This repository includes processed data and code used to generate figures for the paper '*Ten recommendations for reducing the carbon footprint of research computing in human neuroimaging*'. This includes: - **UK Carbon Intensity Data - Time of Day.csv**: These data were used to generate the Figure 1 line graph of mean carbon intensity (grams of carbon dioxide equivalent per kilowatt hour) for each 30 minute period of each working day, using all available data from September 26, 2017, to October 31st, 2023. Data is split by year. Values reflect actual carbon intensity values as provided by the [UK National Grid ESO carbon intensity API][1]. Raw data for this figure can be found through this API. - **UK Carbon Intensity - Time of Day.py**: This script was used to extract the Time of Day data. Input was carbon intensity data for each 30 minute period of each year, as extracted manually from the UK National Grid ESO carbon intensity API. - **UK Carbon Intensity Data - Month Averages.csv**: This file contains the mean carbon intensity for each month from 2018 to 2022, with data taken from the same API as above. These data were used to generate Supplementary Figure 1. Respective standard error of the mean (SEM) values are also provided, used to generate error bars. - **UK Carbon Intensity - Month Averages.py**: This script was used to extract the Month Averages data. Input was carbon intensity data for each 30 minute period of each year, as extracted manually from the UK National Grid ESO carbon intensity API. - **fMRIPrep Junk Files Data.csv**: These data include values used to generate the Figure 3 pie chart, reflecting the proportion of files (gigabytes) form a given fMRIPrep preprocessing pipeline that can be classified as 'Used in analysis' and 'Junk files' (not needed for analysis, and appropriate to delete). We present the size of data (GB) for each subject that falls into different categories of interest. This includes Target Data (processed output files that we intended to keep and may be used in subsequent analysis), and Overall Junk (files that will not be used in subsequent analysis, and can safely deleted). We also provide a breakdown of how this junk data divides into log files, un-needed derivatives (output), scratch space (working directory), and intermediary FreeSurfer files. These data were obtained as part of an [ongoing preregistered study][2] exploring the relationship between estimated carbon emissions and preprocessing performance in fMRIPrep. Specifically, this data was taken from the 'baseline' pipeline, on a stop signal task performed by 257 subjects (from the [UCLA Consortium for Neuropsychiatric Phenomics LA5c Study][3]). - **fMRIPrep Junk Count.py**: This script was used to extract the above data. Input was the output of preprocessing in fMRIPrep, employing a file structure setup by the author. This is an older version of what came to be the public tool [fMRIPrepCleanup][4], available on GitHub. This tools provides fMRIPrep users with the ability to flexibly delete junk files derived from fMRIPrep, and does not rely on a particular file structure. Figure 2 was generated using unprocessed original data from the 2022 v1.0 release of Country Specific Electricity Factors (2022) from www.carbonfootprint.com. Given that all data for this figure is accessible through this link, it is not reproduced here. [1]: https://carbonintensity.org.uk [2]: https://osf.io/839pa [3]: https://openneuro.org/datasets/ds000030/versions/1.0.0 [4]: https://openneuro.org/datasets/ds000030/versions/1.0.0