What the Clock Shows
Fossil-fuel burning and deforestation are the main drivers of global warming. The CO2 they give off makes up more than 75 percent of annual climate pollution. The Bloomberg Carbon Clock is a real-time estimate of the global monthly atmospheric CO2 level.
The following methodology is a nontechnical explanation of how the carbon clock works. The full version, which includes all the math and science underpinning the project, can be found HERE
The graphic draws on CO2 data released from the NOAA Mauna Loa Observatory. The Scripps Institution of Oceanography pioneered CO2 monitoring in March 1958 at the observatory in Hawaii. The National Oceanic and Atmospheric Administration started a parallel effort there in May 1974. Today, NOAA maintains a global network of observatories, sampling towers, flights, and flasks to measure the composition of the atmosphere.
To estimate real-time atmospheric CO2 levels between data releases, and forecast them, we analyze the three most recent years of data and use an average of the most recent four weekly data releases. That analysis is then fed into an algorithm. Each new weekly data point starts a new analysis that yields updated daily clock values and a trend line (shown in yellow on the graphic).
Two projections are made each week, a four-week daily forecast that runs the clock, and an annual forecast that projects the current trend one year into the future. The latter is appended to the graphic where the data end.
The carbon clock projections are the result of two mathematical procedures:
1. The "wavelet”: This is an equation that "learns" the long-term trend line of CO2 and adds on the seasonal peaks and troughs—the squiggles that pass above or below the trend line every half-year or so. It calculates the long-term trend from monthly data over the previous three years, which it uses to derive an initial rough daily forecast for one month into the future.
2. The Singular Spectrum Analysis (SSA) algorithm: This is a statistical tool that improves on the wavelet. It calculates the probable future trend of the data by running possible forecasts over and over until they start to converge. When they do, it quits, and outputs its best estimate for every day of the month. The final step is to use linear interpolation–basically an advanced mathematical method for connecting the dots—to turn the daily values into the second-by-second readings seen on the Clock. The clock displays eight decimal digits, determined by the model.
The shaded areas adjacent to the yellow trend line are “uncertainty bars,” which represent an average of the difference between the wavelet- and the SSA-determined trends. The year-ahead forecast on the graphic has shade bars that show where the projected path of CO2 is likely to fall with 95 percent confidence.
The background atmospheric CO2 concentration is uniform around the world. Daily, weekly, monthly, and annual averages all differ superficially because of short-term variation—basically, weather—that can mask the long-term upward trend. Because the Bloomberg Carbon Clock is projected from the average of the four most recent NOAA weekly estimates, it may be slightly lower or higher than shorter-term measures at any given moment.
The Scripps CO2
program maintains a helpful graphic on its website
that displays CO2
data averaged over several time periods. The hourly, daily, and weekly averages each show decreasing levels of variability. The long-term trend becomes more focused monthly and annually.
The Scripps CO2 time-series is known as the Keeling Curve, after the scientist who initiated and maintained it for almost a half-century, Charles David Keeling.
The animated graphic below the Clock is a combination of several CO2 time series. Moving from the top right, to the bottom left, the Curve is assembled from these sources:
The Year Ahead: The model projects forward one year, to give a visual estimate of the trajectory of CO2. The annual forecast carries a 95% confidence band. The forecast trend is shown as an extension of the yellow historic trend; they are determined the same way, by the average of the difference between the wavelet and the SSA algorithm results.
May 1974 to the Present: Mauna Loa Observatory average CO2 record, maintained by NOAA.
March 1958 to April 1974: Scripps Institution of Oceanography Mauna Loa averages.
Ice Core Record: Fossilized air trapped in Antarctic and Greenland ice has allowed scientists to estimate atmospheric CO2
content going back 800,000 years. The highest value in this record is 298.6 ppm, seen about 330,000 years ago. These records are available online as the Antarctic Ice Core Revised Composite CO2 Data
Satellite images of the Earth were made by the Japan Meteorological Agency
weather satellite, Himawari-8. The imagery is processed at Colorado State University in cooperation with the National Oceanic and Atmospheric Administration and the Japan Meteorological Agency. The images were assembled into video by Blacki Migliozzi, with advice from Dan Delany
. The image archive can be found here
Several scientists either read the technical working paper in draft or provided helpful conversations about the methods described here. They include:
* Michael Ghil, Department of Atmospheric and Oceanic Sciences, UCLA
* Dmitri Kondrashov, Department of Atmospheric and Oceanic Sciences, UCLA
* Mahé Perrette, Potsdam Institute for Climate Impact Studies
* Michael Mann, Earth System Science Center, Penn State University
* Andrew Robertson, International Research Institute for Climate and Society, Earth Institute, Columbia University
* Gavin Schmidt, NASA Goddard Institute for Space Studies
* Pieter Tans, Earth System Research Laboratory, NOAA
Data modelling by: Jan Dash & Yan Zhang
Published: December 1, 2015
Data updated: July 1, 2016