus). Data frame based on WGS84 datum.

Operational implementation

OOF-NIP is a high-resolution ocean forecasting system implemented over the OceanAtlantic part of the Northern Iberia Peninsula it is run operationally on a daily basis producing 24hour hind-cast and 96 hour forecasts providing a full 3D representation of the ocean (temperature,salinity, currents and sea surface height). This system is forced using WRF, SWAT and Mercatorglobal ocean forecast fields.

Atmospheric forcing from WRF model - shortwave radiation revision

In some particular cases we detected significant differences in the forecast of the SeaSurface Temperature (SST) forecast of the OOF-NIP when compared against in-situ and satelliteobservations. To improve the SST forecast we use observed satellite short-wave radiation form OSI-SAF on the hind-cast day of the simulation. Since the simulations are continued from the end of thehind-cast day the introduced BIAS it is much smaller and this has greatly helped to improve theSST prediction on the OOF-NIP model.

Products publication

OOF-NIP outputs (real-time and archived products) are freely available in a dedicatedthredds server for research, educational and commercial use. Available data:

1. Model output interpolated to fixed depth layers. NetCDF files with ocean variables at fixeddepths (1-hourly output). This variables are derived from the raw output from OOF-NIPmodel.

2. Model raw output. NetCDF files with original output (900-seconds window output) fromOOF-NIP model.

Figure 3. MeteoGalicia thredds server.

Results

In this section we show the results from OOF-NIP model compared with four differentsources of observations: data measured at moored stations belonging to the Puertos del Estado buoynetwork (Figure 4), ODYSSEA (ODYSSEA Sea Surface Temperature Analysis), Argo and RadarHF.

Figura 4. Map with the locations of the ocean stations.

Table 1. Statistical parameters BIAS, MAE and RMSE (time period between 7th of May 2014 to25th of March 2015).

OOF-NIP SILLEIRO OOF-NIP BARES OOF-NIP VILLANO

TEMP SAL U – VEL V – VEL TEMP SAL TEMP SAL

BIAS 0,70 -0,01 -0,01 -0,01 0,54 0,05 0,66 0,04

MAE 0,73 0,09 0,17 0,18 0,66 0,08 0,78 0,06

RMSE 0,89 0,12 0,23 0,24 0,81 0,11 1,01 0,09

→ Ocean temperature and salinity at 3 meters. Sea surface height

Figures 5 to 7 show the time series from the 7th of May 2014 to 25th of March 2015 of theOOF-NIP ocean temperature and salinity at 3 meters depth compared with Silleiro, Bares andVillano-Sisargas measured data. The OOF-NIP was quite succefull in capture the decrease-increasein temperature and salinity that occurred during this past 9 months. Although with some errorduring the rainy months the OOF-NIP was also able to reproduce the sudden salinity dropsassociated with river plumes.

Evaluating the entire period, Table 1, the average magnitude of the difference between theforecast and observations, the mean absolute error and the root mean square error show that OOF-NIP was able to produce lower scores when looking at salinity. There were also occasions when theOOF-NIP model show a tendency to overestimate the temperature at 3 meters depth. The bias andrmse observed in temperature were greather than 0.5 ºC but less than 1ºC.

As can be seen in Figure 8 the sea surface height forecast obtained by the OOF-NIP modelprovides a good adjustment to the data observed by the tide gauge Vigo2 both in terms of phase andamplitude.

Figure 5. Time series of the temperature and salinity at 3m depth time period between 7th of May2014 to 25th of March 2015 for Silleiro Bouy (red line) and OOF-NIP model (green line)

Figure 6. Time series of the temperature and salinity at 3m depth time period between 7th of May2014 to 25th of March 2015 for Estaca de Bares Bouy (red line) and OOF-NIP model (green line)

Figure 7. Time series of the temperature and salinity at 3m depth time period between 7th of May2014 to 25th of March 2015 for Villano-Sisargas Bouy (red line) and OOF-NIP model (green line)

Figure 8. Observed sea surface height in Vigo2 tide gauge (read line) and sea surface heightforecast by OOF-NIP model (green line).

→ Sea surface temperature

Figures 9a and 9b shows the monthly average (3 month average) of the sea surfacetemperature forecast from the OOF-NIP model (above figure) compared with the analysed seasurface temperature from ODYSSEA (middle figure) and the estimated error standard deviation ofanalysed sea surface temperature (below figure). Figure 9a shows the monthly average fromNovember 2014 to February 2015 and Figure 9b from June to September 2015. The most visibledifferences in the sea surface temperature occur along the north and south boundaries. The presentresults show slight tendency of OOF-NIP model to overestimate the sea surface temperature in thesouthern and underestimate over the northern probably influenced by the radiation lateral boundarycondition with nudging.

Despite not being able to state with total certainty because of the estimated error standarddeviation of analysed sea surface temperature near the shore the OOF-NIP model was able toreasonable forecast the sea surface temperature.

The spatial structures present in the satellite sea surface temperature observations are alsopresent in OOF-NIP sea surface temperature forecast.

Figure 9a. Three month (November 2014 to February 2015) horizontal average comparisonbetween the sea surface temperature forecast from the OOF-NIP model (above figure) and theanalysed sea surface temperature from ODYSSEA (middle figure). In the below figure theestimated error standard deviation of analysed sea surface temperature.

Figure 9b.Three month (June to September 2015) horizontal average comparison between the seasurface temperature forecast from the OOF-NIP model (above figure) and the analysed sea surfacetemperature from ODYSSEA (middle figure). In the below figure the estimated error standarddeviation of analysed sea surface temperature.

→ Argo temperature and salinity vertical profiles

Argo profiles located within the model domain were downloaded on a daily basis from theIFREMER FTP site. On a daily basis the Argo temperature and salinity profiles were extracted andcompared with model temperature and salinity profiles from the same location and closesttimestamp. Float locations (Figures 10, 11) and profile plots (Figures 10a to 10d, 11a to 11d) as wellas quantitative model skill metrics for each profile (bias, mae and RMSE) were calculated toevaluate the performance of the model. The temperature error increase with depth at all latitudes.Average errors below 1000 m are small less than 1.0ºC for the 40°S - 44°N region. The errors insalinity are smaller than 0.2.

Figure 10. Float Argo Trajectory for platform 6900701 (Float Cycles 102, 107, 122 and 122).

Figure 10a. Temperature and salinity profiles from ARGO at 42.11ºN and -9.89ºW compared to theOOF-NIP model on August 9, 2014. (Salinity is flagged as bad)

Temperature profile: BIAS = -0.81; MAE = 0.92; RMSE = 1.22

Figure 10b. Temperature and salinity profiles from ARGO at 41.62ºN and -10.50ºW compared tothe OOF-NIP model on September 28, 2014. (Salinity is flagged as bad)

Temperature profile: BIAS = -0.85; MAE = 1.2; RMSE = 1.7

Figure 10c. Temperature and salinity profiles from ARGO at 41.11ºN and -9.60ºW compared to theOOF-NIP model on November 17, 2014. (Salinity is flagged as bad)

Temperature profile: BIAS = -0.43; MAE = 0.89; RMSE = 1.16

Figure 10d. Temperature and salinity profiles from ARGO at 42.10ºN and -12.51ºW compared tothe OOF-NIP model on February 25, 2015. (Salinity is flagged as bad)

Temperature profile: BIAS = -0.90; MAE = 0.95; RMSE = 1.35

Figure 11. Float Argo Trajectory for platform 6900538 (Float Cycles 10, 14, 22 and 26).

Figure 11a. Temperature and salinity profiles from ARGO at 40.89ºN and -9.76ºW compared to theOOF-NIP model on January 25, 2015.

Temperature profile: BIAS = 0.08; MAE = 0.62; RMSE = 0.87

Salinity profile: BIAS = 0.06; MAE = 0.14; RMSE = 0.17

Figure 11b. Temperature and salinity profiles from ARGO at 41.19ºN and -10.18ºW compared tothe OOF-NIP model on December 16, 2014.

Temperature profile: BIAS = -0.43; MAE = 0.68; RMSE = 1.01

Salinity profile: BIAS = -0.11; MAE = 0.13; RMSE = 0.18

Figure 11c. Temperature and salinity profiles from ARGO at 41.58ºN and -10.99ºW compared tothe OOF-NIP model on September 27, 2014.

Temperature profile: BIAS = -0.74; MAE = 0.78; RMSE = 1.03

Salinity profile: BIAS = -0.09; MAE = 0.14; RMSE = 0.18

Figure 11d. Temperature and salinity profiles from ARGO at 41.35ºN and -11.10ºW compared tothe OOF-NIP model on August 18, 2014.

Temperature profile: BIAS = -0.57; MAE = 0.77; RMSE = 1.19

Salinity profile: BIAS = -0.05; MAE = 0.13; RMSE = 0.17

→ Ocean currents at 3 meters

Figures 12 to 14 show the time series from 2th of July 2014 to 3th September 2014 andFigures 15 to 16 show the time series from 7th of January 2015 to 18th March 2015 of the oceancurrent at 3 meters depth compared with Silleiro, Bares and Villano-Sisargas measured data.

The OOF-NIP in the first period from 2th of July 2014 to 3th September 2014 show atendency to overestimate in Silleiro, Bares and Villano-Sisargas the intensity of the ocean current at3 meters depth. On days 11 and 27 of August 2014 the OOF-NIP was almost able to reproduce theobserved increase in the ocean current at 3 meters in Villano-Sisargas buoy.

In the second period from 7th of January 2015 to 18th March 2015 the OOF-NIPunderestimate the intensity of the ocean current at 3 meters depth in Villano-Sisargas particularly onthe last two months.

Figure 17 show the time series from the 7th of May 2014 to 25th of March 2015 of theOOF-NIP model ocean current North/South component (v) and East/West component (u) at 3meters depth compared with Silleiro measured data. The OOF-NIP was successful in capture thedecrease-increase that occurred during this past 9 months. When evaluating the entire period, Table1, the average magnitude of the difference between the forecast and observations, the mean absoluteerror and the root mean square error show that OOF-NIP was able to produce lower scores.

From the 7th of May 2014 to 25th of March 2015 the current rose at 3 meters depth (Figure18) show that the frequency of the ocean currents observed in Silleiro buoy is between 180º and240º. The OOF-NIP model show good agreement since the frequency of the ocean currents arebetween 200º and 270º.

Silleiro Buoy Puertos del Estado July 2, 2014 to September 3, 2014

Figure 12a. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 2th of July to 3th of September 2014 forSilleiro Bouy.

Figure 12b. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 2th of July to 3th of September 2014 forOOF-NIP model.

Estaca de Bares Buoy Puertos del Estado July 2, 2014 to September 3, 2014

Figure 13a. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 2th of July to 3th of September 2014 forEstaca de Bares Bouy.

Figure 13b. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 2th of July to 3th of September 2014 forOOF-NIP model.

Villano-Sisargas Buoy Puertos del Estado July 2, 2014 to September 3, 2014

Figure 14a. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 2th of July to 3th of September 2014 forVillano-Sisargas Bouy.

Figure 14b. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 2th of July to 3th of September 2014 forOOF-NIP model.

Silleiro Buoy Puertos del Estado January 7, 2015 to March 18, 2015

Figure 15a. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 7th of January to 18th of March 2015 forSilleiro Bouy.

Figure 15b. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 7th of January to 18th of March 2015 forOOF-NIP model.

Villano-Sisargas Buoy Puertos del Estado January 7, 2015 to March 18, 2015

Figure 16a. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 7th of January to 18th of March 2015 forVillano-Sisargas Bouy.

Figure 16b. Time series of the ocean current at 3m depth (velocity sticks, speed and east [blueline] north [green line] velocity) time period between 7th of January to 18th of March 2015 forOOF-NIP model.

Figure 17. Time series of the ocean current North/South component (v) and East/West component(u) at 3m depth time period between 7th of May 2014 to 25th of March 2015 for Silleiro Bouy (redline) and OOF-NIP model (green line).

Figure 18. Silleiro buoy and OOF-NIP model current rose at 3 meters depth (time period from 7thof May 2014 to 25th of March 2015).

→ Sea surface currents. HF radar

Real-time monitoring systems such as HF radar were used for validation of the real-timeprediction system. Daily averaged OFF-NIP model current fields for two typical observed patternsEast to West above Figure 17 and North to South below Figure 17 were compared with HF radarcurrents. The comparison show good qualitative agreement to HF radar current near the coast,however offshore the OFF-NIP model exhibit smaller ocean current magnitude than the radar HF.

Figure 17. Radar HF (left figure) and OOP-NIP model (right figure) surface ocean currentdirection averaged for 14th December 2014 (above figure) and 21th January 2015 (below image).

Conclusions

An ocean modelling system was developed for the Northern Iberia Peninsula based on theROMS model. Forced with meteorological, hydrological and tidal data. A statistical study wasperformed for 10 months (7th of May 2014 to 25th of March 2015) comparing OOF-NIP modelagainst moored stations belonging to the Puertos del Estado buoy network, ODYSSEA sea surfacetemperature analysis, Argo buoys and radar HF. The OOF-NIP model have considerable good skillat reproducing the temperature and salinity near the coast. The forecast of the sea surfacetemperature is coherent although exhibiting a tendency to overestimate in the southern boundary.The vertical structure of ocean temperature and salinity of the OOF-NIP model is in agreement withthe observations. For example the errors in salinity are smaller than 0.2 psu and in temperature isalways less than 1ºC. In general the OOF-NIP model slight overestimate the ocean currents at 3meters depth.

Finally, it is important to mention that the aim of this report is to give an overview of thechosen operational setup and present some results.

References

Blayo E., Debreu L. 2005. Revisiting open boundary conditions from the point of view ofcharacteristic variables. Ocean Model. 9: 234-252. Flather, R.A. 1976. A tidal model of the north-west European continental shelf. Mem. Soc. R. Sci.Liège 6: 141-164.Fairall C.W., Bradley E.F, Rogers D.P., Edson J.B., Young G.S.1996. Bulk parameterization of airsea fluxes for tropical ocean global atmosphere Coupled Ocean Atmosphere Response Experiment.J. Geophys. Res. 101: 3747-3764.Haidvogel, D.B., Beckman A. 1999. Numerical Ocean Circulation Modeling. Imperial CollegePress, 318 pp.Marchesiello P., McWilliams J., Shchepetkin A. 2001. Open boundary conditions for long-termintegration of regional oceanic models. Ocean Modelling, 3, 1-20.Mason E., Molemaker J., Shchepetkin A., Colas F., McWilliams J., Sangrà P. 2010. Procedures foroffline grid nesting in regional ocean models. Ocean Modelling. Elsevier. 35:1-15Shchepetkin A.F., McWilliams J.C. 2005. The Regional Ocean Modeling System: A split-explicit,free-surface, topography following coordinates ocean model. Ocean Model. 9: 347-404.Song Y., Haidvogel D.B. 1994. A semi-implicit ocean circulation model using a generalizedtopography-following coordinate system. J. Comp. Phys. 115: 228-244.