Calculus for measure
There are things you think you are going to understand, eventually – and things you have to ask about because after a few months, it is pretty clear you won’t magically seize the big picture well enough to make sense of the details.
Ever since I joined ecoinfo last fall, I learned a lot, but still had a slight feeling of disarray when it came to discussions about assessing the environmental footprint of a digital process or item. So I asked if someone could explain the whole thing to me from scratch, and was happy to be offered an online tutorial by Françoise Berthoud to address the basics of the relationship between all things digital and greenhouse gas emissions. In this blogpost, I present what she explained to me (and to a few others who joined in) in the first session – there are some more to come!
Understanding carbon, emissions and footprint
The figure that is usually invoked when talking about the environmental footprint of the digital world is that it makes for 3 to 4 % of greenhouse gas emissions worldwide. In a first step, Françoise explained how CO2 is produced, and more importantly to me at that point, what is the difference between emissions and footprint. The Paris agreement mentions figures relative to the carbon emissions of the different countries. Now these emissions only include what is being produced in the country. All the Chinese hardware we are importing is not included in our emissions, for instance. They are, on the other hand, in our footprint. This difference explains why the average individual carbon emissions of a French individual is around 5T a year, and their footprint 10T. For obvious reason, the situation is reversed in a country like China.
Calculating all things digital
We then went into calculation methods. Because there are only things we can calculate, not so much things we can measure. And there are some needs in terms of how they are being calculated, especially standardizing needs.
The first halt we made on our way to discussing how to assess the footprint of the digital world was to define what is included in said digital world. First, there is hardware. Some equipment or devices clearly are easily identified: PCs, printers,… but some others are sometimes included in this perimeter, and sometimes not, for instance connected objects, or TVs. In other words, even when it comes to hardware, not everybody has the same reference of what should be taken into account as belonging to the digital world. Then, you should add to hardware the emissions coming from data centers, e.g. their equipment. And thirdly, all that concerns network (smartphone, internet access etc.). There three areas are taken into account in coarse-grained calculations such as those from the SHIFT project.
Even with these three dimensions, many aspects remain below the radar. For instance, when the emissions of a data center are calculated, you don’t really consider the emissions connected to building the data center, or to repairing failing equipment. Also, the satellites are not taken into account.
Still, these are the conditions in which we are today calculating carbon emissions of the digital world.
Scoping greenhouse gas emissions
In the next step, Françoise went into the method of calculation for carbon emissions in general, in the more fine-grained perspective that can be applied at individual level. She described the three different scopes that are taken into account (see here Scope 1, 2 and 3 in the Greenhouse Gas Protocol section), choosing examples from our activity as researchers (Scope 1 is when you have a leaking AC in your lab and assess the gas leakage, Scope 2 is the greenhouse emission produced by the electricity you use at your lab, Scope 3 includes hardware, travels, any consumables of your lab, etc.).
Now that we had gone through the basics, we turned to the question: how do I calculate the footprint of a simple digital object, a laptop for instance?
The circle of liiiiiife (not)
So we moved to Life-cycle assessment. In order to realize a so-called LCA, you have to take into account the whole lifecycle of an object, starting with the extraction of the metal needed to build it.
Having identified components, you then have to rely on existing data for every single step of the analysis, but then again, all the data considered are subject to LCA themselves. This is somewhat abysmal, but ok. Still, again, there are variations in the perimeter of the data you take into account when analyzing the Lifecycle of a digital object. And sometimes, there exist no data about one component or the other… so, there are some blurry zones here too, but let’s try to analyze the lifecycle of our laptop.
The general idea is to break it down to every single component, then check the database that presents information on the component (energetic needs, type of pollution it causes,…), and then build a stratified assessment component by component. Graphically, you represent each component with a box, with input and output arrows, and then compact the boxes together for the overall LCA.
There are not that many databases that can be used for LCAs to this day; the main one is ecoinvent (Françoise also mentioned eime for electronics). For each component, the LCA for our laptop will be based on average values known for similar components . Which means that there is always an uncertainty margin when analyzing the lifecycle of an equipment. This is especially true when it comes to computers since the computer types usually included in the database are rather older models. But basically, you can check for the closest monitor size, processor type, etc. to your actual computer and refer to the values indicated in the database. They will include all components and references to database values.
The LCA result is then composed by a series of indicators. There are actually a lot of indicators available, of which only a handful are usually presented in an LCA result. One indicator for instance is the “climate change” impact, meaning the equivalent of CO2 mass produced when producing the component. Other indicators are acidification or water toxicity for instance; such an indicator as water toxicity can then be listed only as a “potential” impact because it basically depends on there being water or not for it to actually deploy its toxicity potential.
The information an LCA provides varies a lot depending on the indicator. For some indicators like climate change or metal depletion, they can be considered a rather reliable estimate, but for other indicators, they are to be considered with some caution. And the end of the lifecycle (waste phase) is generally not taken into account as accurately as it should.
All in all, even the LCA of simple digital equipment, as complex as it may seem when you consider the long pages of tables and wide array of indicators, leaves a lot unassessed…
Looking forward to the next session, on the LCA of a digital service (it promises to be more complex than our laptop example)…. Stay tuned!
I would like to thank Françoise Berthoud as well as the other participants of the online session (especially Olivier Ridoux and Benjamin Ninassi) for their input and pedagogy (and also their patience!).