Evaluation of a dynamically downscaled atmospheric reanalyse in the prospect of forcing long term simulations of the ocean circulation in the Gulf of Lions

https://doi.org/10.1016/j.ocemod.2009.07.004Get rights and content

Abstract

The paper evaluates atmospheric reanalysis as possible forcing of model simulations of the ocean circulation inter-annual variability in the Gulf of Lions in the Western Mediterranean Sea between 1990 and 2000. The sensitivity of the coastal atmospheric patterns to the model resolution is investigated using the REMO regional climate model (18 km, 1 h), and the recent global atmospheric reanalysis ERA40 (125 km, 6 h). At scales from a few years to a few days, both atmospheric data sets exhibit a very similar weather, and agreement between REMO and ERA40 is especially good on the seasonal cycle and at the daily variability scale. At smaller scales, REMO reproduces more realistic spatio-temporal patterns in the ocean forcing: specific wind systems, particular atmospheric behaviour on the shelf, diurnal cycle, sea-breeze. Ocean twin experiments (1990–1993) clearly underline REMO skills to drive dominant oceanic processes in this microtidal area. Finer wind patterns induce a more realistic circulation and hydrology of the shelf water: unique shelf circulation, upwelling, temperature and salinity exchanges at the shelf break. The hourly sampling of REMO introduces a diurnal forcing which enhances the behaviour of the ocean mixed layer. In addition, the more numerous wind extremes modify the exchanges at the shelf break: favouring the export of dense shelf water, enhancing the mesoscale variability and the interactions of the along slope current with the bathymetry.

Introduction

One major challenge for long term coastal modelling is the use of atmospheric forcings which are able to drive the complex coastal dynamics and which are consistent over long period. Beside global atmospheric reanalysis, the use of regional downscaling appears as a possible solution. In this paper, we compare the spatial and temporal variability scales of two atmospheric data sets in the prospect of driving a coastal ocean circulation model in the Gulf of Lions (a large continental shelf located in the North Western Mediterranean) over the period of a decade (1990–2000): one is produced by the REMO regional climate model (Jacob et al., 2001), and the other is the recent global atmospheric reanalysis ERA40 (Uppala et al., 2005). The ocean circulation in the Gulf of Lions is characterized by numerous physical processes which are intermittent in both space and in time: coastal up-and-downwellings (Millot, 1979, Johns et al., 1992), winter dense water formation on and off the shelf (MEDOC – MEDiterranean OCcidental survey, 1970; Fieux, 1974, Mertens and Schott, 1998), dense water cascading down the shelf break (Bougis and Ruivo, 1954, Bethoux et al., 2002, Shapiro et al., 2003, Dufau-Julliand et al., 2004), freshwater river plumes (Simpson, 1997, Estournel et al., 2001, Reffray et al., 2004), eddies and meanders of the North Mediterranean Current (NMC, hereafter) (Albérola et al., 1995, Flexas et al., 2002, Jordi et al., 2005), local and peculiar shelf circulation patterns(Estournel et al., 2003, Petrenko et al., 2005), inertial and diurnal motions (Millot and Crépon, 1981, Van Haren and Millot, 2003). With no significant tidal forcing, the ocean circulation on the shelf and at the shelf break is mainly driven by the atmospheric forcing with a large range of variability scales (Huthnance, 2002).

The seasonal atmospheric variability dominates the coastal hydrology changes in time, involving the formation of a seasonal thermocline, the NMC strength and variability, and episodic cold and dense water formation events. At time scales of few days, the wind is the main driving mechanism for the up-and-downwelling system, the positioning of Rhône river plume, freshwater spreading and the exchanges across the shelf break. At even shorter time scales, solar radiation and thermal breeze generate strong inertial and diurnal motions in the mixed layer with a noticeable impact on the Sea Surface Temperature (SST).

Spatially, the surrounding land orography channels the northern continental winds. This generates a sheared wind system on the shelf (Fig. 1): Mistral landwind (channelled between Alps and Massif Central mountains) and Tramontane landwind (channelled between Massif Central and Pyrenees mountains) may blow together or separately, generating highly variable curl patterns on the shelf and driving very intermittent current patterns on the shelf (Estournel et al., 2003).

In the long term, the mean state and the interannual variability of the circulation on the shelf is the result of the integration over time of the interaction of the previously introduced processes, which occur in a wide range of frequencies and spatial scales and exhibit a strong intermittent character. Understanding the ocean circulation in the Gulf of Lions and its variability at the interannual time scale therefore requires the action of fine and intermittent processes and as well as an accurate description of the atmospheric forcing at fine temporal and spatial scales. Regional dynamical downscaling of global atmospheric reanalyses, such as the REMO experiment (Jacob et al., 2001) evaluated in this study, is one way to approach such a description.

The major long-term reanalyses commonly used to drive global ocean models are those performed at NCEP/NCAR (NCEP1, Kalnay et al., 1996) and at ECMWF (ERA40, Uppala et al., 2005). Despite their coarse resolution (over 100 km), they provide a reliable basis for the definition of the atmospheric forcing of ocean general circulation model simulations over the past decades (Large and Yeager, 2004, Röske, 2006, Griffies et al., 2008, Brodeau et al., 2009). Their resolution is however insufficient to match the requirements of an accurate forcing of the ocean dynamics of the Western Mediterranean Sea and the Gulf of Lions (Demirov and Pinardi, 2007, Hermann et al., in press) and higher resolution products have to be applied where possible. Such products can be provided by dynamical downscaling of global reanalyses with regional atmospheric models (Giorgi, 1990, Sotillo et al., 2005, Hermann and Somot, 2008). In Sotillo et al. (2005) and Hermann and Somot (2008), dynamical downscalings based on spectral nudging have been performed over the Mediterranean Sea, with spatial resolutions close to 50 km.

At the Max Planck Institute for Meteorology, the ERA15 reanalysis and ECMWF operational analyses have been dynamically downscaled with the regional atmospheric circulation model REMO (Jacob and Podzun, 1997, Jacob et al., 2001) over a period of 25 years (1979–2003) with a horizontal resolution of approximately 18 km on a model domain covering Europe and the entire Mediterranean region. The objective of this paper is to evaluate the hourly outputs of the above REMO downscaling as a possible atmospheric forcing of model simulations of the ocean circulation inter-annual variability in the Gulf of Lions between 1990 and 2000. The paper compares the atmospheric surface variables and fluxes entering the forcing function of an Ocean General Circulation Model (OGCM) produced by REMO with those of ERA40. The impact of the two atmospheric datasets on the oceanic circulation is also studied through the comparison of OGCM outputs.

Three issues are discussed in this paper. They are linked with the evaluation of the REMO atmospheric downscaling over the period of interest (1990–2000), but are also relevant to long term simulations of coastal regions. The first one is to evaluate what is gained by using the downscaling in term of spatial resolution of the forcing fields. The second major issue for simulations during a full decade concerns the mean value and low frequency variability of the forcing over the whole basin. Mainly, we investigate if the seasonal to interannual variability of the total amount of heat, freshwater and momentum inputs in the gulf during these 11 years is reproduced between the two datasets. The third issue addressed by our analysis concerns the intermittency and the high frequency content of the atmospheric variables in the REMO analysis. Because REMO provides hourly outputs (against 6 hours integrated outputs for ERA40), it is likely that it resolves the diurnal cycle of the solar radiation and the sea/land breeze. Because those processes have a considerable impact on the ocean mixed layer during the summer, their description in REMO has to be evaluated.

In the prospect of forcing coastal ocean circulation, our assessment makes use of simulations of the ocean circulation in the Gulf of Lions performed with a high resolution (1/64°, i.e. about 1.25 km) regional model (hereafter GoL64) (Langlais, 2007), during a 4 year period (1990–1993).

The paper is organized as follows. Section 2 introduces the REMO downscaling and the ERA40 reanalysis, and describes the atmospheric variables that are required to build the forcing of the ocean model. Section 3 compares the spatial patterns of crucial atmospheric variables in the two datasets and provides an assessment of the benefit brought by the downscaling procedure. Section 4 investigates the seasonal to interannual variability of the forcing variables, comparing the climatic relevance of the two datasets. Section 5 identifies high frequency (daily and shorter) coastal atmospheric processes resolved by REMO. Section 6 summarises the major findings regarding discrepancies and similarities in the data sets, and briefly discusses their impact on the solution of a numerical ocean model of the Gulf of Lions. Conclusions are drawn to assess the skill of the REMO downscaling as an ocean model forcing for the considered decadal period.

Section snippets

Atmospheric reanalyses and ocean forcing

This section describes the two atmospheric forcing datasets used in the paper: one extracted from the REMO dynamical downscaling, and one extracted from the global reanalysis ERA40. The study area where both data sets are compared is presented in Fig. 2a. As the final objective is to find a relevant atmospheric forcing for long term coastal simulations of the ocean circulation in the Gulf of Lions, the atmospheric outputs are interpolated over the grid of the GoL64 ocean model (Langlais, 2007)

Spatial patterns

In this section, we compare the spatial patterns of the atmospheric and flux variables in the Gulf of Lions as described by the two atmospheric datasets. The analysis focuses on the reproduction of the main atmospheric situations relevant to the ocean dynamics on the shelf. In order to underline the instantaneous and the long term impacts of the spatial resolution, snapshots and time means will be compared.

Interannual and seasonal variability

This section compares the variability of REMO and ERA40 at the interannual to seasonal scales. The near surface atmospheric parameters and the air–sea fluxes are spatially integrated over the two regions (shelf and open-sea) defined in Section 3.3. This separation is relevant for the ocean as these two regions are characterized by very different dynamical regimes and are separated by the NMC (which flows along the shelf edge). Annual mean and monthly averaged values of the forcing input

Daily to hourly content

We now examine the high frequency content of the input forcing parameters. The weather in the Gulf of Lions is known to be dominated by strong and intermittent events, especially wind bursts of a few hours to a few days. These events are known to have a strong impact on the ocean circulation in short and even in long term, since the ocean is integrating their effects (Hermann and Somot, 2008, Langlais et al., in preparation). Therefore, a high temporal resolution of the atmospheric forcing is

Discussion and conclusion

We have compared two atmospheric data sets that can be used to drive a regional ocean circulation model of the Gulf of Lions (the GoL64 model, with a 1/64° resolution) over periods of a decade or longer: the REMO dynamically downscaled analysis, and the recent global atmospheric reanalysis ERA40. Compared to ERA40, REMO has a finer grid-resolution (18 versus 125 km), and a better time sampling (1 versus 6 h). The surface atmospheric parameters which enter the surface boundary condition of the

Acknowledgments

The authors acknowledge the support from Ministère de l’Education nationale et de la Recherche and from Centre National de la Recherche Scientifique. We acknowledge the support from the French national programme GMMC, from the French space agency CNES. The ocean model simulations have been performed on the supercomputers of Institut du Développement et des Ressources en Informatique Scientifique IDRIS. We warmly thank Sergey Gulev who connected the French scientists in LEGI and LSEET with the

References (53)

  • J.H. Simpson

    Physical processes in the ROFI region

    J. Mar. Syst.

    (1997)
  • H. Van Haren et al.

    Seasonality of internal gravity waves kinetic energy spectra in the ligurian basin

    Oceanol. Acta

    (2003)
  • C. Albérola et al.

    On the seasonal and mesoscale variabilities of the northern current during the PRIMO-0 experiment in the western Mediterranean Sea

    Oceanol. Acta

    (1995)
  • Baltic Sea Experiment BALTEX, 1995. Initial implementation plan. International BALTEX secr., N° 2, Geesthacht,...
  • P. Bougis et al.

    Sur une descente des eaux superficielles en profondeur (cascading) dans le sud du golfe du Lion

    Bull. Inf. COEC

    (1954)
  • K.E. Brainerd et al.

    Diurnal restratification and turbulence in the oceanic surface mixed layer 1. Observations

    J. Geophys. Res.

    (1993)
  • Brodeau, L., Barnier, B., Penduff, T., Treguier, A.M., Gulev, S., 2009. An ERA40 based atmospheric forcing for global...
  • H.C. Davies

    A lateral boundary formulation for multi-level prediction models

    Quart. J. R. Meteor. Soc.

    (1976)
  • E.K. Demirov et al.

    On the relationship between the water mass pathways and eddy variability in the Western Mediterranean Sea

    J. Geophys. Res.

    (2007)
  • C. Dufau-Julliand et al.

    Three-dimensional modeling of the Gulf of Lions’ hydrodynamics (northwest Mediterranean) during January 1999 (MOOGLI3 experiment) and late winter 1999: western Mediterranean Intermediate Water’s (WIW’s) formation and its cascading over the shelf break

    J. Geophys. Res.

    (2004)
  • Estournel, C., Durrieu de Madron, X., Marsaleix, P., Auclair, F., Julliand, C., Vehil, R., 2003. Observation and...
  • M. Fieux

    Formation d’eau dense sur le plateau continental du Golfe du Lion

    Colloques Internationaux du CNRS

    (1974)
  • F. Giorgi

    On the simulation of regional climate using a limited area model nested in a general circulation model

    J. Climate

    (1990)
  • S. Griffies et al.

    CoordinatedOcean-iceReference Experiments (COREs)

    Ocean Modell.

    (2008)
  • V. Guénard et al.

    An observationnal study of the mesoscale Mistral dynamics

    Bound. Layer Meteor.

    (2005)
  • Hermann, M., Somot, S., Sevault, F., Estournel, C., Déqué, M., in press. Modeling the deep convection in the...
  • Cited by (20)

    • Impacts of high-resolution atmospheric forcing and air-sea coupling on coastal ocean circulation off the Pearl River Estuary

      2022, Estuarine, Coastal and Shelf Science
      Citation Excerpt :

      Atmospheric forecasting skill improves as the atmospheric model's resolution increases, and more specific topographic features (Cavaleri and Bertotti, 2004; Ágústsson and Ólafsson, 2007; Albert et al., 2010) and mesoscale processes (Béranger et al., 2010) are resolved. High-resolution atmospheric forcings improved the ocean circulation simulations due to better representing the oceanic mixed-layer characteristics and dynamics (Estournel, 2003; Langlais et al., 2009), turbulent heat flux (Akhtar et al., 2017), oceanic response to large wind variability (De Mey et al., 2017), and thermohaline circulation (Castellari et al., 2000; Artale et al., 2016). The effects of atmosphere-ocean interaction and feedback become more important in an ocean simulation as the grid resolution increases.

    • Effects of a freshening trend on upper-ocean stratification over the central tropical Pacific and their representation by CMIP6 models

      2022, Deep-Sea Research Part II: Topical Studies in Oceanography
      Citation Excerpt :

      As the FWF exerts a significant influence on the modulation of ENSO events by affecting SSS and buoyancy flux (Zhang et al., 2012; Gao et al., 2020), the ENSO-related effect of the FWF trend should also be discussed. Meanwhile, in the context of global warming, not only do the temperature and salinity in the upper ocean have distinct trends, but the dynamic height, ocean circulations, vertical convections and properties of the deep ocean may also be affected continuously, which are all significant to ocean creatures, atmospheric dynamics and climate systems (Langlais et al., 2009; Miller et al., 2009; Lisa and Nadine, 2015). With the increase in observational data, secular trends in more oceanic fields should be considered.

    • The role of temporal resolution in modeling the wind induced sea surface transport in coastal seas

      2019, Journal of Marine Systems
      Citation Excerpt :

      The relevant role and the influence of the near-surface ocean winds on the marine circulation, at both offshore and coastal scale, was clarified in several studies (Gan and Allen, 2005; Alves, 2006; Jiang et al., 2008; Seo et al., 2008; Athié et al., 2009; Deng et al., 2009). As a consequence, the use of winds as forcing mechanism in ocean model simulations is a common practice as emphasized in Jiménez et al. (2008), Langlais et al. (2009) and Brodeau et al. (2010). Reliable winds are usually provided by meteorological forecasting systems that, based on global and/or regional atmospheric numerical models, are hosted by international and governmental bodies (e.g. the European Centre for Medium-Range Weather Forecast – ECMWF).

    • Circulation of the Mediterranean Sea and its Variability

      2012, The Climate of the Mediterranean Region: From the Past to the Future
    View all citing articles on Scopus
    View full text