[4.1]In the frame of Clean Seas the main objective of the modelling tasks is to provide support to the understanding of what can be directly observed or extracted from earth observation data. By focusing model characteristics on specific parts of the physics, and by comparison with remotely sensed data, it is then possible to reinforce assumptions on the origin of what is observed. In addition to this natural use of the modelling, the exercise has been used in the perspective of adding a temporal dimension to the earth observation measurements. That is to say that for all cases studied a method to follow a pollution pattern has been set up in order to assess the future of the pollution and the impacted area. Suitability and potential use of such surveys are discussed at the end of this section. The modelling exercise has been performed for the three sites for several test period. For each site a significant kind of pollution has been chosen, for which EO offers capabilities of detection and survey. These pollution types, along with the test periods, are summarised below for the three test sites.
North Sea: Recent studies (Hessner et al., 1997) have shown that some specific signatures (fronts of concentration) of the Rhine plume have been recorded on ERS2/SAR images. It was decided to focus on suspended matter temporal surveys that should allow comparison with frontal patterns observable under particular conditions on SST, colour or SAR images. This is a good starting point for the monitoring of heavy metals loaded by the plume.
For the modelling exercise, the suspended matter is assumed to be composed of non-cohesive sediment whose unique contribution is from the Rhine outflow. The shore erosion due to wave attacks will not be considered.
According to a general bibliography on the North Sea dynamic (Otto et al., 1990) the hydrodynamics are mostly driven by tide motion (residual currents can be considered as negligible - especially on the short time scales we are dealing with). The selected area is the zone covering the Rhine plume extent. From a bibliographic survey (mainly from papers by De Kok or Ruddick) it has been finalised as a rectangular zone extending 20 km southward and 100 km northward from the Rhine mouth.
Three time periods have been used: 30th March-7th April 97, 7th-21st August 97 and 10th-18th May 1998. During these periods several interesting hydrodynamic and/or turbidity patterns were observed in SST, colour and SAR images (plume signature depicted by SST structures and dark radar backscatter area).
Baltic Sea: The main objective of the modelling is to help the analysis of the toxic algal blooms that occur during summer. In particular, preliminary responses are provided about the links and correlation with other meteo-oceanographic data to help understanding of the appearance and dynamics of such blooms. The second objective of the modelling is in the end to provide a more complete temporal picture.
A bibliographic survey has been carried out at ACRI in order to identify robust biological modelling that could be used in the frame of the project. Due to the numerous approaches described and the wide variety of parameters to consider, the result of this bibliography survey is that models examined so far are not sufficiently mature to be used as investigative tools in the frame of the Clean Seas project. The extent of the modelling exercise has therefore been focused on the hydrodynamics of the selected area, and in particular the dynamics of the sea surface, to follow observed algal concentrations. The selected area is located between Sweden and Finland around the Åland area - in a region comprised of a rectangle covering 18°-22ûE and 59°30'-60°30'N. The time period is the 15th July 1997, for which there is an interesting combination of remote sensed data to focus on. In particular the model should help to make the correlation between SST data and sea surface patterns observed by SAR.
Gulf of Lion: Numerous oil spills have been detected on SAR images. Modelling of oil spill dynamics that are currently used are based on a global estimates of spill speed with respect to the wind speed and natural dispersion (vertical and horizontal) of the oil. The coastal area has been selected to be centred on Barcelona, extending 100 km south, west and east. The selected date is the 11th January 98 for which a SAR image gave a very clear picture of an oil spill. In this case the role of the model is to provide information, given an oil damping event, on a potential date and location of the impact at the coast.
It is important to mention that, as for all models that are set up for pollution monitoring, solving of the governing equations is performed in two steps. First the hydrodynamic or meteorological modelling is carried out independently of the kind of pollution we want to reproduce, driven by oceano-meteorological conditions. The second step deals with the pollution survey derived from a pre-computed hydrodynamic field. If the "pollutant" is passive this is simply done by following up of the Lagrangian convection (and diffusion) of a given concentration. Obviously, in the case of active pollutants, this survey should take into account the physical (as for sediment) or chemical/biological (as for phytoplankton) mathematical governing equations that drive the behaviour of the pollution. It is naturally assumed that the pollutant does not impact on the hydrodynamics.
Hereafter are summarised the three methodologies that have been specifically set up for each test area in order to reach regional objectives.
Required output : hydrodynamic field, SST field and sediment load
The vertically integrated (2DH) version of the TIDAL® model has been set up and the boundary conditions of sea level elevation have been derived from the available in situ data (10 measurements over the test site - courtesy of RIZK). The Rhine flow has not been taken into account for the hydrodynamic computation.
The validation of the hydrodynamic field has been done by comparison with French Navy (SHOM) average current supplied for navigation purpose. The agreement is correct and has allowed us to fix a value for the bottom friction coefficient used in the model.
The sea surface temperature has been computed using the assumption of a 2DH field, i.e. it is assumed to be uniform over depth H, by solving the classical advection/dispersion law :
[4.1]
in which T is the water temperature, G T is the effective thermal diffusivity coefficient (taken as uniform) and ST is heat source/sink. The objective is to assess to what extent the SST allows us to validate hydrodynamic fields and by comparison with satellite derived SST data to better understand the part of the thermal information that is directly transported by the flow c.f. that which is linked to other sources, e.g. solar radiative flux.
Lastly, to compute sediment flux, we express the sediment flow rate at the sea bottom as (Tanguy, 1991):
[4.2]
rs is the sediment specific mass, DS is the elementary horizontal surface through which the rate is passing, a is a coefficient between 0 and 1 linked to the turbulence level, and Ceq is the equilibrium concentration (w.r.t. convection effect only) given by :
[4.3]
for which dref is the reference depth (say 1 m), D* is the sediment diameter and
[4.4]
In the processing, D50=0.05 mm (as supplied by RIZK) and the evolution of sediment load is computed by solving a similar expression as [4.1] in which C replaces T and the right handed source/sink term is now Qs given in [4.2].
Baltic Sea modelling
Required output: Sea surface concentration dynamics
Hydrodynamic fields are very complex in the Åland area because of strong changes in the bathymetry especially between the Swedish coast and Åland. To get a hydrodynamic field it has been necessary to have access to the results of an external model, namely that of the Swedish Meteorological Hydrological Institute (SMHI). Results have been kindly supplied for the 14th, 15th and 16th July (at 0h and 12h UTM). Attempts have been made to increase the resolution of the model (actually 5’ in longitude and 3’ in latitude) by using a real flow model adapted to the site specifics. Despite heavy efforts, this step has not produced convincing results (adaptation of boundary conditions and of bathymetry to be able to use the SMHI model as a forcing for a sub model would however lead to restrictive assumptions that are in balance with the expected improvements). Therefore results of the SMHI model have been used directly and interpolated where requested. The output is the concentration field dynamics over the test site which is an artificial mean to check whether surface signatures are fully convected by the surface current or if they are influenced by other phenomenon (vertical autonomous motion, evaporation etc.). For that purpose we use a first digitalisation of the image (say grey scale projected on a pre-defined grid) that will provide discrete values of the "physical quantity". This step is followed by a convection of each concentration and a detection of the grid cell in which this quantity "falls". It is thus possible to re-build a series of successive images by convection of the colour scale.
Gulf of Lion modelling
Required output: evolution of an oil spill
The modelling has been developed thanks to the use of a meteorological data interpolation software (CALMET®) that allows a wind velocity field to built up at a higher resolution from some in situ on ground measurements and output profiles from meteorological models. This allows us to easily introduce (i.e. without re-computing the whole aerodynamic fields) thermal effects (e.g. marine breezes).
The resulting wind field has been computed for the 11th January (mainly directed towards the north-west). In addition to classical drift driven by wind, an evaporative component has been added to the present model that allows us to follow the evolution of the chemical composition of the oil and its specific mass. The evaporative flux (mole/s for example) of the ith component of an oil spill is given by (Mackay and Leinonen):
[4.5]
in which :
, with U wind speed in cms-1
Ci : Concentration of the ith component (mole fraction)
: Pure component vapour pressure at the oil temperature (Pa)
R : Universal gas constant
T : Oil temperature (K)
The vapour pressure is given by (Van Kramen and Van Nes):
[4.5]
with Bi: Boiling temperature in °C at 760 mmHg of the ith component.
The following table gives an example of what is obtained by consideration of one day of evaporation.
| Quantity on water surface (m3) | Evaporated quantity (m3) | Oil density average (g cm-3) |
| 402.3 | 26.7 | 0.82 |
The oil considered in the modelling is a classical mixing of component Cj-i (j=5,..,24)
In this section, we describe the kind of tests that have been performed to check both validity and usefulness of the model runs. For each site we present specific results in comparison to EO measurements and perspectives for application as tool to assess coastal impact.
It is rather natural to consider SST as a passive tracer in the open ocean, but this could cause some problems in coastal areas for several reasons. In particular the radiative solar flux can cause temperature differences of up to several degrees between day and night in shallow water. While being useful for comparison and first calibration of hydrodynamics fields, it is recommended that such data be used with caution for assimilation into a model.
The model SST evolution for the second test period (7th-21st August 1997) is represented in Figure 4-5. The whole SST field has been initialised with data extracted from time series of AVHRR temperature (Figure 4-6) and the temporal variations of thermal flux have been considered by applying a boundary conditions at the SW limit (left) that gives the value of the temperature directly from interpolated AVHRR data.
Figure 4-5. SST for 9th August 1997.
The time series of ATSR data (Figure 4-7) may then be considered as an independent dataset to validate the model. Comparisons with signatures observed by ATSR and AVHRR are possible. In particular the spatial patterns of temperature can be compared.
As can be seen by comparison with the model surface temperature fields (Figure 4-5) the model reproduces the general distribution of temperature well. The warm water ‘pool’ situated off the Rhine at the start of the time series (8th August) is dispersed, principally NE along the coast. The region .of warm water is also seen to become more ‘trapped’ against the coast. However, the model fails to reproduce some of the finer scale features seem in the SST images. In particular, there are very distinct ‘lobes’ and filaments, oriented south-east — north-west, in the ATSR SST signals (and the AVHRR fields) which have not been reproduced by the model. These structures are also evident in the SAR image of 11 August (Figure 4-8) and the visible channels of AVHRR (Figure 4-9). It is possible these structures are related to local heating of the water column over shallower areas, as radiative heat processes are not included in the model and would not be recreated in the model results.
Figure4-6. Time series of AVHRR SST for August 1997.
Figure 4-7. Time series of ATSR SST for August 1997
Figure 4-8. SAR image for 11 August 1997
The results below (Figure 4-11) are sediment concentrations for this test period. From these results we can appreciate the extent of the Rhine outflow and the areas where the sediment is discharged and removed. In particular we can appreciate that the sediment motion and disposal will be located near the shore, particularly northward from the Rhine mouth. It is felt that improvements of the model will come by considering waves action, both through the current induced by wave breaking and associated erosion energy. Nevertheless, such sediment load should be compatible with visible signatures from ocean colour data.
Figure 4-9. Time series of ATSR Channel 1 (visible) for August 1997.
Despite the limitations on the surface temperature field derived from the model, the sediment fields (Figure 4-12) appear to be rather well represented by the model. The general distribution shows the sediment travelling NE along the coast creating a broad band of high sediment, almost as far north as Noordwijk. This is reflected in the MOS image of 19 August (Figure 4-10), and the visible channels of AVHRR (Figure 4-9). There is also a distinct sediment maximum close to the mouth of the Rhine, forming a ‘hook’ to the south-west of the Rotterdam Seaway, also seen in the AVHRR image for 19th August. Some of the features absent from the model sediment field, such as the higher sediment concentrations near Ijmuiden and further north along the coast may attributed to missing sediment sources (the Ijmuiden outlet) and missing processes (resuspension of sediment by wave action) in the model. These could be rectified in am adapted model, if required.
Figure 4-10. MOS ‘chlorophyll concentration’ for 19 August 1997.
The model still needs some refinement, most notably in the initialisation fields, and radiative aspects of the temperature distribution. Inclusion of additional sediment sources and effects of wave action would also increase the usefulness of the model.
The sediment load results are also presented in Figure 4-12 for the third test case.
Figure 4-11. Sediment concentrations for 9th August 1997.
The perspective will be to use the model to better understand the origin of the patterns observed in the SST or ocean colour images and to follow their evolution in time. It is now possible to use the model to assess, by integration in time, the amount of sediment that will be extracted/disposed and therefore the amount of heavy metals trapped along the coast.
Gulf of Lion
An oil spill has been extracted from a SAR image of the 11th January 98 and has been projected on the model grid. The EO data is then used directly as input of the model. For demonstration purposes the spill has been located near the coast to show how the spill could reach the coast. The evolution of the spill for the meteorological conditions of the considered day is presented in Figure 4-13, Figure 4-14 and Figure 4-15. The computation takes into account the evaporation of the oil during its travel toward the coast. Problem of validation of such computation is not solved (despite the fact that this kind of computation is classical and should not lead to surprise) since we did not get two successive SAR images showing a spill temporal evolution. The amount of oil at the coast is something difficult to appreciate since it is strongly linked to residual agitation and induced current (that are, from the modelling point of view, very hard to estimate). The information that one could extract from this simulation is the location of impacted area along with expected time for the spill to reach the coast.
Figure 4-12. Sediment concentration for 11th May 1998.
Figure 4-13. Discrete oil spill extracted from a SAR image (only the two spills at the bottom were studied)
The immediate perspective of this modelling is to use it for a survey once a spill is detected and possibly to derive statistics of impact on the coast from offshore statistics.
Moreover, the motion of the spill has been designed to be computed via a Lagrangian description for transport and dispersion. This will allow us to easily take into account a differential dispersion (as could be extracted by spectral analysis of SAR images).
Figure 4-14. Temporal evolution of the spill (theoretical) at t0+1 and t0+12hours.
Figure 4-15. Temporal evolution of the spill (theoretical) at t0+18 and t0+20hours
The water surface state as observed in visible imagery of 15 July 97 has been studied for the Baltic sea (around Åland island). The model was used to propagate spatial information of surface patterns (assumed to be "concentration like") extracted from Landsat TM data in order to obtain something that could be directly comparable to the IRS/WIFS data measured some two hours later. Once again, such propagation is done using a Lagrangian description of tracer motion of an hydrodynamic field. The results are presented in Figure 4-16. The computation has been continued over 48 hours with the same hydrodynamics field to show accumulation/dispersion trends. We present results after 6, 12 and 48 hours. Several interesting features appear, in particular the straight front (oriented NW/SE) whose origin is therefore mainly due to the current field in the area. Potential uses of such tool (automatic extraction and propagation) will be very useful to fill temporal gaps between two images and to check the part of horizontal motion against vertical mixing, along with the influence of wind over the surface (which is not directly used in the computation). Furthermore, this will be a very interesting tool to supply boundary conditions of surface accumulation for coastal small-scale models.
Figure 4-16. Evolution of surface pattern from the 15 July at 10:30 (T0) to T0+48hrs.
