University of Illinois at Chicago
Browse
AHMED-THESIS-2018.pdf (5.94 MB)

Tornado Path Prediction Using Data Driven Techniques

Download (5.94 MB)
thesis
posted on 2018-11-27, 00:00 authored by Madiha Ahmed
Each year around 800-1400 tornadoes occur in the United States. Although a very small number of these tornadoes result in injuries and fatalities, the destruction caused by these events can be catastrophic. Despite having advance warning mechanisms and alert systems, the entire prevention of tornado casualties is still quite challenging (1). Studies have indicated that the rate of tornado-related fatalities and injuries are higher when effective warnings are not issued and people do not have access to suitable storm shelters (2). Tornado warnings are issued based upon the atmospheric disturbances being detected on Doppler radars. This system, however, provides the national average lead time of 13 minutes which is insufficient when families are expected to evacuate the areas under danger (3). The National Oceanic and Atmospheric Administration (NOAA) is currently working on developing a warn-on-forecast system to potentially improve the lead time for warnings before the formation of tornadoes (4). Upon studying numerous tornadoes, however, it became evident that the issues pertaining to tornado casualties could not simply be improved by extending warning times. The lack of communication between individuals and officials creates confusion about appropriate actions that individuals must take. Officials believe that by generating the warnings they have performed their job and the responsibility to take appropriate actions were then on the individuals. Individuals, on the other hand, seek to obtain more information regarding the arrival of a tornado, rather than simply relying on current warning mechanisms such as sirens and warnings broadcasted on television and radio. This communication dilemma results in preventable casualties (5). For the present model, the tornado path was predicted for El-Reno, Oklahoma tornado that had occurred on May 31, 2013. For Model 1, the exact locations were determined by calculating the latitude and longitude points based upon the previously predicted locations. The data was divided into one and two-minute intervals. For Model 2, the same locations were predicted based on the previous original locations that would be provided by the radar in the actual scenario. Results showed that the distance between the predicted and the original locations were reduced significantly when the locations were predicted from the prior original locations in Model 2. Furthermore, the predicted path of the tornado was divided into different circles of diameter equal to the width of tornadoes which are usually around 0.17 miles to 0.28 miles (6). Results indicated that 90% of the locations were predicted successfully when the diameter/width of the tornado was around 0.3 miles. To improve the communication gap between officials and individuals when providing warnings to people in a tornado’s path, the usage of UAVs (unmanned aerial vehicles) is proposed. UAVs equipped with cameras and Global Positioning Systems (GPS) can continuously map the land features of the area stricken by the disaster and provide updates on the evolution of tornado. Moreover, with UAVs, emergency relief-teams can have real-time access to areas under a tornado warning and can closely monitor victims, hence, ensuring the evacuation of houses and businesses in a tornado’s path. This will eliminate the need for the emergency personnel to visit disaster locations for evacuation purposes. UAVs can also replace telecommunication structures which can be impacted by server weather hazards (7). UAVs can be connected to Doppler radars and receive updates regarding previously hit tornado locations. By using current and exact locations, UAVs will be able to use the prediction model to predict the next location of a tornado and can warn residents that are in a tornado’s path. The prediction of tornado path along with warning individuals by UAVs will be helpful for emergency managers in allocating their services as well. When tornadoes or other natural disasters occur, it is the responsibility of emergency relief-teams to provide affected individuals with immediate relief. To provide such services, they are in constant need of up to date information pertaining to the disaster location (7). Due to time limitations, however, a team of UAVs will be required to serve the purpose of providing warnings and leading individuals to tornado shelters. Tuna et al. suggested the using a formation control system to control the position of multiple UAVs relative to each other. The use of such a system will ensure that multiple UAVs will not warn the same location and can potentially prevent collisions among UAVs (8). Lastly, UAVs can be extremely helpful in a post-tornado scenario. A detailed information regarding the damaged infrastructures can be determined from UAV photogrammetry. To assess the damage incurred by a tornado, it is essential to collect necessary data about the damaged locations. However, due to blocked roads by fallen trees or other debris, these areas are not accessible. The images collected by UAVs will be useful in conducting a post-event survey (9). Another issue is to rescue people that are still alive but stuck under the rubbles. By attaching the infrared cameras to UAVs, it will be able to detect any human or animals that are stuck under the damaged infrastructure. Finally, due to the small size and light weight of UAVs, they could be easily damaged by the heavy winds of tornadoes. Hence, it is recommended that UAVs do not go near the eye of the tornado. Upon receiving information regarding the houses and businesses in a tornado’s path, they must travel away from the tornado and providing warnings to the houses and businesses.

History

Advisor

Houshang, Darabi

Chair

Houshang, Darabi

Department

College of Engineering

Degree Grantor

University of Illinois at Chicago

Degree Level

  • Masters

Committee Member

Williams, Quintin Sharabiani, Ashkan

Submitted date

August 2018

Issue date

2018-07-13

Usage metrics

    Categories

    No categories selected

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC