Featured Article: Simulating StORMS and Hurricanes

math

Early in October of 2017, Hurricane Nate made landfall near the mouth of the Mississippi river, making it the fourth hurricane to make landfall in the United States. The last time the U.S saw as many hurricanes in a single year happened way back in 2005. While the landfall of Harvey and Irma were already historic events, making it the first time that two Atlantic hurricanes struck the U.S in the same year. Puerto Rico taking a direct hit from Maria meant that the U.S has now endured landfall from three category 3 hurricanes [1]. The devastating effects of these natural disasters further the need to bring together practitioners from different research areas to help tackle the problem. Operations Researchers, in the recent past, have become more involved in this effort. In October 2017, National Science Foundation awarded $5.3 million in 59 grants to study hurricanes, among them a significant number were aimed towards the prediction and management of the storms.

So, where does simulation-modeling come in? The mathematical simulation of hurricanes is the most accepted approach for estimating wind speeds for the design of structures and assessment of hurricane risk [2]. The approach is used for developing coastal flood estimates, setting flood insurance rates and minimum floor elevations for buildings along the hurricane coastline of the United States, and the approach is routinely used in the banking and insurance industries for setting insurance rates. As such, any structure that is to be built in hurricane prone areas is tested with a combination of wind tunnel data and hurricane hazard models. When we pull up weather.com to keep track of any storms in our vicinity, it is important to understand that almost all the data that we see is a result of some sort of extrapolation based on other factors. For instance, a powerful hurricane would make it difficult to set up a weather station at the center tracking wind speeds and surge levels. Scientists use data obtained from a set of other locations and simulate the overall effect and strength of a hurricane.

Larry Russell described one of the first uses of simulation in 1968 in his Ph.D. thesis titled “Probability distribution for Texas gulf coast hurricane effects of engineering interest”. The basic approach used in this work (and many subsequent studies) is similar in with respect to the fact that it is particularly suitable for sites where insufficient data does not permit getting the effects of a hurricane directly [3]. “First, site-specific statistics of key hurricane parameters including central pressure deficit, radius to maximum winds, heading, translation speed, and the coast crossing position or distance of closest approach are obtained. Assuming that the statistical distributions of these key parameters are known, a Monte Carlo simulation is used to generate a mathematical representation of a hurricane along a straight line path satisfying the sample data.” [4]. The essential idea, like any simulation model, is to use existing data and information about what the general trends of the data look like to extrapolate and simulate further.

A common visual that we are exposed to during hurricane season are spaghetti plots. These are the outputs of different simulation runs on employing different parametric values. The model can then be stepped forward in time and the results are compared and used to gauge the amount of uncertainty in the forecast. If there is good agreement and the contours follow a recognizable pattern through the sequence, then the confidence in the forecast can be high. Conversely, if the pattern is chaotic i.e. resembling a plate of spaghetti then confidence will be low. The use of spaghetti models is commonly employed when describing the outputs of hurricane track models.

The most common types of simulation models today are related to disaster management from the point of view of humanitarian logistics. In the event of a disaster, there is hardly any time to try out different evacuation scenarios. This is where simulation models come in. Testing out different policies and scenarios in a mathematical abstraction of reality allows the government to be prepared when reality strikes, so to speak. If one knows with a high probability there will be major flooding in a certain area, the residents may be instructed to leave their homes. This plan is better than trying to rescue them with helicopters or boats once the actual disaster strikes. If long-range forecasts for severe storms are available, then people in the path of the storm can be given reliable advice about how to protect themselves. In fact, a majority of the grants given out by NSF in October are pertaining to this facet of simulation modeling.

NASA recently published an animation depicting this year’s rough hurricane season. They did this by tracking what is carried on the wind. Tiny aerosol particles, such as smoke, dust, and sea salt are transported across the globe, making visible weather patterns and other normally invisible physical processes. The mathematical models use information from NASA satellites, and sea salt evaporated from the ocean to compute and observe more details of physical processes. Observing how these aerosols interact with the storm systems allows scientists to study the formation of storms [5].

The behavior of the atmosphere and weather is governed by physical laws, which can be expressed in the form of mathematical equations. These represent metrics like temperature, wind speed, humidity, etc. The ultimate goal of using simulations is to solve these equations and to have a forecast for the weather. The best forecasts are made by combining the forecasts from two or three different ones into a “consensus” forecast. The Hurricane Forecast Improvement Program’s (HFIP) at the National Hurricane Center has a goal of having a 50 percent improvement within tropical cyclone track forecasting and intensity guidance by 2020. In addition, another goal of researchers working in this field is to be able to reduce the length of the forecast period, thus being able to make predictions in a matter of hours as opposed to days.

References

 [1] https://weather.com/storms/hurricane/news/2017-10-07-four-us-hurricane-landfalls-nate-maria-irma-harvey

 [2] Peter J. Vickery, Forrest J. Masters, Mark D. Powell, Dhiraj Wadhera, 2009 “Hurricane hazard modeling: The past, present, and future,” Journal of Wind Engineering and Industrial Aerodynamics, 97(7–8), 392-405.

 [3] Russell, L.R., 1968, Probability distribution for Texas gulf coast hurricane effects of engineering interest, Ph.D. Thesis, Stanford University.

 [4] Vickery, P. J., P. F. Skerjl, and L. A. Twisdale, 2000 “Simulation of hurricane risk in the U.S. using empirical track model,” J. Struct. Eng., 126, 1222–1237.

 [5] https://www.youtube.com/watch?v=h1eRp0EGOmE