Project Description

De MEOPAR
Aller à : Navigation, rechercher

Problems being addressed and state-of-the-art overview

The Estuary and Gulf of St.Lawrence (EGSL) is a seasonally ice-infested region with densely inhabited shores. It supports important socio-economic activities such as fisheries, aquaculture, tourism and transportation. Rapidly changing meteorological conditions, complex ocean circulation, long fetch distances, cold temperatures and seasonal sea ice presence all contribute to increase the risk of accidents and the difficulty to respond efficiently. Recently, Canadian federal and provincial (Québec, Newfoundland and Labrador) governments pushed the exploitation of offshore oil and gas resources up in the political agenda [¡4]. Strategic Environmental Assessments that have been successively launched or realized [2,12] all suggest that potential impacts on socio-economic activities and on the ecosystem remain largely undocumented, calling for a massive investment towards increasing our understanding of the system and the development of cutting-edge response technologies.

The Canadian Coast Guard (CCG) is the prime responsible for responding to emergencies in Canadian waters. In the event of an emergency call, CCG Search Coordinators gather key informations about the target (last known position, shape, etc.) and retrieve the latest wind and current forecasts. For the ESGL, these forecasts are produced by the Canadian Meteorological Center and are made directly available to the CCG through Fisheries and Oceans Canada by the Maurice-Lamontagne Institute (MLI). Depending on the exact location of the incident, current forecasts either come from a 400-m resolution model of the Estuary area (called STLE) that predicts currents exclusively from tidal harmonics [27], or from a 5-km resolution coupled atmosphere-ocean forecasting and analysis system of the Gulf area (called MOGSL) [26,31].

With this information in hand, possibly including updates from the rescue team, simulations of surface trajectories are launched to determine optimal search areas. These simulations take into account all informations about the target and time-dependent forces as well as their respective uncertainties. There are two main types of models that are used whether the target is a fluid (e.g. an oil patch or a cloud of toxic algae) or a solid object (e.g. a human or a life raft). When the target is completely submerged in the water, like in the former case, only surface currents are considered and simple Lagrangian or Eulerian advection-diffusion-reaction models are used [13,23]. In the latter case, the wind acting on the emerged part of the object also influences the drift. The shape of the object becomes a central parameter and Leeway drift models [3,6] are then used. In Canada, all this is done with the SpillSims software at MLI [8]. Even though waves play a significant role in the surface drift [4,22] they are not explicitly considered in the present system, nor in most national operational systems. Leeway drift models partly include the effect of waves in the so-called leeway coefficients that must be specified as an input parameter to the model [3,6].

If an emergency happens during winter in ice-infested waters, the surface drift will be highly influenced by sea ice. In the present model sea ice is modeled as a floating newtonian fluid, i.e. it behaves like a passive dye tracer released in the ocean, and is initialised with ice charts provided by the Canadian Ice Service. This choice has been made mainly because traditional viscous-plastic sea ice models [e.g. 17] used in most operational forecasting systems around the world do not represent well the ice drift in loosely ice-infested conditions like the ones often prevailing in the EGSL, where sea ice is relatively thin (<1.5m) on average and made of relatively small floes (<10km). In this context, sea ice is best modeled as a collection of floes interacting with each other than a viscous-plastic continuum. However, in certain regimes of concentration and strain rates, collisional flows behave quite differently from newtonian flows [29]. The emerging class of collisional models [e.g. 9,10,15,28], that consider floe size as a key state variable, may constitute a very interesting option for improving ice drift forecasts.

The strategy outlined above is in operation since 2002 and has responded to thirty requests per year on average since then, and 41 requests in 2012-2013, including exercises. Like any operational system, it is not perfect and a number of weaknesses can be identified and improved if research and development efforts are invested. The most important one is the lack of validation of surface currents against field measurements. The recent deployment of four high-frequency radars (HFR) along the shores of the Estuary now allows for extensive validation. Figure 1 shows a comparison between surface currents predicted by the 5-km operational model and those measured by the HFR array at the same time. This figure highlights two possible weaknesses that remain to be fully characterized. The first one is that the horizontal resolution of the model (5~km) is not sufficient to resolve the mesoscale dynamics in the LSLE, while HFR capture well the mesoscale and partly the submesoscale. It is unclear at this point whether unresolved subgrid scale processes are correctly parameterized. The model uses the widespread Smagorinsky scheme [30] to hopefully represent the net effect of unresolved lateral eddies, but this has not been tested in the EGSL and represents one of the largest uncertainty in model parameters. The new installation of HFR now provides an invaluable means to test and perhaps adapt or reconsider the use of the Smagorinsky scheme for parameterizing mesoscale and submesoscale eddies in this system.

The second aspect is that surface currents in the model represent a vertical average over the first grid cell (5-m thick in this case) while HFR typically measures the displacement of the first 1-2 meter of the water column [33]. Surface drift in the ocean is driven by two main processes, Ekman currents coming from the balance between friction and rotation, as well as the Stokes mass transport arising from non-linear surface gravity wave effects. The model only includes Ekman currents while HFR measurements take into account both contributions [22]. HFR thus offer a great potential for advancing our understanding of surface drift and therefore significantly improving models.

Comparison between HF Radars and the model

Other models limitations in providing reliable surface drift data can also be overcome when real-time observations from HFR are available. For example, [1] used an operational oil spill model driven by wind and waves from an operational forecast system, and optionally by currents from HFR, to simulate the trajectory of a surface drifting buoy deployed in the radars coverage area west off the Galician coast. They found a 62% reduction in the simulated 24-hour Search and Rescue (SAR) area when computed using the HFR currents. [20] compared the use of surface currents observed by HFR and predicted by a 3-D operational ocean forecast model to simulate the trajectories of surface drifters deployed on the Middle Atlantic Bight shelf. They found similar tracking skill at the 95% confidence level.

Recognizing that this problem can't be handled for the entire Canadian coastline, we believe that progress made at regional scales can be significant and contribute in the longer-term at the national scale.

Research approach

The proposed research builds on existing expertise and in a large part on ongoing research activities. It responds directly to the first target area identified in the first MEOPAR's open call for proposals (2013-2016). The overarching goal of this project is to improve surface drift forecasts by making use of an innovative combination of high frequency radar data and a suite of atmosphere, ice, wave and ocean models. We plan to make progress towards this goal by pursuing three specific objectives:

  1. To assess the performance of operational ocean models against new field remote and in situ measurements;
  2. To extract ice drift from high-frequency radar data and improve the sea ice drift model;
  3. To examine whether the explicit inclusion of the wave-induced mass transport (Stokes drift) improves surface drift forecasts.

The project is planned for a one-year duration, but it is designed with the idea of a longer term extension to the project (See §1.10). Proposed activities are separated in two main work packages: one core package focusing on summer conditions (ice-free waters) and one exploratory package focusing on winter conditions (ice-infested waters). This separation acknowledges differences in the availability of data and in processes controlling the surface drift. The former will benefit from an extensive radar coverage and in situ current measurements from various sources and is based on a solid literature, while the latter includes several exploratory aspects pertaining to ice drift detection from radars and sea ice drift modeling in seasonally ice-covered waters.

Summer drift package

At the start of the project in October 2013, a multi-sensor dataset of open water current and drift will be readily available. This includes two CODARs (Coastal Ocean Dynamics Applications Radars) on the south shore and two WERAs (WavE RAdars) on the north shore (see Fig. 1) continuously measuring surface currents since December 2012, every 20 minutes for the WERAs and every hour for the CODARs (CC) over an area approximately 45 km (the estuary width) by 80 km.

An important issue with HFR is that they measure the component of surface currents in the direction of the radar (called radial current). Therefore, measurements from at least two radars must be combined to obtain the vectorial current field, and results are sensitive to the method used for combining radial currents. [19] compared the standard unweighted least-squares (UWLS) method used by [1] with a more recently developed optimal interpolation (OI) method [18], and obtained a 50\% reduction in the simulated 24-hour SAR areas with the OI method relative to the UWLS method. Had [20] used the OI method, they might have found better tracking skills when using observed rather than predicted currents.Here, we propose to test the impact of different methods of combining radial currents from HF radars (namely UWLS, OI and the recently-developed variational method of [34]) on the tracking skills of surface drifters in the lower St-Lawrence Estuary. The four-radar configuration used provides much better geometric constraints when combining radial currents than a standard two-radar configuration, as well as redundancy in case of a radar malfunctioning, which would be a critical feature for an operational SAR system.

The dataset also includes a 12-hour and a 24-hour surface drifter trajectories obtained during an oceanographic expedition from 14 to 19 May 2013 in the LSLE (CC) and Acoustic Doppler Current Profiler (ADCP) surveys of surface (0-20m) current velocity profiles to be obtained in August 2013 (DB}). Time series of current profiles, wind, water and atmospheric temperatures measured by the oceanographic and meteorologic IML-4 buoy will also be added to the Summer dataset. In May 2014, many surface drifters (~12-15) will be repeatedly deployed at different locations within the HFR coverage area. This experiment will be carried out from ISMER's F.J. Saucier research boat over a two-week period, wheather permitting (DB, SS). This will constitute the core dataset against which improved models will be tested.

The assessment of model performance (objective 1) will begin right at the start of the project, in October 2013. Surface current forecasts from the coupled atmosphere-ocean model MOGSL will be archived starting in June 2013 (DL). The STLE tidal model will similarly be run in its present version. Model outputs will be directly compared to measurements of eulerian currents (ADCP and HFR), while buoy trajectories will be compared with the Lagrangian drift computed from models and HFR data. This step will be done with SpillSims (DL) that uses a fourth-order Runge-Kutta scheme, but also by adding stochasticity (random walk) to simulate unresolved dispersion. The idea here is to test the operational system in its current version against HFR and in situ data. Readily, the computation of the Lagrangian drift directly from HFR data constitutes an alternative method to adjust the object position or even make some complementary predictions [5,11].

Based on obtained results, models will be improved with two main research questions in mind. The first one is how resolution affect surface currents. In other words, is it worth investing significant resources to increase model resolution and if yes, by how much? For this, we will make use of the 400-m resolution STLE tidal model and turn on all other relevant forcings. We will then compare directly the eddy-resolving solution with the eddy-permitting solution of the 5-km companion model of the GSL. It is expected that by resolving the internal radius of deformation the high-resolution model will better represent the mesoscale circulation and its variability.

The second question we want to address is to what degree surface gravity waves affect the surface drift? Answers to this question will be sought by comparing HFR, ADCP and buoy data that all integrate the effect of currents on different vertical scales. Wave parameters (significant wave height, peak period and mean direction) will be estimated using a spectral wave model to be developed, based either on the SWAN (Simulating WAves Nearshore [16]) or the WAVEWATCH III code, and supplemented by in situ wave measurements (UN). This will allow for the separate estimation of the Stokes drift, which is maximal at the very surface and decays rapidly over the first few meters, and Ekman currents that are obtained independently from hydrodynamic models. Those two contributions will be combined and compared to buoy trajectories in order to examine whether waves should be explicitly taken into account in operational forecasting systems (objective 3).

Not only the current operational models will be validated against a thorough dataset, but an original combination of models and observations will be developed and proposed for a major system update.

Winter drift package

Forecasting the surface drift in ice-infested conditions is a very difficult task with very little previous experience to learn from. There is very limited surface current data available for the winter months in the LSLE, and the level of complexity of sea ice drift models is kept minimal. In principle, HFR can be used to extract valuable information of sea ice drift. To our knowledge, there has only been one published attempt at extracting sea ice drift from HFR measurements [21], but no independent measurements were available for validation. Here, we propose to collect independent current and drift data during winter months in order to develop and validate new HFR ice drift algorithms (CC,DD). An Acoustic Wave and Current Profiler (AWAC) will be deployed offshore Rimouski, within the range of HFRs, in November 2013 (UN). It will record waves, current profiles and ice thickness, if present. Beacons will also be acquired and deployed in forming ice from CCG icebreakers (e.g. CCGS Martha L. Black) or on nearshore floes from ice canoes. UQAR has a sport team of trained ice canoe athletes that have offered their help for deploying ice beacons. Since their retrieval is less obvious in winter than in summer, these beacons will be deployed at strategic locations between Québec and Rimouski in order to maximize data acquisition. Buoys will be transmitting for a 6-month period during which they are expected to drift in the HFR coverage zone. They will be recovered if they beach at accessible sites before their batteries die.

To better simulate the drift of ice floes, we propose to implement a collisional model in which a large number of circular floes of difference sizes are submitted to external forces and are interacting with each other (BT,DD). This model will explicitly take into account floe size-dependent forces affecting drift speed such as air and water skin and form drag forces and the wave radiation pressure [32]. The wave radiation pressure will be simply parameterized as a function of wind speed and direction and ice concentration. The model will be initialized with ice charts or satellite images (e.g. NASA MODIS Subsets) and parameters like the floe size distribution, ice concentration and drag coefficients will be tuned to best fit buoy trajectories. Sea ice thermodynamics will not be simulated.

This exploratory work is highly innovative since it will provide an exhaustive multi-sensor dataset in ice-infested waters against which new models and algorithms (objective 2) will be developed and tested.

Expected outcomes of the research

It is expected that by the end of the project, we will have a clear idea as to how the current operational drift forecasting system should be updated to significantly improve its performance. Such improvement is expected to happen partly during the project, but will be continued by interested government partners beyond the project. There will also be a massive amount of data that will eventually make the subject of primary publications in oceanography journals.

Contribution(s) to MEOPAR's Vision and Strategy Objectives

This project is a timely contribution for advancing observation and prediction technologies in Canadian coastal areas. It will make a valuable contribution to both MEOPAR’s observation and prediction cores. All model improvements and relevant data will be transferred to and shared with Environment Canada for improvements of their operational systems, consolidating the link between academic and government research.

Social/policy issues that will be addressed

This project will directly address safety and security issues in Canadian coastal seas, will contribute to better evaluate environmental impacts associated with oil spill and harmful events, and will better inform policies and intervention protocols.

User involvement and external partnerships

The partnership with Fisheries and Oceans Canada and the Maurice-Lamontagne Institute (MLI), the provider of real-time surface drift forecasts in the event of an emergency in the EGSL, is instrumental. It will provide the research team an in-depth comprehension of the operational environment that is currently in place and the occasion to update and improve it if found necessary. This partnership naturally links the project to the main user of drift forecasts, namely the Canadian Coast Guard Environmental Response Office and to the provider of operational services, namely the Environment Canada Canadian Meteorological Center. They will both provide in-kind support to help with the deployment of buoys over sea ice and to provide model output, respectively.Project-related results and information will be archived on a wiki site (http://demeter.uqar.ca/meopar/index.php/Accueil), that has been used so far for proposal writing. This site will become publicly accessible if the project is funded and augmented as the project evolves. Project data will be archived in ISMER's Ocean Data Management System (ODMS, to be operational in September 2013) and made publicly available through the St. Lawrence Global Observatory (http://slgo.ca).

Timeline and key deliverables

Five key deliverables are proposed as interconnected but independent outcomes of the work. It is expected that the performance of current operational forecasting system (D1) will be assessed quite early in the project based on already existing datasets. In parallel to this assessment, the set-up of new models (wave and sea ice) the update of the STLE model and the development of the HFR ice drift retrieval algorithm will take place and all be ready for testing in March 2014 (D2 -- D3). Although some buoy trajectory data are readily available (as of May 2013), the complete database will be complete in June 2014 (D4). A final report containing metadata information and detailed model validation will be produced (D5) in order to facilitate the technology transfer to partners.


Date Activities Key deliverables
Oct 2013 Inclusion of all forcings in the STLE model. Set-up of the wave model. 2013 Summer data analysis.
Nov Deployment of the AWAC.
Jan 2014 D1 -- Performance assessment of the current operational system.
Feb Deployment of buoys on ice floes
Mar D2 - New and updated models

2a -- wave model 2b -- updated STLE model 2c -- collisional ice drift model D3 - New HFR processing algorithms

May Deployment of buoys in open water. In situ measurements from a small boat.
Jun D5 -- Core buoy drift database.
Sep End of the one-year project D5 -- Model validation report.

Expected milestones and performance metrics for each year of the project

In this project, buoy trajectories constitute the ground truth against which model and HFR measurements will be compared. A significant milestone will thus be to acquire these trajectory data both in summer open water and winter ice-infested conditions. Model and HFR performances will be assessed by comparing ensembles of virtual trajectories computed from these gridded products. The final distance between the mean virtual and real trajectories as well as the inclusion of the final real position within the computed search area form the set of metrics that will be used to assess performance and decide whether modification to model or processing algorithm has improved the forecast or not.

Longer-term perspectives

At the end of this one-year project, we expect that the research team will be in a position to propose a longer-term research project involving a slightly larger community of partners, both national and international. Many research topics have already been identified. They include:

  • Applying data assimilation methods (e.g. the Ensemble Kalman Filter) of HFR data [e.g. 7] or hyper-ensemble statistics based on a combination of atmosphere, ocean and wave models, following [24,25], to further improve surface drift forecast skills. The strong collaborative links that the leader of this proposal maintains with specialists in operational modeling and data assimilation at the Nansen Environmental and Remote Sensing Center in Norway are expected to be exploited. This will add up to existing Canadian expertise in this field.
  • Advancing research in wave-current coupled physics and translate new knowledge into innovative coupled models. This aspect requires developing collaborations with key international researchers who are currently working on these issues.
  • Including wave-ice interactions in operational models, a research area to which DD has significantly contributed and continues to do so in other national and international projects. One example is the Beaufort Regional Environmental Assessment (BREA) project Enhancing the Canadian METAREAs coupled atmosphere-ice-ocean analysis and forecasting system for fine scale applications in the Beaufort Sea, led by DFO.
  • The development of a short-term forecasting system based wind forecast and a few hours background of HFR data only, following the approach of [5], [11] or using neural networks. Such forecasting system become interesting when HFR array can be rapidly deployed in the surroundings of an incident.
  • The deployment of HFR arrays in other strategic locations in the Gulf of St. Lawrence, for example where oil and gas fields are expected to be exploited. In this context, oceanographers from other disciplines (chemistry and biology) will be included in order to tackle the problem of the fate of oil spill in cold environments.

Finally, we believe that the strong emphasis that we place on highly qualified personnel training in this project is of fundamental importance to building and maintaining a long-term capacity and leadership.