Presentation on theme: "1 Proposed Hazard Module Expert Panel Open Meeting Austin, Texas March 13, 2014."— Presentation transcript:
1 Proposed Hazard Module Expert Panel Open Meeting Austin, Texas March 13, 2014
2 Agenda Introductions Overview Proposed Hazard Module Wind Wave and Surge Flooding Timing and Delivery of Hazard Module Future Work Q&A
3 Introductions Sam Amoroso, Ph.D. P.E., S.E.Forte & Tablada, Inc. Bob Bailey, Ph.D., P.E.Exponent, Inc. Bill Coulbourne, P.E.Coulbourne Consulting Andrew Kennedy, Ph.D. University of Notre Dame Doug Smith, Ph.D., P.E.Texas Tech University
4 Overview 1 st Open Meeting Austin, August 22, 2013 Develop Framework Plan 2 nd Open Meeting Corpus Christi, December 10, 2013 3 rd Open Meeting Austin, March 13, 2014
5 1 st Open Meeting Panel Member Backgrounds The TWIA expert panel has been appointed under Insurance Code §2210.578 and 28 Texas Administrative Code §§5.4260- 5.4268. The panels purpose is to develop ways of determining whether a loss to TWIA-insured property was caused by wind, waves, tidal surges, or rising waters not caused by waves or surges. After the panel completes its work, the commissioner will consider the panels findings and publish guidelines that TWIA must use to settle claims.
6 2 nd Open Meeting Present Preliminary Overall Methodology Initial Focus: Residential Slab Only Claims
8 Proposed Hazard Module Wind Dr. Doug Smith Objectives Methodologies Model Classes and Investigations Recommendations Wave and Surge Flooding Dr. Andrew Kennedy Wave and Surge Definitions Model Requirements Technical Requirements Coupled Wave and Surge Models Wave and Surge Measurements
10 Objectives for Hazard Module Goals Site Specific: wind speed time history wind direction time history surge time history wave time history Minimum of error Used to predict damage to individual structure as the storm passes
12 Hazards Module Flow Chart Steps Collect Hurricane Wind Field Data over Life of Storm. Develop Global Hurricane Wind Field. Use the hurricane wind field as input to surge model and wave model. Obtain time correlated histories. Refine the time histories. Final time histories for vulnerability module.
13 Wind Hazard Component Two purposes Input for surge and wave module Input for wind damage prediction module Wind speed and direction at mean roof height (MRH) Adjusted for upwind terrain MRH
14 Methodologies Physical Measurements Wind Surge Models Wind field model Surge model Wave model Combination of measurement and modeling
16 Model Classes Parametric Easy to use numerical model. Requires minimal computational power. Observational Wind field constructed using objective analysis from surface wind speed measurements. Requires minimal to moderate computational resources. Dynamical Numerical Weather Prediction Wind field constructed from sophisticated equations sets. Requires significant computational resources.
17 Model Classes ClassExamplesTime Resolution Horizontal Resolution Dir. & Speed Accuracy ParametricHolland Modified Holland Willoughby User selected User Selected ObservationalHRD H*Wind 1 LNSS WFA 2 OWI IOKA 3 3 hrs 10 minutes 30 minutes 4-5 km 500 m 5 km±20° ±2m/s Numerical Weather Prediction HWRF 4 GFDL 5 6 hrs 9 km 5 km 1 Hurricane Research Division Hurricane Wind 2 Letchford Norville Schroeder Smith and Associates Wind Field Analysis 3 Oceanweather, Inc Interactive Objective Analysis System 4 Hurricane Weather Research and Forecasting 5 Geophysical Fluid Dynamics Laboratory
18 Model Investigation Models Holland (symmetric) with translation Modified Holland (asymmetric) with translation H*WIND LNSS WFA Evaluation Criteria Model output wind speed compared to surface observations Desire minimum error Hurricane Ike
33 Recommendations Use an observational wind field for the hurricane wind field model. High spatial and temporal resolution Produce wind speed and direction time histories with minimum of error Increase the number of surface observations contributing to the observational wind field (e.g. using pre-positioned or mobile platforms) to improve accuracy. Embed the observational wind field in a larger synoptic model for surge and wave modeling. Use local observational wind field analysis to refine wind time histories if necessary.
35 Wave and Surge Hazard Texas has experienced numerous catastrophic wave and surge events during hurricanes. The two most notable are: 1900 Great Galveston Hurricane Hurricane Ike (2008) Thousands of buildings were destroyed in each of these storms. Particularly for post-storm slab cases, knowledge of wave and surge hazard levels (storm surge, waves heights) is needed to help separate wave/surge and wind damage.
36 Models and Measurements Both direct measurements and models will be important in establishing wave heights and surge elevations. For TWIA, both are most important in built up regions with insured properties. Models and measurements will both be used to determine hazard levels (waves and surge) at properties. Resulting hazard levels will aid in adjusting claims.
37 Wave and Surge Definitions Separate inundation into waves and surge Surge is the slow variation of average water levels over periods of 10 minutes or longer Waves are the faster variation of water levels Like waves at the beach Typically 5-20s periods dominate in hurricanes At a given location wave and surge properties will define the hazard. When combined with the structural properties, these will allow damage estimates.
38 Model Requirements (Waves and Surge) To be most useful to TWIA, models will need to have certain properties relating to accuracy and usability. There may be more than one choice of model that meets requirements. Models will need to be set up well before any storm so they may be run quickly after the storm. Models should give predictions of waves and surge that are as accurate as possible for a hurricane impacting the Texas Coast. The region most important for accuracy is on normally dry ground, at insured properties. Models should be validated against historic storms.
39 Technical Requirements (Wave and Surge Models) Models should not just include the Texas Coast, but will need to encompass the Gulf of Mexico. Grid resolution needs to be as fine as possible in the vicinity of insured properties (20-80m). Wave and surge models should be tightly coupled. Models should be easily modified to account for post- storm changes/erosion to dunes and shoreline. Model should have an accurate wind field model that is consistent with the wind hazard module. Surge model should include tides.
40 Example Topographic Detail – Coarse Grid Source: Bailey, 2010, Probable Maximum Surge and Seiche Flooding, Presentation to the Nuclear Regulatory Commission.
41 Example Topographic Detail – Fine Grid Source: Bailey, 2010, Probable Maximum Surge and Seiche Flooding, Presentation to the Nuclear Regulatory Commission.
42 Example of High Resolution Hurricane Ike Simulation using SWAN+ADCIRC SWAN computes waves, ADCIRC computes surge. May not be the only model choice that could give good results. 18 million element simulation using hindcast winds (high accuracy reconstruction). Resolution down to 20m in some complex areas. Runs on parallel computer (like all high resolution simulations). Tides included. Comparisons against high water marks, NOAA gauges, rapidly installed gauges.
43 Grid Resolution on the North Gulf Coast Maximum Surge Levels Source: Hope et al. (2013). JGR-Oceans, 118, 4424-4460.
44 Model Comparison with Ike High Water Marks Two shades of green indicate model predictions to within ±0.5m (1.6 feet) Very good accuracy around Texas Coast Source: Hope et al. (2013). JGR-Oceans, 118, 4424-4460.
45 Measured SWAN STWAVE Model Comparison with Wave Height All data are offshore of land Overland accuracy lower Scatter Index: 0.2603 (lower is better) Source: Hope et al. (2013). JGR-Oceans, 118, 4424-4460.
46 Loss of Accuracy with Fewer Grid Points 18 million vs 827,000 element simulations R 2 decreases from 0.823 to 0.758 (higher is better) RMS error from 0.284m (0.9ft) to 0.356m (1.2ft) Medium Resolution High Resolution Two shades of green indicate model predictions to within ±0.5m (1.6 ft) Source: Kerr et al. (2014a). JGR-Oceans, 118, 4633-4661.
47 Loss of Accuracy with Very Few Grid Points 827,000 (ADCIRC) vs 185,000 element (SLOSH) R 2 decreases from 0.716 to 0.555 (higher is better) RMS error from 0.486m (1.6ft) to 0.961m (3.2ft) Low Resolution Medium Resolution Different Models Used (ADCIRC vs SLOSH) Source: Kerr et al. (2014b). JGR-Oceans, 118, 5129-5172.
48 Coupled Wave and Surge Models Inclusion of wave radiation stresses into surge model is necessary to achieve maximum accuracy. Gives Wave Setup in nearshore. In high resolution model Ike simulation, R 2 decreases from 0.832 to 0.785 if waves are not included. RMS error increases from 0.284m (0.9ft) to 0.334m (1.1ft).
49 Wave and Surge Measurements Direct measurements of waves and surge near insured properties will help to improve evaluations of hazard levels. Continuously recording pressure measurements with high temporal resolution will provide the best combination of accuracy and cost efficiency. Instruments will likely need to be rapidly installed pre-storm, based on the storm track. High Water Marks can give good data post-storm. It may prove best to couple with external partners to collect data.
50 Example USGS Measurements Bolivar Peninsula, Direct Oceanfront. Installed directly before storm. Processed data provides evaluation of both surge elevation and wave height over time. Source: East et al. (2008). USGS Open-File Rep. USGS 2008-1365.
51 Source: Doran et al. (2009), USGS Open-File Rep 2009-1061. Post-Storm Update of Model Topography Can be large changes to dune and other elevations that can affect model performance. Updated bathymetry essential – likely with airborne lidar surveys. Should be quick – within the first week post-storm. Pre-Ike Post-Ike Difference
52 Summary: Wind and Storm Surge Features Use an observational wind field for the hurricane wind field model. Increase the number of surface observations contributing to the observational wind field to improve accuracy. Embed the observational wind field in a larger synoptic model for surge and wave modeling. Wave and surge models should include the Texas Coast and the Gulf of Mexico. Grid resolution needs to be as fine as possible in the vicinity of insured properties (20-80m). Wave and surge models should be tightly coupled. Models should be easily modified to account for post- storm changes/erosion to dunes and shoreline. Surge model should include tides.
53 Timing and Delivery of Hazard Module Other major components (damage, economic loss) rely on outputs from hazard module. The panel understands that, by law, TWIA must act on all claims within 60 days from when they are filed. The hazard module must therefore produce results in a time period much less than 60 days. This deadline imposes a very aggressive schedule and will dictate choices as work performed in a short time frame differs from what can be done with less time pressure. The module will need to be set up and tested before the start of hurricane season, and be ready to ingest storm data.
55 Next Step – Damage Estimation Module Proposed Module… Based on Probabilities of Component Failure (Wind) or Probability of Structural Collapse (Surge and Wave) Coupling of Component Damages Time Histories of Damage estimated from Hazard Time Histories Present proposed module during next Open Meeting (May 2014).
56 Future Work Calibration and Validation of Model Compare Ike damage with model predictions for same location. Adjust model as necessary. Conduct randomized model validation using claim data from Hurricane Ike. Present findings in a future Open Meeting (June 2014) Finalize recommendations and present to TDI. Continue with development of a method or model to estimate damage to commercial properties starting with slab-only cases.
57 From the Expert Panel Outline of the Overall Methodology dated 18 November 2013: Need contract in place to acquire post-storm LIDAR imagery. Need contract in place to conduct high wind and storm surge modeling post-storm. Need methods in place to collect high quality real- time storm surge and wind field data during storm. TWIA Action Items