Performance Benchmarking Tsunami Models for NTHMP’s Inundation Mapping Activities

The coastal states and territories of the United States (US) are vulnerable to devastating tsunamis from near-field or far-field coseismic and underwater/subaerial landslide sources. Following the catastrophic 2004 Indian Ocean tsunami, the National Tsunami Hazard Mitigation Program (NTHMP) accelerated the development of public safety products for the mitigation of these hazards. In response to this initiative, US coastal states and territories speeded up the process of developing/enhancing/adopting tsunami models that can be used for developing inundation maps and evacuation plans. One of NTHMP’s requirements is that all operational and inundation-based numerical (O&I) models used for such purposes be properly validated against established standards to ensure the reliability of tsunami inundation maps as well as to achieve a basic level of consistency between parallel efforts. The validation of several O&I models was considered during a workshop held in 2011 at Texas A&M University (Galveston). This validation was performed based on the existing standard (OAR-PMEL-135), which provides a list of benchmark problems (BPs) covering various tsunami processes that models must meet to be deemed acceptable. Here, we summarize key approaches followed, results, and conclusions of the workshop. Eight distinct tsunami models were validated and cross-compared by using a subset of the BPs listed in the OAR-PMEL-135 standard. Of the several BPs available, only two based on laboratory experiments are detailed here for sake of brevity; since they are considered as sufficiently comprehensive. Average relative errors associated with expected parameters values such as maximum surface amplitude/runup are estimated. The level of agreement with the reference data, reasons for discrepancies between model results, and some of the limitations are discussed. In general, dispersive models were found to perform better than nondispersive models, but differences were relatively small, in part because the BPs mostly featured long waves, such as solitary waves. The largest error found (e.g., the laboratory experiment case of a solitary wave on a simple beach) was 10 % for non-breaking wave conditions and 12 % for breaking conditions; these errors are equal or smaller than the thresholds (10 % and 20 %, respectively) defined by the OAR-PMEL-135 for predicting the surface profile; hence, all models examined here are deemed acceptable for inundation mapping purposes.