The People’s Climate March on September 20th brought over 300,000 people to the streets of New York City to voice support for policies that reduce the man-made effects of climate change across the globe. It couldn’t be any more timely that an international research team led by scientists at Arizona State University released the Fossil Fuel Data Assimilation System. It’s a global database of CO2 emission estimates at the 0.1 decimal degree resolution (about 8-10 kilometers in the continental U.S., depending on latitude) containing hourly and yearly data from 1997 to 2010. You can visualize the data by year here, and it’s available for download in a few different file formats, including text and csv. It’s quite an incredible database, and the first of its kind at that resolution. The CO2 is estimated using a combination of existing data sources such as population, remotely sensed nighttime lights and the location of power plants. The methodology can be found in Rayner et al. (2010) and Asefi-Najafabady et al. (2014).
My first question when discovering this dataset was “Where might co2 emissions have increased or decreased since 1997?”. My hunch was that overall emissions have increased, but that the spatial distribution might show some interesting trends across the United States. For example, the spatial distribution may have been affected by the fact that in 1997 the U.S. was in a period of strong economic growth, while in 2010 the country was still recovering from the Great Recession.
Data Processing
For this exercise, I converted the text files into polygons and sampled at the county level to get estimates of CO2 change between 1997 and 2010. Alternatively, the NetCDF files could be used to generate rasters for visualization of the data. QGIS has a NetCDF browser plugin for doing just that. Since the dataset is at the 0.1 degree resolution, it lends itself to creating a raster quite easily. Originally, I planned to vectorize the raster dataset to display it in CartoDB (as of now, CartoDB does not support raster files). However, I thought it would be more interesting and useful to aggregate the data to a more common unit of analysis that people would understand, such as counties.
There are a few ways to go about this. I’ll describe how to do this in ArcGIS using the ET GeoWizards plugin, but it’s also possible to do this analysis in QGIS. ET Geowizards is available as a free ArcGIS plugin. There’s also a paid version with some more advanced features that requires a license.
First, I had to convert the text files containing the coordinates and CO2 emission value into a vector dataset. Originally thinking I was going to create a raster visualization, I used the Point to Raster tool in ArcGIS to create a raster surface of the data, but later decided to calculate the data at the county level, and vectorized the raster data with the Raster to Polygon tool. This produced polygons of raster cells at the 0.1 degree resolution. The data itself is measured in Kilograms of Carbon (kgC) per square meter, so I transformed the data to total kgC, then converted to metric tons. Since the vector polygons are smaller than counties, I was able to resample the data using the Transfer Attributes tool in ET GeoWizards. This tool applied a proportion of the CO2 emissions total to each county polygon based on the proportion of the CO2 emissions polygons that overlapped each county. If the CO2 emissions polygon was entirely inside the county, the total amount was applied. Next, I summarized CO2 emissions by county. Quite conveniently, the Transfer Attributes tool will do all this.
If you don’t have access to ArcGIS, the entire attribute transfer process can be accomplished using the QGIS Intersect and Dissolve tools. The county and smaller CO2 emissions polygons can be intersected in QGIS, and the area of the resulting polygons can be divided by old polygon areas to get a ratio. That ratio can then be applied to the CO2 emissions polygons and summarized at the county level.
These processes work under the assumption that the value for the CO2 emissions polygons are a total amount of emissions for the entire area of the polygon. However, CO2 emissions are rarely uniformly distributed — therefore, using this coarse resolution CO2 at any geographic level smaller than a county is probably not appropriate.
Below you’ll find a map of estimated CO2 emissions change in percentage between 1997 and 2010 for every county in the United States, visualized in CartoDB.
Increase in Carbon Dioxide Emissions
In the Tableau visualization at the bottom of the page, you can also view the ten counties with the largest increase in CO2 emissions in metric tons, compared to their population change. Most notable are the increases in Collin and Denton County, Texas. Both counties are in suburban Dallas. Other counties are in similarly fast growing areas in California and Florida, along with Allegheny County, Pennsylvania (home to Pittsburgh). It makes sense that rapidly growing areas would see such an increase in CO2 emissions. Allegheny County is a different story, since it lost population during the data’s time frame. Overall, all major Texas metros saw significant increases, as did Florida — with the notable exception of Miami. Southern California, Las Vegas and Phoenix saw increases. There’s also a noticeable trend of increase across the central Midwest, from Illinois through Ohio.
Decrease in Carbon Dioxide Emissions
The counties with the largest decrease in carbon dioxide emissions are mostly in parts of the country that haven’t done too well economically over the past 20 years, such as the Detroit area. New England and most of the Northeast in general also saw a decrease, with the cities of Boston, Philadelphia and Baltimore. CO2 emissions decreased across the Northwest, including Portland and Seattle. Perhaps the most stark trend visible is the decrease across the Great Plains, surely related to the decrease in population in that part of the country.
Largest Counties in the U.S.
Of the ten largest counties in the U.S., most saw increases in CO2 emissions. The exceptions here being the two counties (boroughs) in New York (Kings and Queens) and Miami-Dade, Florida. Both King and Queens in New York City saw only small population increases, but Miami-Dade saw a rather significant increase in population. Cook County, Illinois was the only county in the top ten to see an increase in emissions but a decrease in population. We find that in Maricopa County, home to Phoenix, and Harris County, home to Houston, CO2 emissions percentage growth has outpaced population growth in the time period.
Further Study
There are certainly opportunities to try to better understand some of what the data is indicating. For example, we might expect areas with rapidly growing populations to experience an increase in CO2, but what about counties that declined during the time period, such as Allegheny County, PA? There’s also some interesting trends to try to better understand. This is a great start, but I hope to see a higher resolution version of this dataset released in the future.