The power of patterns – embrace the hurricanes.

Recent hurricane Harvey has become one of the most devastating and dangerous of its kind along the history. However, the question arises if our science has the tools to do something to help in such cases. Of course, currently there Is no technology that can physically eliminate the threat of hurricane by neutralizing it or changing its path but one can think of a more mathematical way of predicting its behavior.

Figure 1. Forecast 2005-2015 Katrina

Now we all know, that predicting weather is not an easy task which is very resource consuming and above all is unstable and its accuracy decrees dramatically over time. There are two main factors deciding on safety of inhabitants of a potentially endangered area.
First is the time. Due to meteorologists observations we can track hurricanes quite well, so even thought hurricane can change its characteristics (e.g. wind speed, direction etc.), we know where it currently is localized. The satellites and aerial observations makes a decent job here. Therefore the speed and vector of the storm may change, but we have means to track it down.

Figure 2. Forecast for Hurricane Harvey August 24.

However, the question is: where will it hit next? This is how we came to the second factor which is determining the evacuation area. As it is not possible to predict the exact area that the hurricane will strike the evacuation must be urged over vast areas that potentially can be in danger. This results in millions of people urged by the authorities to evacuate. While evacuation always covers all areas likely to be destroyed by the storm, the predictions might help to narrow the area of hurricane impact. This gives us a chance to relocate rescue units and resources so that the evacuation could be more accurately supervised in areas that will suffer from the actual hurricane hit. Additionally the better forecasting brings the potential to estimate the losses and future demands of supplies and help for the most endangered areas.

Patterns to the rescue🔗

Due to the revolution of AI today we can use a fresh take on hurricane forecasting. Contingency plans must be ready for any occasion, but in a final call they must meet up the dynamic changes in weather, accordingly. Millions of decisions must be made along with forecasting machine learning pipelines constantly refining, based on data conclusions while eliminating noise and zero-in on the important data.

The Hurricane Harvey was target to a machine learning systems that were based on global hurricane forecasts like the Canadian Model, the European Centre, and the U.S.’ National Hurricane Centre model. The system used the pattern recognition with to bias-correct forecasts originating from the historical performance of forecasts along last 30 years.

Figure 3. This figure shows the overall recorded track of Hurricane Harvey as of August 30th.

As the weather forecast becomes more “intelligent” it is extremely important to work on this kind of application further due to enormous possibilities to save not only economy but also potentially many human lives.

Figure 4. This figure from Weather Analytics’ Beacon Hurricane platform shows a forecast of Harvey two days prior to making landfall.

Let us hope that upcoming cat.5 Irma Hurricane will be a well foretasted one.

Side notes🔗

There is also an interesting news 1, 2 and article on how to InsureTech companies use visual intelligence to transform claim response times using unstructured data. The insurers use machine learning to approximate a claim size likely due to the hurricane event. The visual inspection can be additionally processed based on aerial and satellites images. This can be also useful for deciding if the claim would be directly related to hurricane.

Edit Unfortunately this has again came up to a point where the predictions failed when today (as of 10th , Sept. 2017) Hurricane Irma has striked the south of Florida. It was forecast that it will focus on Miami, FL area, instead it rushed towards the west coast of Florida moving towards Tampa.