5. Discussion.
The results of the simulations conducted in this study provide the most extensive understanding of the consequences of a radioactive leak from Moruroa Atoll to date. This study has made use of some of the best ocean modelling tools currently available including output from a global eddy-permitting ocean model. The duration and depth of tracer release to the water column have shown important differences in the ultimate fate of the radioactive material. The inclusion of natural variability into the simulations has been particularly illuminating, highlighting the difference in both pathways of tracer advection and magnitude of tracer concentration under different climate scenarios (section 5.1). Previous studies of radioactive tracer dispersion from Moruroa Atoll have assumed constant velocity fields, and generally adopted coarse resolution models with limited domains. Our experimental case adopting instantaneous release of tracer at the surface from Moruroa Atoll allows a direct comparison with two previous studies (RT90 and LR99; see section 5.2).
5.1 Influence of seasonal and interannual ocean current variability on tracer advection and concentration
This study has demonstrated the importance of the inclusion of seasonal and interannual variations in ocean model velocity fields for the fate of radionuclide pollutants in the South Pacific Ocean. No previous study considered this factor beyond the regional scale, despite evidence for seasonal and interannual ocean current variability.
When the seasonal cycle of current velocity is incorporated into the instantaneous surface release experiments from Moruroa Atoll, the variable ocean currents greatly reduce the easterly transport. The cycling of the direction of velocity fields a few degrees to the east of the source acts to maintain the maximum surface concentration of radioactive material within French Polynesia. The transport towards South America seen in the annual-mean scenario is still apparent with the seasonal cycle, however it is much reduced. This result is in contrast not only with the annual-mean release from our experiment, but also with those of previous studies by LR99 and Mittelstaedt et al. (1999) who found maximum radionuclide concentrations to be in the eastern South Pacific. The importance of the time of release within this seasonal cycle also influences the ultimate fate of the radioactive material. In a winter release, when south-easterly surface currents are weaker at the release site, a greater amount of the tracer is able to be vertically mixed to depths of 25 -75 meters. This reduces the amount of tracer that is contained within the surface current cycle, allowing a greater proportion to advect to the east. The weakening of the currents in the upper six model depth levels seen with the ENSO velocity fields is also important for the resultant radionuclide advection as easterly transport is much slower while the ENSO fields are active. The subsurface westerly advection of tracer is also influenced by the inclusion of a mean-monthly cycle of velocity fields. The retention of larger quantities of tracer in French Polynesia results in a greater degree of vertical mixing to the levels of westward currents at depths below 260m in the central South Pacific. Radioactive material therefore reaches the Australian continent one year sooner than in the annual-mean scenario.
The results from the scenario of instantaneous release to the depth of the karst layer also show sensitivity to the interannual current variability, however seasonal variability, as would be expected at this depth, has much less effect. Little difference can be seen between the releases under monthly cycle advection and annual-mean advection until the strong and variable currents off the east coast of Australia are encountered. Despite the lack of direct current observations below the surface mixed layer (0-150m) in the central South Pacific, this aspect of the model agrees with the general picture of the wind-driven subtropical anti-cyclonic gyre. The gyre occupies the upper 400m of the ocean and is decreasingly responsive with depth to seasonal variations, such as trade winds and cyclones. Below this depth, the westward flowing Antarctic Intermediate Water is found, which is not influenced by local atmospheric conditions, but driven by longer-term and larger-scale thermohaline forces and geostrophic effects (Tomczak and Godfrey, 1994).
When tracer is released under ENSO velocity fields and compared with the annual-mean simulations, a different scenario is revealed. Tracer advected under ENSO velocity fields is transported much more slowly from the source when compared with the annual-mean release. This can be partially explained by the reduction in the anticlockwise momentum of the subtropical gyre under ENSO conditions, most obviously demonstrated by the dramatic reduction in the velocity of the equatorial currents (Tomczak and Godfrey, 1994). This reduction is due to weaker winds caused by reduced meridional SST gradients during ENSO years.
Whilst the direction and speed at which radionuclides are transported from Moruroa Atoll differs when ocean current variability is included in the instantaneous release experiments, the maximum concentration signal decays at similar rates. This is due largely to the relatively weak ocean currents in the central South Pacific, meaning that background mixing determines to a large extent the size of the maximum tracer signal.
The influence of ocean current variability on the gradual release scenarios from Moruroa has a direct effect on the local concentration of radioactive material. When current velocities in the release region are higher, concentration is reduced. This is a simple relationship between the rate at which tracer is advected and mixed away from the source to the rate of input. At all times within the seasonal cycles, maximum tracer concentrations are greater than those of the annual-mean scenarios. The greatest concentrations experianced are associated with the weakest local currents, these occur in the first two years of the ENSO scenarios.
The simulation of the instantaneous release of radionuclides to the surface model layer from Moruroa Atoll can be compared with two previous studies that have modelled the same scenario (RT90, and LR99). RT90 estimate the quantity of radioactive material contained within Moruroa to be equivalent to 3.2 x 1017 Bq of 137Cs. They then simulate a 'worst case' scenario in which this entire radioactive inventory was instantaneously released to the surface layer, the model was then run for 10 years. This simulation was repeated by LR99 using a higher resolution model with a larger domain. The results of our study, with 1 Bq m-3 release concentration can be scaled so that the initial tracer release is equal to that of the previous studies.
Figure 12 shows a schematic diagram of the direction of propagation of the maximum tracer concentration in the three simulations using the annual-mean velocity fields. It can be seen that there is general agreement between our study and that of LR99, while those of RT90 are in the opposite direction. RT90 found the direction of maximum tracer concentration to be westward, while LR99 and our simulations indicate primarily south-eastward surface transport.
The dramatic disagreement in tracer advection direction between our study and that of RT90 can be attributed to the very coarse resolution of their model. The horizontal resolution in RT90 is 3.75o longitude by 4.5o latitude compared with 2o by 0.5o - 1.7o (LR99) and 0.25o by 0.25o in our model. The RT90 low resolution is unable to capture many of the features of the ocean circulation, with flow at the surface near Moruroa Atoll simply a broad slow flux in the subtropical gyre. Their model uses only 12 depth levels - the surface level of which is 50m thick - compared with the 20 levels in this study (with surface level 25m thick). The coarse RT90 vertical resolution in the upper ocean is unable to realistically represent the vertical sturcture of the gyre circulation in the South Pacific Ocean. An additional factor is the surface forcing fields, which in RT90 are derived from a coarse grid atmospheric model. Our velocity fields, in contrast, come from a model forced by ECMWF model reanalyses, which capture storm tracks and weather-scale variability in the atmosphere. This leads to increased horizontal structure in the surface flow fields.
The results of our study when compared with those of LR99 show that the location of the tracer concentration peak after 10 years is 30o further to the east and 10o to the south. Both models show the tracer peak to be located within the convergence region in the eastern South Pacific, however the higher resolution of the POCM shows some subtle currents persisting with eastward flow that are not apparent in the LR99 circulation model. The subsurface advection of tracer that results in upwelling off the east coast of Australia is seen in both models.
The maximum tracer concentration over time shows a large difference between the three models. Figure 13a. shows the general agreement between the results for maximum tracer concentration in LR99 and RT90, despite the massive difference in tracer transport direction. The results from our study are consistently a factor of 100 greater than those of LR99 and RT90. It should be noted however that initial concentrations are different in the three results due to the different model resolutions, such that larger volume elements result in lower initial concentrations. This difference in initial concentration will clearly influence the resultant maximum concentration of tracer over time. The initial difference in release concentration is a factor of 10 and 100 lower in the LR99 and RT90 studies respectively when compared with our results. The small size of Moruroa Atoll is not specifically resolved by any of the three models, so it may be assumed that in all cases the initial concentration is underestimated as it is instantaneously mixed over the source grid box. The RT90 and LR99 studies therefore impose a much greater degree of dilution to the tracer prior to any incorporation of physical ocean processes. In fact the final maximum concentration in our study is only a factor of 10 less than the initial concentration in RT90.
The initial dilution difference between the three model case does not explain all of the difference seen in maximum concentrations. When maximum concentration (Qmax) is divided by the initial concentration (Q0) plotted against time, our model run still shows significantly higher concentrations thoughout the simulation compared with LR99 and RT90 (Figure 13b). The horizontal resolution of the three models provides a second important explanation for the large difference Qmax/Q0. Neither RT90 nor LR99 were of a sufficient horizontal resolution to resolve many of the smaller-scale features of ocean circulation such as narrow jets and quasi-stationary eddies. Thus, the majority of water mass mixing is parameterised and advection fields are relatively broad and sluggish, particularly in the case of RT90. The finer resolution of the POCM is better able to capture the advective components of ocean dynamics, and thus reduce the need for their parameterisation. It is likely that our horizontal mixing coefficient is a little too low for particularly high energy areas of the ocean such as the EAC. However the low energy environment of the majority of the South Pacific Ocean is reasonably well parameterised by our mixing term. It is likely that horizontal mixing has been overestimated in previous studies, particularly RT90, and the results of our study more accurately simulate tracer dispersion from Moruroa.
The final important difference between the results of the three studies concerns the model domain. Our study has utilised a global ocean model, effectively without artificially imposed boundaries, and therefore does not require the inclusion of side boundary conditions. The RT90 and LR99 models were both restricted to South Pacific Ocean domains. The results of our study show the tracer being advected beyond the boundaries of both these earlier studies within the 10-year integration (Figure 12), a situation that is impossible to account for in the earlier models.
If the RT90 estimate of 3.2x1017 Bq of 137Cs is an accurate guide to the possible amount of contaminated material contained within Moruroa, our study shows that the resultant contamination to the local marine environment could remain as high as 1000 times greater than the current background concentration (10.8 Bqm-3 Cousteau, 1988) 10 years after such an incident .
This study has demonstrated the importance of the incorporation of ocean current variability and high model resolution in tracer release modelling studies. There are however improvements that could be made. An ideal simulation would combine seasonal and interannual variability into a single simulation. This could involve superimposed monthly and interannual cycles, or real-time forecasting. In the case of the ENSO climate cycle this would account for La Niña, El Niño and the transition between these events. Releases could be simulated at times throughout the interannual and seasonal cycles. Such a study would require many integrations and is beyond the scope of our initial goals. There would likely be more variability in the possible fate of radioactive material in the marine environment with a greater variety of ocean circulation scenarios. Our study has demonstrated the importance of seasonal to interannual variations without covering the full range of possible cycles.
Improvements to the off-line tracer model used in this study could also be made. In particular, the representation of subgrid scale mixing would be greatly improved with the incorporation of a physically based parameterisation of horizontal mixing such as isopycnal mixing and a bolus transport term (e.g., Gent and McWilliams, 1990). This would be particularly useful in improving the prediction of tracer fate in eddy-rich regions such as the ACC and EAC, where current velocities are high and variable. Ideally, the model would also include a simple wind-driven mixing scheme for representing enhanced upper ocean vertical mixing in the presence of strong winds.
http://www.maths.unsw.edu.au/~doughaze Page created by Douglas R. Hazell 7/6/2001 Last update of this page: 1/8/2001