Analysis of Mediterranean ocean variability using five numerical simulations

ANALYSIS OF MEDITERRANEAN OCEAN VARIABILITY USING FIVE NUMERICAL SIMULATIONS Enrique Vidal Vijande directed by Dr. Ananda Pascual Ascaso Palma de Mallorca, 2012 (~ AD DE LAS PALMAS DE GRAN CANARIA CSIC A Ln i \cn it3 ld c lcs --_ ..... T III l" nall'i,¡r\ ImEOEp. D SALVADOR GALVÁN HERRERA, SECRETARIO DEL DEPARTAMENTO DE FÍSICA DE LA UNIVERSIDAD DE LAS PALMAS DE GRAN CANARIA, CERTIFICA, Que el Consejo de Doctores del Departamento en su sesión de fecha .............................. tomó el acuerdo de dar el consentimiento para su tramitación, a la tesis doctoral titulada “ANALYSIS OF MEDITERRANEAN OCEAN VARIABILITY USING FIVE NUMERICAL SIMULATIONS” presentada por el doctorando ENRIQUE VIDAL VIJANDE y dirigida por la Doctora Dª ANANDA PASCUAL ASCASO. Y para que así conste, y a efectos de lo previsto en el Artº 73.2 del Reglamento de Estudios de Doctorado de esta Universidad, firmo la presente en Las Palmas de Gran Canaria, a........................de.............................................de dos mil doce Departamento de Física PROGRAMA DE DOCTORADO EN OCEANOGRAFÍA BIENIO 2007 - 2009 título de la tesis Analysis of Mediterranean Ocean Variability using Five Numerical Simulations [analisis de la variabilidad oceánica del mediterraneo utilizando cinco simulaciones numéricas] Tesis doctoral presentada por Enrique Vidal Vijande para obtener el grado de Doctor por la Universidad de Las Palmas de Gran Canaria. Dirigida por la Doctora Dª Ananda Pascual Ascaso, científico titular del CSIC en el IMEDEA(CSIC-UIB) Enrique Vidal Vijande Dra. Dª Ananda Pascual Ascaso Palma de Mallorca, Abril 2012 Acknowledgements Mi mayor agradecimiento es, sin lugar a duda, para Ananda Pascual, por haber sido una fant´astica directora de tesis y haberme guiado y apoyado durante estos a˜nos. Gracias por su enorme implicaci´on, su amistad y por haber creado un ambiente de trabajo enriquecedor y estimulante. Tambi´en agradezco profundamente su apoyo a mis iniciativas y proyectos personales. Ha sido para mi una verdadera mentora. Mis m´as sinceros agradecimientos a Joaqu´ın Tintor´e por haberme dado la oportunidad de empezar a trabajar en el IMEDEA, y luego por aceptar ser mi director de tesis durante el primer a˜no, y por su apoyo continuado a lo largo de estos a˜nos. Agradezco tambi´en la concesi´on de la beca pre-doctoral I3P del CSIC, la cual me ha permitido realizar esta tesis, y al proyecto EU-MyOcean FP7 por financiar mi asistencia a diversos congresos. Gracias a Alejandro Orfila por su apoyo y ayuda en escribir el proyecto de tesis que me permiti´o la concesi´on de la beca pre-doctoral, adem´as de sus valiosos consejos. IdeeplyacknowledgethehospitalityandgenerosityofBernardBarnierduring my stay in Grenoble, and also for providing access to the various ORCA simulations used throughout this thesis. We had very interesting discussions, and provided me with valuable insights into the inner workings of ocean models. I would also like to thank Jean-Marc Molines for all his technical assistance with the ORCA model data. I would like to thank Nicolas Ferry for providing the GLORYS1V1 data. I would like to thank Samuel Somot for providing the NEMOMED8 data. I would also like to thank Bruno Buongiorno Nardelli for his hospitality and friendship during my stay in Rome, for teaching me how to use EOFs for oceanographic data, and also for all the photographic conversations. Del Instituto de Ciencias del Mar de Barcelona, mis agradecimientos a Josep Luis Pelegr´ı por aceptar ser mi tutor durante los cursos de doctorado y por toda su ayuda y generosidad. De la Universidad de Las Palmas de Gran Canaria, gracias a ´ Angel Rodr´ıguez Santana y al equipo de Tercer Ciclo por su ayuda en los tramites para el dep´osito y lectura de esta tesis. i ii Gracias a mis padres Gerardo y Mar´ıa y a mi hermano Carlos por su incesante apoyo y ´animo, que me ha permitido llegar hasta aqu´ı. Esta tesis no hubiese sido posible sin ellos. A Laura, por estar siempre a mi lado, por haberme dado toda tu energ´ıa y apoyo, y mantener mi cabeza centrada en los momentos complicados, sobre todo en los ´ultimos meses. Abstract The oceans play a fundamental role on the slow evolution of climate and all life on earth, and must therefore be studied and understood. The increase in oceanographic research and available measurements over the past half century have greatly increased our knowledge about the ocean’s behaviour and variability, and highlighted its complexity and ubiquity over a wide range of space and time scales. However observational datasets still remain too short, too superficial or too dispersed in time and space to allow detailed studies of many of the physical processes governing the ocean’s variability. In order to continue advancing our understanding of how the oceans work, it is crucial to complement observational data with ocean numerical modelling studies. Throughout this thesis, we explore di↵erent aspects of the variability and circulation of the Mediterranean Sea, where many of the oceanic processes found throughout the worlds oceans can be studied in a reduced scale. We perform a comprehensive analysis through the validation of a series of ocean numerical models comparing them with observational datasets. The first part is an initial approach at the analysis in the Mediterranean Sea of a 1/4o global ocean simulation (ORCA025-G70). We focus on basin scale mean temperature (T) and salinity (S) temporal evolution at di↵erent depth layers by comparing to the MEDAR database, mean sea surface height temporal evolution and trends by comparing with altimetry, and transport through the Strait of Gibraltar and the Black Sea by comparing to the available literature. The results show that T variability at surface is well reproduced, but trends in intermediate and deep layers are positively biased. Poor results are obtained for S interannual variability, mainly due to the sea surface salinity (SSS) relaxation. Sea surface height interannual variability and seasonal cycle are well reproduced but climatological trends are too positive due to an imbalance in the freshwater fluxes from the atmospheric forcing. Water transport through the main straits is within the range of observed values. The second part is a continuation of the analysis with global models, this time focusing on the Western Mediterranean (WMED) and adding two more simulations, one with higher vertical resolution and improved atmospheric forcing (ORCA025-G85) and one with data assimilation (GLORYS). The results show that the combination of a higher vertical resolution and better time continuity of the atmospheric forcing in G85 o↵er better T trend results at intermediate and deep layers with respect to G70. On the other hand, G85 has a much weaker SSS restoring and displays unrealistic negative S trends. Two exceptional deep convection events in the Gulf of Lions are reproduced by G85 due to the extremity of the atmospheric conditions, while weaker convection events are not reproduced due to the coarse resolution of both the model and atmospheric forcing. GLORYS has a better T and S performance than the ORCA hindcasts. Temporal evolution of sea level interannual variability is well reproduced in all simulations, but the ORCA hindcasts, which do not maintain a water balance show exaggerated positive trends, especially G85 that has a weaker SSS restoring. Through EOF analysis, the simulations recreate the surface circulation variability patterns correctly within their resolution limits. Transports through the main straits appears to be iii iv correct when contrasted against the available literature. In the third and final part of this thesis we use two high resolution simulations, ORCA12 (1/12o) and NEMOMED8 (1/8o) focusing on the WMED. For the standard T and S analysis, the results of both simulations show an improvement in T, especially trends at intermediate and deep layers. In S, ORCA12 gives similar results to the previous simulations, however NEMOMED8 shows a marked improvement due to a better water balance and lack of SSS restoring. When reproducing deep convection in the Gulf of Lions, NEMOMED8 correlates well with observations, however ORCA12 creates a very warm and stable intermediate layer preventing most deep convection events from reaching beyond intermediate layers. The analysis of mean geostrophic currents revealed correct overall circulation scheme but each simulation had slight discrepancies with observations. When decomposing the circulation signal using EOFs, the temporal and spatial patterns of NEMOMED8 are consistent with observations, however ORCA12 had an overly smooth and stable circulation pattern. Overall, NEMOMED8 is found to be a mature and reliable simulation although some room for improvement remains. ORCA12 is an interesting push forward in terms of global high resolution modelling, but being a first iteration, much improvement is expected in following simulations. Resumen Los oc´eanos desempe˜nan un papel fundamental en la lenta evoluci´on del clima y de toda la vida en la Tierra, y por lo tanto deben ser estudiados y comprendidos. El aumento de la investigaci´on oceanogr´afica y los datos disponibles en el ´ultimo medio siglo han mejorado considerablemente nuestros conocimientos sobre el comportamiento del oc´eano y su variabilidad, resaltando su complejidad en un amplio rango de escalas espaciales y temporales. Sin embargo, las observaciones disponibles siguen siendo demasiado cortas, demasiado superficiales o demasiado dispersas en el tiempo y el espacio para permitir estudios detallados de muchos de los procesos f´ısicos que gobiernan la variabilidad del oc´eano. Con el fin de seguir avanzando en nuestra comprensi´on los oc´eanos, es fundamental complementar los datos observacionales con los estudios de modelizaci´on num´erica. A lo largo de esta tesis, se exploran diferentes aspectos de la variabilidad y la circulaci´on del Mar Mediterr´aneo, una cuenca en la que muchos de los procesos oce´anicos que tienen lugar en el oc´eano global se puede estudiar a una escala reducida. Para ello, llevamos a cabo un an´alisis exhaustivo a trav´es de la validaci´on de una serie de modelos num´ericos oce´anicos compar´andolos con datos observacionales. La primera parte es una primera aproximaci´on al an´alisis en el mar Mediterr´aneo de una simulaci´on global de 1/4 o(ORCA025-G70). Nos centramos en: temperatura y salinidad promedio a escala de cuenca y su evoluci´on temporal a diferentes profundidades mediante la comparaci´on con la base de datos MEDAR; evoluci´on temporal y tendencias del nivel del mar mediante la comparaci´on con altimetr´ıa, y, finalmente, el transporte a trav´es del Estrecho de Gibraltar y el Mar Negro, mediante la comparaci´on con las publicaciones disponibles. Los resultados muestran que la variabilidad de temperatura en superficie se reproduce bien, pero la tendencia en las capas intermedias y profundas est´an sesgadas positivamente. En salinidad se no se obtienen resultados satisfactorios en cuanto a la variabilidad interanual, y es debido principalmente a la relajaci´on de salinidad en superficie. La variabilidad interanual y el ciclo estacional del nivel del mar se reproducen correctamente, pero las tendencias climatol´ogicas son excesivamen positivas debido a un desequilibrio en los flujos de agua dulce del forzamiento atmosf´erico. El transporte del agua a trav´es de los estrechos principales se encuentra dentro del rango de valores observados. La segunda parte es una continuaci´on del an´alisis con modelos globales, esta vez centr´andose en el Mediterr´aneo Occidental (WMED) y la adici´on de dos simulaciones m´as similares a G70, una con una mayor resoluci´on vertical y la mejora del forzamiento atmosf´erico (ORCA025G85) y una con los asimilaci´on de datos (GLORYS). Los resultados muestran que la combinaci´on de una mayor resoluci´on vertical y una mejor continuidad en el tiempo del forzamiento atmosf´erico en G85 ofrecen mejores resultados en la tendencia de temperatura en las capas intermedias y profundas con respecto a G70. Por otro lado, G85 tiene una relajaci´on en salinidad mucho m´as d´ebil que G70 y muestra tendencias negativas irrealistas en salinidad. Dos acontecimientos excepcionales de convecci´on profunda en el Golfo de Le´on son reproducidos debido a la extremidad de las condiciones atmosf´ericas G85, mientras que los eventos de convecci´on m´as d´ebiles no se reproducen debido a la baja resoluci´on tanto del modelo v vi como del forzamiento atmosf´erico. La temperatura y salinidad de GLORYS muestran un mejor comportamiento que las simulaciones ORCA. La evoluci´on temporal de la variabilidad interanual del nivel del mar se reproduce bien en todas las simulaciones, pero en las ORCA, que no mantienen el balance h´ıdrico muestran tendencias positivas exageradas, sobre todo G85 que tiene un una relajaci´on en salinidad m´as d´ebil. A trav´es del an´alisis EOF, las simulaciones recrean razonablemente bien (dentro de los l´ımites de su resoluci´on) las caracter´ısticas de los patrones de circulaci´on superficial. Los transportes a trav´es de los estrechos principales parecen correctos cuando se comparan con las publicaciones disponibles. En la tercera y ´ultima parte de esta tesis se utilizan dos simulaciones de alta resoluci´on, ORCA12 (1/12o) y NEMOMED8 (1/8o) centr´andonos en el Mediterr´aneo Occidental. Para el an´alisis est´andar de temperatura y salinidad, los resultados de ambas simulaciones muestran una mejora en temperatura, especialmente las tendencias de las capas intermedias y profundas. En salinidad, ORCA12 da resultados similares a las simulaciones anteriores, sin embargo NEMOMED8 muestra una notable mejora debido a un mejor balance h´ıdrico y la ausencia de relajaci´on en salinidad. Al reproducir la convecci´on profunda en el Golfo de Le´on, NEMOMED8 se correlaciona bien con las observaciones, sin embargo ORCA12 crea una capa intermedia muy c´alida y estable que bloquea la convecci´on m´as all´a de las capas intermedias. El an´alisis de las corrientes geostr´oficas promedio muestra un esquema de circulaci´on general correcto, pero cada simulaci´on tiene peque˜nas discrepancias con respecto a las observaciones. Al descomponer la se˜nal de la circulaci´on utilizando EOFs, los patrones temporales y espaciales de NEMOMED8 son consecuentes con las observaciones, sin embargo ORCA12 muestra un patr´on de circulaci´on excesivamente suave y estable. En general, NEMOMED8 demuestra que es una simulaci´on madura y fiable a pesar de que sigue habiendo cierto margen de mejora. ORCA12 es un avance muy interesante en cuanto a simulaciones globales de alta resoluci´on, pero al ser una primera iteraci´on, es de esperar que los resultados mejoren notablemente en las pr´oximas versiones. 1. Introduction 3 ments) to basin scale structures (e.g. the Gulf stream), including intermediate structures of the order of 10-100 km (e.g. mesoscale eddies). The temporal scales of variability of these structures can go from days to seasons, or even decades. The presence in the ocean of all these structures, having di↵erent scales of variability and interaction between them, makes the real ocean a very complex system. This complexity is induced by several physical mechanisms, of which one of the main components is the atmospheric forcing (namely, the wind stress and the heat and freshwater fluxes), which directly influences the ocean surface. In a simplified view, the seasonal cycle of the heat flux over the ocean causes a seasonal fluctuation of sea surface temperature, while the seasonal change in the speed and direction of the mean winds changes the intensity and directions of the mean ocean currents and gyres. In addition, the average atmospheric forcing shows considerable year to year variability (e.g. severe vs mild winters), as well as several synoptic atmospheric events (e.g. localized strong gusts of wind, storms, etc.), and have a large impact on the local scale (ocean currents, temperature, salinity, etc). These processes altogether alter the seasonal cycle giving rise to an interannual variability of the ocean. The ocean can be considered, for a large range of flow features, a quasiinviscid, stratified fluid under the influence of the earth’s rotation and where dynamical instabilities of the mean flow can appear separately from the direct action of the atmospheric forcing (chapter 13, Gill (1982)). These dynamical instabilities (both baroclinic or barotropic) can grow and generate mesoscale eddies and sharp ocean fronts with an associated mesoscale variability. The natural scale associated with mesoscale variability in the ocean is the Rossby radius of deformation or internal radius of deformation, which is basically the horizontal length scale at which the rotational e↵ects become as important as buoyancy e↵ects. In many cases, the mesoscale eddies appear and evolve independently of the external atmospheric forcing, as well as potentially interact with the mean currents and the bottom topography producing long term responses in the ocean circulation giving rise to an internal interannual variability of the ocean. In the context of interannual variability, with this complex system comprised of so many interlinked processes, it becomes very difficult to discern between the natural variability of the system and the alterations caused by human activity. Despite significant progress, observational datasets remain too short, too superficial (satellites) or too dispersed in time and space (drifters 41. Introduction Figure 1.1: Spatial and temporal scales of the ocean (based on Chelton et al. (2001) and moorings) to allow detailed studies of the above physical processes across their full range of scales (Pendu↵et al., 2006) (figure 1.2 shows the availability of in situ observations since 1958). In order to study the relative importance of each of the mechanisms playing a role on the ocean variability (the external forcing or the internal ocean variability), it is crucial to complement all the information from in-situ data with data from numerical modelling studies. 1.2 Ocean Numerical Modelling In a general sense, ocean circulation numerical modelling involves numerical techniques to calculate approximate solutions of the complex, non-linear, multi-dimensional partial di↵erential equations governing the ocean dynamics. 1. Introduction 5 Ocean models have proved to be a powerful tool to study the ocean dynamics and its variability, specially due to the increase in the computational power achieved during the last two decades. Recent numerical hindcast simulations provide the continuous and complete evolution of the ocean for the simulation period but require careful quantitative assessments and model-observation mismatch evaluations to guide dynamical studies and further model improvements (Pendu↵et al., 2007). 1.3. Observations 27 pressions des XBT ont ´et´e corrig´ees selon la Table 1 de Wijffels et al. [2008]. Environ 8 millions de profils sont ainsi disponibles. Tous les profils sont soumis `a un contrˆole qualit´e utilisant des crit`eres objectifs d´evelopp´es par le Met Office Hadley Centre (Ingleby and Huddleston, 2007). Seules les observations valides sont conserv´ees, c’est-`a-dire nous avons rejet´e toutes les observations ”flagg´ees”. 1958 1968 1978 1988 1998 2007 Figure 1.7 – Couverture spatiale pendant les ann´ees 1958, 1968, 1978, 1988, 1998 et 2007 des profils issus de di↵´erents instruments : bathythermographes=MBT, XBT (en violet), TESAC (en jaune), bou´ees et flotteurs Argo (en orange). Contrairement aux simulations num´eriques et aux produits satellites d´ecrits pr´ec´edemment, la couverture spatio-temporelle de ces profils de temp´erature et de salinit´e est dispers´ee et irr´eguli`ere. La Figure 1.7 montre la r´epartition spatiale du r´eseau hydrographique pendant les ann´ees 1958, 1968, 1978, 1988, 1998 et 2007, et met en ´evidence son h´et´erog´en´eit´e spatiotemporelle. La couverture est g´en´eralement tr`es faible avant l’introduction des bathythermographes. En 1958, les premi`eres collectes de donn´ees r´eguli`eres sont e↵ectu´ees lors de missions oc´eanographiques, mais elles restent confin´ees dans l’h´emisph`ere Nord pr`es des continents dans les Oc´eans Atlantique, Pacifique et Arctique, ainsi qu’en Mer M´editerran´ee. Des mesures sont ´egalement relev´ees le long de quelques trajets des navires marchands au milieu Figure 1.2: Spatial coverage during the years 1958, 1968, 1978, 1998, and 2007 of in situ measurements from di↵erent instruments: bathythermographs=MBT, XBT (in violet), TESAC (in yellow) and ARGO Floats (in orange) (from Juza (2011)). Because ocean models are completely flexible in their configuration, and can be designed in many di↵erent ways to study the various ocean processes, it is important to understand firstly, what exactly their function is going to be, and secondly their limitations in terms of what scales can be represented or resolved 61. Introduction by the model, what scales need to parametrized, and finally what scales are being imperfectly represented. The ocean circulation is a complicated function of the density structure of the water masses composing the ocean basins, the radiative fluxes at its surface, and the atmospheric forcing imposed at the surface. This forcing contains a variety of temporal (ranging from hourly to beyond decadal variations) and spatial scales (ranging from kilometre to basin scales). Due to this extraordinary complexity, and because of computational limitations, ocean models must compromise their configurations in terms of resolution, area covered, equation approximations/simplification and other parametrizations. As such there are many types of ocean models and Kantha and Clayson (2000) present a comprehensive review. For this thesis where we are interested in the three-dimensional ocean variability, the models must be baroclinic, global or regional, have sufficient resolution to resolve the features to be studied, maintain a free-surface formulation (for sea level and mass addition as features to be studied) and be linked to the atmosphere either via external atmospheric forcing or coupled atmospheric models. One of the typical simplifications used in large-scale ocean circulation models is the hydrostatic/Boussinesq approximation. This ignores density changes in the fluid except when gravitational forces are concerned. For numerical models which are aimed at creating realistic simulations of the past (referred to as hindcasts), the use of realistic atmospheric forcing data is necessary to drive the simulation. These include surface radiative, heat and buoyancy (precipitation and evaporation) fluxes, as well as wind stress forcing. What the model can and cannot resolve (both temporally and spatially) is also influenced by the temporal and spatial resolution of these atmospheric forcing variables, which depend on the availability of high quality measurements. In recent years the quantity, distribution and quality of such observations has increased considerably but the further back in time we go, the scarcer the number of measurements. This is especially true for the open ocean, where data on wind, humidity and precipitation fluxes is very limited. Because these fluxes depend on small air-sea temperature and humidity di↵erences, direct measurements of the heat fluxes and freshwater sources with coverage and accuracy required for an ocean model are extremely difficult to obtain over the vast world ocean. While weather prediction models can compute such fluxes, 1. Introduction 7 they do not yet produce the data with the sufficient accuracy and mesoscale resolution needed for ocean models. Moreover, forcing the models with climatological estimates of surface fluxes generally results in unrealistic model Sea Surface Temperature (SST), or Sea Surface Salinity (SSS), which then become inconsistent with the imposed surface flux themselves (Barnier, 1998). Many ocean models now use re-analyses of atmospheric data, which are gridded, quality controlled compilations of observational data such as ECMWF ERA40 (Uppala et al., 2005). In these re-analyses, observations from multiple sources are run through an integrated forecasting system using a three dimensional variational technique to provide analyses every six hours. Newer revisions of ERA40 such as ERA-Interim (Dee et al., 2011) and ARPEGE (Beuvier et al., 2010) are dynamical downscales that aim at improving and increasing the resolution of atmospheric data. Because numerical model atmospheric forcing still needs much improvement, imbalances in the freshwater and heat fluxes can cause simulations to drift from stable conditions. What is commonly done to overcome this is to force the model with a simple relaxation of SST or SSS or both to a climatological mean. This approach presents many known drawbacks to represent ocean variability, since mesoscale features such as eddies and fronts are damped unrealistically rapidly by this restoring. The physical characteristics of the ocean in terms of shape (coastline), topography, channels and straits are also vital for the performance of the model, as they can have important e↵ects on many scales of the circulation. The resolution with which these are resolved can in some cases determine whether the model can simulate realistic circulation patterns at certain locations. Straits and channels are of especial importance since it is here where water and heat flux exchange between seas and basins occur. Another challenge in modelling the topographical characteristics of the ocean are the continental shelves and slopes, which play an important part in ocean circulation, particularly with regards to coastal currents, jets as well as cold water cascading into the deep ocean. Because of the shallow depths involved, circulation on the shelf is forced strongly by winds and astronomical tides, while in the deep ocean it is the density gradients and the winds. Additionally, the transition between shelf and abyssal ocean is quite abrupt with the slope regions having spatial scales of tens of kilometres (compared to thousands in the open ocean). Therefore low or medium resolution models tend to greatly smooth these abrupt topographical features, therefore a↵ecting many of coastal circulation patterns that tend to flow along these features (Kantha and Clayson, 2000). 81. Introduction In general, higher resolution is always preferable although it is limited by computational and storage capabilities and therefore compromises have to be made. High resolution global models (example figure 1.3) are computationally extremely expensive (often requiring the use of”super-computer clusters” for extended periods of time). This is why regional models (example figure 1.4) are capable of operating at higher resolution than global models and thus more practical for regional studies. we used only processors having ocean grid points (186 processors). Each processor thus had 82×87×46 grid points. A one-row overlapping halo (1 row) is shared with the neighboring processors, using explicit communications between processors, (Message Passing Interface library). Figure 1gives a global view of the domain broken-down into individual processors. The time step of the ocean component is 1,440 s (60 time steps per day) and the sea-ice component is called once every five time steps. One year of model simulation requires 2,200 h of CPU on 186 IBM SP4 processors and takes about 12.6 h of elapsed time. Maximum memory is 0.479 Gb per processor and total memory is 84 Gb. 2.3 Sensitivity tests with ORCA-R025 A series of 10-year simulations were run to evaluate the contribution of various numerical choices to the solution. The focus is on the PS representation of topography and the EEN vorticity scheme used in the calculation of momentum advection, which produced the greatest improvements to the model solution. Therefore, we compare (in Section 3) a simulation that does not include these two features, referred to as G04, with a simulation that includes both, referred to as G22, (Table 1). Other numerical options were tested. In particular, a FS simulation using the EEN vorticity scheme was run (simulation G03, Table 1) so the effects of the PS could be separated. Some results from this experiment will be used in Section 3. Before we provide (in Section 3) an assessment of the changes induced by the numerics, we provide a brief overview of simulation G22 (with PS and the EEN momentum advection scheme). Our analysis will remain rather descriptive. A full understanding of how the numerics impact on the physics of the model requires a large number of sensitivity experiments and complex diagnostics, which are currently under way in a North Atlantic configuration of the code (J. Le Sommer et al. 2006, in preparation). Preliminary results from this work are used in the discussion of Section 4to illustrate possible Fig. 2 Simulation G22 (partial step topography, new EEN vorticity scheme for momentum advection): snapshot of the sea surface height (ssh, in color) and ice cover (in white) in Austral winter in year 10 of the simulation Table 1 List of the 10-year-long sensitivity experiments carried out with ORCA-R025 Simulation Vorticity scheme Bottom topography* G03 EEN (new) FS G04 ENS (old) FS G22 EEN (new) PS The present study mainly uses the results from the G04 and G22 simulations. No relaxation to SSS is applied in any of these experiments. A free slip sidewall boundary condition is used in all simulations *FS Full step and PS partial step Figure 1.3: Coverage of a global model (ORCA G22) showing a snapshot of sea surface height and ice cover during winter (from Barnier et al. (2006)). However, regional models are not without their own problems; these models require that their boundary conditions be provided by either nesting into a coarser resolution model or through relaxation to climatology. Nesting within a coarser resolution model can introduce errors due to di↵erences in the parametrization of the physical processes in both models as well as errors due to the numerical techniques used for the nesting, which can propagate these errors into the regional domain having a significant impact on the simulation’s evolution. 1. Introduction 9 Figure 1.4: Coverage of a regional model (MEMOMED8) showing the sea surface height (SSH) for 1993-2007. 1.3 The Mediterranean Sea In the context of ocean variability and climate change studies, the Mediterranean Sea (figure 1.5) is considered a ”miniature ocean” (Bethoux and Gentili, 1999; CIESM Initiative Group, 2002), a kind of ideal, accessible, reduced scale ocean laboratory where many phenomena present in many di↵erent regions of the global ocean can be studied at a smaller scale: deep convection (MEDOC Group, 1970; Leaman and Schott, 1991), shelf-slope exchanges (Bethoux and Gentili, 1999), thermohaline circulation and water mass interaction (W¨ust, 1961), mesoscale and submesoscale dynamics (Robinson et al., 2001), etc. Despite its small size, the thermohaline circulation within the basin is particularly active (Wu and Haines, 1996). Robinson et al. (2001) gives a value of 10-14 km for the internal Rossby Radius of Deformation, four times smaller than the typical value in the open ocean. Physical mechanisms are thus better monitored and understood in this ocean basin, contributing then to the advancement of knowledge of physical interactions and biogeochemical coupling at near-shore, local, sub-basin and global scales (Tintore, 2009). The Mediterranean is a mid-latitude semi-enclosed sea, connected to the Atlantic Ocean through the ⇠300 m deep Strait of Gibraltar. It is comprised of two main basins, the western (WMED) and eastern (EMED) Mediterranean basins, connected by the Sicily Strait. Each of these basins are composed of several sub-basins, each one characterized by di↵erent oceanographic features (di↵erent topography, water masses and atmospheric forcing) and separated by straits and channels (Astraldi et al., 1999). 10 1. Introduction Figure 1.5: The regions and basins of the Mediterranean Sea. The general picture of the Mediterranean circulation is rather complex due to the presence of multiple interacting scales, including basin, sub-basin scale and mesoscale structures (Robinson and Golnaraghi, 1994). In the Mediterranean one can find strong coastal boundary currents, unstable jets that shed vortices, permanent and recurrent sub-basin gyres, and energetic mesoscale eddies. All these structures interact and give rise to a significant variability in a wide range of temporal scales (mesoscale, seasonal and interannual). This complexity and variability of the ocean circulation arises from multiple driving forces: topographic e↵ects, variable atmospheric forcing and internal dynamical processes. Another element which is also important in the Mediterranean because of its huge socio-economic impacts is sea level variability. In the Mediterranean, it depends on local as well as remote forcing (sea level changes are linked to the NAO by the combined e↵ect of atmospheric pressure anomalies and changes in evaporation and precipitation (Tsimplis and Josey, 2001)). Global steric expansion occurring outside of the Mediterranean basin can transmit a signal through the Strait of Gibraltar. Within the Mediterranean, inter-annual and higher frequency variability is dominated by wind and atmospheric pressure changes (Gomis et al., 2006) while steric variations, associated changes in the water budget by rivers, evaporation and precipitation driven by the regional atmospheric forcing, all contribute to sea level changes (Tsimplis et al., 2008). The rates of sea level change over the Mediterranean Sea have shown considerable variability over the last century. Before the 1960‘s, relative sea level was increasing at a rate of about 1.2 mm/yr, between 1960 and 1994 sea level 1. Introduction 11 was decreasing (Tsimplis and Baker, 2000), and after the mid 1990‘s altimetric measurements suggest rapid sea level rise, especially in the eastern Mediterranean. The spatial variability of sea level is also extremely non-uniform, which was clearly highlighted with the advent of satellite altimetry, capable of mapping the open ocean and its geographical variations of sea level change (Cazenave et al., 2001). In the following sections we present some of the main circulation features of the Mediterranean Sea and modelling e↵orts carried out in this region. 12 1. Introduction 1.3.1 Surface Circulation The general circulation scheme of the Mediterranean Sea is complex and composed of three predominant and interacting spatial scales: basin scale (including thermohaline circulation), sub-basin scale, and mesoscale (Robinson et al., 2001). Mesoscale variability is known as the dominant signal in ocean circulation (Pujol and Larnicol, 2005), and is present throughout most aspects of Mediterranean circulation, from the gyres found in the Alboran Sea and Levantine basin to the eddy vorticity associated with the major currents such as Algerian, Northern and Ionian currents (figures 1.6 and 1.7). It also plays an important role in the formation of deep waters in the Gulf of Lions through mesoscale baroclinic instability (Herrmann et al., 2008a). The Mediterranean surface circulation circuit starts as Atlantic Water (AW) enters the Mediterranean Sea through the Strait of Gibraltar at surface layers (100-200 m approx.) and into the Alboran Sea. Circulation in this sea is dominated by one or two sub-basin scale anticyclonic gyres (Vi´udez et al., 1998), a quasi permanent western anticyclonic gyre (the Western Alboran Gyre, WAG) (figure 1.6) and the intermittent Eastern Alboran Gyre (EAG). While the EAG can also have cyclonic phases (Vi´udez and Tintor´e, 1995) it is mostly anticyclonic (Tintor´e et al., 1988). The eastern boundary of the EAG is known as the Almer´ıa-Or´an front and marks the beginning of the Algerian current which flows along the African coast. This narrow coastal current is very unstable and produces many anticyclonic eddies of 50-100 km in diameter which can be clearly seen with infrared satellite images (Millot, 1999). For this reason, the Algerian Basin is characterized by intense mesoscale activity. Some of these eddies grow in size and detach from the coast reaching in some cases the southern Balearic Sea (Ruiz et al., 2002). The Algerian Current reaches the Strait of Sicily where it bifurcates in two branches. One branch enters the EMED, while the other branch remains in the WMED. Circulation at the Strait of Sicily presents a two-layer exchange, at the surface, AW enters the eastern basin, and in the deeper layers, the Levantine Intermediate Water (LIW), which are formed in the Levantine basin, enters the western basin. The AW branch that remains in the WMED follows the Italian, French and Spanish coasts, as the so-called northern current (Millot, 1999), and the LIW follows a similar cyclonic path along the periphery of the WMED at a depth of between 200 and 800 m. The northern current is also fed by AW flowing northward along east of Corsica and eventually flows into the 1. Introduction 19 1.3.3 Numerical Modelling The Mediterranean presents a very interesting region for ocean modelling. As mentioned above, many of the ocean’s dynamical processes take place in the Mediterranean making it a very suitable but challenging region for modelling studies. It is however not without its challenges. The Mediterranean topography is extremely complex (figure 1.9), its water balance is maintained at the Strait of Gibraltar, which is narrow (14.3 km) and shallow (300 m). This requires all but the highest resolution models to make special adjustments at this location to allow for an adequate flow. In addition to Gibraltar, there are many other straits and channels that play vital roles in the water mass circulation around the Mediterranean. The Sicily Strait (145 km, 316 m depth) separates the eastern and western Mediterranean basins, with all the surface and intermediate water mass exchanges happening through it. The Otranto Strait in the Adriatic, the Cretan Arc separating the Aegean from the Levantine basin and the Ibiza and Mallorca Channels are also vital in the regional circulation. Whether a model is capable of adequately representing these topographic features will determine its success at modelling circulation and water mass exchanges. Figure 1.9: Satellite image of the Mediterranean Sea with the bottom topography overlaid (source: Google Earth), and detail of the various channels and straits. 20 1. Introduction Mediterranean dynamics are dominated by mesoscale variability therefore models striving to study processes other than large scale mean circulation must have sufficient resolution to resolve the mesoscale. There are also dynamical processes such as deep convection that are vital for Mediterranean and global oceanography and are generally difficult to model because of their dependence on short and intense atmospheric events as well as on small mesoscale baroclinic instabilities (Herrmann et al., 2008a). Several regional modelling studies have focused on the Mediterranean Sea using high resolution 1/8oand 1/16oregional models such as DieCAST (Fern´andez et al., 2005), OPAMED8 (Somot et al., 2006), EU-MFSTEP (Tonani et al., 2008), NEMOMED8 (Sevault et al., 2009) and MED16 (B´eranger et al., 2010) to name a few. The DieCAST 1/8osimulation used in Fern´andez et al. (2005) is not a hindcast, it is a simulation run to assess circulation and transport variability in the Mediterranean Sea. It is forced by climatological monthly mean winds and uses relaxation towards monthly climatological surface temperature and salinity. The OPAMED8 1/8osimulation by Somot et al. (2006) is a scenario of the Mediterranean Sea under climate change IPCC-A2 conditions and run with an atmosphere regional climate model (ARPEGE) over the 1960-2099 period using a hierarchy of three di↵erent models. Their objective is to obtain a high resolution atmospheric forcing to run the OPAMED8 simulation (based on the OPA model (Madec et al., 1998) of the Mediterranean Sea under a transient climate change scenario. The EU-MFSTEP 1/16osimulation by Tonani et al. (2008) is a very high resolution model based of the OPA code (Madec et al., 1998) and used for operational daily forecasts of the Mediterranean Sea. It is available since 1997. The NEMOMED8 1/8osimulation by Sevault et al. (2009) is a simulation focused on the study of the dynamics behind certain features and variability of the Mediterranean Sea, which has also been used in the recent studies by Beuvier et al. (2010) and Herrmann et al. (2010) that focused on the EMT and the 2005-2006 NW Mediterranean convection events respectively. The MED16 (B´eranger et al., 2010) is a modelling experiment to assess the performance of atmospheric forcing resolution on winter ocean convection in the Mediterranean Sea by running two 4-year simulations 1998-2002 (with 11 years spin-up), one using ERA40 and one using ECMWF analysed surface fields which have twice the resolution of ERA40. Additionally, some recent studies such as Hu et al. (2009) and Schae↵er et al. (2011) have used short, very high resolution models (⇠1 km resolution) to study very specific areas and processes in the Mediterranean (specifically, mesoscale eddies in the 1. Introduction 21 Gulf of Lions). Throughout this thesis we study the performance of a variety of ocean models in the Mediterranean Sea; from medium resolution 1/4oglobal models to regional 1/8oand finally a first iteration of a global 1/12omodel. Chapter 2 Motivation and Objectives The general motivations and objectives of the present thesis are to study the Mediterranean (with a focus on the Western Mediterranean) variability through the analysis and validation of five ocean numerical hindcast simulations by comparing them to di↵erent sources of observations (including in situ data and remote sensing). We attempt to perform a comprehensive analysis of the main features and variables that represent the Mediterranean Sea circulation, water masses and dynamics, compare them to real data, and suggest where the simulations have room for improvement. In Chapters 5 and 6 we use low resolution 1/4ohindcasts with the aim of improving our understanding of long term interannual variability at basin scales. In Chapter 7, we use high resolution 1/8oand 1/12o hindcasts to assess the improvements gained by the higher resolution as well as more detailed studies at sub-basin and meso-scales. Modelling studies demand a large amount of work in their programming, running, diagnostics, etc. Part of the objectives of these studies arise from a collaboration agreement with LEGI (Laboratoire des Ecoulements Gophysiques et Industriels) and MERCATOR-OCEAN to validate their ocean simulations for the Mediterranean Sea region. Chapter 5 is a first approach at the analysis and validation in the Mediterranean Sea of a global ocean simulation; the ORCA025-G70 simulation. This simulation was chosen because it was, at the time, one of the highest resolution global simulations available, it was modern and had been validated in several regions of the world‘s oceans. However, no particular attention had been paid 23 24 2. Motivation and Objectives to its performance in the Mediterranean. Higher resolution regional models were available for the Mediterranean, but no-one had analysed a global model there. The resolution of 1/4ois insufficient to reproduce the mesoscale dynamics, but adequate for the large scale circulation which is the main focus of this chapter (mainly basin scale). Specifically, the scope of this chapter is to assess whether ORCA-G70 is capable of correctly reproducing the state of the ocean with regards to certain key variables; mean temperature and salinity basin scale variability and trends at di↵erent depth layers by comparing the simulation to the MEDAR temperature and salinity database, Sea Surface Height and Steric Height time series analysis of seasonal cycle amplitude and phase, interannual variability, trends, as well as the spatial distribution of trends, variance and the annual amplitude. This is done by comparing the simulation to satellite altimetry data. Finally we check whether transport through the two main water inputs into the Mediterranean Sea (Atlantic waters through the Strait of Gibraltar, and freshwater input from the Black Sea) are correct and water balance is maintained through them by comparing with the available literature. Chapter 6. Following the positive and interesting results obtained when analysing ORCA-G70, the goal of this chapter is to expand on the work carried out in Chapter 5 by adding more simulations with di↵erent configurations to the analysis and asses how their di↵erent characteristics (same 1/4oresolution) a↵ect their capability to reproduce circulation features in the Western Mediterranean. In addition to ORCA-G70, we analyse ORCA-G85 (a simulation with improved atmospheric forcing, greater vertical resolution and weaker sea surface salinity restoring), and GLORYS1V1 (a shorter simulation with data assimilation). The inclusion of GLORYS1V1 aims at verifying whether reanalysis products that include data assimilation help to improve the description and our understanding of ocean variability in the Western Mediterranean Sea. It will highlight which degree of improvement can be expected from an ocean reanalysis in that region with respect to a free run. In addition to the basin scale analysis performed in Chapter 5, we focus on specific aspects of the circulation in the WMED such as the deep water formation in the Gulf of Lions and transport through the Balearic Channels, assessing how the di↵erences in the model configurations a↵ect their performance. Finally, we analyse the large-scale mean geostrophic circulation in the WMED by breaking it down into its principal components using Empirical Orthogonal Functions (EOFs). Chapter 7. In this final study, we go a step further and analyse two high 2. Motivation and Objectives 25 resolution models, one global (ORCA12, 1/12oresolution) and one regional (NEMOMED8, 1/8oresolution). With these higher resolution simulations, the objective is to study the mesoscale variability of the WMED circulation. For consistency we also perform the basic temperature and salinity basin scale analysis, as well as looking at the deep-water formation capabilities of these high resolution simulations, since, as concluded by the previous chapters, resolution (both of the simulation and the atmospheric forcing) is one of the limiting factors when reproducing deep convection events in the Gulf of Lions. However the bulk of this chapter is dedicated to investigate the dynamical characteristics of the WMED, and to do so we focus on the geostrophic circulation of the simulations given by their sea surface height and compare it to the geostrophic circulation calculated from altimetry measurements. To assess the mesoscale variability, we analyse the Eddy Kinetic Energy (EKE). Finally, we attempt to analyse and identify the principal components of the WMED mesoscale circulation by performing EOF calculations in three main regions; the Alboran Sea, the Algerian Basin and the North-Western Mediterranean. Chapter 3 Data and Numerical Simulations All of the models analysed throughout this Ph.D. Thesis (from global to regional scale) are based on the NEMO2 (Nucleus for European Models of the Ocean) (Madec, 2008) modelling system, which presently includes the latest version of the primitive equation, free surface ocean circulation code OPA9 coupled to the multi layered sea-ice model LIM2. The fundamental equations of ocean dynamics and diagnostic sea surface height used in these models can be consulted in appendix A. 3.1 The ORCA Hindcasts The ORCA hierarchy of ocean models are coupled ocean/sea-ice global configurations of the NEMO numerical framework implemented by the DRAKKAR group (Barnier et al., 2006) aiming at the study of ocean variability under realistic atmospheric forcing conditions over the last half century. The models simulate the evolution of temperature, salinity, velocity, sea surface height (SSH), sea-ice characteristics, and oceanic concentrations of tracers (CFC11 and C14) (Barnier et al., 2007). The DRAKKAR collaboration is a European modelling project which provides the framework for joint and coordinated modelling studies between research groups in France, Germany, Russia and Finland. Common characteristics of the ORCA models The model configurations use the ORCA global tri-polar grid at varying 27 28 3. Data and Numerical Simulations resolutions (2o,1 o,1/2 o, 1/4o, 1/12o). E↵ective resolution gets finer with increasing latitudes (the grid size is scaled by the cosine of the latitude, except for the Arctic, figure 3.1), and in the vertical, grid spacing is finer near the surface and increases with depth (max. depth is 5844 m). Bathymetry is derived from the 2-min resolution ETOPO2 bathymetry data of the NOAA National Geophysical Data Centre. Initial conditions for temperature and salinity are derived from the NODC Word Ocean Atlas data set for middle and low latitudes, and for the Mediterranean they are derived from the MEDAR climatology (more details can be found in Barnier et al. (2006) and Molines et al. (2007). The ORCA hindcasts studied here are driven at the sea surface by realistic atmospheric forcing. One of the great difficulties of ”forced” oceanic modelling is the balance of atmospheric fluxes. The uncertainties in air temperature, humidity, winds, rainfall and air-sea fluxes are so large than when one integrates the global heat and freshwater fluxes, there is usually a huge imbalance which drives an unacceptable drift in the ocean model. 3.1.1 1/4oGlobal Hindcasts: ORCA G70 and G85 In chapters 5 and 6 we analyse the ORCA-R025 G70 and ORCA-R025 L75.G85 numerical simulations (G70 and G85 hereafter), which are implemented on a 1/4oresolution grid. E↵ective resolution is ⇠27.75 km at the equator, ⇠20-22 km in the Mediterranean and ⇠13.8 at 60oN/S). At this resolution, the simulations are eddy permitting but not eddy-resolving in large oceans such as the Atlantic, but not even eddy-permitting in the Mediterranean where the first Rossby radius of deformation is around 10-14 km. Grid, masking, and initial conditions are inherited from the global configuration of the MERCATOROcean operational oceanography center, with 1442x1021 grid points, 46 vertical levels for G70 and 75 vertical levels for G85. Some smoothing is applied when adding the bathymetry (some of the straits are widened to two grid points (⇠40 km at Mediterranean latitudes) to allow adequate flow given the coarse resolution of the models). A restoring has been applied towards the Levitus climatology (T,S) at the Gibraltar Strait exit, in the Gulf of Cadiz. This restoring increases from 200 m to 400 m, then remains constant to the bottom. Since this restoring only a↵ects the Mediterranean outflow, its e↵ects are not felt directly within the Mediterranean Sea. 3. Data and Numerical Simulations 35 Figure 3.2: Distribution of Temperature and Salinity profiles in the MEDAR database (Vidal-Vijande et al., 2011). 36 3. Data and Numerical Simulations Salinity Profile Project, Argo and the Arctic Synoptic Basin-Wide Oceanography Project. EN3 contains observations from bathythermographs (XBTs and MBTs), hydrographic profiles (CTDs and predecessors), moored buoys and ARGO drifters. The dataset consists of quality controlled, global, monthly T and S fields on a 1o⇥1ogrid covering the period 1950 to the present. The vertical domain extends down to 5000 m, with data on 42 levels. Data coverage is highly non-uniform, especially at the beginning, with huge gaps in in parts of the tropics and southern hemisphere contrasted with a few highly sampled ship tracks (mostly MBTs). This distribution improves over time, but large areas of the world ocean still remain with inadequate temporal and spatial sampling (Ingleby and Huddleston, 2007). 3.5 Satellite Altimetry Satellite altimetry provides realistic high-resolution sea surface height (SSH) observations making it a useful tool in the validation of sea level and surface circulation numerical models (further details on the altimetry principle are provided in appendix B). However, altimeter measurements also include Earth‘s geoid which varies tens of meters across the ocean and needs to be subtracted from altimetry data to obtain a usable product (Dobricic, 2005). Since the exact shape of the geoid is largely unknown (the best data available comes from the GRACE satellite measurements and they have a resolution of approximately 300 km, which is only sufficient for large scale studies) the typical solution to circumvent this limitation is to subtract a temporal mean of sea surface height at each location leading to only the variable part of sea surface height, known as sea level anomaly (SLA). Satellite altimetry data are available as a global product at 1/4oresolution but in this thesis we use the higher-resolution product specific for the Mediterranean. It is the REF MERGED SLA product which is freely available through the AVISO FTP (http://www.aviso.oceanobs.com). These data consist of gridded, delayed time sea level anomaly fields specific for the Mediterranean Sea merging several altimeter missions (Topex/Poseidon, Jason-1/2, ERS 1/2, ENVISAT) provided weekly with a resolution of 1/8o. The reference product has been selected in order to have a homogeneous time series with the same configuration of satellite altimeters included in the objective analysis scheme. 3. Data and Numerical Simulations 37 The methodology used in AVISO (Ssalto/Duacs system) to process altimetric data is to build up a homogeneous and inter-calibrated dataset based on a global crossover adjustment using Topex/Poseidon as the reference (Le Traon et al., 1998). Sea surface height measurements are then geophysically corrected (tides, wet/dry atmosphere, ionosphere) with the atmospheric correction applied in order to minimize aliasing e↵ects and to recover total sea level. The classical Inverted Barometer correction has been replaced by the MOG2D barotropic correction (Carr`ere and Lyard, 2003) which improves the representation of high frequency atmospheric forcing as it takes into account wind and pressure e↵ects. Then, along-track data are resampled every 7 km using cubic splines and SLA is computed by removing a 7-year mean corresponding to the 1993-1999 period. Measurement noise is reduced by applying Lanczos (cut-o↵ and median) filters. The mapping method to produce gridded SLA fields from along-track data is described in Le Traon et al. (2003). The long-wavelength error parameters are presently adjusted according to the most recent geophysical corrections. For more details on the altimeter data processing the reader is referred to Pujol and Larnicol (2005). This change, as shown in (Pascual et al., 2009), constitutes an important step forward with respect to previous altimeter products (e.g. those used in Brachet et al. (2004) and (Pascual et al., 2007)). This is particularly relevant close to the coast (defining here the coastal ocean as the region over the continental shelf and slope, ranging from 0-100 km from the coast), where altimetric observations often are of lower accuracy or not interpretable due to a number of factors including inaccurate tidal corrections, incorrect removal of atmospheric (wind and pressure) e↵ects at the sea surface (Volkov et al., 2007), as well as the radar signal being contaminated by land up to a distance of around 40 kilometres from the coast. In a small sea with many narrow straits and basins such as the Mediterranean, major features which are normally found further out at sea are usually not a↵ected, however coastal currents can show significant error due to land proximity (Vignudelli et al., 2005). To overcome this problem, several groups from Europe are working on these issues in the frame of projects funded by Space Agencies (such as PISTACH funded by CNES and COASTALT funded by ESA). Essentially, improved altimeter corrections and increasing the data recovery near the coast, are the main tasks within those projects (Bou↵ard et al., 2009; Volkov et al., 2007). Despite its limitations, altimetry has become an invaluable tool for the study of sea level variability across the worlds oceans. For this study, monthly means have been computed from weekly gridded 38 3. Data and Numerical Simulations maps (averaging 4-5 maps for every month). 3.6 Mean Dynamic Topographies An important issue to be analysed in the models described in this work is to evaluate the capability to reproduce the main features of mean circulation. However, as mentioned above, the Sea Level Anomaly altimetry fields only include the variable part of the total sea level signal. Thus, the mean circulation has to be analysed from alternative sources. One of the methods is to use approximations derived from numerical models as a first guess and then make local corrections using in situ observations. Here we use two di↵erent MDTs, one by Rio et al. (2007) and one by Dobricic (2005). Rio et al. (2007) use an average over 1993-1999 of dynamic topography outputs from MFS (Tonani et al., 2008) as a first guess. This is then improved in a second step by combining drift buoy velocities and altimetry using a synthetic method to obtain local estimates of the mean geostrophic circulation which are then used to improve the first guess through an inverse technique to map the Synthetic MDT onto a1/8 ogrid. Dobricic (2005) uses a di↵erent method; they use the statistics from the MFS operational assimilation system, monitoring the long-term bias in the SLA assimilation (using the period Sept 1999 to Sept 2002) and try to eliminate it. Chapter 4 Methods In this thesis we have used a variety of statistical methods to analyse the di↵erent datasets. In this chapter we describe these methodologies and add references to the chapters in which they have been used. 4.1 General Statistics Throughout the works presented in this Ph.D. thesis we use a series of standard statistical analysis methods to analyse the various datasets and variables. These are the main ones used: Means -Insomecasesweusetemporalmeans,creatingtwo-dimensional maps, and in other cases we use spatial means, creating time-series plots of seas, basins or smaller regions. Variance is a measure of the amount of variation of the values of a particular variable. We use variance to assess the spatial variability of certain variables (e.g. sea surface height) by calculating the variance over a certain period for each grid point of the dataset and creating a two-dimensional variance map. This is used in Chapter 5. Trends are useful to describe the long term behaviour of a dataset or timeseries. Since the data used throughout this PhD thesis is of a fairly low frequency, made up of monthly or yearly averages of gridded data, we calculate trends using simple linear regression (by least squares) and then get the confidence levels at the 95% for the slope of the fit: 39 40 4. Methods ˆa=Pn i=1 yi(ti¯ t) Pn i=1(ti¯ t)2(4.1) d=±ˆs pPn i=1(ti¯ t)2Tn2(1 ↵/2) (4.2) where ˆais the estimation of the slope, dis the confidence interval at the ↵ level of significance, ˆsis the estimation of the standard deviation, and Tn2is the inverse of the t-student cumulative probability function with n-2 degrees of freedom (Vargas-Y´a˜nez et al., 2005). The trend can be calculated either with the full signal or with the signal minus the seasonal cycle. The final step is to fit a straight line to the dataset using the function f=ˆax +ˆ b. The Correlation Coefficient is a measure of the statistical relationship between two dependant datasets. Correlations have been calculated using the following formula: r=PmPn(Amn A)(Bmn B) q(PmPn(Amn A)2)(PmPn(Bmn B)2 (4.3) where Ais the matrix mean of A,andBis the matrix mean of B. Annual cycles. Throughout this thesis we use two di↵erent methods to calculate the annual (seasonal) cycle; in Chapter 5 we obtain the annual cycle by fitting two harmonic functions using a least squares method to assess whether the amplitude of the spatial patterns along the annual cycles of model and observations coincide. y(t)=Aacos ✓2⇡ 365.25ta◆+Asa cos ✓2⇡ 182.63tsa◆(4.4) where Aaand aare the amplitude and phase of the annual cycle, and Asa and sa are the amplitude and phase of the semi-annual cycle (Pascual et al., 2008). 4. Methods 41 In most cases when we wish to study the interannual variability of a certain signal, the seasonal cycle, which is usually the dominant signal, has a much higher amplitude than all other signals, therefore making interpretation of interannual variability harder. We therefore remove the seasonal cycle by subtracting it from the total signal. The removal of the seasonal cycle for timeseries analysis is done by removing a monthly climatology from the timeseries data. The climatology is created by calculating the mean of every month of the year (i.e. mean of all Januarys, Februarys, etc). 4.2 Empirical Orthogonal Functions (EOFs) EOFs are used to identify and quantify the spatial and temporal variability consistently reducing the dimensions of large datasets to few significant orthogonal (uncorrelated) modes of variability, and to estimate the amount of variance associated to each mode of variability in percentage terms. Despite the dimension reduction of the datasets performed by the EOF method, the spatial patterns and their expansion coefficients (principal components (PC), i.e., the evolution in time of the spatial modes) are not always easy to interpret, especially referring to PC of secondary modes of variability, often complicated by non-periodic signals. It is important to take into account that the modes of variability contain information about the dataset that is not necessarily related to physical features (Olita et al., 2011). In Chapters 6 and 7 we use EOFs to decompose the complex ocean circulation signal into its Principal Components (PC) of spatial variability Fk(x, y) and their associated temporal components ak(t) making it easier to analyse and describe them separately. The signal O(x, y, t) is decomposed in the using the following formulation: O(x, y, t)=pn1 n X k=1 Fk(x, y)ak(t)(4.5) where |ak|=1and|Fk|6=1 42 4. Methods 4.3 Water Mass Transports Transport through straits and channels is calculated by integrating the zonal velocities provided by the simulations at either North-South, East-West, or diagonal transects. Positive values are considered transport due East (e.g.. Atlantic inflow into the Mediterranean Sea) or North, and negative values were considered transports due West (e.g.. Mediterranean outflow into the Atlantic Ocean) or South. In the case of diagonal transects, which in this study only refers to the Mallorca Channel, positive values are considered for flows with a dominant Northward component and negative values are considered for flows with dominant Southward component. 4.4 Hovm¨oller Diagrams In this thesis we use Hovm¨oller diagrams to plot the time evolution of vertical profiles of temperature and salinity. The Xand Yaxes are time and depth respectively with the value of temperature or salinity represented by the color gradients. These are used in Chapters 6 and 7 to study the time-evolution of deep convection events in the Gulf of Lion. 4.5 Eddy Kinetic Energy (EKE) The Eddy Kinetic Energy is a measure of the degree of variability and may identify regions with highly variable phenomena such as eddies, current meanders, fronts or filaments. It is calculated from the sea level anomaly (altimetry) or sea surface height anomaly (models) (⌘0) by making the assumption of geostrophy : EKE =1 2[U02 g+V02 g](4.6) U0 g=g f @⌘0 @y(4.7) 4. Methods 43 V0 g=g f @⌘0 @x(4.8) where U0 gand V0 gdenote the zonal and meridional geostrophic velocity anomalies. fis the Coriolis parameter, gis the acceleration of gravity and the derivatives @⌘0 @yand @⌘0 @xare computed by finite di↵erences where xand yare the distances in longitude and latitude, respectively (Pascual et al., 2007). 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 51 5.4 Seasonal Steric Signal As opposed to G70, the MEDAR data is only available in annual time steps for the second half of the 20th century. This does not allow for the representation of the seasonal cycle. There is however a monthly MEDAR climatology (monthly averages without interannual variability) for this period, which can be used to determine the di↵erences and similarities of the steric signal for both G70 and MEDAR. The steric height (SH) is the component of sea level driven by the expansion/contraction due to changes in temperature and salinity. It is computed for each grid point of the model as the vertical integration from surface to the bottom of the specific volume anomaly (↵)(respecttothe specific volume at 35 psu and 0oCcausedbychangesinpotentialtemperature (T) and salinity (S): Steric Sea Level =1/g Z0 H ↵dx (5.1) Figure 5.5 shows the computed steric height for the Mediterranean basin from the climatological data (from the surface to the full depth). Amplitudes coincide perfectly (⇠5 - 6 cm), and phases are very similar with only very minor di↵erences. The most notable result is the absolute height di↵erence between both datasets, with G70 being an average of 14.33 centimetres higher than MEDAR. This is because the mean salinity for the entire Mediterranean basin is significantly lower in G70 than in MEDAR, while temperature is slightly higher. A simple test of adding a bias of 0.13 psu to the salinity (which is the mean di↵erence observed for the whole Mediterranean integrated from the surface to the sea floor in Figure 5.3) made the absolute steric height of MEDAR change by ⇠17 cm. Halosteric and thermosteric sensitivity analysis shows that about 85% of the steric signal is due to temperature (amplitude of ⇠8-10 cm) and 15% due to salinity (amplitude of ⇠2 cm) (not shown). The term Dynamic Height is similar to steric height, but the integration is done from the surface to a chosen reference level (usually referred to as the level of zero-motion). The steric signal / dynamic height is very important, not only because of its e↵ect on sea level, but because from it, one can compute an important component of the ocean; the geostrophic currents. These currents exist due to a balance of forces caused by the distribution of horizontal 52 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation Figure 5.5: ORCA G70 (dashed-line) and MEDAR of steric height calculated from a reference (solid-line) climatology depth level of 4000 m. density/pressure gradient fields (computed empirically from measurements of temperature and salinity) and the Coriolis force (for a more complete description, please see Pickard and Emery (1990)). 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 53 5.5 Sea Level: Comparison with Altimetry In addition to temperature and salinity, the Sea Surface Height performance of the model was also analysed and compared to Altimetry data. Since the altimetry dataset is only available since 1993, only the period from 1993-2004 was analysed in this section. This analysis was also performed on the computed steric height of the model. Other studies by Tsimplis et al. (2008) have compared numerical models such as ORCA G70 and OPAMED8 (Somot, 2005) to tide gauges for a longer time period, 1960-2000. 5.5.1 Mean Sea Level We look at the time evolution of the sea surface height variables over the Mediterranean and its two main basins, the EMED and WMED basins. Figure 5.6 shows the Mediterraneans basins mean anomaly time-series for the models SSH, SH and altimetry for the 1993-2004 period, where the main component of this signal is clearly due to the seasonal cycle. From these results, it is clear that the model is perfectly capable of accurately reproducing the phase and amplitude of the seasonal cycle with very good comparison between the models SSH and the Altimetry (which in theory are observing the same processes), with correlations of ⇠0.8 (0.89 with the signals de-trended). A clear example of this is the sea level signal linked with the 1996 negative NAO (North Atlantic Oscillation, Woolfe et al. (2003)), which according to Tsimplis et al. (2008) is the strongest signal of the last four decades (particularly in the WMED). When computing the models steric height, the specific volume anomaly was integrated from the surface to the full depth of the model. Figure 5.6 shows that the steric component accounts for about half of the total sea level signal. The computed steric height of the model shows an annual cycle amplitude of ⇠5 cm, and the full SSH signal as diagnosed by the model of ⇠10 cm. These values coincide with the altimetry data and are confirmed by Bouzinac et al. (2003) and Larnicol et al. (1995). A notable limitation of the model is its ability to reproduce long-term SSH trends. With an average trend of 14.95±1.52 mm/year for the whole Mediterranean, the model overestimates 4-5 times the trend observed by altimetry (3.6±1.54 mm/year). The trend is higher in the EMED than in the WMED, 54 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation Figure 5.6: Timeseries of G70 sea surface height (SSH, red), steric height (SH, green) and altimetry (ALTI, blue) for the WMED (top), EMED (middle) and whole Mediterranean (bottom) but this coincides with Altimetry. A surprising result is that the steric height computed from the models temperature and salinity data does not display the same exaggerated trend, moreover, the trend is almost identical to the altimetry trend (3.66±1.16 mm/year). However, the fact that the steric heights trend coincides with altimetry should be interpreted with caution due to the discrepancies observed in the temperature trends between ORCA and MEDAR (especially at intermediate and deep layers, Figure 5.1). The spatial distribution of trends (not shown) for SSH and altimetry confirms that the models trend overestimation is a global feature of the model (in the Mediterranean 8 - 18 mm/year, average 14.95 mm/year), whereas altimetry displays areas of both positive and negative trends (-16 to +10 mm/year, average 3.6 mm/year). These trends calculated using basin averages are useful to provide a general idea of the basins behaviour but may not be truly representative of many areas 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 55 within the basins themselves. These must therefore be analysed with caution. Tsimplis and Rixen (2002) found that given the strong spatial variability of the trends, a basin average could not be used to realistically assess its behaviour (as a consequence, further studies will include smaller, sub-basin scales). As an example, altimetry in the EMED shows a strong sea level drop in the northern Ionian basin (⇠10 mm/yr) and the opposite in the Levantine basin (⇠14-18 mm/yr). Up to now, no numerical study has been able to reproduce this sea level drop in the northern Ionian basin, instead showing an intense sea level rise over the EMED with similar values to the altimetry trend in the Levantine basin (Tsimplis et al., 2008). In order to study the interannual variability of the signals, the seasonal cycle and trend are removed. Examining figure 5.7 shows that most of the peaks in the model coincide with those in the altimetry data despite some di↵erences in the intensities. Altimetry generally appears to show a more intense interannual variability than SSH however standard deviation calculations revealed that the model and altimetry show similar values over the period studied. Interestingly, the model shows a lower frequency ⇠4 year signal, with a positive trend from 1993 to 1996, negative from 1996 to 2000, and positive again from 2000 to 2004. 56 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation Figure 5.7: Timeseries (with the seasonal cycle and trends removed) of ORCA sea surface height (red) and altimetry (blue) for the WMED, EMED and whole Mediterranean. STD refers to the standard deviation values for each time-series. Corr and Corr NT refer to correlation and de-trended correlation respectively. 5.5.2 Trend As observed in section 5.5.1, the model displays a significant positive trend which is much higher than observed values. In Figure 5.8 we look at the spatial distribution of the trends for SSH and altimetry. While altimetry shows areas of both positive and negative trends with values from -16 to +10 mm/year (average 3.6 mm/year), the model shows a completely positive trend scenarios with values ranging from 8 - 18 mm/year (average 14.95 mm/year). An interesting signal in the Altimetry trend is a negative patch in the Ionian basin. This is most likely due to the circulation changes occurred during the EMT (East Mediterranean Transient), where the areas of deep water formation shifted from the Adriatic Sea to the Aegean (Roether et al., 1996). This signal has not been reproduced by the model. Then again, it is not expected for a global model, especially since this signal has been very difficult to reproduce 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 57 in dedicated studies. Figure 5.8: G70 SSH and Altimetry Trend maps for the 1993-2004 period 58 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 5.5.3 Variance Figure 5.9 shows the maps of variance for the 1993-2004 period of the model’s SSH, the steric height and satellite altimetry. Comparing the model’s SSH and computed steric height (SH) shows a large di↵erence between them, with SH’s variability having a much lower overall intensity than the SSH. Figure 5.9: G70 Steric Height(top), G70 Sea Surface Height (middle), and Altimetry (bottom) variance maps for the 1993-2004 period. The steric e↵ect is an important part of the sea level seasonal cycle, ac- 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 59 counting for roughly half of the total sea level signal. This means that the overall variance of the steric height should behave similarly with respect to the variance from the SSH. However, figure 5.9 shows that the SSH variance is significantly more than double the steric height variance. For example, east of the Balearic Islands, the SSH variance is ⇠70-80 cm2whereas the steric height variance is ⇠25-30 cm2. As identified in sections 5.5.1 and 5.5.2, the SSH signal displays an exaggerated positive trend in SSH. Removing this trend from the variance (figure 5.10) for both steric height and SSH shows little change in the steric height but a significant reduction in the SSH variance, bringing the ratios much closer to the expected values (e.g. east of the Balearic Islands, SSH = ⇠50-60 cm2while SH variance remains similar ⇠25-30 cm2). By comparing the de-trended SSH map with altimetry it is clear that the model is not capable of resolving the intense mesoscale features which are detected by the altimetry, however, ignoring this fact (that was expected given the models spatial resolution), the average variance values of both SSH and altimetry are quite similar. 60 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation Figure 5.10: G70 Steric Height(top) and G70 Sea Surface Height (middle) and altimetry (bottom) de-trended variance maps for the 1993-2004 period 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 67 5.8 Freshwater Balance of the Mediterranean As seen in section 5.5.1, the SSH calculated by the model shows a significant positive trend. In an attempt to identify the possible source of this trend, a water budget calculation was made for the Mediterranean Sea. The water budget is the balance between the water coming into the Mediterranean through its main straits; the Strait of Gibraltar and the Turkish Straits (connecting the Mediterranean Sea to the Black Sea), and the net downward/upward water flux (precipitation minus the evaporation plus the river run-o↵;P-E+R). The transports were calculated in section 5.7, with mean transport values into the Mediterranean Sea for the Strait of Gibraltar and the Turkish Straits of 0.0674 Sv and 0.0189 Sv respectively. The net sea level change due to horizontal water transports was obtained by dividing the total volume of water entering into the Mediterranean Sea by its area (giving a net trend of water inflow of 1.083 x 103±51.7 mm/yr). From this term, the Net Downward Water Flux (NDWF, which includes the salinity restoring term) was subtracted (1.0291 x103±32.5 mm/yr), giving a total net positive sea level change trend for the Mediterranean Sea of 54 mm/yr with an associated error of 44 mm/yr (calculated by a bootstrap method). These trend values, when taking into account the associated error, fall within the trends observed for the models SSH in section 5.5.1 (around 15 mm/yr). The salinity restoring term applied to the model increases evaporation in the NDWF an equivalent to 8.06±4.1 mm/yr, however this is insufficient to compensate for the low evaporation rate of the atmospheric forcing, resulting in an imbalance between the horizontal and vertical water fluxes. Given that the water transport into the Mediterranean Sea is within the range of values obtained in the literature (although transport through the Turkish Straits can show significant error and make a contribution to the sea level trend), the positive trend observed in the models SSH is likely related to an imbalance of the water and heat fluxes of the model. 68 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 5.9 Summary This chapter has focused on studying the interannual and seasonal variability in the Mediterranean Sea by performing a model assessment of the ORCAR025 G70 Simulation, and comparing it with Altimetry and the MEDAR (temperature and salinity) observational database. When comparing the ORCA outputs with the MEDAR database we have found that the mean surface temperature values and the surface layer (0-150 m) over the 1962-2001 period were quite accurately represented with regard to temperature (de-trended correlations of 0.7) However, the sea surface salinity restoring term applied to the model eliminates most of the interannual variability (de-trended correlations of 0.36). Mean temperatures for this layer are slightly higher in the model (0.08 - 0.16oC), very likely related to the atmospheric forcings (ERA40) known underestimation of the total heat loss (⇠3.88 W/m2for ORCA in the Mediterranean as opposed to the well established observation based value of ⇠5 W/m2inferred from heat transport at Gibraltar, meaning that the Mediterranean is gaining heat. However, this result is actually within the range of other observational and modelling studies. Ruiz et al. (2008) puts together a table of the di↵erent heat flux studies and the values range between -11 W/m2and 29 W/m2). Intermediate (150-600 m) and deep (600 mbottom) layers show a clear positive trend that was not seen in MEDAR. This is possibly due to the atmospheric forcings resolution which prevents the formation of deep water resulting in cold, dense waters not reaching the deep ocean which eventually heated up through di↵usion. Our results have shown that the mean surface salinity for the entire Mediterranean basin is significantly lower in G70 than in MEDAR (⇠0.3 psu), which is replicated in intermediate and deep layers to a lesser degree and could be a consequence of a weak sea surface salinity restoring, without sufficient evaporation to compensate for a weak ERA40 water loss flux. The evaluation of G70 with regard to sea level (both in terms of absolute sea surface height (SSH) and its steric component) has revealed that the model reproduces the large scale interannual variability reasonably well, as well as the seasonal cycle when compared to the altimetric data. However the model presents an unrealistic SSH positive trend (⇠15 mm/year). Given that the 5. Basin Scale analysis in the Mediterranean Sea using a 1/4o simulation 69 water transport into the Mediterranean is within the range of values obtained in the literature (although transport through the Turkish Straits can show significant error and make a contribution to the sea level trend), the positive trend observed in the models SSH is likely related to an imbalance of the water budget of the model (E-P-R). As expected with this models 1/4oresolution, which is eddy-permitting but not eddy-resolving, the model is incapable of correctly reproducing most mesoscale features. This is especially notable in the Alboran Sea and Algerian Current where the model is unable to reproduce the gyres and eddies that are formed in these regions. Besides the mesoscale and sea level trends, this global ocean model behaves well in the Mediterranean Sea, taking into account its relatively low resolution for the dynamic features of this semi-enclosed sea. With a few key issues (such as surface salinity restoring and atmospheric forcing) that, once identified, can be improved, the G70 ocean model can provide a very promising tool for the study of the Mediterranean seasonal cycle and inter-annual variability characteristics. Chapter 6 will expand on the knowledge acquired during the model assessment and incorporate analysis of new model runs (from the ORCA series). These new and improved simulations cover a longer time period (up to 2007), have improved the atmospheric forcing (requiring a weaker salinity restoring) and increased the number of vertical levels. Chapter 6 Western Mediterranean Analysis using three 1/4o simulations This Chapter is based on the article (in press) : Vidal-Vijande, E., A. Pascual, B. Barnier, J.-M. Molines, N. Ferry, and J. Tintor´e, 2012: Multiparametric Analysis and Validation in the Western Mediterranean of three global OGCM Hindcasts. Scientia Marina,EOF-2011 Special Edition. 6.1 Introduction In this Chapter expand on Chapter 5 which is based on the article by VidalVijande et al. (2011) and perform the validation in the Western Mediterranean Sea of a series of 1/4oglobal ocean simulations based on the NEMO code (Madec, 2008): two climatic scale hindcasts (ORCA025 G70 and ORCA025 L75.G85) and one shorter simulation with data assimilation (GLORYS1V1). The inclusion of GLORYS1V1 reanalysis in this study aims at verifying that reanalysis products that include data assimilation help to improve the description and our understanding of ocean variability. It will highlight which degree of improvement can be expected from an ocean reanalysis in that region with respect to a free run. This study will focus primarily in the Western Mediterranean, which is a region of significant interest for our research cen71 72 6. Western Mediterranean Analysis using three 1/4osimulations tre (IMEDEA). After carrying out a general assessment of temperature and salinity, particular attention will be paid to the models capability to reproduce the deep water formation in the Gulf of Lions and its transport through the Balearic Sea. The basin average sea level trends will be compared against satellite altimetry and the mean circulation evaluated with two di↵erent mean dynamic topographies. The transport in Gibraltar will be also analysed as well as the exchanges at the Balearic Channels, (Mallorca and Ibiza) where most of the WMED North-South exchanges of heat and water takes place (Pinot et al., 2002). 6.2 Temperature In order to assess the performance of the models with regards to temperature and salinity, the WMED was divided into vertical layers (figures 6.1 and 6.2) following a similar pattern to Chapter 5 which was based on the system used by Rixen et al. (2005). The numerical models have been compared to the EN3 hydrographic database. For these comparisons, both the simulation data for G70 and G85 and the EN3 hydrographic data have been filtered using a 1-year running average in order to remove the intra annual variability and focus on the longer 44-47 year period variability. For GLORYS, since the simulation is much shorter (2002-2009) the data have not been annually filtered. Figure 6.1 shows the temperature timeseries averaged over the WMED at di↵erent depth layers. Figure 6.1(a) is the initial comparison of the three observational datasets (MEDAR, ISHII and EN3). All three datasets behave similarly except for ISHII at deep layers (since it only reaches 700 m) and for salinity in intermediate and deep layers where the variability is much higher than the other two. Of all three datasets, EN3 was deemed the most appropriate due to its compromise between temporal resolution (monthly), depth range (0-5000 m) and period (1950 to present). Figure 6.1(b) shows the comparison between EN3, G70 and G85. Both G70 and G85 have near identical variability at the surface layers, and are in good agreement with EN3. G85 correlates especially well with EN3 (de-trended correlation of 0.87). The major di↵erence found between the three datasets at surface layers is a bias in mean temperature. Taking EN3 as the reference, G85 is 0.44oC cooler and G70 is 0.55oCwarmer. 6. Western Mediterranean Analysis using three 1/4osimulations 73 Figure 6.1: WMED mean temperature anomaly of the interannual variability at di↵erent layers. (a) is the comparison between EN3 (blue), MEDAR (red) and ISHII (green) from 1960 to 2004. (b) is the comparison between EN3 (blue), G70 (red) and G85 (green) from 1960 to 2004. (c) is the comparison between EN3 (blue) and GLORYS (red) from 2002-2009. Mean temperatures (T) and trends (Tr) with their error are given below each image for the di↵erent data streams 74 6. Western Mediterranean Analysis using three 1/4osimulations At intermediate layers (150-600 m), G70 presents an exaggerated positive trend that is reduced in G85 when compared to EN3 (G70 = 0.15oC/yr, G85 = 0.05oC/yr and EN3 = 0.01oC/yr) and mean temperature values (G85 = 13.92 oC, EN3 = 13.47oC). Deep layers also show significant improvement, with G85 and EN3 showing similar trends and closer mean values (although interannual variability is non existent in both simulations). The improvement in temperature trends in G85 can be attributed to two factors; the better time continuity of the DFS4 forcing which has an impact on long term trends, and the increased vertical resolution which has a strong e↵ect on the vertical mixing coefficient of the turbulent closure scheme. The di↵erences in mean temperatures are also result of the atmospheric forcing parameters; DFS3 (G70) is globally imbalanced for heat resulting in a warming ocean, hence higher mean temperatures. On the other hand, DFS4 (G85) maintains a near-zero global heat balance. DFS4 also has stronger wind speeds causing a global cooling of surface waters, which together with the better vertical mixing coefficients due to increased vertical resolution helps propagate the e↵ects of the atmospheric forcing throughout the water column and maintain a better correspondence with the observations. However, as will be discussed in the following section, the inability of the model to reproduce normal winter convection causes temperatures at intermediate and deep layers to warm up, showing a positive temperature bias. As is to be expected from a model with data assimilation, GLORYS performs well when compared with the EN3 observational data (Figure 6.1(c)). At surface layers, GLORYS has slightly weaker peaks in its interannual variability but the overall pattern of variability is very similar with high correlations of 0.77 over the WMED. Intermediate layers also show relatively good performance with slightly lower correlations of 0.66, although the model shows very low interannual variability, similar to the behaviour of the ORCA models. This might probably be due to a reduced number of sub-surface observations included in the data assimilation. At deep layers, the basin mean temperature does not show the spike in 2004 caused by the production of anomalously warm and salty deep water during the winters of 2004-2005 and 2005-2006 (Schroeder et al. (2008a), Schroeder et al. (2010)), however, as seen in Figure 6.6 the simulation does indeed create a strong convection event in the 20052006 winter. The behaviour of the models with respect to observations at di↵erent frequency bands has also been analysed using spectral analysis (not 6. Western Mediterranean Analysis using three 1/4osimulations 75 shown). WMED spatially averaged power spectra were computed showing that for the simulations a predominant peak is registered in surface temperature at 10-12 months corresponding to the annual cycle, and the same occurring for EN3. A weaker semiannual peak at 6 months also appears in the simulations which is not clearly found in the observations. 76 6. Western Mediterranean Analysis using three 1/4osimulations 6.3 Salinity Figure 6.2: WMED mean salinity anomaly of the interannual variability at di↵erent layers. (a) is the comparison between EN3 (blue), MEDAR (red) and ISHII (green) from 1960 to 2004. (b) is the comparison between EN3 (blue), G70 (red) and G85 (green) from 1960 to 2004. (c) is the comparison between EN3 (blue) and GLORYS (red) from 2002-2009. Mean salinity (S) and trends (Tr) with their error are given below each image for the di↵erent data streams While EN3 salinity trends are always either slightly positive or close to zero, both G85 and G70 simulations show negative trends at surface and intermediate layers, with G85 displaying trends 5 to 8 times stronger than G70. This is due to the salinity relaxation in the G85 simulation that is 1/6th of that imposed in G70. This lower relaxation applies less evaporation to the simulation causing a very strong and unrealistic negative trend in salinity, also rendering the mean values over the period studied much lower (at surface lay- 6. Western Mediterranean Analysis using three 1/4osimulations 83 Figure 6.7: Hovm¨oller diagrams showing the temporal evolution of temperature (top) and salinity (bottom) of a 1ox1 oarea at the center of the Ibiza Channel for the G85 simulation. 6.5 Mean Sea Level Performance regarding sea level in the simulations is analysed by comparing them to altimetry over the available common period (1993-2004/2007/2009)(Figure 6.10). Detailed comparison of mean sea level from altimetry and sea surface height from G70 can be found on Chapter 5 and will not be repeated here. However, the main result in this regard was that the model was capable of correctly reproducing the seasonal cycle in both phase and amplitude as well as the interannual variability but SSH from the model showed a exaggerated positive trend of 14.96±1.46 mm/yr when the trend for altimetry was 3.62±1.32 mm/yr. G85 also displays a correct seasonal cycle but the trend has increased even further to 20.27±1.37 mm/yr due to the e↵ect of a reduced salinity relaxation term, which implies lower evaporation and in consequence an increase in sea level. These positive trends of the ORCA models are not an isolated feature of the Mediterranean, but global due to an imbalanced freshwater bud- 84 6. Western Mediterranean Analysis using three 1/4osimulations Figure 6.8: Hovm¨oller diagrams showing the temporal evolution of temperature (top) and salinity (bottom) of a 1ox1 oarea at the center of the Ibiza Channel for the EN3 database. get. The freshwater budget is of vital importance since it contributes to the mean sea level budget closure, impacting sea level rise and density driven circulation (Ferry et al., 2010), however its di↵erent contributors (precipitation, evaporation, river run-o↵and glacier melt) have large uncertainties making the achievement of a balance very difficult. If the trends are removed, the power spectra (Figure 6.11) of G85 and altimetry are very similar in terms of peaks and energy. Essentially both spectra present a marked annual cycle at 10-12 months and a less well defined semiannual peak, but it trends are not removed, higher energy is seen in G85 at lower frequencies. In GLORYS, the net water budget is artificially set to zero in each time step becoming perfectly balanced and therefore it does not a↵ect the mean sea level (MSL) of the model. The consequence is that any changes in MSL are due to the assimilated data provided by altimetry (Ferry et al., 2010). As a result, the comparison between GLORYS and altimetry data yields almost 6. Western Mediterranean Analysis using three 1/4osimulations 85 Figure 6.9: TS Diagrams of a 1ox1 oarea at the center of the Ibiza Channel for the G85 simulation for the period of January to April of 1996 (a), 1997 (b), 2003 (c) and 2006 (d). SW stands for Surface Waters, the rest of the acronyms are detailed in the text. identical results, with no drift in the models SSH and a trend very similar to altimetry. 86 6. Western Mediterranean Analysis using three 1/4osimulations Figure 6.10: ORCA G70 (red), G85 (green) and GLORYS (GL, black) SSH plotted against altimetry mean sea level anomaly (blue) for the WMED. (top) full signal (middle) with the seasonal cycle removed (bottom) with seasonal cycle and trends removed (the numerical values indicate the trend that has been removed). 6.6 Mean Surface Circulation We contrast the mean circulation of G70 and G85 derived from mean SSH and surface velocity currents over the 1993-2004 period with the available literature. The mean circulation in the Western Mediterranean is well known. Atlantic waters enter the Strait of Gibraltar into Alboran Sea (Astraldi et al. (1999), Beranger et al. (2005), Tsimplis and Bryden (2000), S´anchez-Rom´an et al. (2009), Garc´ıa-Lafuente et al. (2002), Gomis et al. (2006)) and form the Western Alboran Gyre (WAG) and Eastern Alboran Gyre (EAG) (Vi´udez et al., 1998). The circulation continues out of the Alboran Sea as the Algerian Current (AC) (Millot, 1999). The AC shows significant mesoscale activity as it continues towards the Sicily Channel. Part of it crosses the channel and 6. Western Mediterranean Analysis using three 1/4osimulations 87 Figure 6.11: Altimetry (top) and G85 (bottom) power spectra for mean SSH for the period 1993-2007. the rest recirculates into the Tyrrhenian Sea, moving along the Italian Coast and becoming the Northern Current as it passes into the Ligurian, through the Gulf of Lion and into the Balearic Sea. The Northern Current recirculates before it reaches the Ibiza Channel and becomes the Balearic current (Pinot et al. (1995), Alvarez et al. (1994), Astraldi et al. (1999), Ruiz et al. (2009)). The ORCA models show the same general circulation pattern (figure 6.12). In more detail, both G70 and G85 show the presence of WAG and EAG, with G70 showing a more prominent WAG. Given the coarse resolution, both models aggregate the Algerian current and the area of intense mesoscale activity above the Algerian current into one meandering, high-intensity band, with G85 reaching further north than G70. Both models show a low minima to the SE of the Gulf of Lions and between Sardinia and Sicily, with G70 being more intense in the former and G85 in the latter. In the Balearic Sea, G70 reproduces the recirculation of the Northern Current further south than G85, closer to Ibiza channel. 88 6. Western Mediterranean Analysis using three 1/4osimulations Figure 6.12: Mean Sea Surface Height maps from G70 (top) and G85 (bottom) with surface velocity vectors added. 6.7 EOF Analysis of SSH We further evaluate the main patterns of variability of G85 by using Empirical Orthogonal Functions (EOFs) and comparing with altimetry. Figures 6.13 and 6.14 present the first EOF mode of altimetry and G85 SSH respectively. The first EOF mode of altimetry, which explains 72.6% of the covariance, shows the strong seasonal variability of the Alboran Gyres with maximum amplitude in autumn (Larnicol et al., 2002). Also seen is the autumn/winter intensification of the cyclonic circulation in the Gulf of Lions. In the first EOF mode of G85 (86.2% cov.) the seasonality of the cyclonic gyre in the Gulf of Lions is also observed to a certain degree, weaker and more towards the north. The signature of the Alboran anticyclonic gyres is almost negligible in G85. Along 6. Western Mediterranean Analysis using three 1/4osimulations 89 the path of the Algerian Current, G85 does not show the mesoscale variability seen in altimetry as expected, although it does exhibit a North-South gradient. Figure 6.13: The first EOF mode for altimetry SLA (sea level anomaly) with the amplitude (top) and pattern (bottom). Figures 6.15 and 6.16 show the first EOF mode of both datasets with the seasonal cycle removed. As is to be expected, altimetry displays more intricate mesoscale patterns and better defined eddy structures such as the WAG and EAG. Despite the weaker Alboran gyres seen in G85 and the stronger pattern in the area of the Gulf of Lion, the simulation does show similar (but more di↵use) mesoscale patterns along the path of the Algerian current; particularly the eddy east of the Almer´ıa-Or´an front and the two eddies southeast of the Balearic Islands. The normalized amplitude timeseries of the first EOF mode 90 6. Western Mediterranean Analysis using three 1/4osimulations Figure 6.14: The first EOF mode for G85 SSH with the amplitude (top) and pattern (bottom). show important disparities, although there are coincidences in the position of some peaks such as 1996. The di↵erence in the explained covariance of both datasets can be attributed to resolution. In altimetry, subsequent modes (not shown) correspond mainly to marked mesoscale eddy features which the model cannot resolve. 6. Western Mediterranean Analysis using three 1/4osimulations 91 Figure 6.15: The first EOF mode for altimetry (without seasonal cycle) with the amplitude (top) and pattern (bottom). 6.8 Transports 6.8.1 Strait of Gibraltar Table 6.1 shows the transport values for the di↵erent simulations analysed in this study. G70 has an inflow of 1.076±0.08 Sv, an outflow of 1.008±0.09 Sv and a net inflow of 0.067±0.06 Sv. G85 has a slightly less intense exchange with 1.006±0.09 Sv inflow, 0.948±0.1 Sv outflow and a weaker 0.05±0.06 Sv net inflow. GLORYS shows a more intense exchange (1.296±0.12 Sv inflow, -1.243±0.10 Sv outflow) but similar net flow to G85 (0.052±0.1 Sv). In this 92 6. Western Mediterranean Analysis using three 1/4osimulations Figure 6.16: The first EOF mode for G85 SSH (without seasonal cycle) with the amplitude (top) and pattern (bottom). regard, all three models fall within the established observational values which in themselves show a large uncertainty due to the lack of long-term direct measurements, the complexity of horizontal and vertical profile of the flow, and the presence of strong tidal currents (Tsimplis and Bryden (2000), Gomis et al. (2006)). To date, no observational e↵orts have been maintained long enough to calculate a long-term mean. The longest study to date is by SotoNavarro et al. (2010) who combined atmospheric data from reanalysis, satellite and experimental observations to calculate a 4 year time series of the Atlantic inflow using the net flow estimated from the Mediterranean water budget and the Mediterranean outflow derived from current meter observations. They ob- 6. Western Mediterranean Analysis using three 1/4osimulations 99 the mesoscale although through EOF pattern analysis, some larger mesoscale features are indeed reproduced. Transport through the Strait of Gibraltar is within the established observational estimates (Candela (2001), Tsimplis and Bryden (2000), Garc´ıa Lafuente et al. (2002)), showing correct inflow (G70: 1.078 Sv, G85: 1.006 Sv), outflow (G70: 1.008 Sv, G85: 0.955 Sv) and net values (G70: 0.067, G85: 0.050 Sv). Only GLORYS appears to have a more intense exchange (1.296 Sv in, 1.243 Sv out) but correct net flow (0.052 Sv). The Ibiza and Mallorca Channels play a vital role in the WMED as it is through these channels that a large part of the North-South exchange of heat and water occurs. All simulations display the correct seasonal variability at the Ibiza Channel, with predominant southward transport in winter and northward in the summer. G70 has correct mean values but does not reproduce the 1996-1998 high southward transport events. These events are reproduced by G85 and coincide with the observations by Pinot et al. (2002) during the CANALES Experiment. It is important to note that according to G85, these years are exceptional. GLORYS produces very intense southward flow (1.5 to 2.5 Sv), far exceeding the other simulations. Observations are not available for the GLORYS years so they cannot be directly compared, however available observational data of previous years show weaker transports (0.6 to 1.3 Sv). This study contributes to the improvement of the ORCA hierarchy of simulations and points out the strengths and weaknesses of these simulations in the Mediterranean Sea. Chapter 7 Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) This Chapter is based on the an article which is currenly in preparation with the following authors: Vidal-Vijande, E., A. Pascual, S. Somot, B. Barnier, J.-M. Molines, and J. Tintor´e, 7.1 Introduction In previous chapters, we have analysed simulations that due to their limited resolution, could not resolve or even produce Mediterranean mesoscale features. Given a Rossby Radius of Deformation of 10-15 km, numerical modelling studies attempting to study dynamical features and processes in the Mediterranean must be of high enough resolution to resolve them. G70, G85 and GLORYS having a 1/4ocoarse resolution were simulations mainly focused on large scale or basin scale processes, not the mesoscale features that dominate the Mediterranean dynamics. This chapter builds upon the previous work by incorporating two higher 101 102 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) resolution simulations, NEMOMED8 (1/8o) and ORCA12 (1/12o). To assess their behaviour in relation to the 1/4osimulations, we start with similar basin scale analysis as seen in the previous chapters, and then we expand on the dynamic features such as the mesoscale circulation and eddy kinetic energy. Before proceeding to the results, we should recall that unlike NEMOMED8, which is a mature simulation specifically fine-tuned to the Mediterranean Sea, ORCA12 is the first hindcast produced by the DRAKKAR group at this resolution and is aimed at assessing the capabilities of the global ORCA models at this resolution. And thus, much improvement is to be expected after this first attempt. 7.2 Temperature We assess the basin scale mean temperature and salinity for the WMED comparing EN3 to NEMOMED8 and ORCA12. Since ORCA12 is a shorter simulation than previous models, we focus on the altimetry period 1993-2007. Due to the shorter period studied, in this case we have used the available monthly data instead of annual means seen in figures 5.1, 5.3, 6.1 and 6.2. The seasonal cycle has been removed. As is to be expected, monthly data yield higher variability in both the simulations and the hydrography. As mentioned in previous chapters, caution must be taken when considering the hydrography variability since sustained monthly observations are scarce and not uniform in space. Figure 7.1 (a) shows the temperature variability at surface, intermediate and deep layers in the WMED. At surface layers, the three datasets show good agreement as has been the case in previous chapters, with high correlations of 0.76 for NM8 and 0.69 for O12. Both the NM8 and O12 simulations behave very similarly and follow EN3 closely. EN3 displays higher variability at certain peaks. There is a significant di↵erence in mean surface layer temperature between EN3 and the simulations, with EN3 at 16.84oC, O12 15.84oCand NM8 at 15.65oC. At intermediate layers (150-600 m) EN3 shows a higher degree of variability than the simulations. NM8 follows the overall pattern set by EN3 without defined peaks (correlation of 0.47), and O12 displays a positive trend up to 2005 with upwards step pattern every winter. The mean temperatures of the 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)103 Figure 7.1: WMED mean temperature (a) and salinity (b) anomaly at di↵erent layers. Comparison between EN3 (blue), ORCA12 (red) and NEMOMED8(green) 1994 to 2007. three datasets are much closer in this case, with the simulations essentially the same and EN3 just ⇠0.2oClower. At deep layers (600 m to bottom), the simulations have very low variability and negligible trend (although slightly positive). EN3 on the other hand shows considerable variability, which is harder to positively rely on due to the low number of actual measurements that reach into this layer. The few CTD measurements that reach beyond 600 m would certainly be scarce and not uniform throughout the WMED. Regarding mean temperatures, ORCA12 is significantly colder than both EN3 and NEMOMED8. 104 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) 7.3 Salinity Figure 7.1 (b) represents the salinity at di↵erent layers. As with the previous ORCA simulations, salinity is a parameter that is hard for simulations to reproduce because it depends on several atmospheric forcing factors which have considerable error over the open ocean (such as precipitation, wind stress and evaporation). In this regard, O12 does not give very good results at surface layers, with its variability not coinciding at all with EN3 and having statistically insignificant or even negative correlations, it is however a better result than seen in the 1/4oG70 and G85 simulations since the trend remains stable. On the other hand NM8 is a considerable improvement; although the variability is lower than EN3, NM8 does follow the same lower frequency pattern as the hydrography and actually has a significant positive correlation of 0.57. This becomes less so at intermediate and deep layers, where NM8 and O12 do follow the overall trend but do not show a particularly good agreement with EN3 variability. We must again be cautious to consider the EN3 variability due to the frequency and spatial distribution of the actual measurements, which in the case of salinity are even lower than those of temperature. One of the problems of numerical models and salinity is that some of them require surface salinity restoring in order to avoid significant drift. Salinity relaxation can be useful in many of the world’s large oceans where the variability is dominated by the seasonal cycle, thus restoring salinity to a climatology helps keep realistic salinity values. However, the Mediterranean variability is extremely complex, and while there is a dominant seasonal cycle, many other important factors play into the total variability signal. For this reason, a standard salinity restoring to a climatology does not work well in the Mediterranean. In fact, the salinity forcing and its interannual variability is essential for the reproduction and understanding of certain key Mediterranean processes such as the EMT. Therefore SSS relaxation as well as any other modelling trick limiting the spatial and temporal variability of the 3D thermohaline structure of the Mediterranean should be avoided (Beuvier et al., 2010). In this case, O12 is using a fairly strong surface salinity restoring (more details in Chapter 3) whereas NM8 actually does not apply any salinity restoring to its simulation. 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)105 7.4 Deep Water Formation in the Gulf of Lion We analyse the deep water formation capabilities of NM8 and O12 in the Gulf of Lions and compare it with EN3. It must be noted that NM8 has already been extensively tested in this area by Herrmann et al. (2010), however we include our analysis for completeness using hovm¨oller diagrams for comparing with the ORCA simulations (which is not done by the aforementioned authors). As in section 6.4, we analyse the temporal evolution of the vertical temperature and salinity profiles (mean of a 1ox1osquare centred at 42.6oN, 4.4oE). Figure 7.2 shows the EN3 reference dataset. The seasonal summer warming and winter cooling at the surface is clearly visible, with warming reaching down to a maximum depth of about 100 - 150 meters and winter cooling which causes vertical mixing and convection normally reaching the intermediate layer at ⇠300 - 400 meters. The deep convection events display a marked interannual variability, with years of intense and deep convection and years of very reduced vertical mixing. The NEMOMED8 simulation (Figure 7.4) clearly displays the interannual variability in the strength of vertical mixing, with intermittent events of deep convection where the mixing reaches below 1000 meters. As opposed to the simulations studied in previous chapters (G70, G85, GLORYS), deep convection is not limited to the very intense atmospheric forcing years of 2005-2006, but also occur with less intensity in previous years, more in agreement with the observational datasets. In salinity this is also visible, with freshening of the water column during these deep convection events, being more intense in during the 2004-2005 and 2005-2006 winters. A notable di↵erence between NM8 and EN3 is the intermediate layer temperature, with NM8 being slightly warmer ⇠13.8oC with regards to EN3 ⇠13.4oC. Likewise in salinity, the intermediate layer salinity for NM8 is saltier (⇠38.65 psu) than EN3 (⇠36.50 psu). The ORCA-12 simulation (Figure 7.4 has a very di↵erent vertical temperature profile than EN3 and NM8. Whereas in both NM8 an EN3, the bottom limit if the intermediate waters is not well defined, cooling progressively from the maxima at around 400 meters down to about 800 meters, the intermediate waters in O12 have their lower limit at a very marked thermocline between 400 and 500 meters depth. They are also significantly warmer than NM8 and EN3 at between 14.2 and 14.7oC. Deep layer temperatures are similar to the observations, creating a very strong thermocline. 106 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) Figure 7.2: Hovm¨oller diagrams showing the temporal evolution of temperature (top) and salinity (bottom) of a 1ox1 oarea at the center of the Gulf of Lion for the EN3 hydrographic dataset. Interannual variability in the winter convection events is also clearly observed in O12 although the water in these events is warmer than EN3 and NM8 and does not appear to significantly cross the thermocline at the deep limit of the intermediate waters. It therefore appears that deep convection does not occur, probably related to the strength of the thermocline at ⇠500 meters. The warmer (less dense) waters mixing down from the surface are not dense enough to cross the thermocline which is too stable for winter vertical mixing to break it up. This appears to be the case even in the severe 2004-2005 and 2005-2006 winters. 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)107 Figure 7.3: Hovm¨oller diagrams showing the temporal evolution of temperature (top) and salinity (bottom) of a 1ox1 oarea at the center of the Gulf of Lion for the NEMOMED8 simulation. 108 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) Figure 7.4: Hovm¨oller diagrams showing the temporal evolution of temperature (top) and salinity (bottom) of a 1ox1 oarea at the center of the Gulf of Lion for the ORCA-12 simulation. 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)115 Figure 7.9: Mean altimetry EKEg timeseries for the Alboran Sea (top), Algerian Basin (middle) and North Balearic Sea (bottom). 116 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) Figure 7.10: WMED mean Geostrophic Eddy Kinetic Energy (EKEg) map calculated from NM8 for the period 1993-2007. Figure 7.10 shows the mean EKEg calculated from NM8 for the 1993-2007 period. NM8 has a similar overall distribution of high EKEg regions, with maxima in the Alboran Sea and Algerian Basin. In this case, the strongest signal is seen in the EAG, which is very clearly defined. The WAG is also visible, with a maxima at the eastern edge, but with the presence of the entire gyre unlike in altimetry. This denotes a more variable WAG with respect to altimetry SLA observations. Higher EKEg is also found at the start of the Algerian Current but not so much to the south-west of Sardinia. Overall the EKEg levels are similar to altimetry in the 0 to 300 cm2/s2range. What is present in NM8 but not clearly visible in altimetry is the flow through the Ibiza Channel and to a lower extent the Mallorca Channel, as well as their continuation as the Balearic Current. The temporal evolution of NM8 EKEg (Figure 7.11) also presents the occurrence of singular events with the potential to influence the mean spatial EKEg patterns. In the Alboran Sea a very intense peak at the end of 1993 with a value of ⇠700 cm2/s2which likely represents a very intense EAG is 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)117 Figure 7.11: Mean NM8 EKEg timeseries for the Alboran Sea (top), Algerian Basin (middle) and North Balearic Sea (bottom). 118 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) certainly a↵ecting the distribution of EKEg values in the spatial means. In the Alboran Sea, there are seasonal maxima but no discernible outlying peaks. In the N-Bal region, a few higher EKEg peaks, likely related to increased flow through the Ibiza Strait are leaving an impression on the spatial means. Figure 7.12: WMED mean Geostrophic Eddy Kinetic Energy (EKEg) map calculated from O12 for the period 1993-2007. Figure 7.12 shows the mean EKEg calculated from O12 for the 1993-2007 period. Immediately visible are the reduced levels of EKEg, which are less than half those of altimetry and NM8. This is due to a relatively stable mean circulation with little variability outside of the mean, therefore when subtracting the MDT to leave the SLA, a larger part of the variability is gone with it. Despite the reduced EKEg values, the locations of maximum EKEg remain similar to NM8 albeit with some di↵erences. In the Alboran Sea, the WAG and EAG are partially visible, but not as clearly as in NM8 or altimetry. The high EKEg region at the start of the AC is the most intense region in the entire domain. The flow through the Ibiza Channel is also very intense, showing that this location has high variability (not removed with the MDT). In the Algerian Basin, there is also a high EKEg region to the south-west of 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)119 Sardinia as with NM8 and altimetry. Figure 7.13: Mean O12 EKEg timeseries for the Alboran Sea (top), Algerian Basin (middle) and North Balearic Sea (bottom). The temporal evolution of O12 EKEg (Figure 7.13) shows interannual variability in all three regions being studied, without any particular outliers which could obviously skew the mean spatial patterns. The peaks display a seasonality in the Algerian Basin (with maxima in winter) and the North Balearic Sea (with maxima in summer). In the Alboran Sea, the seasonal maxima are hard 120 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) to pinpoint since they do not appear to happen regularly in any particular season. 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)121 7.7 Mean Circulation Component Analysis using EOFs We analyse the mean circulation in three di↵erent regions of the WMED (Alboran Sea, Algerian Basin and North WMED) using Empirical Orthogonal Functions (EOFs) on the Absolute Dynamic Topography (ADT = MDT + SLA) of altimetry, and the sea surface height (SSH) parameter of NEMOMED8 and ORCA-12. We calculate EOFs on the full SSH /ADT signal (hereafter EOF-F) and also on the SSH/ADT anomaly by removing a temporal mean (1993-2007) at each grid point (hereafter EOF-A). The analysis of both EOF-F and EOFA is useful because in the case of EOF-Fs, the di↵erent components of the mean circulation can be analysed and quantified separately. EOF-As allow the study of di↵erent variability signals which are obscured by the dominating mean circulation. For the following analysis, we divide the WMED into subregions to better account for the variability in each subregion. For ease of understanding the analysis, we recall the season definitions : Winter (January - March), Spring (April - June), Summer (July - September), Autumn (October - December). 122 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) Figure 7.14: 1st EOF-F mode of total SSH for the Alboran Sea with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)123 Alboran Sea 1st EOF-F Mode (Figure 7.14) For altimetry, the first EOF-F mode accounts for most of the variance (85.9%) and explains the seasonality of the region. Both the WAG and EAG gyres are present but with clear predominance of the WAG (reflected by the gradient and covariance explained by this mode), which is present all year round with a minima in winter and maximum towards the end of summer. There is never an inversion of the anticyclonic circulation (amplitude is always positive, never negative). Note that this first EOF-F mode is very similar to the MDT (Rio et al., 2007). For NM8 the seasonal variability coincides with altimetry in phase and amplitude, with more interannual variability. The EOF-F pattern does not reproduce the WAG, only the Atlantic Jet with some curvature. Conversely, the variability of the EAG is the predominant signal as well as an intense cyclonic gyre located north-east of the EAG marking a sharp front between both gyres, clearly defining the Almer´ıa-Or´an front. The south-east boundary of the cyclonic eddy may correspond to the beginning of the AC. The explained covariance is lower than for altimetry but is still high explaining 62.7% of the covariance. O12 displays a di↵erent spatial pattern, in this case the WAG appears displaced eastward, a situation that has been observed by authors such as Vi´udez et al. (1998), (Vargas-Y´a˜nez et al., 2002) and Flexas et al. (2006) where the WAG can migrate northwards and be ”pushed” eastwards by the Atlantic Jet, although this is not the known predominant state of the Alboran Sea. East and West of the WAG are two cyclonic gyres caused by the meandering flow of the strong Atlantic Jet. There is no presence of the EAG or the Almer´ıaOr´an front. This first EOF-F mode explains 99.5% of the covariance and has little seasonal and interannual variability suggesting this spatial pattern is very stable. 124 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) Figure 7.15: 2nd EOF-F mode of total SSH for the Alboran Sea with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) 2nd EOF-F Mode (Figure 7.15) The second altimetry EOF-F mode has a similar pattern in the WAG area 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)131 Figure 7.19: 2nd EOF-F mode of total SSH for the Algerian Basin with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) 132 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) Figure 7.20: 1st EOF-A mode of total SSH for the Algerian Basin with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)133 feature to the south-east of the Balearic islands. This pattern is maximum in winter and inverts in the summer. The Algerian Current is hardly present in this first EOF-A mode. NM8 shows a very similar spatial pattern and amplitude to altimetry with the dipole mentioned above. In this case, the gradient between both cyclonic and anticyclonic features is stronger as their nuclei are closer together and slightly more intense. The explained covariance (58.9%) is also of a similar magnitude to altimetry. In this case, there is a fainter appearance of the AC. The first EOF-A mode of O12 is clearly dominated by the mesoscale features found within the path of the Algerian Current but well detached from the coast, again with maximum in winter. These mesoscale features could correspond to the typical ”coastal eddies” associated with the AC (Millot, 1999). This mode explains a high 79.2% of the covariance but does not match what is seen by altimetry and NM8 in their respective first EOF-A modes. 2nd EOF-A Mode (Figure 7.21) The second EOF-A mode of both altimetry and NM8 show another strong dipole to the south-west of Sardinia formed by a cyclonic eddy to the west and an anticyclonic eddy to the east (the anticyclonic eddy is sometimes referred to as the Southeast Sardinia Gyre (Astraldi et al., 2002)). In this case it is far more defined, with a stronger horizontal gradient. They both explain a similar covariance (altimetry = 5% and NM8 = 4.4%). In the case of altimetry the signal is mainly due to isolated events in 1997, 2003 and 2006, whereas in NM8 they are more regular. In NM8 there are also a number of weaker mesoscale signals along the North-African coastline associated to the AC variability. For O12, this mode is related to some variability of the AC, in this case closer to the African coastline, revealing some seasonal and interannual variability. North-Western Mediterranean Basin (N-WMED) The N-WMED domain used for this analysis covers part of the Ligurian Basin, the Gulf of Lions, the Balearic Sea and part of the Liguro-Proven¸cal Basin. 1st EOF-F Mode (Figure 7.22) 134 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) Figure 7.21: 2nd EOF-A mode of total SSH for the Algerian Basin with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)135 Figure 7.22: 1st EOF-F mode of total SSH for the N-WMED with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) N-WMED AL TI SSH (ADT) EOF Mode 1 Amplitude . 1993-2007 (total signal) -0.5 -1 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Expl. COy: 69.5% N-WMED NMB SSH EOF Mode 1 Amplitude· 1993-2007 (total signal) -0.5 -1~'-:93C-C:'9:':-94:-:-19:':9C-5 -:1c:'99:::6-'~99:C:7C-C:19:':-98:-:-19:::9::-9 -=2c:'OOOC--:2~OO:C'-2=OO'-:2:CC-200:"::::-3 -=2c:'OOCC4-:20~0:::5-2=006~CC200:':-7~ Expl. COy: 88.6% N-WMED 012 SSH EOF Mode 1 Amplitude . 1993-2007 (total signal) 0.5 -0.5 -1~'-:93C-C:19L94:-:-19:':9C-5 -:1c:'99:::6-'~99:C:7C-C:19L98:-:-19:::9::-9 -=2c:'OOOC--:2~OO:C'-2=OOL2:CC-200:"::::-3 -=2c:'OOCC4-:2~OO:::5-2:::006~CC200:':-7~ Expl. COy: 99.8 % N-WMED ALTI SSH (ADT) EOF Mode 1 1993-2007 (total signal) 15 10 - 5 -o --5 -10 -15 -20 -25 Expl. COy: 69.5% N-WMED NM8 SSH EOF Mode 1 1993-2007 (total signal) -5 -10 --15 --20 --25 -30 -35 Expl. COy: 88.6% N-WMED 012 SSH EOF Mode 1 : 1993-2007 (total signal) -50 -52 -54 -64 -66 -68 -70 Expl. COy: 99.8 % 136 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) The first EOF-F mode of altimetry in the North-WMED (69.5% explained covariance) is dominated by the cyclonic circulation in the Gulf of Lions and the Northern Current which reaches the Balearic Sea (NW of Mallorca) and recirculates into the Balearic Current (Lopez Garc´ıa et al. (1994), Ruiz et al. (2009)). This mode shows a clear seasonal cycle with maximum intensity in winter and minimum in autumn. Since the sign of the temporal amplitude never becomes negative, the pattern weakens but never reverses. In the case of NM8, there is a very strong cyclonic signal in the Gulf of Lions, with an intense Northern Current. However the recirculation happens mainly further North than in altimetry (to the north-west of Menorca) with a weaker signal actually reaching north-west of Mallorca. Some recirculation at that location still occurs but is less predominant. The phase of the seasonal cycle coincides with altimetry. The explained covariance is higher than altimetry at 88.6%. For O12, the main feature is the cyclonic circulation in the Gulf of Lions, with no protrusion of the Northern Current into the Balearic Sea. The recirculation happens even further north than NM8 and an anticyclonic mesoscale signal appears to block the Northern Current from reaching further south. The phase of the temporal amplitude is similar to both altimetry and NM8 but has lower variability as well as a very high explained covariance (99.8%), again suggesting a stable pattern of the first EOF mode. 2nd EOF-F Mode (Figure 7.23) The second altimetry EOF-F mode is also associated with the variability of the cyclonic circulation in the Gulf of Lions as well as the weaker signal of the Northern Current’s recirculation in the Balearic Sea. The phase is inverted with respect to the first EOF-F amplitude, therefore this mode is related to the weaker circulation of the Northern Current in the summer. As with the AC discussed above, the second modulates the first, reducing part of its maximum intensity in winter and increasing its minimum in summer. Also visible is the surface transport through the Ibiza Strait, which is northward in the summer and southward in the winter in agreement with previous in situ observations (Pinot et al., 2002). For NM8, the second EOF-F mode shows a similar pattern and temporal amplitude to altimetry with the cyclonic circulation in the Gulf of Lions. In this case, the transport through the Ibiza Strait is more intense, clearly con- 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)137 Figure 7.23: 2nd EOF-F mode of total SSH for the N-WMED with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) N-WMED AL TI SSH (ADT) EOF Mode 2 Amplitude . 1993-2007 (total signal) -1 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Expl. COy: 23.6% N-WMED NMB SSH EOF Mode 2 Amplitude· 1993-2007 (total signal) -0.5 -1~'-:93C-C:'9:':-94:-:-'9:':9C"5 -:,c:'99:::6-'~99:C:7C-C:19:':-98:-:-19:::9::-9 -=2c:'OOOC--:'~OO:C'-'=OO'-:2:CC-'OO:"::::-3 -=,c:'OOCC"4 -:20~0:::5-2=006~CC200:':-7~ Expl. COy: 6.8% N-WMED 012 SSH EOF Mode 2 Amplitude . 1993-2007 (total signal) -1~'-:93c-c::"::::-cc:':-~~~=-=:-:-19:::9::-9 -=,c:'OOOC--:'~OO:C'-'=OOL2:CC-'OO:"::::-3 -=2c:'OOCC"4-:2~OO:::5-2:::006~CC200:':-7~ Expl. COy: 0.1 % N-WMED ALTI SSH (ADT) EOF Mode2: 1993-2007 (total signal) 20 15 -10 -5 -5 Expl. COy: 23.6% N-WMED NM8 SSH EOF Mode 2: 1993-2007 (total signal) 20 15 - 10 - 5 -o -5 -10 Expl. Coy: 6.8% N-WMED 012 SSH EOF Mode 2 : 1993-2007 (total signal) -2 -2 -4 Expl. COy: 0.1 % 138 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) tinuing north-east as the Balearic Current and eventually becoming part of a mesoscale feature west of Sardinia. Again, for O12 this mode explains a very low percentage of the covariance (0.1%) therefore is not statistically significant. Its main features are the transport through the Ibiza Strait and two anticyclonic mesoscale eddies to the west and north of Mallorca. 1st EOF-A Mode (Figure 7.24) The first EOF-A mode of altimetry is associated with an intense northsouth front centred at 42oN usually referred to as the North Balearic Front (Lopez Garc´ıa et al. (1994), Millot (1999), Olita et al. (2011)). This front is caused by the wind channelling e↵ect of the Pyrenees, especially in the summer where the more stable weather conditions in the N-WMED are disturbed by this e↵ect. NM8 also displays the occurrence of the North Balearic Front, albeit with lower intensity. This mode also corresponds to the transport through the Ibiza Channel seen in the second EOF-F mode and mesoscale activity in the LiguroProven¸cal Basin. The first EOF-A mode of O12 is similar to the second EOF-F mode suggesting that the pattern was so stable in the first EOF-F mode that removing a temporal mean at each grid point essentially removed this first mode. O12 displays a di↵erent pattern to both altimetry and NM8, with a very intense northward flow through the Ibiza Strait (maximum at the end of summer), which continues north closer to the Iberian Peninsula than the Balearic Islands. This is not the location where northward flow is known to occur (Millot (1999), (Pinot et al., 1995), (Ruiz et al., 2009)). Northward flow is usually associated with the Balearic Current which flows very close to the Balearic Islands. In this case, two anticyclonic mesoscale features appear to be constraining the flow further northward. 2nd EOF-A Mode (Figure 7.25) The second EOF-A mode for all three datasets is dominated by discrete mesoscale features in the northern Balearic Sea. In altimetry, a the main signal is an anticyclonic eddy that occurred north of Mallorca in 1998 and studied in depth by Pascual et al. (2002). O12 shows a similar eddy although it happens 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o)139 Figure 7.24: 1st EOF-A mode of total SSH for the N-WMED with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) N-WMED AL TI SSH (ADT) EOF Mode 1 Amplitude . 1993-2007 (anomaly) 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Expl. COy: 81.5% N-WMED NM8 SSH EOF Mode 1 Amplitude 1993-2007 (anomaly) Expl. COy: 76.5% N-WM ED 012 SSH EOF Mode 1 Amplitude . 1993-2007 (anomaly) Expl. Cov: 86.3 % N-WM ED ALTI SSH (ADT) EOF Mode 1 : 1993-2007 (anomalyl 12 -11 -10 -9 Expl. COy: 81.5% N-WMED NM8 SSH EOF Mode 1 : 1993-2007 (anomaly) 18 16 -14 10 Expl. Cov: 76.5% N-WMED 012 SSH EOF Mode 1: 1993-2007 (anomaly) 16 15 14 13 -12 -11 10 Expl. COy: 86.3 % 140 7. Western Mediterranean Analysis using two higher resolution simulations: NEMOMED8 (1/8o) and ORCA12 (1/12o) Figure 7.25: 2nd EOF-A mode of total SSH for the N-WMED with amplitude (left) and pattern (right). Altimetry (top), NEMOMED8 (middle) and ORCA12 (bottom) N-WMED ALTI SSH (ADT) EOF Mode 2: 1993-2007 (anomalyl 10 N-WMED AL TI SSH (ADT) EOF Mode 2 Amplitude . 1993-2007 (anomaly) -, -1 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Expl. COy: 3.1% -, -6 Expl. COy: 3.1% N-WMED NM8 SSH EOF Mode 2 : 1993-2007 (anomaly) N-WMED NM8 SSH EOF Mode 2 Amplitude 1993-2007 (anomaly) -, -0.5 Expl. COy: 6.1% -, -, Expl. COy: 6.1% N-WMED 012 SSH EOF Mode 2: 1993-2007 (anomaly) N-WMED 012 SSH EOF Mode 2 Amplitude . 1993-2007 (anomaly) 0.5 Expl. COy: 3.7 % -2 Expl. COy: 3.7 %