Emissions from Animal Agriculture: Not 18%, Not 14.5%, Not 16.5%, Not 15.6% but presently incalculable
- 14.5% continues to be cited in the media and peer reviewed science in spite of data being 16 years old
- FAO confusion over commensurability of GWP values renders new percentage figure void
- FAO briefly posted a new figure of 15.6% but then it was removed from their web-site
If you delve into the issue of emissions from animal agriculture you will find yourself in something of a minefield. I have followed this debate for 15 years but recently researched it more closely for a book I’m writing. This blog post is an important accompaniment to an article (Twine 2021) I recently had published on the issue in the journal Sustainability, the research for which was done in January. This post amends and extends the analysis of the article. Readers have the option of going to that article first (including the comment/amendment) and then returning to this blog. Some of this blog post will get quite technical because there is unavoidable complexity here, but I have aimed to keep this accessible.
If you have taken a passing interest in this debate (it would be a hope that anyone with an interest in climate change has done so) then you will know that the United Nations Food and Agricultural Organisation (FAO) produced two reports on the issue, each providing an estimated percentage of the contribution of animal agriculture to overall greenhouse gas (GHG) emissions. The first of these, Livestock’s Long Shadow, was published in 2006 and used its own data from 2001-2004 to estimate total CO2-eq emissions from animal agriculture and data from WRI CAIT from 2000 for global total CO2-eq emissions. The well quoted figure of 18% for emissions from animal agriculture as a percentage of total emissions was simply arrived at by expressing the former as a proportion of the latter.
The second FAO report entitled Tackling Climate Change through Livestock was published in 2013 (whilst the title of the second report seemed to use different language: global ‘livestock’ production no longer cast a long sinister environmental shadow, but was now the way to tackle emissions – the truth of the matter is that both reports advocated for technical efficiency changes to ‘livestock’ production as the way to address the emissions from the sector, rather than tackling levels of production or consumption – for more on this see Twine 2021). This time the FAO revised down their estimate from 18% to 14.5%. This was symbolically and politically important as the media message could become “this issue is not as bad as we thought at first”, but still, of course, finding that animal agriculture was responsible for 1 in 7 of all emissions. This second report drew upon FAO data for emissions from animal agriculture from the year 2005, and for global total emissions 2004. Also for this latter data they switched their source from WRI CAIT to the IPCC.
Citing percentages as factual and authoritative when they are really dated and provisional
A first point to make here refers to citation. I wonder how many researchers and journalists who routinely still cite either of these figures (18% or 14.5%) realise how old the data sources are? [Here is a recent example of 18% being cited Mylan et. al. 2019 and here is an example of 14.5% being cited Lazarus et. al. 2021 – there are further examples in the article]. The most recent data source – of the second report – is now 16 years old. Reflect upon that for a moment. How many of those who cite these figures also know that they are in important senses provisional and dated constructs? For example, both the WRI CAIT and IPCC data for global total emissions tend to undergo periodic revisions due to more accurate estimates or changes employed to the weightings of the global warming potential (GWP) of gases such as methane and nitrous oxide (more on this later because ambiguity over these is why the article had to he amended). Furthermore, when bodies like the IPCC produce annual global total emissions data it often comes with a plus or minus of several gigatonnes because the estimates are so difficult to be sure of [as an example of this, the IPCC Fifth Assessment Report says that global total emissions data for 2010 was 49 Gt CO2 equiv. + or – 4.5 Gt CO2 equiv.. (p.42). I’ll return to the significance of this 49 figure later but such pluses or minuses say a lot about the provisionality and uncertainty of any percentage figures for animal agriculture that may be generated from them].
Therefore, when trying to estimate a percentage figure of emissions from animal agriculture, as the FAO did in both these reports, there is a lot of uncertainty and margin of error that could change the percentage figure quite markedly. With the second report the FAO arrived at 14.5% by expressing their own data on emissions from animal agriculture (7.1 Gt CO2 equiv. from the year 2005) as a proportion of the IPCC global total emissions for the year 2004 of 49 Gt CO2 equiv. (as reported in the IPCC Assessment Report 4 published in 2007). Not only are the data from slightly different years but the FAO, as I mentioned above, switched their source of global total emissions data from WRI CAIT in their first report (Livestock’s Long Shadow that yielded a percentage figure of 18%) to the IPCC in their second report. Whilst there may be very good reasons for aligning with IPCC data, it is noteworthy that a brief check of the recent historical data shows that IPCC estimates for global total emissions data are consistently higher than those of WRI CAIT data. A larger total figure is more likely to produce a lower percentage figure for GHG emissions from animal agriculture. Furthermore by the time the IPCC produced their next Assessment Report (number 5 in 2014) they had revised down their estimate of global total emissions data for 2004 from 49 Gt CO2 equiv. to 45 Gt CO2 equiv.. Now this will have partly been because the IPCC AR5 used different GWP weightings for methane and nitrous oxide than IPCC AR4 but it still underlines the provisonality of such data, which are used to yield these headline percentages that then carry highly significant political and cultural symbolism in climate change and other debates.
As I point out in the article there is also uncertainty about the FAO 7.1 Gt CO2 equiv. figure. This figure stayed the same between the two FAO reports, although it was calculated differently. After the first report the FAO introduced a new methodological tool for the calculation of emissions from animal agriculture which they called the Global Livestock Environmental Assessment Model or GLEAM for short. The second FAO report, Tackling Climate Change through Livestock (2013) is essentially using the data from GLEAM 1.0. As I explain in the article (pp3-5) there are some unanswered questions over how the second report limits its accounting of emissions from manure and land use changes. For example I point out that the reports use specific and recent time periods for deforestation and that “The second report limited its analysis of feed crop expansion to soybean cultivation in Brazil and Argentina, whereas the first included all feed crop expansion in Brazil and Bolivia. This temporal, spatial and feed-crop type delimiting fails to account for the global scale of carbon sink loss via deforestation related to animal agriculture” (p.5). I also mention that other, more detailed subsequent analyses (Searchinger et al 2018; Poore &Nemecek 2018), have questioned the extent to which the FAO adequately accounts for land use changes. Incidentally the peer-reviewed findings of Poore & Nemecek (2018) [see also their related erratum to that paper here] put the contribution of animal agriculture to overall emissions at 28.1% based on 2010 data for both emissions and total global emissions (using the EDGAR database).
The lesser known third FAO ‘report’
Many of the researchers who continue to cite the FAO 14.5% figure (or even the 18%) doubtless remain unaware that the FAO actually produced a third analysis (GLEAM 2.0) in 2017, which is also explained in the article. The interesting thing here was that on first analysis the FAO seemed to have not sought publicity for this – it’s on their website but there was no formal report as with the previous two – and there was seemingly no declaration of a newly updated percentage figure for the contribution of emissions from animal agriculture (although see later for ambiguity about this). What they did make clear was that their analysis had shown that compared with the second report emissions from animal agriculture had grown 1 Gt from 7.1 Gt CO2 equiv. to 8.1 Gt CO2 equiv.. This data was from 2010, so a 5 year update on the 2005 data used in the second report published in 2013.
Since I wanted to discover what new percentage figure this new analysis might reveal I sought out IPCC data for global total emissions in 2010, finding a total figure of 49 Gt CO2 equiv. in IPCC Assessment Report 5. Expressing 8.1 as a percentage of 49 yielded a new percentage estimate (for 2010) of 16.5%, hence the title of the article. This, as we shall see shortly, was an error, albeit one that was difficult to detect. I have subsequently also noted another publication that used 16.5% quoting FAO Gleam 2.0 as the source of that figure – this was the Chatham House report on Food system impacts on biodiversity loss (Benton et. al. 2021) published in February 2021. Whether or not they made the same error is presently unclear.
In the article itself I do nevertheless express some suspicion around this new estimate due to it being old data and very likely an insufficient accounting of the land-use contribution of animal agriculture to emissions. So certainly the article, as I will explain in a moment, should not now be seen as a call to instigate 16.5% as a new percentage, but as an exposure of the unsatisfactory nature of the debate and some of the less than transparent practices of the FAO, as well as a critique of the longevity of 14.5% as a much cited figure.
The issue of GWP incommensurability
So how was the error made? It relates to the issue of global warming potential (GWP) weightings that are applied to methane and nitrous oxide. The quick explanation for non-experts is that methane and nitrous oxide have a more powerful warming effect than CO2 and therefore to arrive at a figure of CO2 equivalence different weightings are applied to methane and nitrous oxide (and several other gases too). However, there is consistent disagreement amongst climate scientists over the levels of these weightings, and which values that are chosen, tend to fluctuate over time. The IPCC for example has not been very good at agreeing upon a consistent GWP weighting for methane and nitrous oxide (two gases which figure prominently in emissions from animal agriculture). Incidentally weightings are usually calculated over a 100 year period (although some have argued that if you calculated them over a 20 year period it would mean more policy attention would have to be given to methane and nitrous oxide emissions, and thus to animal agriculture).
What is certainly true is that if you want to generate an accurate percentage figure the data comparison needs to be using the same GWP weightings, otherwise they will be incommmensurate. The first FAO report (Livestock’s Long Shadow) had made it’s GWP weightings commensurate with IPCC data from its 2001 Third Assessment Report (23 for methane, and 296 for nitrous oxide), also presumably used at that time by WRI CAIT who were the then source of total global emissions data for Livestock’s Long Shadow. The second FAO report (Tackling Climate Change through Livestock) had made it’s GWP weightings commensurate with IPCC data from its 2007 Fourth Assessment Report (25 for methane and 298 for nitrous oxide). My assumption was that GLEAM 2.0 data was also commensurate with IPCC data but the evidence on this is ambiguous at best, and this leads to a problem for arguing 16.5% as a new estimated percentage figure.
Thinking the data commensurate was reasonable given that the FAO website says “Three gases are considered in GLEAM: carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O). The latest available global warming potential from IPCC (2014) are used to convert all emissions into CO2 equivalent (298 for N2O and 34 for CH4)“. This implies commensurability with IPCC data but it’s not the case when you check the IPCC Fifth Assessment Report (2014) to which it refers. If you go to that link and look at the graph (Figure TS.1) on p.42 you can see a total global emissions figure of 49 Gt CO2 equiv. for the year 2010. This was the total that I used to generate the 16.5% figure. However this data were actually still using GWP weightings from the IPCC Second Assessment Report (1995), 21 for methane and 310 for nitrous oxide. It is perhaps a measure of the uncertainty of the IPCC over GWP weightings that its AR5 used more than one set of weightings data. And when the FAO claimed it was using the weightings of the IPCC AR5 this was not strictly true because the AR5 used more than one weighting.
For the IPCC AR5 also offers a total emissions figure for 2010 (not in Figure TS.1) of 52 Gt CO2 equiv.. Specifically it says, “Using the most recent GWP100 values from the AR5 [WGI 8.7] global GHG emissions totals would be slightly higher (52 GtCO2eq/yr)..”(p.45). When you go to the IPCC AR5 Chapter 8 on Anthropogenic and Natural Radiative Forcing, which includes the data in table WGI 8.7 (p.714), you find a table with several new GWP weightings. The table introduces a further level of complexity because it lists two weightings for methane and nitrous oxide, one with and one without inclusion of climate–carbon feedbacks. The weightings for methane and nitrous oxide with climate carbon feedbacks are listed as 34 and 298, and without climate carbon feedbacks, 28 and 265. As you’ll recall, the 34 and 298 weightings are those of the FAO Gleam 2.0 analysis.
So what are climate-carbon feedbacks? Climate-carbon feedback “refers to the effect that a changing climate has on the carbon cycle, which impacts atmospheric CO2, which in turn changes further the climate” (Gasser et. al. 2017: 236), and the IPCC AR5 took the position that “it is likely that including the climate–carbon feedback for non-CO2 gases as well as for CO2 provides a better estimate of the metric value than including it only for CO2” (p.714). Research since IPCC AR5 has attempted to clarify the issue of GWP100 weightings on the issue of climate-carbon feedback (Gasser et. al. 2017).
The FAO’s vanishing percentage figure of 15.6%
Here is where things get further interesting. As I have just indicated, there seems to be a GWP commensurability between GLEAM 2.0 weightings and the IPCC AR5 weightings inclusive of climate carbon feedbacks (both using 34 and 298 for methane and nitrous oxide respectively). It is possible then that these are the values the FAO meant when they claimed in GLEAM 2.0 that they had used the latest GWP weightings from IPCC AR5 (even though, as I have highlighted, IPCC AR5 uses several different GWP weightings in different places).
Using the internet archive, which shows older versions of websites, it is possible to see that for a while FAO GLEAM 2.0 did in fact conclude that emissions from animal agriculture constituted 15.6% of all greenhouse gas emissions (an update to the 14.5% figure), see here the sentence “Total GHG emissions from livestock supply chains are estimated at 8.1 gigatonnes CO2-eq per annum for the 2010 reference period. That amount represents 15.6 percent of all human-induced emissions, estimated at 52 gigatonnes CO2-eq for the year 2004 (IPCC, 2014)“. However, this sentence was subsequently removed from the FAO website, apparently due to uncertainties over GWP commensurability, although no public explanation was provided. The sentence was just removed.
Emissions from Animal Agriculture – not presently calculable
What can we conclude from all of this? I think three things. The FAO GLEAM 2.0 managed to produce an analysis of emissions from animal agriculture that was ultimately incommensurable with IPCC data and so could not constitute an adequate sequel to Livestock’s Long Shadow (2006), or Tackling Climate Change Through Livestock (2013) in the sense of generating a new estimated percentage figure. Secondly, the FAO lacked transparency in terms of its process, for example, wrongly implying commensurability, and removing the estimated percentage figure of 15.6% without explanation. Thirdly, the conclusion must be that something resembling an accurate estimated percentage for GHG emissions from animal agriculture is not presently calculable (at least using the FAO data) due to the incommensurability issue.
This is itself a significant finding for the climate science and policy community: we do not currently have an accurate estimate of percentage contribution for one of the leading and most contentious causes of the climate crisis. It would be better to be honest about this rather than allowing unknowing researchers or journalists to continue to cite 14.5% (or even still 18%). And as my article makes clear, FAO authors were still using 14.5% after the GLEAM 2.0 analysis had been published.
This ‘not presently calculable’ situation could change when the FAO report new data but will clearly depend on that data being commensurable with IPCC (or other reputable) data. The percentage figure should be resolved on the basis of considerably more recent data than 2010 so as to better inform policymakers. The points in my article which are critical of the lack of transparency from the FAO are strengthened by this amendment. The further critical points pertaining to the longevity and influence of the 14.5 percent figure, and possible framing/epistemological bias from the FAO remain unaffected by this update.
I gratefully acknowledge the input from Daniel Braune, Head of Research at ProVeg International, who brought both the points about incommensurability and the FAO’s brief assertion of the 15.6 percent figure to my attention.