ECOHAB-GOM Modeling Team Meeting
Dec 8, 2000 Attendees: Signell, McGillicuddy, Stock, Hetland.
There was unanimous agreement that coupled physical/biological simulations of Alexandrium in a Gulf-wide domain is a high priority for future work. At present, we have proof-of-concept simulations in such a domain using ECOM-3D which consist of the following.
T,S from the Dartmouth climatology
Alexandrium = 0.
Observed cyst distribution
Germination as a function of temperature, light, endogenous clock
Growth as a function of temperature and salinity
Winds, heat fluxes, river outflows from 1993
M2 tide from the Dartmouth climatology
This simulation shows promising results.
Also recently developed are some hydrodynamic simulations in a similar domain using ROMS. These also look promising, and may have some distinct advantages over ECOM-si, such as a better advection scheme.
From the coupled modeling point of view, we want to base our runs on the most realistic fields that are available. However, it is equally important for us to understand the limitations of those hydrodynamic simulations in order to interpret the coupled results. To the extent possible, we want to guide our research with the coupled model toward the more robust aspects of the hydrodynamic simulations and try to steer clear of the more troublesome aspects.
This brings up a key question: how are we going to evaluate the fidelity of the large-scale hydrodynamic simulations?
One approach would be to force the model with climatology and compare the results with the mean of all observations. This would require a large effort to synthesize the measurements into data products to which the model could be compared.
Another possibility would be use the forcing from a particular year and compare model results with observations from that same time period. Rich suggested that 1994 is a particularly good choice, because the GOM was heavily instrumented during that time. He has already gathered all the data and set up the software to subsample the model output at the space/time locations of the available observations. Given the fact that the infrastructure for this particular comparison is already in place, we will proceed with using 1994 as a test case for both ECOM-si and ROMS. This will be carried out in the near term so that a decision about which model to use for the large-domain coupled simulations can be made by late spring/early summer 2001. Thus, we should be actively pursuing coupled simulations in the Gulf-wide domain by summer 2001. Until then, the rough distribution of tasks is:
1) Evaluation of ECOM-si and ROMS hydrodynamic simulations for 1994. Rich/Rob
2) Finish 1993 RMRP retrospective in domain II Charlie/Dennis
3) Write manuscript on cross-shelf transport mechanism Dennis/Rich
Plans for coupled simulations in the Gulf-wide domain
The experimental design will be very similar to the proof-of-concept runs described above. Simulations will be constructed for the 1993/1994 RMRP data, as well as the 1998/2000 ECOHAB surveys.
Evaluation of the coupled simulations will pose a difficult challenge. Whereas the physics of domain II was dominated by the wind and river plume dynamics, the Gulf-wide domain plays host to more complex hydrodynamic phenomenology. Without data assimilation we have no hope of accurately simulating the detailed flow in any particular year. However, we hope that the overall character of the hydrodynamic simulations is realistic (of course this will be tested using data available for those particular years). Given that the hydrodynamic simulation will not provide a one-to-one match with the currents from these particular years, we must develop error metrics for the coupled simulations that do not depend on the details of the circulation. A number of possibilities were discussed, including
-- feature-oriented comparisons (e.g. can we simulate the overall pattern of the bloom's retreat into the Bay of Fundy in 1998?)
-- water-mass comparisons (e.g. does the simulated center of mass of cells occupy the same temperature and salinity characteristics as the observed distribution?)
Clearly it will be a challenge to develop quantitative criteria for evaluating these simulations.