Climate Change

Climate change is no longer some far-off problem; it is happening here, it is happening now.
- Barack Obama
Climate change has an assorted list of causes but the scientific consesus stays that human influences remain a major factor.The US has been long at the international negotiation table of climate change policies. One would think that US withdrawing out of Paris Climate Accords would garner more attention than it did, but apparently with an EPA marred with scandals and a President with a twitter account is enough of a distraction. This makes increasing awareness among the masses a paramount necessity. But what does Climate change have anything to do with Tornadoes? Apparently, there is no clear evidence of a correlation and thus, not guilty until proven. Although recent studies indicate that the frequency of US outbreaks with many tornadoes is increasing, and is increasing faster for more extreme outbreaks. It is difficult to identify long-term trends in tornado records, which only date back to 1950 in the United States. Also, it is just as difficult a phenomenon to model. Thank the almighty that it is not as difficult to visualize.
Team "R you Shiny"

A team of graduate students at the University of Illinois at Chicago.

Analysis and Munging

Data

The core data about the tornadoes and their tracks, from 1950 to 2017, is available here with a description of the fields here. In addition, we also used the National FIPS County Codes from here.

  • As of this writing, there are 63,690 records that represent 62,530 tornadoes.
  • Multiple records are used to represent tornadoes that cross state borders and/or more than four counties. In this dataset, there are at most 4 records/up to 16 counties (1 instance) per tornado. Six instances have 3 records/up to 12 counties, and 107 instances have 2 records/up to 8 counties.
  • The times in the data are normalized to CST for all but 37 records. For the purpose of this experiment, we will assume that they are all in CST.
  • Data for loss is represented in one of three ways.
    • 1950-1995: 5 followed by the number of zeroes as the values in this cell (for example, a value of 4 represents a loss of up to $50,000),
    • 1996-2015: value in millions, and
    • 2016-2017: actual dollar value.
    These have been normalized to the actual loss value. Inflation is not taken into account.
  • There are many instances where a continuation record is not placed right after the initial record. These, and other modifications, including generating year- and state-specific IDs, eliminating unnecessary columns, and the dollar loss value calculation (above) have been applied using a C# application. Look in the Munging folder in the codebase to view the list of modifications applied.
  • Descrepancies noted and corrections applied:
    • yr:om = 2001:56, 2002:506, 2010:252, and 2015:610626 have an erroneous extra records, which have been removed.
    • For yr:om = 2015:582218, it also interesting to note that it crosses statelines between TX and OK multiple times.
    • For yr:om = 2001:592, NoOfStates was incorrectly set to 0, probably because FIPS1 is 0. This has been updated to 1 State, and FIPS2 has been moved to FIPS1.
    • For yr:om = 2011:291657, ns:sn:sg 1:0:-9 changed to 1:1:1 since there is only 1 record for this instance.
    • In 26,363 instances (one instance from 2 states, the rest from single state tornadoes), elat and elon values were set to 0. These have been updated to the slat and slon values respectively.
    • Some counties are repeated multiple times within the same tornado instance. These have been deduplicated.
Insights

Interesting thing found through the visualization

How To Use Application

The user can navigate the application through the control panel, with additional ability to interact using the maps.

You Spin Me Round main layout of the application

Different tasks that the user can perform using this application are as follows:

Viewing state and county data

  • The user can see the tornado tracks by state and using the controls can compare the data from two states in the map as well as in tabular and chart form.
  • Further, the user can view the data on per county basis by selecting the counties from dropdowns provided in the control panel.
    • Controls for filtering by state and county

  • In addition to using the control panel, user can switch to the State or County mode and then directly make the selection on the map for looking at the tornado tracks.
    • Direct Interaction with map for viewing state/county data


  • The user can view the heat map showing where it is more or less safe to be regarding torandoes on per county basis.
    • Heat Map showing where it is more or less safe to be regarding torandoes


  • The tornado data can also be viewed as parallel co-ordinates chart which allows user to find correlations between different parameters like magnitude, injuries, fatalities, loss and year.
    • Parallel co-ordinates representation of tornadoes


  • In addition to that the data can be viewed as bar chart and table which represents injuries, fatalities, loss and number and percentage of tornadoes which in combination with filters mentioned below will show data by year, month or hour of the day.
    • Bar Chart showing number of tornadoes for each year

    Table showing number of tornadoes for each year


High Level Filters

  • Using the top level filters the user can refine the data by Year | Month | Hour | Distance from Chicago and County. User can also step through time year by year using the slider play button.

Detailed Filters

  • Also,the user has been provided with controls to further detail down the data using these filters. Some of the options include filter by: Magnitude | Width | Length | Distance from Chicago | Loss | Injuries and Fatalities
    • Filters available to the user of the application

  • The user has the ability to change the tornadoes color and width on the basis of the filter criterias mentioned above.
    • Color and Width By option

Map Types

  • The user can alter the map background to any of the five different types. Different map types used are:
    • Dark (Default theme): Provides better contrast and allows the user to focus on the data
    • Light: Some users prefer light themed applications, so this map type can be helpful for them
    • Satellite: Can be used to see the correlations between the geography and tornado occurrences
    • Minimal: Users can view the tornado tracks without being distracted by loads of labels. Minimalistic labels have been provided.
    • High Contrast: In contrast to Minimal map type, this map type along with being high contrast also has lots of labels which is helpful for users who are unaware of the geography. In addition to that, the colors used in this map type are color blind safe.
    • 5 different map types for users to choose from

Preferences

  • The time measurements can be switched from am/pm to 24hr time format and also between metric and imperial measurements.
    • Measurements and Time format preferences

About

  • The information icon shows information about the project including liks to the resources used.
    • Information about the project