Predicting Tropical Storms Power Outages with Multidisciplinary Information for the Continental United States
Abstract
Hurricanes, Typhoons, and other tropical storms can cause widespread damage to infrastructure which is difficult to predict because of the complex interactions between the weather hazard, the infrastructure, and the surrounding environment. While previous research has demonstrated that machine learning methodologies can be applied to predict impacts like the number of power outages caused by tropical storms, they have also shown that the quality and availability of data can limit the accuracy and generalizability of these data-driven models. A more comprehensive approach to data-driven impact modeling is needed. We will present such an approach applied at a national-scale with a machine-learning model designed to predict the number of customers affected by power outages caused by tropical storms in the Continental United States. Using a record of 16 tropical storm events (2015 to 2019), we generate comprehensive and multidisciplinary information about the storms as well as the economy, demographics, and environment of each county in the model to ensure that the major aspects of the hazard, vulnerability, and exposure are included. The data sources include: ERA 5 Reanalysis for weather and dynamic land surface characteristics, MODIS for a Leaf Area Index Climatology, National Land Cover Database (NLCD) Tree Canopy coverage, USGS gridded population density, county-level demographic and housing data from the American Community Survey (ACS), and county-level business and employment data from County Business Patterns (CBP) dataset. Based on our results, we believe this or similar approaches could be applied to create a range of useful data-driven impact models that could be applied to different hazards and infrastructural systems.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMNH0070007W
- Keywords:
-
- 0520 Data analysis: algorithms and implementation;
- COMPUTATIONAL GEOPHYSICS;
- 0555 Neural networks;
- fuzzy logic;
- machine learning;
- COMPUTATIONAL GEOPHYSICS;
- 1914 Data mining;
- INFORMATICS;
- 4333 Disaster risk analysis and assessment;
- NATURAL HAZARDS