US IOOS Modeling TestbedLeadership Teleconference

May 3, 2011

Inundation, Storm Surge and Waves

Rick Luettich, Univ North Carolina at Chapel Hillrick_luettich@unc.edu

252-726-6841 x137

People Actually Doing Work

• BIO – Will Perrie, Bash Toualney, Fumin Xu• LSU – Chunyan Li• NOAA – Jesse Feyen, Jamie Rhome, Cristina Forbes, Amy Haase, Anne

Kramer• RENCI – Howard Lander• SURA – Gary Crane, Linda Akali• USF – Lianyuan Zheng, Bob Weisberg • UF – Don Slinn, Justin Davis (CI)• UMassD – Changsheng Chen, Qichun Xu• USACE – Jeff Hanson• UND – Joannes Westerink, Aaron Donahue, Corbitt Kerr• VIMS – Harry Wang, David Forrest (CI)• WHOI – Bob Beardsley

• Others……………………

Outcomes and Scientific Insights GainedOutcomes:• Customization of the Interactive Model Evaluation and Diagnostics System

(IMEDS) for use by the surge, waves & inundation team – partnership with US Army Corps of Engineers, new features added by Cyberinfrastructure team

• Developed the ability to compare SLOSH results with other model results using the same meteorological forcing

• Established a SURA testbed server as key repository for archiving and exchanging data.

• Population of SURA testbed server with model grids, model forcing and extensive observational datasets for model evaluation.

• Acquisition of HPC time on LONI (LSU) and Teragrid (UT) HPC resources by SURA IT staff has been critical for project success.

• Port of FVCOM and SELFE to the LONI and Ranger (UT) HPC systems. Of particular note, the UT IT staff helped debug FVCOM so it can now run model with large grid size and large number of processors.

• Excellent collaboration and working relation on model development, testing bug-reporting, etc. for the WAVEWATCH III model

Outcomes and Scientific Insights GainedOutcomes (continued):• Development and initial validation of fully coupled wave-current interaction

plus inundation capabilities for FVCOM, ADCIRC, SELFE and SLOSH models for extratropical and tropical storms.

• A collaboration with the NOAA National Weather Service Weather Forecast Office (WRO) in Taunton has been established. Several meetings have been held and a plan was made in April on methods to transfer forecast products to the Taunton WFO.

Outcomes and Scientific Insights Gained

Insights: • The group has gained better perspective of the importance of resolving inlets,

bays, rivers, channels, barrier islands, wetlands, etc, to accurately model tides and storm surge, particularly in flat coastal areas.

• Model results at select points are highly dependent on bathymetry data. Bathymetry / resolution can determine whether an observational point such as a NOAA tide gauge is in a “wet” or “dry” grid element.

• Significant differences in storm surge results can occur using the same wind forcing, between 2D and 3D simulations, particularly in coastal regions near the storm track. We are continuing to evaluate causes of this.

• It is important to have dynamically coupled wave-surge models:– Waves appear to contribute ~ 1m to the surge during Hurricane Ike on the LATX

shelf in the Gulf of Mexico– In Scituate experiments, wave-current interaction significantly increases water

transport into the harbor and intensifies an anti-cyclonic eddy in the entrance, which in turn pushes water against the coast and thus causes flooding. Without waves, currents are generally oriented parallel to the local isobaths. With waves, currents turn coastward, which can produce a significant onshore water flux.

Outcomes and Scientific Insights Gained

Insights (continued):• The Nor’easter storm-induced inundation occurs in a short time period; the

maximum flooding is at or near high tide.• Model response is sensitive to the wind averaging time used in the wave

and surge models. A “wind gust factor” is sometimes used to convert between different wind averaging intervals.

• For tides, ADCIRC, FVCOM and SELFE give similar results in the Gulf of Mexico. Differences are confounded by different model ramping, run lengths and the specific software used to perform harmonic analysis.

• Model comparisons with the extensive observational dataset for Hurricane Ike indicate the importance of a proper frictional balance on the shelf and the need to fully model the LATX shelf (to capture the large scale geostrophic setup “forrunner” as well as the near track surge)

• Using the same wind forcing, ADCIRC, FVCOM and SELFE in 2D appear to have a similar surge response for Ike. SLOSH seemed to miss the forrunner and has a significantly lower surge than the other models.

Summarize Testbed Products Completed by June

• Interactive Model Evaluation and Diagnostics System (IMEDS)– GUI driven toolbox for inundation model evaluation– Robust suite of error metrics and graphical displays– Skill scores synthesize model performance across space and time• Cyberinfrastructure (Justin Davis) has provided NOAA MDL with code that:– (vers. 1) created point time series in NetCDF from SLOSH native Rexfiles– (vers. 2) will convert fields of SLOSH Rexfiles to NetCDF• Fully coupled wave-current interaction plus inundation modeling systems for

FVCOM+SWAVE, ADCIRC+unSWAN, SELFE+WWM and SLOSH+SWAN models for extratropical and tropical storms. (SLOSH does not include tides.) Initial validation of each system in both extratropical and tropical domains.

• Final unstructured model grids for LATX shelf / Gulf of Mexico (tropical) and Scituate Harbor / Gulf of Maine (extratropical) domains – see figures.

Summarize Testbed Products Completed by June

Summarize Testbed Products Completed by June

Summarize Testbed Products Completed by June• Model forcing– Tropical: tidal open boundary conditions, gridded model wind fields for hurricanes Ike,

Rita– Extratropical: tidal + large scale open boundary conditions, gridded model wind fields

for 2 extratropical storms (2005, 2007)• Observational data– Tropical: 10 tidal constituents at 59 stations in Gulf of Mexico, for Ike & Rita - water

level hydrographs at over 100 stations, high water mark data for numerous stations, wave data from NOAA buoys in Gulf & USACE in shallow areas.

– Extratropical: tidal timeseries for August 2010 in Scituate Harbor, 1 high water mark, wind and wave data from NOAA buoys in the Gulf of Maine

• Model results– Tropical: final ADCIRC, SELFE, FVCOM - 2D & 3D tides for different bottom friction

parameterizations. Initial ADCIRC, SELFE, FVCOM, SLOSH surge/inundation runs for hurricane Ike with and without waves

– Extratropical: final ADCIRC, SELFE and FVCOM tides. WWIII & SWAN model wave results for both storms. Initial SELFE, SLOSH surge/inundation runs for both storms. The 3D, FVCOM Scituate system was run for the 2007 storm for 3 cases: 1) with waves only, 2) with currents only, 3) with wave-current interaction.

Summarize Testbed Products Completed by June• Model results (continued):– We also assessed the contributions of 1) the air pressure gradient and 2) wind

gustiness by re-running the model with inclusion of these forcings. For the 2005 experiment, we have completed all experiments for the case without considering the effect of wind gusts. All model products were saved as NetCDF format and uploaded into the SURA server.

• The system was also applied in forecast mode for the 2010 Dec. 27 Nor’easter storm, and showed success in predicting the coastal inundation that occurred within Scituate Harbor.

Anticipated Progress During NCE (Jun-Dec)

• Model results– Tropical: Final ADCIRC, SELFE, FVCOM, SLOSH tide/surge/inundation runs for

hurricanes Ike & Rita with and without waves (w/o tides for SLOSH)– Extratropical: Final ADCIRC, SELFE, FVCOM, SLOSH tide/surge/inundation runs

for 2 storms (w/o tides for SLOSH). Final WWIII & SWAN model wave results for 2 storms

• Various model sensitivity studies, e.g., quantify the improvement to wave simulations using 29 frequencies and 36 directions (Δθ = 10o) rather that 25 frequency bins and 24 directions (Δθ = 15o); reduce Δt so as not to exceed 10 CFL so that we can use high-order propagation scheme in SWAN; explore different shallow water physics options in SWAN

• Additional observational studies, e.g., output 2-d wave spectra at locations of buoys where observed 2-d wave spectra are available, for model-data comparisons, in 1-d and 2-d.

• Comprehensive inter-comparison and validation of inundation modeling results using IMEDS and other proposed tools, techniques. UMassD-WHOI to lead extratropical model comparisons; UND-USF to lead tropical model comparisons

Challenges to Progress and Lessons Learned• Each modeling team entered the project with their own way of doing business

(including model, assumptions, compute resources, I/O formats, utility programs, error metrics, validation approaches). This has caused many “gotchas” and “do overs” along the way. This is particularly problematic for comparing models. Standardization is clearly important! However, it has been difficult to prioritize standardization over generating model results.

• Existing tropical grid has rather poor resolution in many inland areas, causing poor hydraulic conductivity in these areas and precluding accurate inundation results.

• Including wave-current interaction significantly increases the computational power needed for an inundation simulation. When using the same number of nodes, the time required to run the coupled current-wave system can vary from 2x to 10x more than the case without waves (appears to be model dependent).

• Significant computing and logistical challenges are associated with computing storm surge/waves/inundation on high resolution grids. This is a particularly true for real time model runs.

Challenges to Progress and Lessons Learned• Challenges remain to get all models to run on all compute platforms.• Specific science questions, e.g., in the Gulf of Maine, modeled winds compare

well to buoy observations, yet waves are notably underestimated, tidal harmonic analysis is not as robust as anticipated.

• The wind gust effect. We found that in the 2010 Nor’easter event, the using the gust wind instead of the mean wind to compute the surface wind stress and energy input increases the wind energy input by 24%. Since inundation usually occurs at the maximum wind and high tide, correctly incorporating the gust winds is a critical issue in making accurate inundation hindcasts/forecasts.

• The rapid project time line and milestone / products focus has at times been challenging for participants.

How Can TAEG Help You?How can SURA Mgmt Help You?

• Thus far SURA has been extremely helpful organizing the testbed, providing reasonable structure for the testbed and providing resources (particularly computing) to execute the storm surge / inundation modeling.

• It would be most helpful if SURA is able to secure a No Cost Extension until 12/31/2011 and then provide additional funding to continue the test bed past the end of the NCE. I believe we have developed a good vision and chemistry within the group that should continue to provide productive model advancement for at least 1-2 additional years.

• Data transfer between compute platforms and the SURA data archive remains a problem. Continued help to resolve this problem would be very useful. We also expect the need for additional computational resources if the project is extended past the end of the proposed NCE.

• SURA might pursue dialogue and funding from an expanded set of federal partners. Ideally we would diversify our funding from solely IOOS.

• The surge/inundation has benefited from ongoing feedback and encouragement from TAEG member Bruce Ebersole who has participated in most of our conference calls and our PI meeting.

How Can TAEG Help You?How can SURA Mgmt Help You?

• As we start to develop more extensive quantitative comparisons between model results, I believe there will be an increased opportunity for critique and suggestions from the TAEG.

• The TAEG could strongly encourage the need for standardization, both in terms of model I/O, skill metrics and tools available to work with standardized model products. We note the need for more easily used and advanced tools for model assessment. There is some concern that the use of Matlab to compute statistics (e.g. means, stds, vertical averages) on very large data sets may be very slow. To improve efficiency, analysis tools might be written in Fortran for implementation on parallel machines.

Presentations at Scientific Meetings

• Past:– Multiple presentations - Storm Surge/Inundation PI meeting, March 7-8,

2011– R. Beardsley and C. Chen. Progress of Development of the Scituate

Inundation Forecast System. Meeting with NWS, Taunton, March 18, 2011.

• Planned:– Gordon Conference on Coastal Modeling; June 26-July 1, 2011– 12th International Workshop on Wave Hindcasting and Forecasting and 3rd

Coastal Hazards Symposium; October 30-Nov 4th, 2011– 12th International Estuarine and Coastal Modeling Conference -ECM12,

November 5-7, 2011.– American Meteorological Society; January 22-26, 2012 - maybe– 2012 Ocean Sciences Meeting February 20-24, Salt Lake City, UT - maybe