Strategy for Development
Wednesday, February 13, 2002 (7:00-8:30 pm)
Hawaii Convention Center - Room 314
1801 Kalakaua Avenue
Honolulu, Hawaii 96815
About 75 scientists attended the Town Meeting, which was advertised in broad emails, an ad in EOS, and announcements and signs at the Ocean Sciences meeting. The meeting was hosted by Chris Sherwood of the U.S. Geological Survey (USGS) and Rocky Geyer of the Woods Hole Oceanographic Institution (WHOI) as part of a NOPP-funded project to plan for a larger community modeling project. Partnering principal investigators and institutions on the NOPP project include Hernan Arango and Scott Glenn of Rutgers University, Alan Blumberg and Pravi Shrestha of Hydroqual, Inc., Tom Gross of the National Oceanographic and Atmospheric Administration, John Hamrick of TetraTech, Inc., and Courtney Harris of Virginia Institute of Marine Science.
The objectives of the meeting were as follows:
- Outline goals of the community modeling effort.
- Update community on progress to date.
- Discuss test cases and other modeling issues.
- Introduce and discuss plans for an autumn workshop.
Sherwood and Geyer summarized the long-term contributions we think this project could provide, as follows:
- Good open-source code(s) suitable for addressing important scientific and societally relevant problems.
- An active community for model development and application.
- Web site with code, tools, test cases, discussions, scientific papers, to serve as a resource to the community.
- Model(s) that provide a realistic sedimentary environment for use by other disciplines, such as geochemistry, toxicology, or benthic ecology.
- Forums for scientific exchange, including meetings.
- Guidance for research programs to improve sediment-transport components.
Geyer emphasized that, although many of these goals were to provide an infrastructure for community models, there were also important scientific issues to be tackled. It is vitally important, he argued, to engage both observationalists and theoreticians in this project and to link model development and testing with observations made in laboratories, in the field, and at coastal observatories.
Progress to Date
Sherwood outlined the process that has led to this town meeting. The establishment of full-featured, well-tested, and widely used coastal sediment-transport models would have clear benefits for ongoing and future USGS projects that require coastal modeling. For this reason, the USGS funded a moderate, in-house effort to promote and support a community modeling effort, led by Sherwood, Brad Butman, and Rich Signell. The ongoing project also includes funds for field studies to improve our capability to measure key model parameters, and to apply models in the context of USGS regional projects.
The USGS sponsored a workshop in Woods Hole in June 2000, attended by more than 60 scientists and engineers. A full report of workshop conclusions is described in an EOS article (Sherwood, C. R., R. P. Signell, C. K. Harris, and B. Butman, 2000, Workshop Discusses Community Models for Coastal Sediment. EOS, Transactions of the American Geophysical Union, 81(43):502) and the results are summarized below.
- Attendees endorsed the community modeling concept and suggested that the USGS would be an appropriate institutional host for the project.
- Attendees agreed there was a need for suite of test cases against which new and existing models could be compared.
- Attendees emphasized the need for a better understanding of the physics geochemistry of sediment transport, and argued that model development should occur in parallel with improved observations and theoretical advances.
NOPP awarded us funds for a planning project in FY2002. The objective of the NOPP project is to broaden the community, identify partnerships, establish a structure for evaluation of sediment transport models, and to evaluate new and existing models. This project is a team effort with academic, industry, and government participants and the long-term goal is to promote the development of a node in the "commons for ocean information" that would offer sediment transport models and modeling capabilities. The tasks funded in the planning proposal are as follows:
- Establish a conceptual framework for sediment-transport model development.
- Enhance and maintain community model web site.
- Sponsor a special session and Town Meeting at AGU/ASLO Ocean Sciences Meeting.
- Host planning workshop to write NOPP proposal for fiscal year 2003.
The first task is ongoing and, as part of that task, we have developed a strategy for model development that is discussed in greater detail below. The second task is also ongoing, but web sites have been established at the USGS (http://woodshole.er.usgs.gov/project-pages/sediment-transport/) and Rutgers University (http://marine.rutgers.edu/po/). Modeling information will soon be hosted at a non-institutional site (http://www.ocean-modeling.org/). These web sites currently host information about USGS and NOPP projects, source code for some models, specifications for the first test case, and links to other relevant sites. In the near future, the specification and discussion of test cases will be expanded. Progress on the third task includes the Ocean Sciences 2002 special session, "Application and Assessment of Coastal Sediment Transport Models", convened by Harris and Sherwood, with more than 35 presentations; and this Town Meeting. The final task is to convene a meeting and lay the groundwork for a larger-scale project; details of such a meeting were discussed at the end of the town meeting and summarized below.
Two model codes were recently made available to the general community, a major step toward meeting the objectives of this project. Pravi Shrestha of Hydroqual, Inc., announced in the Ocean Sciences 2002 special session convened by Harris and Sherwood that their ECOM-SED code will be made available as open source code. Details of this announcement and model code will soon be available on the Hydroqual web site (http://www.hydroqual.com). Also, a research and development version of TOMS/ROMS, with preliminary sediment-transport capabilities developed by John Warner of the USGS is available from the Rutgers site (http://marine.rutgers.edu/po/models/roms/index.php).
Two other important models may soon be generally available to the research community. These include Delft3D, which is a full-featured model developed by WL|Delft Hydraulics Laboratory in The Netherlands, and EFDC, also a full-featured model developed principally by Dr. John Hamrick of TetraTech, Inc. with support from the US Environmental Protection Agency. Negotiations for a licensing agreement between WL|Delft and the Office of Naval Research are hopefully nearing completion; the agreement would likely provide access to the Delft3D code to researchers working on ONR and NSF projects. (Contact Dr. Linwood Vincent (firstname.lastname@example.org) at ONR for more information on this). According to Hamrick, the EPA hopes to release a public version of EFDC next year, and in the interim, beta versions of the code may be available to researchers (Contact John Hamrick (email@example.com) for more information).
Strategy for Community Model Development
Sherwood outlined the strategy we have adopted for advancing coastal community sediment transport models. Our approach is summarized as this:
- Start with existing model(s) (rather than from scratch).
- Make them available and encourage their use by providing an infrastructure for exchanging ideas, code, test cases, and modeling tools.
- Compare model results with test cases and evaluate against good lab and field data; and apply models to real problems to gain experience and identify weak components.
- Work on weakest model components by optimizing code, improving nesting, parallelization, or assimilation, improving algorithms, integrating processes...whatever is needed most.
- Enhance tools for model input, adjoint models, and model post-processing.
- Improve understanding of basic physics by observation and experiment, and incorporate this understanding in model algorithms.
Following these introductory remarks, Shrestha initiated a discussion of test cases by outlining the specification of Test Case 1 and briefly providing results for that case using ECOM-SED. (Specifications are on the USGS web site). John Warner showed results from the same test case using Delft3D and the modified version of TOMS/ROMS, and outlined two more test cases under consideration, an estuary with a freshwater input and buoyancy-driven circulation simple estuary, and a lake with wind-driven circulation issues.
Presentation of the test cases launched a general discussion of test cases and other modeling issues. Several attendees commented on the need for a range of test cases, including test cases with either analytical solutions or answers gleaned from good lab or field measurements. Dano Roelvink of WL|Delft Hydraulics Laboratory presented a table of test cases used at Delft, and there was general agreement that a similar hierarchy of test problems, ranging from solutions to analytic formulations to real-world cases was desirable.
Other comments/suggestions during the discussion touched on the following.
- The goals and scope of the modeling project should be specified and published. This means that the intended domains and time scales should agreed upon, and a document (white paper) outlining the objectives should be drafted and ratified.
- One component should be documentation and a forum for questions about the models.
- Model code should come with pre-defined input, boundary, and forcing conditions for demonstration runs in generic environments. These could serve as starting points that users could emulate and customize for their own applications.
- Some models do not contain the same physics; this needs to be emphasized when comparing model results (i.e. bedform development, etc).
- We need to be specific about model algorithms and make sure models are doing what we think they are. For example, there are many implementation of the Mellor-Yamada level 2.5 turbulence closure with slightly different physics and/or implementations.
- Reduced-dimension models are very useful tools for exploring parameterizations or working on a subset of key processes, and should be included in this project wherever appropriate. (A good example is the General Ocean Turbulence Model).
- Models (and model input/outputs) should be standardized for ease in use and to facilitate communication among models or model components. A formal standard list of parameters should be generated.
- We need to maintain a stable public version of the code and migrate research (back room) versions to the public on a continuing basis.
- Intellectual property issues associated with models developed in academia and potential difficulties of placing copyrighted code in the public domain should be addressed.
- Liability associated with general distribution and application of model code should be considered.
- The most important problems involve cohesive sediments, and the most successful approach to predicting behavior of cohesive sediments is local calibration of empirical constants such as threshold for erosion. We should consider establishing standard methods (e.g., approach of Lick and co-workers) for determining these parameters.
- Our community should develop a list of observations required to initialize, force, or test sediment-transport models. The list should be made available to scientists involved in the design of coastal observatories and other long-term or large-scale measurement program.
- Modelers should be involved in the design of observational programs.
- Researchers in the UK have been developing models to predict morphologic change in estuaries as sea level rises; we should stay abreast of their progress and may want to adopt strategies from them.
- Start an email list for fall meeting early so people can prepare for the meeting.
The final task to be completed in the NOPP planning project plan further work that will result in a community model that meets the goals outlined above. However, we also think it is important that the current project produce useful scientific results so, in the final segment of the Town Meeting, we proposed a draft agenda for an Autumn Workshop. The workshop, tentatively scheduled for Washington, D.C. or Williamsburg in August or September, 2002, could include the following:
- Day 1: Exchange scientific results achieved using community models
- Day 2 AM: Develop list of high-priority critical modeling and model infrastructure activities.
- Day 2 PM: Develop list of high-priority observational or experimental science to advance models.
Day 1 would be a scientific forum with presentations concerning evaluation of test cases, comparison of model results with observations, development and testing of new algorithms, or application of models to address problems of scientific and/or societal relevance. We proposed the following ground rules: 1) Any model results must be based on code that is available to the general research community; 2) As space and time will be limited, attendees will be selected on the basis of abstracts submitted in advance. We hope that each talk will form the basis of a refereed publication, and we will try to interest journal publishers in a special volume.
Day 2 would provide the foundation for a proposal to NOPP or other appropriate funding agency.
There was a general consensus among Town Meeting attendees that this workshop was an appropriate next step. One issue that arose [among the hosts after the meeting] is whether a portion of the fall meeting should be devoted to broaden the constituency of the community model by defining its potential for scientific and management impact. This could be done by devoting a part of the meeting to review, revision, and ratification of a white paper outlining the project objectives and plans, and highlighting the advantages to other disciplines (e.g., benthic ecology, geochemistry). Another issue raised was about timing: the fall meeting should not overlap with the Physics of Estuaries and Coastal Seas conference scheduled for Hamburg, Sept. 17-20. [Also, many attendees will be at a meeting at Friday Harbor in September, and mobilization for EuroSTRATAFORM must be considered.] An overwhelming majority of the attendees present at the close of the Town Meeting [maybe 35 people remaining] indicated they would be interested in attending such a meeting.
[These are editorial remarks from Sherwood] Overall, the Town Meeting was apparently successful in meeting its main objectives: to heighten awareness and broaden involvement in the community model project among scientists and engineers; to solicit input and opinions on topics including test cases and to publicly announce the next step towards formulating a larger project with agency funding. The attendance of 6-8 funding agents from ONR and NSF indicate potential interest from these agencies and NOPP, but the Town Meeting was not an ideal forum for presenting the project. Now that the general goals of a community sediment-transport model have been articulated and we have reached out to a wide group of interested scientists and engineers, the next step is to more outline concrete plans for future work. Several people have suggested that we a scientific advisory committee and develop a document (white paper) that outlines the broader need for a coastal sediment model (i.e., justifies the project in terms of end users and other disciplines), the advantages of our proposed direction, and our specific objectives and plans. We welcome any input or suggestions on these topics.
C. R. Sherwood, (firstname.lastname@example.org) USGS (508) 457 2269
C. K. Harris, (email@example.com.EDU) VIMS (804) 684 7194
W. R. Geyer, (firstname.lastname@example.org) WHOI (508) 289 2868