Links to the dashboards:
The National Oceanic and Atmospheric Administration (NOAA) in the USA releases the Storm Events Database that records the occurrence of storms and other significant weather phenomena having sufficient intensity to cause damage to property and/or crops. Our primary objective was to explore three major types of storms that cause widespread economic damage in the USA: tropical cyclones and floods (including hurricanes), severe local storms (including tornadoes), and wildfires and droughts. Our project focused on exploring varied independent variables like population, economic activity, and associated weather statistics that affect the total damage caused by these natural disasters in the USA.
While it is not possible to predict where and when the next natural disaster will strike, it is possible to develop models that can predict the amount of damage from such disasters. These predictive models can be useful to plan for better emergency management as the nation's increasing population grapples with the increase in both the frequency as well as the intensity of storms caused by global warming. In addition, local governments can use the models to plan for disruptions by using these models for what-if analysis.
This project aimed to answer the following questions:
- What are the various factors that contribute to the amount of damage from storms in the USA? These factors include population, economic activity in a county, and the weather related statistics such as precipitation and temperature.
- How to collect and visualize the input features related to natural disasters for the period 2000-2021.
- Is there a meaningful correlation between the input features and the damage caused by the storms?
- National Oceanic and Atmospheric Administration (NOAA) Storm Events Database: This database includes details on the occurrence of storms that cause loss of life, significant property damage, and disruption to commerce in the counties and forest zones in the USA.
- US Census.gov: In order to explore the socio-economic factors that impact the damage from the natural disasters, we needed to access several secondary datasets. The Census Bureau’s Population Estimates Program dataset provided the population estimate in the county affected by the storm for the year of the event (between 2009-2019). The Decennial Census provided the population for 2000 and 2010 and we interpolated the data for the years 2001-2008. The Census Bureau’s County Business Patterns dataset provided the number of businesses, their number of employees and their total payroll in the county. Its Non-Employers dataset provided the number of small businesses and the total revenue generated by them. Finally, its Economic Census provided the economic activity for the top 3 industries in the county.
- US Drought Monitor (USDM): The US Drought Monitor publishes a map showing the severity and location of drought throughout the country. We use the drought severity data for the county where the wildfire event occurred as input to the model for predicting the damage from wildfires.
- Weather.gov: The primary dataset is missing the county FIPS, latitude and longitude of storm events that occur in a forest zone (a National Weather Service forest zone covers more than one county). We used the zone county correlation dataset to get the FIPS, latitude and longitude of a given zone FIPS ID for an event in the storm events dataset.
- National Center for Environmental Information (NCEI): In order to get the weather records for a county, we needed the list of weather stations in and around an affected county. From the list, we find the nearest weather station that has daily weather summaries for the past 10 years from the date of the event. We obtain the precipitation, snowfall and temperature data for the affected county.
