Mason Archival Repository Service

Mapping and Predicting Community Vulnerability to Hurricane Florence in Coastal North Carolina Using Machine Learning

Show simple item record

dc.contributor.advisor Sun, Donglian
dc.contributor.author Dahal, Om
dc.creator Dahal, Om
dc.date 2019-11-18
dc.date.accessioned 2020-06-02T17:07:53Z
dc.date.available 2020-06-02T17:07:53Z
dc.identifier.uri http://hdl.handle.net/1920/11786
dc.description.abstract Extreme record breaking hurricanes followed by heavy rainfall and flooding claimed dozens of lives and damaged billions of dollar worth of property every year in the Atlantic coastal areas of the United States indicating that they are most vulnerable areas to hurricane hazards. Nevertheless, all the communities are not equally vulnerable due to their varying degrees of exposure and coping abilities. Thus, it is of vital importance to study the extent of vulnerability in different communities for the purpose of prevention, preparedness, response, and recovery efforts. This study attempted to predict and categorize vulnerability of communities to the hurricane Florence in the New Hanover County, North Carolina considering hurricane and subsequent disasters as a composite event. The Random Forests, a data driven machine learning method, was used to predict and categorize vulnerability of communities in census blocks. The explanatory variables were created from distance features and raster datasets. The training features were selected from crowdsourced data, disaster emergency evacuation locations, and satellite imagery collected during hurricane events. The regression results showed 0.93 percent R2 value with tweets, roads, elevation, NDVI, and waterbodies as top five important variables. The classification results showed the accuracy per variable ranging from 0.96 to 1.00 with NDVI, roads, elevation, SPI, and tweets as top five important variables. The results demonstrated that the Random Forests ensemble learning method can be a valuable tool for categorical prediction and mapping of vulnerable communities from hurricanes. Furthermore, the results from both regression and classification models revealed that demographic variables are among the least important variables however they are not insignificant. It is recommended to combine all the three types of variables in prediction modeling for community vulnerability to hurricanes. The novel method used in this study may be used to identify the categories of vulnerable communities from various types of natural disasters in the other communities. It is also likely that predictions for vulnerability of buildings in the communities can be made using this method. en_US
dc.language.iso en en_US
dc.subject Hurricane Florence en_US
dc.subject machine learning en_US
dc.subject regression model en_US
dc.subject community vulnerability en_US
dc.subject random forests en_US
dc.subject classification model en_US
dc.title Mapping and Predicting Community Vulnerability to Hurricane Florence in Coastal North Carolina Using Machine Learning en_US
dc.type Thesis en_US
thesis.degree.name Master of Science in Geoinformatics and Geospatial Intelligence en_US
thesis.degree.level Master's en_US
thesis.degree.discipline Geoinformatics and Geospatial Intelligence en_US
thesis.degree.grantor George Mason University en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search MARS


Browse

My Account

Statistics