RisingTemperatures

A Java application for analyzing climate change CSV data.

Note: Fork of Model class architecture (PoC): SecresCSV - A GUI for opening/viewing, saving, printing, editing and/or refreshing multiple CSV files at a time in tabular format.

About the Data:

• Aims to analyze the temperature changes across 243 regions that cover approximately 197 countries of Planet Earth from 1750 to 2015
• Procured from Berkeley Earth, which is a California-based independent non-profit agency focused on land temperature data analysis for climate science
• Consists of five large datasets which have monthly average land and average ocean temperatures from 1750-2015 for 243 regions around the earth organized by state and major cities

Methods/Design:

• Dependency Management/Build Automation:
  • Eclipse with Maven
  • Dependencies:
    • JFreeChart
    • FlatLaf
    • OpenCSV
  • Git to manage workflow
  • After every build, commit to git remote repo
• Wrapping Build with Installer:
  • jpackage --> wraps JAR file with installer to generate installer extension (i.e. exe/msi, dmg/pkg, deb/rpm) for platform dependence
  • Executable JAR file (created with Maven Shade Plugin) also used to maintain platform independence
• Reading Data:
  • OpenCSV parser in Model class
  • View class contains JTable; charts read from JTable's DefaultTableModels, not dataset files
• Optimization:
  • Optimizing Loading of Data
    • 3.5 hr --> 10 sec --> 3 sec
  • Optimizing Charts
    • 20 sec --> 2 sec
    • Used HashMap to iterate once through dataset for "change in temp. bar charts" to filter through data
• Visualization:
  • JFreeChart
  • Types of Visualizations:
    • TimeSeries line charts
    • Scatter plots with varying degree regression models
    • Line charts with regressions
    • Various types of bar charts to illustrate temperature and change in temperature
    • Thermometer Plot
• General Code:
  • Follows Modified-MVC Architecture
  • Model consists of CSV data and adds to JTable's DefaultTableModel

Discussion

• Visualizations progressed similarly at different intervals of time
  • Mid-Late 1700s - early 1800s --> Large uncertainty, data points scattered
  • Late 1800s - 1900s --> distinctive trends with discernible increase and decrease in temperature, adherence to regression models
  • Late 1900s - 2000s --> Unnatural increase in temperature, data points moving away from regression models, trends appear less distinct
  • All of the above --> Recent increase is unnatural, data has been moving away from regression line with positive slope, so temperature is increasing
• Other
  • Comparing linear and power regressions: global & regional temperature increasing linearly, not exponentially
  • Bar graphs identify temp. & change in temp. --> all regions have increased in temp. (1912-2012) except Kyrgyzstan
  • Effects: Rising temperatures, more droughts, intense hurricanes

Future Applications/Research

• Fork of Model class architecture (PoC): SecresCSV - A GUI for opening/viewing, editing, printing, and/or saving (beta) multiple CSV files at a time in tabular format.
• Models calculated (linear & power) can be applied for future predictions - Ex. heatmap overlay from 2020-2030
• Future projects could fork and easily add new visualizations based on existing architecture (add new class that extends AbstractGraph with model and view methods, create instance in GraphCharts and coordinate with View class
• Applying sorting and filtering algorithms could prove useful in future charts/maps etc.