We present a hazard assessment tool for wave and storm surge warning areas based on precomputed simulations using synthetic tropical cyclone events.
In September 2013, two simultaneous tropical cyclones made landfall in Mexico within a 24-h window: Ingrid in the Gulf of Mexico and Manuel in the Pacific. These events generated exceptional rainfall (Pedrozo-Acuña et al. 2014) that resulted in 192 deaths and estimated economic losses of $5.7 billion (U.S. dollars) (Impact Forecasting 2014). Only 4 months later, on 16 January 2014, the Mexican president announced the creation of a National Hurricane and Severe Storms Center (CNHyTS) tasked with increasing the prevention of hydrometeorological hazards in Mexico. The simultaneous events of 2013 almost certainly served as a catalyst to the development of CNHyTS; however, Mexico has always faced disasters associated with tropical cyclones.
Mexico is exposed to tropical cyclones from two cyclogenesis regions (North Atlantic and eastern North Pacific), creating different challenges at the federal level toward emergency response. Tropical cyclones in the eastern North Pacific represent approximately 18% of global events (Frank and Young 2007) and typically strike Mexico or have a direct impact when traveling along the Mexican coast, even without landfall. One of the world’s tropical cyclone “hotspots” is located ∼500 km south of Los Cabos (tip of the Baja California Peninsula) and 500 km southwest of Cabo Corrientes (southern end of the bay of Bahia Banderas, where Puerto Vallarta is located). Between 1949 and 2000, 83 named storms occurred within this area, corresponding to approximately 8 times the density in the Atlantic just east of the Florida Peninsula (Rosengaus-Moshinsky et al. 2002).
Mexico experienced the fourth highest number of landfall events between 1970 and 2009, surpassed only by China, the Philippines, and Japan (Gibney 2010). The destructive effects of tropical cyclones threaten the ∼11,000 km of Mexican coastline, with approximately half of the population (∼55 million) exposed to the direct effects of tropical cyclones, representing a sizable fraction of the total population under tropical cyclone risk in Region IV (North America, Central America, and the Caribbean) of the World Meteorological Organization (WMO).
The Servicio Meteorológico Nacional (SMN), which is analogous to the National Weather Service (NWS) in the United States, is responsible for weather and meteorological analysis in Mexico. Working under the WMO framework, SMN has provided tropical cyclone forecasts and guidance to other authorities and to the general public since before the announcement of the CNHyTS. Both coastlines of Mexico also fall under the responsibility of the WMO National Hurricane Center [NHC; also known as the Regional Specialized Meteorological Center (RSMC)-Miami], which provides frequent forecast information regarding track, intensity, and wind field extension (6 hourly when a tropical cyclone has been declared and 3 hourly when such a storm threatens the coastline). Moreover, in agreement with SMN, it issues coastline alerts (watches and warnings) using graphical products.
The NHC products include uncertainty estimates on the forecasted track, but forecasts of the destructive effects of tropical cyclones (i.e., wind and rainfall fields over the continent and wave and storm surges along the coastline) are not among its international responsibilities. Such estimates are critical for hazard management preparedness and response, and each country under the threat of tropical cyclones needs to develop their own operational tools for hazards warning. In the case of Mexico, a preliminary CNHyTS was embedded within the existing SMN institutional structure in May 2015 and included a 24-h team of forecasters tasked with interpreting NHC bulletins and providing more detailed tropical cyclone guidance related to hazard management. The CNHyTS seeks to provide forecasts of the destructive effects of tropical cyclones by developing its own operational tools, starting with rapid wave and storm surge forecasting.
In an operational setup, wave and storm surge forecasts should be disseminated within minutes of receiving NHC forecasts. In the United States, where roughly half of all fatalities are related to storm surges (Rappaport 2014), considerable efforts have been devoted to storm surge forecasting since the establishment of the Storm Surge Unit by the NHC in 1980 (Rappaport et al. 2009). The Meteorological Development Laboratory (MDL) of the National Oceanic and Atmospheric Administration (NOAA) developed the Sea Lake and Overland Surge from Hurricanes (SLOSH) model (Jelesnianski et al. 1992) as an aid to forecast storm surges. The initial conception of SLOSH was to guide forecasters in the development of weather bulletins at the NWS, although more recently it has been used to delineate storm surge levels in coastal areas (Glahn et al. 2009). The SLOSH model provides information for evacuation planning and advisories based on the maximum envelopes of high water (MEOW) and the maximum of MEOWs (MOMs). A MEOW is a map composed of the maximum storm surge level obtained at each grid cell for a set of simulations of a particular storm category, forward speed, trajectory, and initial tide level, where the uncertainty of landfall location is given by the run of the same storm with different parallel tracks. An MOM is composed of the maximum storm surge values obtained from different MEOWs for a particular storm category, so that it represents the worst case scenarios for such a storm category, independent of the storm forward speed, trajectory, and initial tide level. More recently, in 2007 the MDL implemented the P-surge model (Taylor and Glahn 2008) in experimental mode to provide probabilities of a given storm surge level. The P-surge model uses SLOSH to run an ensemble of hypothetical storms based on an NHC advisory. The ensemble is based on permutations of the forecasted storm, including different tracks, speeds, and wind intensity, with historical forecast errors and uncertainty for each parameter incorporated in order to assign a weight for each track. The SLOSH model calculates the maximum storm surge derived from each storm at every grid cell, and the probability error is included as the weight for each storm to create a probabilistic storm surge map.
Acting above the storm surge, waves are another important hazard from tropical cyclones. In the United States, wave guidance is provided by the Environmental Modeling Center (EMC) of the NWS, which has a group of experts dedicated to wave forecasting. The EMC has provided wave guidance based on nine grids covering their area of responsibility, ranging from a global resolution of 0.5° to fine-resolution grids of up to 1/15° covering U.S. coastal areas (Chawla et al. 2013). Recently, the EMC developed the Nearshore Wave Prediction System (van der Westhuysen et al. 2013) for local weather forecast offices, allowing them to perform high-resolution wave forecasting consistent with their wind forecasts, where EMC only provides the wave boundary conditions.
The use of operational wave and storm surge models in the United States is possible in part because of long-term federal funding to the storm surge and wave forecasting programs at NOAA (e.g., the Hurricane Forecast Improvement Project started with a $13 million amendment to NOAA’s budget; Gall et al. 2013). In contrast, developing countries under the NHC area of responsibility have limited funding for the development of operational forecasts. For instance, the SMN has not been involved in operational modeling of ocean hazards and has only had limited funding available for research projects in academic institutions. Furthermore, the use of high-fidelity operational waves and storm surge models requires high-performance computing power and associated support systems (e.g., secure energy, telecom bandwidth, and redundancy), which is not as readily available in developing countries as in developed countries. However, developing countries will still benefit from less sophisticated tools that can be developed despite tight time constraints and budgets. These types of tools could be used by forecasters with knowledge of waves and storm surges, independently of their modeling capabilities.
As one of these tools, one could precompute the waves and storm surges of a large set of realistic track, intensity, size, and translation speed combinations for tropical cyclones and then, under the real-time threat of a tropical cyclone, choose the most similar as a proxy for wave and storm surge forecasting. However, the brief historical records available do not allow for an analog tropical cyclone set because the probability of finding a proper analog would be too low. Instead, this type of tool could be based on synthetic tropical cyclones, with sufficient precomputing of synthetic track/intensity cases to enable users to identify analogous synthetic events to be forecast in real time. The creation of the CNHyTS provided the opportunity for the development of such a tool, although the tight schedule of the project (6 months from initial development to implementation) was only possible because of the experience of the working group.
Here, we present version 1.0 of an operational tool for forecasting tropical cyclone waves and storm surge hazard areas. The forecasting tool satisfies CNHyTS requirements by demanding low computing power and the ability to be applied in real time under the threat of a tropical cyclone over Mexico. The tool 1) allows forecasts to be achieved within minutes of receiving the track/intensity/extension forecast of the NHC and 2) does not stress the limited material and human resources of a forecasting office that must provide forecasts every 3 or 6 hours. The tool was developed under the assumption that users do not need a modeling background but do need a strong understanding of the physical processes driving wave and storm surge generation and propagation. This requirement was important because forecasters at CNHyTS are not necessarily modelers, as the implementation of deterministic/probabilistic numerical models is not a short-term goal of the organization. The operational tool was implemented in test mode in early 2015 and has already been used during a hurricane season.
SYSTEM DATABASE.
The Quick Assessment Tool for Waves and Storm Surges under Tropical Cyclones (QATWaSS-TC) is based on synthetic tropical cyclone wind fields and precalculated wave and hydrodynamic simulations gathered into a catalog of events (i.e., the system database), which was generated following a series of steps outlined in Fig. 1.
Flow diagram used for the generation of the system database.
Citation: Bulletin of the American Meteorological Society 98, 3; 10.1175/BAMS-D-15-00170.1
Synthetic events and associated wind fields.
Using historical events to develop the system database would have resulted in a limited scope, as only just over 100 events have made landfall in Mexico since 1980. As an alternative, we used synthetic events to create a robust database of 3,100 events making landfall over Mexico (1,550 each along the Pacific and Gulf of Mexico/Caribbean Sea coasts). The events represent a variety of storm conditions related to track, forward speed, intensity, and landfall location. The generation of synthetic events was based on Emanuel et al. (2008), with warm-core vortices randomly seeded across the ocean. These vortices may develop or decay according to the ocean temperature climatology, and, if developed, they are stirred by a beta–advection model driven by large-scale wind fields obtained through NCEP–NCAR reanalysis. The seeded vortices are not considered tropical cyclones unless they develop wind speeds of at least 21 m s−1. A detailed description of the generation of synthetic events can be found in Emanuel et al. (2006, 2008). Only the seeded vortices that became tropical cyclones making landfall along the Mexican coastline were included in the system database.
Numerical modeling.
The atmospheric pressure and wind fields for each synthetic event were used to drive a third-generation wave model and hydrodynamic model. Both models are based on unstructured meshes and were constructed for each basin with a coarse resolution offshore (∼10 km) gradually diminishing to a finer resolution along the coast (∼1 km). In the case of the hydrodynamic model, a few coastal locations were given resolutions of up to ∼250 m. The Pacific domain was limited to latitude 12.5°–33.5°N and longitude 92°–120°W. The Gulf of Mexico/Caribbean Sea domain included boundaries at the Florida Strait (25°N, 80.5°W to 23°N, 80.5°W) and at the Caribbean Sea between Central America (15°N, 83.3°W) and Cuba (20.7°N, 78.3°W). Bathymetry data included local surveys of select areas and 1-minute gridded elevations/bathymetry for the world (ETOPO1) data (Amante and Eakins 2009); both domains and their bathymetries are described in Meza-Padilla et al. (2015). Owing to the scarcity of topographic information and to computational time constraints, meshes were bounded by the shoreline, and flooding and drying were not considered in the simulations. It is important to mention that some synthetic events were generated outside the model domain, so that the swell generated in the eastern Caribbean Sea or the western and southern Pacific was not considered in the simulations. While this can be considered a critical flaw in an operational wave forecast system, it was considered acceptable for this rapid forecast tool, where the main goal is to provide early warnings to coastal areas in the vicinity of a possible landfall. However, forecasters should use other sources of information (e.g., global wave models) to account for swell, since the severity of the waves generated by a local tropical cyclone is also dependent on underlying sea conditions (Ochi 2003). In cases where swell is present in the area of interest, this could be accounted for by the forecaster.
The MIKE 21 Spectral Wave (SW) wave model was used to obtain the wave field corresponding to each synthetic tropical cyclone, and the MIKE 21 Hydrodynamic (HD) Flexible Mesh (FM) model was used to obtain the storm surge generated by each event (i.e., surface elevation). The MIKE 21 SW model is a third-generation spectral model based on the wave action equation used to simulate growth, decay, and transformation of wind-generated waves (Sørensen et al. 2004). The MIKE 21 HD FM model solves the momentum, continuity, temperature, salinity, and density equations with turbulent closure scheme equations. It is based on the incompressible Reynolds-averaged Navier–Stokes (RANS) equations, which are subject to Bousinessq and hydrostatic pressure assumptions. The spatial discretization of the equations for both models is based on a centered finite-volume method over unstructured meshes. Further information about these models can be found in DHI (2016a,b). The models can run in coupled mode so that the feedback between waves, currents, and water levels are considered; however, in this implementation, the models were run uncoupled to reduce computational time. Both models were run with a constant water level equal to mean sea level (i.e., no tides were included in the simulations). Since tidal phase during landfall is not considered, forecasters will need to manually account for it in advisories, taking into consideration that the tidal phase may nonlinearly increase water levels created by the storm surge (Rego and Li 2010).
As in many other developing countries, Mexico has very few measuring stations for waves and sea level. For instance, there are 36 tidal gauges along Texas alone (NOAA 2016a), compared with 38 tidal stations for the whole of Mexico (UNAM 2016). Similarly, the only online wave data for Mexico are from the Mexican Institute of Transport [Instituto Mexicano del Transporte (IMT) 2016], with data only available as graphic displays for recent dates. This makes it difficult to calibrate numerical models for Mexican waters or even to assess the accuracy of operational tools with historical cases. Nonetheless, the hydrodynamic model was calibrated based on Hurricane Ike (2008) using tidal gauges near Galveston Bay, while the wave model was calibrated based on different NOAA buoys in the Gulf of Mexico, as presented by Ruiz-Salcines (2013) for historical hurricanes. The final model setup is described in full by Meza-Padilla et al. (2015).
Catalog of events.
The results from each model (winds, waves, and storm surge) were analyzed in order to obtain the maximum values during the lifetime of each synthetic storm, giving a total of 9,300 matrices of maximum envelopes (3,100 for maximum wind intensity, 3,100 for maximum significant wave height, and 3,100 for maximum surface elevation, with each basin containing 1,550 synthetic tropical cyclones). The maximum value matrices together with the synthetic tropical cyclone tracks and intensity information composed the main database of the QATWaSS-TC. As an example of the information in the database, Fig. 2 shows the maximum envelope for the different parameters for event 903 in the Gulf of Mexico/Caribbean Sea. It is important to note that the wind speed maximum envelopes are not to be used for forecasting the track or intensity of the storm but are part of the system as an aid to the forecaster to select the synthetic events to include in the wave and storm surge forecast. The QATWaSS-TC database was then populated with all of the event information (tropical cyclone tracks and parameters) as well as the maximum envelope maps, as shown in Fig. 3.
Examples of maximum envelopes for synthetic event 903 in the Gulf of Mexico/Caribbean Sea, showing (top left) maximum wind speed (m s−1), (top right) maximum water level (m), (bottom left) maximum significant wave height (m), and (bottom right) maximum wave power (kW m−1).
Citation: Bulletin of the American Meteorological Society 98, 3; 10.1175/BAMS-D-15-00170.1
QATWaSS-TC database structure, where lon = longitude, lat = latitude, Vm = maximum sustained wind speed, Rmw = radius of maximum winds, Pc = central pressure, and Pn = neutral pressure.
Citation: Bulletin of the American Meteorological Society 98, 3; 10.1175/BAMS-D-15-00170.1
CHARACTERISTICS OF QATWaSS-TC.
Figure 4 shows the conceptual model of the QATWaSS-TC tool, which comprises the catalog of events incorporated into a database accessed through Google Maps. The tool aims to help forecasters to delineate vulnerable areas along the coast in relation to waves and storm surge hazards by providing maximum envelope maps for wind, waves, and storm surges based on official advisories from the NHC and on the selection of synthetic events by the forecaster.
QATWaSS-TC flow diagram, where Hs = significant wave height, WLs = water level, and Ws = wind speed.
Citation: Bulletin of the American Meteorological Society 98, 3; 10.1175/BAMS-D-15-00170.1
The storm parameters represented by tropical cyclones in the database (e.g., translation speed, trajectory, landfall location, storm size, and intensity) are limited to those of the 3,100 events. However, as the synthetic events used as proxies will most likely differ in one or more characteristics from the event being forecasted, additional uncertainty may affect the accuracy of significant wave height and water levels. For instance, storm size could be more important than storm intensity for generating higher storm surge values (Irish et al. 2008), and slower moving storms may create lesser storm surges (Irish et al. 2008; Rego and Li 2009) but higher flooded volumes (Rego and Li 2009; Appendini et al. 2014). Such uncertainty could be reduced by increasing the number of synthetic events; however, the time constraints for implementation did not allow for more simulations. While the parameters of the forecasted storm are not considered in the automatic selection of synthetic events, the user can manually deselect all tropical cyclones that do not comply with the characteristics of the event to be forecasted. The hazard assessment areas are then only a result of the events selected by the forecaster based on his or her knowledge of tropical cyclones, waves, and storm surge. Other uncertainties involved in the use of this tool reflect the actual forecasts and wind fields used in the wave and storm surge modeling. For example, Cardone and Cox (2009) showed that the real-time estimates of wind speed and storm size produced by warning center advisories may create up to 20% uncertainty in storm surge estimates.
IMPLEMENTATION OF QATWaSS-TC.
To illustrate the use of QATWaSS-TC, we present the case of Hurricane Patricia, which formed in the eastern Pacific on 20 October 2015. Though initially estimated to make landfall as a category 5 hurricane, postanalysis of data estimated landfall at approximately 67 m s−1 (i.e., a category 4 hurricane; Kimberlain et al. 2016). We selected Patricia both because it represents an extraordinary event and because 2015 was the first hurricane season to be monitored by the preliminary CNHyTS group. Based on the best track data (NOAA 2016b), Patricia had a maximum intensification rate of 54 m s−1 in 24 h, passing from a tropical storm to a category 5 hurricane during this period. Patricia presented the strongest winds ever recorded in the NHC responsibility area and the lowest pressure on record in the Western Hemisphere (Kimberlain et al. 2016), second only to Super Typhoon Tip (1979) on a global level. Event analysis by Kimberlain et al. (2016) indicated that the strongest 1-min-averaged sustained winds were ∼95 m s−1, and there was a minimum pressure of 872 mb, occurring 11 h before landfall.
Only 24 h before Patricia made landfall, the uncertainty cone from the NHC covered the coastline from north of San Blas, Nayarit, to Melaque, Jalisco (approximately 400 km of coastline), which put the cities of Puerto Vallarta and Manzanillo, in addition to many rural areas, at risk of high seas and storm surges. Only 12 h before landfall, the main cities were still under hurricane warning and voluntary evacuations were taking place. Fortunately, Patricia made landfall in an area of low population density and wind speeds above category 3 force were limited to a concentrated area around the eye, resulting in localized damage.
We used QATWaSS-TC to determine the wave and storm surge warning areas. The first step was to provide the system with the actual position of the tropical cyclone as well as a forecast location. We selected the location of the event as provided by the NHC in advisory 14, approximately 14 h before landfall (with estimated landfall at 2315 UTC) and corresponding to the time when Patricia achieved maximum intensity winds. The location was used together with a search radius to identify all synthetic events whose tracks passed through both radii. When QATWaSS-TC is initialized, a display shows an empty map and input dialog boxes (Fig. 5) related to the type of event (e.g., tropical storm, minor category 1 and 2 hurricanes, and major category 3–5 hurricanes), the event center position (present location of the tropical cyclone taking place), the forecast position (this could be a landfall location or any other location of interest), and the search radius for both positions. The default search radius for the present position is set to 30 km, which we found to be a reasonable value after several sensitivity tests. For the forecast location, the default is set to a 3-day uncertainty cone radius as determined at the beginning of each hurricane season. Both search radii can be modified by the user during searches, without restarting the system.
Criteria for synthetic events, including (a) fitting search criteria for tropical cyclones, (b) fitting search criteria for hurricanes, and (c) events selected by the user. Figures are screenshots of the QATWaSS-TC graphic interface; therefore, the text is in Spanish. Please see the appendix for a list of translated terms.
Citation: Bulletin of the American Meteorological Society 98, 3; 10.1175/BAMS-D-15-00170.1
After the user inputs search information, a map is displayed showing all synthetic events that meet the search criteria (Figs. 5a,b), from which the user can manually select the events to use for the warning assessment (Fig. 5c), and a flag is introduced to the map at every landfall location showing coordinates and the wind speed during landfall. The text is given in Spanish since the system is to be used by an official Mexican institution (for an English translation please see the appendix). The resolution of the output map interface corresponds to the numerical modeling mesh.
The user can interact several times with the search of events as well as inspect the individual maximum envelope maps for the different parameters (significant wave height, storm surge, and wind speeds) of the events selected and listed. This allows the user to select the events most suitable for the warning assessment. The system database also contains information on the wave power for each synthetic event, which can be used to assess swell at a particular location far from the storm (Innocentini et al. 2014). Based on Patricia advisory 14 and user-selected synthetic events, QATWaSS-TC generates maximum envelope maps that include individual events’ maximum envelope maps (Fig. 6a), maxima of maximum envelope maps from several events (Fig. 6b), and the normalized maxima of maximum envelope maps (Fig. 6c).
Maximum envelope maps of significant wave height (m) for (a) individual event 1605; (b) maxima of maximum envelope for events 1940, 2044, 2029, 1734, and 1605; and (c) the normalized maxima of maximum envelope for the same events. Figures are screenshots of the QATWaSS-TC graphic interface; therefore, text is in Spanish. Please see the appendix for a list of translated terms.
Citation: Bulletin of the American Meteorological Society 98, 3; 10.1175/BAMS-D-15-00170.1
Based on the individual plots (e.g., the maximum envelope map of significant wave height; Fig. 6a), the user can select and deselect events that passed the first selection filter (Figs. 5a,b) and then decide which events to use for the warning area assessment (Fig. 5c). The user criteria are critical at this stage since the accuracy of the warning area forecast is based on the events included in the maxima of maximum envelope maps. After several interactions, and when the user is satisfied with the choice of events, the system can plot the maximum values at each element mesh considering all selected synthetic events (i.e., the maxima of maximum envelope; Fig. 6b).
While QATWaSS-TC is composed of 3,100 events, it is likely that the user will have to use synthetic events with different characteristics (e.g., intensity, storm size, and storm speed) in order to assess hazard areas and uncertainty for a given storm. For instance, only the Mexican coastline of the Caribbean Sea and near the United States–Mexican border is covered by synthetic events from all tropical cyclone categories when making landfall (Fig. 7). For other areas, the forecaster will need to select a combination of storms with different categories to cover the uncertainty cone. In such maps (e.g., Fig. 5b), the maxima of maximum envelopes will be dominated by the most intense storm, and direct interpretation will provide an inaccurate estimate of potentially affected areas. In this case, forecasters will have to rely on their own understanding of storm surge and wave processes and take into consideration the uncertainties imposed by the use of a combination of events. To aid the forecaster, we implemented a normalized plot for the maxima of maximum envelope maps, in which the values for each storm (i.e., waves, surface elevation, and wind speed) are normalized by the maximum value. In this manner, all events in the normalized maxima of maximum have the same scale, with the highest intensity set to 1 (unity) for each individual event. For example, if the user selects a tropical storm, a category 2 event, and a category 5 event, the normalization will set the maximum values of each to 1, so that the user can infer the warning areas. If the user is aware of the normalization process and a category 3 hurricane is approaching land, he or she will know, based on the events used for the mapping, that the potential areas under threat may differ because none of the events in the system was a category 3 hurricane. Furthermore, in reality, a storm’s behavior also depends on a variety of other parameters (e.g., bathymetry, coastal morphology, storm size, and translation speed). The user should be aware of the real conditions of the forecast position, and the event that is being assessed, to provide a sound estimate of the warning areas. It is important to note that all maps derived from QATWaSS-TC are subject to misinterpretation and are intended for use by trained forecasters only. The maps are not suitable or intended for release to the public.
Wind speed category at each track location for the 3,100 synthetic events, where blue corresponds to tropical depressions, green to tropical storms, orange to minor hurricanes, and fuchsia to major hurricanes.
Citation: Bulletin of the American Meteorological Society 98, 3; 10.1175/BAMS-D-15-00170.1
The QATWaSS-TC is conceived as a qualitative tool to aid forecasters and not to provide estimates of wind speed, significant wave height, or surface elevation. However, under an operational environment it is desirable to have a minimum threshold for the hazard parameters to determine the warning areas. Therefore, the background of the forecaster and his or her knowledge of the area are critical. For instance, in the Mexican Pacific, the mean annual significant wave height in deep water is around 1.5 m, while extreme waves (based on 12 h of exceeding the 99th percentile) are above 3.5 m (Reguero et al. 2013; Cox and Swail 2001), which could provide the threshold for the warning areas. In the particular case of Patricia, the maxima of the maximum map of significant wave height (as obtained using QATWaSS-TC) showed values above 4 m between the locations of San Blas and Manzanillo (Fig. 6b), which would trigger a warning of hazardous waves for this coastline. For this assessment, there were synthetic events making landfall as major hurricanes at the uncertainty cone limits, so that the normalized map (Fig. 6c) does not provide an asset to the forecaster. In the case that there was only one major hurricane in the synthetic database, the individual maximum envelope map for that event would provide the baseline for the significant wave height values that can be obtained and then the normalized maxima of the maximum map will provide the extent of the area under risk. We do acknowledge this is a rough approximation but one that can provide accurate estimates to delineate warning areas when forecasters have a background in the physical processes underlying waves and storm surge generation and propagation as well as in the local characteristics of the area.
To test the accuracy of estimates from QATWaSS-TC for Patricia-generated waves, we compared the results to those of the WaveWatch III model of NOAA’s EMC/National Centers for Environmental Prediction (not shown). The wave model provided similar results to QATWaSS-TC, with estimates of significant wave height between 4 and 8 m along the coast from San Blas to Manzanillo. Nevertheless, the system should still be considered a qualitative aid for the estimation of warning areas and should not be used for quantitative estimates.
One of the main advantages of QATWaSS-TC is that it does not rely on high-performance computing, which would allow computation of waves and storm surges using data from NHC advisories. To compare results that could be obtained using real-time forecast models to the results from QATWaSS-TC, we computed wave maximum wave fields from advisory 14 and best track data for Patricia (Fig. 8). The significant wave height values near the coastline from the precomputed synthetic events (QATWaSS-TC) showed similar intensities to the analogous events (uncertainty cone tracks), although the values near San Blas were overestimated by the synthetic events. For the operational forecast, this suggests that warning areas would be similar whether QATWaSS-TC or real-time models based on the advisory were used. Here, we only included two synthetic events, which covered the extremes of the uncertainty cone, so high waves could be expected anywhere in between. Comparing the results from both QATWaSS-TC and the simulations using the advisory information to the results using the best track data, we found that the wave warning area for waves above 4 m was equal to those from both the simulations based on the advisory and QATWaSS-TC, with the exception of the area south of San Blas; however, the values at the coastline were smaller for the best track simulation.
QATWaSS-TC database poststorm assessment of waves generated by Hurricane Patricia (2015), showing (a) the best track (larger dots), synthetic events 1605 and 2029 (smaller dots), the forecast track with the 5-day uncertainty cone during NHC advisory 14, and the location of Manzanillo and San Blas; significant wave height maximum envelope maps for synthetic events (b) 1605 and (c) 2029, and for (d) the north track, (e) the south track, (f) the central track of advisory 14 uncertainty cone, and (g) for the best track data.
Citation: Bulletin of the American Meteorological Society 98, 3; 10.1175/BAMS-D-15-00170.1
Finally, we performed a qualitative assessment of QATWaSS-TC using the poststorm damage survey conducted by CNHyTS and NWS/NOAA. The results of the survey show property damage and flooding up to 3.5 m above mean sea level resulting from the combined effects of waves and a storm surge ∼120 km southeast of the landfall point (Playa Paraíso). This area was part of the extension of the coastline identified as under risk by QATWaSS-TC, lending additional credibility to the system.
CONCLUSIONS.
In this study, we developed a quick wave and storm surge warning tool for tropical cyclones (QATWaSS-TC), which is the first operational tool for the recently announced National Hurricane and Severe Storms Center in Mexico. The tool was developed on a tight budget and within limited time constraints: 6 months from conception to implementation. Based on prerun high-fidelity models, the tool allows forecasters to provide rapid estimates of wave and storm surge warning areas related to tropical cyclones along the Mexican coastline. When tested using Hurricane Patricia (2015) as an example, the tool provided accurate estimates for warning areas.
Despite the advantages presented by QATWaSS-TC, the approach has several limitations. First, the system database contains only 3,100 synthetic events, so events for use as proxies will likely differ in at least one characteristic (e.g., track, translation speed, maximum wind speed, or storm size) from the event being forecasted. Second, finescale bathymetry is only available in some localized areas, and topography has not been included, thus no overland flooding is calculated. Finally, quantitative estimates can only provide aids for the qualitative assessment of warning areas as the limitations discussed above result in high uncertainties related to quantitative estimates in nearshore areas.
To reduce the uncertainty imposed by the limitation of events, the database could be updated with additional synthetic events. For example, high-fidelity models could be run outside of hurricane season to produce more precomputed scenarios, which in the case of Mexico could add at least 3,100 events per year, considering the same meshes are used. To increase the quantitative precision of the system, high-fidelity models should also include more accurate bathymetric data and topography. However, with approximately 11,000 km of coastline, Mexico is unlikely to perform surveys to gather precise bathymetric and topographical data; although this could be feasible for other Latin American countries and the Caribbean islands with considerably shorter coastlines (i.e., Cuba and the Bahamas have about 40% of the coastline of Mexico, and 2/3 of the countries in the area of NHC responsibility have less than 5% of Mexico’s coastline). In the case of smaller countries, the use of 3,100 events could provide a sufficiently large dataset to reduce the uncertainty imposed. Furthermore, these updates would allow greater automatization of the tool, enabling a more quantitative usage, and in particular would reduce discrepancies that may arise owing to different interpretations by different forecasters.
QATWaSS-TC could be easily adopted in countries with limited numerical modeling capabilities and without a complex forecasting system (e.g., many countries in the Caribbean and Central America), although forecasters would be required to have knowledge of waves and storm surge generation and propagation. While high computing resources are needed to precompute scenarios for the system, none are needed during the tropical cyclone season, when forecasts of warning areas can be done in minutes. The system can be developed and implemented under low budgets and tight schedules, both of which are common in many developing countries.
ACKNOWLEDGMENTS
The authors thank the Comisión Nacional del Agua (CONAGUA) for providing support under Project CNA-SGT-GASIR-14/2014. The authors are very grateful to Professor Kerry Emanuel for supplying the synthetic events and for allowing their use in this study, Gonzalo Martin for IT support, and two anonymous reviewers who greatly helped to improve the manuscript. The CNHyTS and NWS/NOAA storm damage survey was conducted by Orlando Bermudez (NWS), Pedro Restrepo (NWS), Humberto Hernandez Peralta (SMN), and Michel Rosengaus (advisor to CONAGUA). The views and opinions expressed in this manuscript do not reflect the opinion of the donor institute.
APPENDIX
GLOSSARY.
| Altura de ola significante | Significant wave height |
| Buscar | Search |
| Categoría | Category |
| Envolvente | Envelope |
| Evento(s) | Event(s) |
| Eventos seleccionados | Selected events |
| Generar envolvente normalizado | Generate normalized envelope (normalized maxima of maximum envelope) |
| Generar envolvente | Generate envelope (maxima of maximum envelope map) |
| Gráficar máximos | Plot maxima (maximum envelope map) |
| Huracán mayor | Major hurricane |
| Huracán menor | Minor hurricane |
| Incluir | Include |
| Máximo | Maximum |
| Medida | Measured (here, parameters to display are wind, waves, and storm surge) |
| Nomalizada | Normalized |
| Oleaje | Waves |
| Posición actual | Present position |
| Posición esperada | Expected position |
| Quitar | Remove |
| Tormenta tropical | Tropical storm |
REFERENCES
Amante, C., and B. Eakins, 2009: ETOPO1 1 arc-minute global relief model: Procedures, data sources and analysis. NOAA Tech. Memo. NESDIS NGDC-24, 19 pp. [Available online atwww.ngdc.noaa.gov/mgg/global/relief/ETOPO1/docs/ETOPO1.pdf.]
Appendini, C. M., A. Pedrozo-Acuña, and A. Valle-Levinson, 2014: Storm surge at a western Gulf of Mexico site: Variations on Tropical Storm Arlene. Int. J. River Basin Manage., 12, 403–410, doi:10.1080/15715124.2014.880709.
Cardone, V. J., and T. Cox, 2009: Tropical cyclone wind field forcing for surge models: Critical issues and sensitivities. Nat. Hazards, 51, 29–47, doi:10.1007/s11069-009-9369-0.
Chawla, A., and Coauthors, 2013: A multigrid wave forecasting model: A new paradigm in operational wave forecasting. Wea. Forecasting, 28, 1057–1078, doi:10.1175/WAF-D-12-00007.1.
Cox, A. T., and V. R. Swail, 2001: A global wave hindcast over the period 1958–1997: Validation and climate assessment. J. Geophys. Res., 106, 2313–2329, doi:10.1029/2001JC000301.
DHI, 2016a: MIKE 21, spectral wave module. DHI Scientific Documentation, 62 pp.
DHI, 2016b: MIKE 21 & MIKE 3 flow model FM: Hydrodynamic and transport module. DHI Scientific Documentation, 54 pp.
Emanuel, K., and R. Rotunno, 2011: Self-stratification of tropical cyclone outflow. Part I: Implications for storm structure. J. Atmos. Sci., 68, 2236–2249, doi:10.1175/JAS-D-10-05024.1.
Emanuel, K., S. Ravela, E. Vivant, and C. Risi, 2006: A statistical deterministic approach to hurricane risk assessment. Bull. Amer. Meteor. Soc., 87, 299–314, doi:10.1175/BAMS-87-3-299.
Emanuel, K., R. Sundararajan, and J. Williams, 2008: Hurricanes and global warming: Results from downscaling IPCC AR4 simulations. Bull. Amer. Meteor. Soc., 89, 347–367, doi:10.1175/BAMS-89-3-347.
Frank, W. M., and G. S. Young, 2007: The interannual variability of tropical cyclones. Mon. Wea. Rev., 135, 3587–3598, doi:10.1175/MWR3435.1.
Gall, R., J. Franklin, F. Marks, E. N. Rappaport, and F. Toepfer, 2013: The Hurricane Forecast Improvement Project. Bull. Amer. Meteor. Soc., 94, 329–343, doi:10.1175/BAMS-D-12-00071.1.
Gibney, E., 2010: Which countries have had the most tropical cyclones hits? NOAA Atlantic and Meteorological Laboratory, Hurricane Research Division, accessed 4 April 2016. [Available online at www.aoml.noaa.gov/hrd/tcfaq/E25.html.]
Glahn, B., A. Taylor, N. Kurkowski, and W. A. Shaffer, 2009: The role of the SLOSH model in National Weather Service storm surge forecasting. Natl. Wea. Dig., 33, 3–14. [Available online at www.nws.noaa.gov/mdl/pubs/Documents/Papers/Role_of_SLOSH_Model_August2009.pdf.]
Holland, G. J., 1980: An analytic model of the wind and pressure profiles in hurricanes. Mon. Wea. Rev., 108, 1212–1218, doi:10.1175/1520-0493(1980)108<1212:AAMOTW>2.0.CO;2.
Impact Forecasting, 2014: Annual global climate and catastrophe report: Impact forecasting—2013. Aon Benfield, 66 pp. [Available online at http://thoughtleadership.aonbenfield.com/Documents/20140113_ab_if_annual_climate_catastrophe_report.pdf.]
IMT, 2016: Red Nacional de Estaciones Oceanográficas y Meteorológicas. Accessed 4 April 2016. [Available online at http://imt.mx/SitioIMT/DIPC/ServiciosTecnologicos/Reneom/reneomDesarrollo.php.]
Innocentini, V., E. Caetano, and J. T. Carvalho, 2014: A procedure for operational use of wave hindcasts to identify landfall of heavy swell. Wea. Forecasting, 29, 349–365, doi:10.1175/WAF-D-13-00077.1.
Irish, J. L., D. T. Resio, and J. J. Ratcliff, 2008: The influence of storm size on hurricane surge. J. Phys. Oceanogr., 38, 2003–2013, doi:10.1175/2008JPO3727.1.
Jelesnianski, C., J. Chen, and W. Shaffer, 1992: SLOSH: Sea, lake, and overland surges from hurricanes. NOAA Tech. Rep. NWS 48, 73 pp.
Kimberlain, T. B., E. S. Blake, and J. P. Cangialosi, 2016: National hurricane center tropical cyclone report: Hurricane Patricia. NOAA/NWS Rep. EP202015, 32 pp. [Available online at www.nhc.noaa.gov/data/tcr/EP202015_Patricia.pdf.]
Meza-Padilla, R., C. M. Appendini, and A. Pedrozo-Acuña, 2015: Hurricane induced waves and storm surge modeling for the Mexican coast. Ocean Dyn., 65, 1199–1211, doi:10.1007/s10236-015-0861-7.
NOAA, 2016a: Tides and currents. Accessed 4 April 2016. [Available online at http://tidesandcurrents.noaa.gov/map/.]
NOAA, 2016b: Patricia 2015 best track data. Accessed 4 April 2016. [Available online at ftp://ftp.nhc.noaa.gov/atcf/archive/2015/bep202015.dat.gz.]
Ochi, M. K., 2003: Hurricane Generated Seas. Elsevier Ocean Engineering Series, Vol. 8, Elsevier, 154 pp.
Pedrozo-Acuña, A., J. A. Breña-Naranjo, and R. Domínguez-Mora, 2014: The hydrological setting of the 2013 floods in Mexico. Weather, 69, 295–302, doi:10.1002/wea.2355.
Rappaport, E. N., 2014: Fatalities in the United States from Atlantic tropical cyclones: New data and interpretation. Bull. Amer. Meteor. Soc., 95, 341–346, doi:10.1175/BAMS-D-12-00074.1.
Rappaport, E. N., and Coauthors, 2009: Advances and challenges at the National Hurricane Center. Wea. Forecasting, 24, 395–419, doi:10.1175/2008WAF2222128.1.
Rego, J. L., and C. Li, 2009: On the importance of the forward speed of hurricanes in storm surge forecasting: A numerical study. Geophys. Res. Lett., 36, L07609, doi:10.1029/2008GL036953.
Rego, J. L., and C. Li, 2010: Nonlinear terms in storm surge predictions: Effect of tide and shelf geometry with case study from Hurricane Rita. J. Geophys. Res., 115, C06020, doi:10.1029/2009JC005285.
Reguero, B. G., F. J. Méndez, and I. J. Losada, 2013: Variability of multivariate wave climate in Latin America and the Caribbean. Global Planet. Change, 100, 70–84, doi:10.1016/j.gloplacha.2012.09.005.
Rosengaus-Moshinsky, M., M. Jiménez-Espinosa, and M. T. Vázquez-Conde, 2002: Atlas climatológico de ciclones tropicales en México. Centro Nacional para la Prevención de Desastres, Instituto Mexicano de Tecnología del Agua, 106 pp. [Available online at http://cambioclimatico.inecc.gob.mx/descargas/atlas_ciclones.pdf.]
Ruiz-Salcines, P., 2013: Campos de viento para hindcast de oleaje: Reanálisis, paramétricos y fusión. M.E. thesis, Department of Ciencias y Técnicas del Agua y del Medio Ambiente, Universidad de Cantabria, 84 pp.
Sørensen, O. R., H. Kofoed-Hansen, M. Rugbjerg, and L. S. Sørensen, 2004: A third-generation spectral wave model using an unstructured finite volume technique. Proc. 29th Conf. on Coastal Engineering, Lisbon, Portugal, ASCE, 894–906.
Taylor, A. A., and B. Glahn, 2008: Probabilistic guidance for hurricane storm surge. 19th Conf. on Probability and Statistics, New Orleans, LA, Amer. Meteor. Soc., 7.4. [Available online at https://ams.confex.com/ams/pdfpapers/132793.pdf.]
UNAM, 2016: Servicio Mareográfico Nacional. Accessed 4 April 2016. [Available online at www.mareografico.unam.mx/portal.]
van der Westhuysen, A. J., and Coauthors, 2013: Development and validation of the nearshore wave prediction system. 11th Symp. on the Coastal Environment, Austin, TX, Amer. Meteor. Soc., 4.5. [Available online at https://ams.confex.com/ams/93Annual/webprogram/Manuscript/Paper222877/AMS2013_Westhuysen-etal_ext_abstr_paper4-5.pdf.]
