Data and code

The data and code used in Clisense are archived in this Google Drive folder. For the app to work, you need to download all the files to the same folder. When using R Studio, you will see a “Run App” button at the top right corner of the editor, where “Run” would normally be.

Please notice that not all of the authors have consented to the uploading of the input data into Google Drive. For the app to work fully, you need to download those other files from the original source (specified further below) and put them in the same folder as everything else.

Atmospheric temperature

Air temperature data comes from three sources.

First, Version 4.6 of the HadCRUT record. Here is a link with some explanations on the data. I believe I’m allowed to upload the file to Google Drive, but per the licensing terms I have to write © Crown copyright, Met Office (if you’re from the UK) or © British Crown copyright, Met Office (if you’re not).

Second, the Cowtan & Way modification of the HadCRUT record. It has been argued that, because HadCRUT does not estimate temperatures in areas without coverage (e.g. the Arctic), it is biased low. The Cowtan & Way record is an attempt to fix that; here you can find a better explanation. I must thank lead author Kevin Cowtan, who gave permission for the uploading of the temperature file.

Third, a series kept by the National Oceanic and Atmospheric Administration (NOAA). In this case data starts in 1880; here there are explanations on the series and here is the specific file used by Clisense.

Note: in June 2019 the HadCRUT component that tracks sea surface temperatures received a significant update, HadSST4. This in turn led to a new version of Cowtan & Way. Since the global HadCRUT series hasn’t received the updated sea-surface figures yet, I’ve preferred not to include the new Cowtan & Way numbers either; once HadCRUT has added the new sea-surface temperatures, I will update it together with Cowtan & Way.


The forcing figures come Lewis & Curry 2018. The estimate of the forcing resulting from a doubling of atmospheric CO2 (3.8 watts per square meter) also comes from that paper, for reasons of consistency. Since the forcing data series ends in 2016, that’s the last year for which Clisense can offer results.

Lead author Nicholas Lewis kindly allowed me to use the forcing data for this app. He also offered useful input on how to handle the figures.

At first sight it’s hard to know what several of the forcing labels mean. To get a better idea, I recommend reading both the Lewis & Curry paper and, if you have time, chapter 8 of the IPCC’s Fifth Assessment Report.

Ocean heat

Data on ocean heat content comes from three sources. The first, Zanna et al 2019, estimates full-depth ocean heat content from 1870 to 2018.

The second source is the dataset kept by the Institute of Atmospheric Physics, also known as the Cheng series. In this case the data on ocean heat content go back to 1940, so the first year for which Clisense and estimate energy imbalance is 1941. I’m thankful to lead researcher Lijing Cheng for allowing the archival of the data in Google Drive. Here is his website and here is the actual data file Clisense uses.

Finally, NOAA offers two series. One is a 5-year running mean starting in 1955, so the first year for which it offers results is 1957 (i.e. the 1955-1959 average). While yearly data would be preferable, that series only goes back to 2005, so Clisense uses instead the pentadal data. Further info on NOAA’s ocean warming data can be found here.

The Cheng and NOAA series only cover warming down to a depth of 2000 meters. To account for ocean heat uptake below that level, I add 0.07 watts per square meter from 1992 on. This follows the IPCC’s Assessment Report 5, which in page 264 (page 10 of this PDF) stated the believed rate of warming below 2000 meters was 35 terawatts. Notice that there is great uncertainty about this warming rate, and zero warming was assumed for the pre-1992 period.

%d bloggers like this: