Carbon footprint / energy use reporting guidelines

[ampersandmcd]

Summary Sentence

Computational research is an inherently energy-intensive practice. As a leading venue for the presentation and publication of computational research in climate, earth, and environmental science, it is our responsibility to encourage or require reporting of the estimated carbon footprint / energy use of each submission, following the lead of Strubell et al. [1], Lacoste et al. [2], Anthony et al. [3], Henderson et al. [4], and other thinkers in this space.

We must, of course, consider this responsibility without discouraging submission or implying judgement towards researchers who are doing incredibly important work!

[1] Strubell et al., 2019, Energy and Policy Considerations for Deep Learning in NLP
[2] Lacoste et al., 2019, Quantifying the Carbon Emissions of Machine Learning
[3] Anthony et al., 2020, Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models
[4] Henderson et al., 2020, Towards the systematic reporting of the energy and carbon footprints of machine learning

What needs to be done?

• Decide whether to encourage or require carbon footprint / energy use reporting
• Compile list of tools to facilitate such reporting
  • CodeCarbon
  • carbontracker
  • Live Grid Carbon Intensity Map
• Prepare template / guidelines for reporting

Who can help?

• Andrew
• Anastase

---

Update after the issue was opened

[Add details]