You are on page 1of 10

IBP1822_12 SURFACE CURRENT MONITORING IN SUPPORT TO OFFSHORE INDUSTRY Hlne Etienne1, Marc Lucas 1, Marion Sutton1, Fabien Lefevre1,

Laurent Guerlou1

Copyright 2012, Brazilian Petroleum, Gas and Biofuels Institute - IBP


This Technical Paper was prepared for presentation at the Rio Oi & Gas Expo and Conference 2012, held between September, 1720, 2012, in Rio de Janeiro. This Technical Paper was selected for presentation by the Technical Committee of the event according to the information contained in the final paper submitted by the author(s). The organizers are not supposed to translate or correct the submitted papers. The material as it is presented, does not necessarily represent Brazilian Petroleum, Gas and Biofuels Institute opinion, or that of its Members or Representatives. Authors consent to the publication of this Technical Paper in the Rio Oil & Gas Expo and Conference 2012 Proceedings. .

Abstract
There is growing recognition in the offshore industry of the need for support to management of operations such as drilling or the need to put in place adequate contingency plans in order to respond efficiently to oil leakage events. To this end, CLS has developed a state of the art surface current prediction system massively based on satellite (such as altimetry, sea surface temperature, ocean color) and in-situ observations (ADCP, surface drifters). Depending on the area of interest (shallow water, coastal area, deep ocean, waters around islands, waters in a river plume) currents can be computed/observed by different means. The development of the Hybrid Current methodology aims at using the best currents depending on the location considered. The different types of currents are then merged (in space and time) by the mean of a specific hydrodynamic model (Shallow Water, hereafter SW). Tidal currents, calculated with a global tidal model, are then added to the SW currents to produce the total Hybrid Surface Currents (hereafter HSC). This system provides daily near real time maps of the surface currents at high resolution (4 km, hourly data) and has a 48 hours forecast ability. The currents can be used for day to day operations (direct use, through a web interface or a bulletin) or as inputs to existing oil drift models. Additionally, the system can also provide accurate data for the analysis of past events and the development of contingency scenarios. Finally, the system produces surface currents for design and development purposes.

1. Introduction
Observing ocean currents in real time is still nowadays a significant challenge. Most traditional methods (Current meters, ADCP, drifting buoys, HF Radar) provide only very limited spatial and temporal coverage. Providing even a relatively small temporal spatial coverage is often prohibitively expensive and complicated in terms of installation, maintenance, recovery and processing. As a result, model data are often used as a substitute. However, model data have some significant drawbacks. These include a resolution which is often too coarse to properly address the needs of operational processes (in term of ocean currents) and unavoidable parameterizations which often means that the model drifts away from the observation. One of the solutions developed during the last decades, thanks to the launch of ocean observation satellites, has been the implementation of data assimilation techniques. This involves using observed data to correct the model dynamics. However, this approach is very expensive numerically and scientifically complex, all the more so in the case of the assimilation of observed ocean current velocity. Satellite data offers numerous advantages over traditional ocean observing techniques. Among them is the spatial coverage which is almost global and unaffected by issues such as the remoteness of the location or the local weather conditions (although cloud cover can be an issue). Time coverage is also excellent, as the data if often continuous, with time series going back as far as 20 years. The resolution of satellite data is also very high. Indeed, it can reach down to a few meters for SAR imagery although it is often in the range of a few kilometers for most products. Furthermore, the delivery of the data is fast, i.e nowadays down to a few hours (near real time), far smaller than the characteristic time scales of many ocean phenomena.

______________________________ 1 Dr, oceanographer - CLS

Rio Oil & Gas Expo and Conference 2012 It is therefore relevant to find a substitute to a full 3D ocean model by focusing only on surface currents. To this end, CLS created a HSC system that is easily deployable, relocatable, computationally cheap and that incorporates the most relevant and available satellite data. This system is designed to answer the need for ocean surface currents in hindcast as well as in forecast situations.

2. Hybrid Surface Current Methodology


The HSC methodology is essentially based on a regional high resolution 2D SW model with atmospheric forcing and time evolution. Considering the SW hypothesis, the upper ocean can be modeled as a mobile layer overlying a static layer. The mobile layer depth is time and space dependent. We assume that the velocity shear between the two layers is negligible. The layer of interest is defined as the upper ocean above the main thermocline. The model has thus a 2D spatial representation of the ocean. This is a computationally cheap model, easy to operate in case of specific emergency needs. Data are introduced into the model using a spectral nudging techniques and damping coefficients. Depending on the available data, each model variables (here zonal velocity U, meridional velocity V, model layer height H) are damped on different spatial scales. The high resolution velocity of the model comes from satellite measurements using Surface Quasi-Geostrophic (hereafter SQG) method and the correlation between Sea Surface Hight (hereafter SSH) and Sea Surface Temperature (hereafter SST) gradients. Hence high resolution satellite SST can be used, as well as Sea Surface Color (hereafter SSC) which is a good proxy for SST. 2.1. Spectral Nudging This method is a simple technique for direct data insertion into the model. As the SW model is an approximation, the spectral nudging technique is introduced and provides a control of model error. It is used in the HSC to combine the multiple sources of ocean surface currents considered as valuable inputs for the final product. This technique is based on the implementation of relaxation terms in the model. This is possible insofar as the ocean current data have already been mapped. Depending on the data selected and the represented dynamics, we can filter the scales retained in the final (combined) current product. Spatial frequencies of interest are added in the tendency equations to constrain both the surface velocity and the surface layer depth. Several damping can be used at the same time. For instance: - Large scale velocity damping - Large scale height damping - Small scales velocity damping 2.2. Surface Quasi Geostrophy In the absence of coastal altimetry, coastal radar or very high-resolution performing models, the only possible source of current at high resolution comes from high resolution satellite remote sensing data such as SST data or SSC. These products can achieve resolution of up to of 5km (Donlon et al., 2011) with global coverage or resolutions of 1.1 km with multi sensor and regional coverage (Barre et al., 2006). The SQG method is based on the validity of geostrophy applied to small scales, typically between 30km and 300km (Isern-Fontanet, Chapron et al., 2006), and the correlation between SSH and SST gradients. We can also use the SSC content maps. Indeed this later parameter is often a good proxy (linear) of the SST or the Sea Surface Salinity (SSS) (Sarmiento et al., 2004) Using SSC is particularly useful for getting information about the circulation in region with marked frontal features such as river outflow areas or shelf breaks. In this case, the high resolution velocities are associated with sea surface chlorophyll A concentration gradients. This method requires a conversion parameter between the SST (or the SSC) and the SSH which can then be used to compute a surface velocity. This can be made using in situ data or by adjusting the derived energy spectra close to that of the available OGCM output or altimetry products. However, the geostrophy hypothesis implies that the theory may be weakly valid near the equator. Another important requirement is that there is no cloud coverage over the area that generates gaps in the SST or SSC maps. Thus, satellite maps have to be filled. Gaps due to cloud cover are removed by linear interpolation of the neighbouring. This is a good approximation if the cloud spatial coverage is not large and spurious. Many gridded products use optimal interpolation methods to fill in the cloud cover (Martin et al., 2007). 2.3. Tidal Component 2

Rio Oil & Gas Expo and Conference 2012 A global tide model (MOG2D) is used to add the tidal component to SW model outputs. This model is based on studies undertaken by the French Scientific Research Agency (Carrere and Lyard, 2003). This model is operated by CLS and provides tidal heights and tidal currents with a global coverage. MOG2D is based on the solution of the tidal barotropic equations on a global finite element grid. It solves barotropic, non-linear, shallow water hydrodynamic equations over time. This model is validated by academic works and industrial studies. It is also used operationally to correct altimetry measurements from tidal component. Tidal heights and tidal currents are provided in the past and in the future at a classical hourly rate. 2.4. Use of In-Situ Data: Data Fitting and Genetic Algorithm In situ observations are a critical component of the CLS HSC methodology. They are used to tune the various parameters of the model or to correct the surface currents. Even if the HSC is able to make the current dynamics evolved in time, if the initialisation (restart) and information at the boundaries are not correct, the results will drift from the observed ocean dynamics. This is even more acute if some restoring data are used. In this case, some spurious mesoscale or small scale structures are introduced into the model. This issue can be solved by fitting the co-localised model data (data fitting) to the observations or with a dedicated Genetic Algorithm either in a re-analysis (hindcast) mode (using past data) or in a real time mode. This insures that the models estimates are solidly anchored with in-situ data. Drifters or any current measurement data can be used in order to correct the produced current forecast. This is an a-posterior treatment. This is done by a Successive Correction Method using a Gaussian Filter to model the covariances (Hayden and Purser, 1988; Hayden and Purser, 1995.).The recursive filter has been especially designed to provide a computationally efficient interpolation method capable of producing realistic results for datasets with spatial inhomogeneities. The Genetic Algorithm approach (hereafter GA) is an optimization method based on techniques drawn from the world of genetics and natural evolution: in a given population, only the best-adapted individuals pass their genes on to the next generations thanks to inheritance, mutation, selection, and crossover. With every generation, the specie is thus better adapted to its environment, either through the sharing of genetic material or through random mutations. The GA is the computing equivalent of this natural process, the parameters of the model having the role of genes and observations the role of the natural environment. Commonly, the algorithm terminates when either a predefined maximum number of generations is reached, or a satisfactory fitness level is find for solution. If the algorithm calculation has terminated due to a maximum number of generations, a satisfactory solution may or may not have been reached. This mostly depend on observation data coverage and availability, which is the critical point of the methodology. A study had shown the feasibility of applying a GA in the fields of ocean currents and of adjusting a model solution to theoretical data. But this method is not used for operational purposes as not enough in situ data are available on a daily basis.

3. The Example of the Campos Configuration


This Hybrid Current methodology has been applied over various regions of interest such as the south Brazilian coast, the Gulf of Mexico or the Gulf of Guinea. Some validation results are presented here, showing the ability of the method to provide accurate information on surface current. The SW equations rely on several approximations which are valid over a region of homogeneous physical properties such as mixed layer depth and characteristic scales (temporal and spatial). As a result, in practice, the model is deployed over a reduced area (at most 15 by 15) and with the understanding that some parts of the model will yield much better results than others. The results presented here are for a configuration deployed offshore Brazil (Figure 1), hereafter called the Campos configuration. Note that this configuration does not use the data fitting post treatment. This region is divided into two dynamically distinct areas: in the south, the Campos Basin known for its highly non linear circulation, due to the presence of many mesoscale eddies and a northern region with more linear dynamics. The Brazilian Current (hereafter BC) flows southward between 20S-28S adjacent to coast. This is a western boundary current which transports between 5-13 Sv. It separates from the coast at about 38.5W+/- 1.1 (De Souza and Robinson, 2004). Large frontal meanders are observed in the vicinity of 22S and the presence of a standing eddy between 23S and 22S has been established (Da Silveira et al., 2008). At the surface, the maximum speed is of 0.4-0.7 m/s according to Da Silveira et al. (2008). Another current in the region is the Brazilian Coastal Current (hereafter BCC) which flows northward near the coast (to approximatively 23S) and has a distinct SST temperature signature (the 20C isotherm isolates the BCC waters from the BC waters). There is also clear correlation with productivity as established by Suny and Servain (1998).

Rio Oil & Gas Expo and Conference 2012 This current is driven by large scale phenomena, notably the oscillation in the subtropical front in the South Atlantic. The BCC average speed is 0.11 m/s (std dev of 0.06 m/s) and the mean direction is 27 and has about 95% of its energy in eddies and small scale perturbations.

Figure 1. Campos configuration bathymetry (in meters). 3.1 Description of the Campos Configuration: The configuration presented here extends from 29S to 14S and from 50E to 35E along the Brazilian coast. Figure 1 shows the model domain bathymetry and the land-sea mask as represented in the model. The bathymetry of the Campos configuration comes from a 1/60 CLS database. It has been altered by removing coastal lakes and lagoons which can cause numerical instabilities. The time step is 600 seconds. For storage reasons, the output frequency was set to 1 hour (in theory, it could be set to 10 minutes). The Campos configuration is run from March 2010 to March 2011. This is a regional configuration set up with open boundaries (surface and lateral) where fluxes (ocean and atmosphere) must be set. The open boundary conditions (u,v, h) are provided by the Mercator PSY4v1 operational system (http://www.mercator-ocean.fr/fre/produits-services/produits-numeriques/produits-PSY4V1) with a time resolution of 1 day and a spatial resolution of 1/6; and by DUACS delayed and near real time daily Absolute Dynamic Topography (http://www.aviso.oceanobs.com/en/data/products/sea-surface-height-products/global/index.html) Inside the Campos domain, the wind driven part of the circulation is calculated from ECMWF 6 hourly fields. The large scale and coastal circulation are altered through a restoring to the daily Mercator PSY4V1 u and v fields. SSC data from the CLS generated fields provides information on the smaller scale features such as eddies and fronts by use of computed SQG velocites. In this area, some algae bloom can be seen in the SSC field and produce SSC gradient without any relation with horizontal current pattern which is why the correlation between geostrophic currents (from altimetry) and SQG currents has to be checked regularly. 3.2 Data Sources: One of the key features of the system is the access to multiple sources of data. Depending on the specific area chosen, different products will be used to represent specific scales of the dynamics. Two aspects of the data determine its suitability for use by the HSC, namely it quality and its availability in real, near real time and in forecast mode (up to 3days for the HSC forecast purpose.) CLS is a data provider for the Sea Level Thematic Assembly of the My Ocean project (http://www.myocean.eu.org/) and produces satellite merged product that are distributed daily to users around the world. Hence, if consistent with the HSC need, priority is given to these data sets in operational mode. The atmospheric forcing, namely wind stress come from atmospheric models in forecast mode, or merged observed products for hindcast activity. When available, a regional configuration is preferred but the outputs from global models (such as ECMWF) provide excellent fields for offshore features. The altimetry data comes from observed Absolute Dynamic Topography fields. The resolution depends on the region of the world chosen. It can vary from 1/3 to 1/6 in the case of regional altimetry data. The AVISO (http://www.aviso.oceanobs.com/en/home/index.html) data is particularly well suited as it is updated daily and maintained operationnally. The choice of the velocity data for large scale usually comes down to currents derived from altimetry (geostrophic currents) or currents obtained from OGCM models. AVISO and MyOcean provide efficient and reliable 4

Rio Oil & Gas Expo and Conference 2012 access to such data although there are other purveyors of such data. But only an OGCM can provide current forecast which is why they are used for daily operational purposes. For small scale velocity data from the SQG method, there is nowadays a wealth of SST and SSC data available captured through passive and active sensors. Again, the determining criteria will often be operational availability. In many case, it is better to use a product that combines a range of sensors from different satellites, in particular to resolve the cloud cover issues. Some SST and SSC data are operationally produced at CLS. For the SST, maps from InfraRed (IR) and MicroWave (MW) sensors are merged to provide SST map 3 times a week. Using these two kinds of instruments allows high resolution view (from IR) and good coverage (less cloud, from MW) of the ocean surface. The available SST data are on a global domain and at 2km resolution. The SSC products are also from a merged of different satellite missions (MODIS/AQUA, MODIS/TERRA, MODIS ENVISAT). SSC is observed in the visible band. This is a limitation as there is no night observation, no measurement under cloud cover and an increasing uncertainty near the coast. SSC is produced 3 times a week with a 2km resolution at a global scale. To validate the HSC system, the surface currents are compared to drifters observation as well as in situ ADCP data when available. CLS has developed and maintains a database of all available drifters data from Coriolis (http://www.coriolis.eu.org/) or ARGO (http://www.argo.ucsd.edu) measurement programs. It is important to note that the quality of the drifting buoy data is being questioned in view of the tendency for the buoys to loose their drogue after a couple of months at sea (Grodsky et al., 2011). It must be noted that tools used for the comparison with drifters observation dont allow for a proper analysis of small features which are the principal contribution of the HSC system. Indeed, buoys data are filtered using a Lanczos filter to remove inertial movements, which corresponds more or less to a 3 days averaging of the data. It is therefore only possible to validate the large features of the models. Comparisons are made along the drifters tracks and at the drifters time step. In the following validation, prior to any comparison, the HSC model outputs are interpolated at the drifters hourly positions, averaged over 3 days and compared to the 3 days filtered drifter data at these positions. 3.3 Results The Campos outputs are compared to drifters observations. Statistics are computed along the drifters tracks and at the drifters time step. The following Figure 2 gives the trajectories of drifters passing through the area during the one year period 01 March 2011-29 February 2011.

Figure 2. map of the speed (left, in cm/s) difference and direction (right, in degree) between model and surface drifter trajectories in the model area between march 2010 and march 2011. The south-western part of the domain is not sampled at all by the drifters during this period. As a result, the BCC area wont be validated here. The mean speed difference is about -2.8 cm.s-1, indicating that the HSC slightly underestimates the surface current strength. Figure 2 shows that some regions, such as the coastal entrance of the domain (north of 18S) or some particular local features (eddy trajectories centred around 22S-43W or 22S-41W) have an absolute velocity value difference with drifters greater than 35 cm.s-1. Considering the current direction, the mean difference with drifters (Figure 2) is about -3.4 with an RMS difference equal to 73. The area where direction 5

Rio Oil & Gas Expo and Conference 2012 differences are greater than 45 corresponds mainly to the northern open boundary. The same differences can be found in the input data set used to specify the surface current trough the open boundary (not shown). In other words, the difference with the drifter data is not only due to the HSC but rather to the boundary forcing data used. In the interior, the features showing greater direction differences with drifters correspond to mesoscale pattern, mostly eddies. They are in most cases slightly shifted from their observed positions. Figure 3 shows that the maximum speed of the drifters is about 0.7 cm.s-1, whereas the HSC reaches 0.68 cm.s1. Generally, the underestimation is not very important for the higher speed range. It can reach 16% of the drifters velocity for HSC velocities between 0.2 and 0.4 cm.s-1, which is a reasonable result.

Figure 3. Quantile-Quantile (in m.s-1) plot of HCS current (quantile every 1%) The drifters indicate a major current direction in the southwest quadrant (Figure 4). The HSC indicates a major direction to the north. The northeast to southeast part of the current is quite well estimate by the HSC, but the southweastward component is underestimated. The southward current, mainly corresponding to the Brazil current (Figure 5), is also well represented by the HSC. The northwestward part of the current distribution is correct, even if the velocity is underestimated. The main difference in the current distribution between drifters and model occurs in the northward and westward parts. The HSC misses the westward component of the observed current, mostly in the domain interior. The overestimation of the northward component of the current is also seen in the OGCM input forcing and can partly explain this pattern.

Figure 4. mean current roses over the one year period for drifters (left) and HCS (right) currents

Rio Oil & Gas Expo and Conference 2012

Figure 5. Mean surface current over the March 2011- December 2011 period for HCS Figure 6 shows two maps of SQG currents produced with the SSC. A strong SSC gradient along the coast is obtained, mostly due to land impact on the biology (export of nutrients from the continent due to natural and human activities). In this case SSC is not related to a river discharge or buoyancy flux from an upwelling. Spurious high velocities are generated that are filtered out from the spectral nudging input data. The mean patterns of the SQG current show the presence of coastal standing eddies north of 20S and around 25S. In this last area, the SSC patterns allow a representation of the northward BCC and the eddy activity along coast. This shows the ability of the SQG SSC to improve the near shore resolution of current features, even though we cannot validate precisely the velocities due to a lack of in situ observations in the vicinity of the BCC, south of 25S.

Figure 6. SQG from SSC data computed for 2011, on 1st of March and on 6th of December.

4. The Services

Rio Oil & Gas Expo and Conference 2012 The above products are then distributed to the users via a dedicated web access or through personalized services. Both accesses have been operationally implemented. The WEB service delivers 7 days / 24 hours surface current data: Currents forecast are updated every day (Figure 7) Temporal resolution is every 3 hours on the charts, and hourly at the rig location (Figure 8) Spatial resolution is ~3kms Ocean Current is available up to 7 days in the past Ocean Current is available 3 days in forecast Like any WEB service current information can be accessed from anywhere as long as you have Internet access.

Figure 7. Screen capture of the CLS offshore web site for the Campos configuration

Figure 8. Screen capture of the CLS offshore web site at one rig location for the Campos configuration. Moreover, dedicated bulletins are prepared by one oceanographer on duty (Figure 9). Bulletins can be provided twice or three times a week (on demand) to give the customer warnings and expert advice on incoming ocean features. 8

Rio Oil & Gas Expo and Conference 2012 These bulletins contain: Monitoring of eddies with dedicated warning levels Time series of current forecast at the rig location(s) Analysis performed by one of our experts

Figure 9. Example of available personalized bulletin.

5. Conclusion
9

Rio Oil & Gas Expo and Conference 2012 To answer the need for higher accuracy surface currents in the offshore industry, a new method massively based on satellite observations has been developed. This method combines shallow water equations and restoring to observed and calculated velocity fields. Initial results show a significant improvement to traditional modeling method, mostly in the coastal area. However, the tuning of the parameters, a prerequisite to high quality results, needs evenly distributed observations in both space and time, and the HSC uncertainties are linked to the input data set errors. New satellite data products such as sea surface salinity can easily be incorporated in the system and thus further improve the representation of currents, in particular in the vicinity of river plumes. One of the principal strengths of this approach is the low numerical cost. The simulations can be performed quickly in view of the reduced number of grid points.

6. References
BARRE, N., PROVOST, C., SARACEANNO, M. Spatial and Temporal Scales of the Brazil extendashmalvinas Current Confluence Documented by Simultaneous MODIS Aqua 1.1-km Resolution SST and Color Images. Advances in Space Research, v. 37, n. 4, p. 770-786, 2006 CARRERE, L. , LYARD, F., Modeling the barotropic response of the global ocean to atmospheric wind and pressure forcing - comparisons with observations. Geophys. Res. Let., v. 30, n. 6, p. 81-84, 2003. DA SILVEIRRA, I. C. A., LIMA, J. A. M., SCHMIDT, A. C. K., CECCOPIERI, W., SARTORI, A., FRANSISCO, C. P. F., FONTES, R. C. F., Is the meander growth in the brazil current system off Southeast Brazil due to baroclinic instability. Dynamic of Atmospheres and Oceans, v. 45, p. 187-207, 2008 DE SOUZA, R. B., ROBINSON, I. S., Lagrangian and satellite observations of the Brazilian Coastal Current. Continental Shelf Research, v. 24, p. 241-262, 2004 DONLON, C. J., MARTIN, M., STARK, J. D., ROBERTS-JONES, J., FIEDLER, E., WIMMER, W. The Operational Sea Surface Temperature and Sea Ice analysis (OSTIA). Remote Sensing of the Environment. doi: 10.1016/j.rse.2010.10.017 2011, 2001 GRODSKY, S. A., LUMPKIN, R., CARTON, J. A., Spurious trends in global surface drifter currents. Geophys. Res. Lett., v. 38, L10606, doi:10.1029/2011GL047393, 2011 HAYDEN, C. M., PURSER, R. J., Recursive Filter Objective Analysis of Meteorological Fields: Applications to NESDIS Operational Processing. J. Appl. Meteor., v. 34, p. 3-15, 1995 HAYDEN, C. M., PURSER, R. J., Three-dimensional recursive filter objective analysis of meteorological fields. In: Eighth Conf. on Numerical Weather Prediction, Baltimore: Amer. Meteor. Soc., p. 185190, 1988 ISERN-FONTANET ,J., CHAPRON, B., LAPEYRE, G., KLEIN, P., Potential use of microwave Sea Surface Temperatures for the estimation of ocean currents. Geophys. Res. Lett, v. 33, L24608, doi:10.1029/2006GL027801, 2006 MARTIN, M. J., HINES, A., BELL, M. J., Data assimilation in the FOAM operational short-range ocean forecasting system: a description of the scheme and its impact. Q.J.R. Meteorol. Soc., v. 133, p. 981-995, 2007 SARMIENTO, J. L., SLATER, R., BARBER, R., BOPP, L., DONEY, S. C., HIRST, A. C., KLEYPAS, J., MATEAR, R., MIKOLAJEWICZ, U. , MONFRAY, P., SOLDATOV, V., SPALL, S. A., STOUFFER, R., Response of ocean ecosystems to climate warming. Global Biogeochem. Cycles, v. 18, GB3003, doi:10.1029/2003GB002134. 2004 SUNYE, P. S., SERVAIN, J., Effects of seasonal variations in meteorology and oceanography on the Brazilian sardine fishery. Fisheries Oceanography, v. 7, p. 89-100, 1998

10

You might also like