Editing IPCC:AR6/WGIII/Chapter-2 (section)

=== 2.2.1 Uncertainties in GHG Emissions ===

<div id="h2-1-siblings" class="h2-siblings"></div>

Estimates of historical GHG emissions – CO 2 , CH 4 , N 2 O and F-gases – are uncertain to different degrees. Assessing and reporting uncertainties is crucial in order to understand whether available estimates are sufficiently robust to answer policy questions – for example, if GHG emissions are still rising, or if a country has achieved an emission reduction goal ( [[#Marland--2008|Marland 2008]] ). These uncertainties can be of scientific nature, such as when a process is not sufficiently understood. They also arise from incomplete or unknown parameter information (e.g., activity data, or emission factors), as well as estimation uncertainties from imperfect modelling techniques. There are at least three major ways to examine uncertainties in emission estimates ( [[#Marland--2009|Marland et al. 2009]] ): (i) by comparing estimates made by independent methods and observations (e.g., comparing atmospheric measurements with bottom-up emissions inventory estimates) ( [[#Saunois--2020|Saunois et al. 2020]] ; [[#Petrescu--2020|Petrescu et al. 2020]] a and 2020b; Tian et al. 2020); (ii) by comparing estimates from multiple sources and understanding sources of variation ( [[#Macknick--2011|Macknick 2011]] ; [[#Andres--2012|Andres et al. 2012]] ; [[#Andrew--2020|Andrew 2020]] ; [[#Ciais--2021|Ciais et al. 2021]] ); and (iii) by evaluating estimates from a single source ( [[#Hoesly--2018|Hoesly and Smith 2018]] ), for instance via statistical sampling across parameter values (e.g., [[#Monni--2007|Monni et al. 2007]] ; Robert J. [[#Andres--2014|Andres et al. 2014]] ; [[#Tian--2019|Tian et al. 2019]] ; [[#Solazzo--2021|Solazzo et al. 2021]] ).

Uncertainty estimates can be rather different depending on the method chosen. For example, the range of estimates from multiple sources is bounded by their interdependency; they can be lower than true structural plus parameter uncertainty, or than estimates made by independent methods. In particular, it is important to account for potential bias in estimates, which can result from using common methodological or parameter assumptions, or from missing sources (systemic bias). It is further crucial to account for differences in system boundaries – that is, which emissions sources are included in a dataset and which are not, otherwise direct comparisons can exaggerate uncertainties ( [[#Macknick--2011|Macknick 2011]] ; [[#Andrew--2020|Andrew 2020]] ). Independent top-down observational constraints are, therefore, particularly useful to bound total emission estimates, but are not yet capable of verifying emission levels or trends ( [[#Petrescu--2021a|Petrescu et al. 2021a]] , 2021b). Similarly, uncertainties estimates are influenced by specific modelling choices. For example, uncertainty estimates from studies on the propagation of uncertainties associated with key input parameters (activity data, emissions factors) following the IPCC Guidelines (IPCC 2006) are strongly determined by assumptions on how these parameters are correlated between sectors, countries, and regions ( [[#Janssens-Maenhout--2019|Janssens-Maenhout et al. 2019]] ; [[#Solazzo--2021|Solazzo et al. 2021]] ). Assuming (full) covariance between source categories, and therefore dependence between them, increases uncertainty estimates. Estimates allowing for some covariance as in Sollazo et al. (2021) also tend to yield higher estimates than the range of values from ensemble of dependent inventories ( [[#Saunois--2016|Saunois et al. 2016]] , 2020).

For this report, a comprehensive assessment of uncertainties is provided in the Supplementary Material (2.SM.2) to this chapter based on [[#Minx--2021|Minx et al. (2021)]] . The uncertainties reported here combine statistical analysis, comparisons of global emissions inventories and an expert judgement of the likelihood of results lying outside a defined confidence interval, rooted in an understanding gained from the relevant literature. This literature has improved considerably since AR5, with a growing number of studies that assess uncertainties based on multiple lines of evidence ( [[#Saunois--2016|Saunois et al. 2016]] , 2020; Tian et al. 2020; [[#Petrescu--2021a|Petrescu et al. 2021a]] , 2021b).

To report the uncertainties in GHG emissions estimates, a 90% confidence interval (5th–95th percentile) is adopted – that is, there is a 90% likelihood that the true value will be within the provided range if the errors have a Gaussian distribution, and no bias is assumed. This is in line with previous reporting in IPCC AR5 (Blanco et al. 2014; Ciais et al. 2014). Note that national emissions inventory submissions to the UNFCCC are requested to report uncertainty using a 95% confidence interval. The use of this broader uncertainty interval implies, however, a relatively high degree of knowledge about the uncertainty structure of the associated data, particularly regarding the distribution of uncertainty in the tails of the probability distributions. Such a high degree of knowledge is not present over all regions, emission sectors and species considered here.

Based on the assessment of relevant uncertainties above, a constant, relative, global uncertainty estimates for GHGs are applied at a 90% confidence interval that range from relatively low values for CO 2 -FFI (±8%), to intermediate values for CH 4 and F-gases (±30%), to higher values for N 2 O (±60%) and CO 2 -LULUCF (±70%). Uncertainties for aggregated total GHG emissions in terms of CO 2 -eq emissions are calculated as the square root of the squared sums of absolute uncertainties for individual gases (taking F-gases together), using GWP100 to weight emissions of non-CO 2 gases but excluding uncertainties in the metric itself.

This assessment of uncertainties is broadly in line with AR5 WGIII (Blanco et al. 2014), but revises individual uncertainty judgements in line with the more recent literature ( [[#Saunois--2016|Saunois et al. 2016]] , 2020; [[#Janssens-Maenhout--2019|Janssens-Maenhout et al. 2019]] ; [[#Friedlingstein--2020|Friedlingstein et al. 2020]] ; Tian et al. 2020; [[#Solazzo--2021|Solazzo et al. 2021]] ) as well as the underlying synthetic analysis provided here (e.g., Figures 2.2 and 2.3 in this chapter; and [[#Minx--2021|Minx et al. 2021]] ). As such, reported changes in these estimates do not reflect changes in the underlying uncertainties, but rather a change in expert judgement based on an improved evidence base in the scientific literature. Uncertainty estimates for CO 2 -FFI and N 2 O remain unchanged compared to AR5. The change in the uncertainty estimates for CH 4 from 20% to 30% is justified by larger uncertainties reported for EDGAR emissions ( [[#Janssens-Maenhout--2019|Janssens-Maenhout et al. 2019]] ; [[#Solazzo--2021|Solazzo et al. 2021]] ) as well as the wider literature ( [[#Kirschke--2013|Kirschke et al. 2013]] ; [[#Tubiello--2015|Tubiello et al. 2015]] ; [[#Saunois--2016|Saunois et al. 2016]] , 2020). As AR6 – in contrast to AR5 – uses CO 2 -LULUCF data from global bookkeeping models, the respective uncertainty estimate is based on the reporting in the underlying literature ( [[#Friedlingstein--2020|Friedlingstein et al. 2020]] ) as well as Working Group I (Canadell et al. 2021). The 70% uncertainty value is at the higher end of the range considered in AR5 (Blanco et al. 2014).

Finally, for F-gas emissions top-down atmospheric measurements from the 2018 World Meteorological Organization’s (WMO) Scientific Assessment of Ozone Depletion (Engel and Rigby 2018; [[#Montzka--2018|Montzka and Velders 2018]] ) are compared to the data used in this report ( [[#Crippa--2021|Crippa et al. 2021]] ; [[#Minx--2021|Minx et al. 2021]] ) as shown in Figure 2.3. Due to the general absence of natural F-gas fluxes, there is a sound understanding of global and regional F-gas emissions from top-down estimates of atmospheric measurements with small and well-understood measurement, lifetime and transport model uncertainties (Engel and Rigby 2018; [[#Montzka--2018|Montzka and Velders 2018]] ). However, when species are aggregated into total F-gas emissions, EDGARv6.0 emissions are around 10% lower than the [[#WMO--2018|WMO 2018]] values throughout, with larger differences for individual F-gas species, and further discrepancies when comparing to older EDGAR versions. Based on this, the overall uncertainties for aggregate F-gas emissions is judged conservatively at 30% – 10 percentage points higher than in AR5 (Blanco et al. 2014).

<div id="_idContainer012" class="Basic-Text-Frame"></div>

[[File:5b9a37d6792965b3d0d736bd32459e16 IPCC_AR6_WGIII_Figure_2_3.png]]

'''Figure 2.3''' '''|''' '''Comparison between top-down estimates and bottom-up EDGAR inventory data on GHG emissions for 1980–2016. Left panel:''' Total GWP100-weighted emissions based on IPCC AR6 (Forster et al. 2021a) of F-gases in Olivier and Peters (2020) [EDGARv5FT] (dark-red dotted line, excluding C 4 F 10 , C 5 F 12 , C 6 F 14 and C 7 F 16 ) and EDGARv6 (bright red dashed line) compared to top-down estimates based on AGAGE and NOAA data from [[#WMO--2018|WMO (2018)]] (blue lines; Engel and Rigby (2018); [[#Montzka--2018|Montzka and Velders (2018)]] ). '''Right panel:''' Top-down aggregated emissions for the three most abundant CFCs (–11, –12 and –113) and HCFCs (–22, –141b, –142b) not covered in bottom-up emissions inventories are shown in dark blue and yellow. For top-down estimates the shaded areas between two respective lines represent 1 σ uncertainties. Source: [[#Minx--2021|Minx et al. (2021)]] .

Aggregate uncertainty across all GHGs is approximately ±11% depending on the composition of gases in a particular year. AR5 applied a constant uncertainty estimates of ±10% for total GHG emissions. The upwards revision applied to the uncertainties of CO 2 -LULUCF, CH 4 and F-gas emissions therefore has a limited overall effect on the assessment of GHG emissions.

GHG emissions metrics such as GWP100 have their own uncertainties, which has been largely neglected in the literature so far. [[#Minx--2021|Minx et al. (2021)]] report the uncertainty in GWP100 metric values as ±50% for methane and other short-lived climate forcers (SLCFs), and ±40% for non-CO 2 gases with longer atmospheric lifetimes (specifically, those with lifetimes longer than 20 years). If uncertainties in GHG metrics are considered, and are assumed independent (which may lead to an underestimate) the overall uncertainty of total GHG emissions in 2019 increases from ±11% to ±13%. Metric uncertainties are not further considered in this chapter, but are referred to in Cross-Chapter Box 2 in this chapter, and [https://www.ipcc.ch/report/ar6/wg3/chapter/chapter-2 Chapter 2] Supplementary Material on GHG metrics (2.SM.3).

The most appropriate metric to aggregate GHG emissions depends on the objective (Cross-Chapter Box 2). One such objective can be to understand the contribution of emissions in any given year to warming, while another can be to understand the contribution of cumulative emissions over an extended time period to warming. In Figure 2.4 the modelled warming from emissions of each gas or group of gases is also shown – calculated using the reduced-complexity climate model Finite Amplitude Impulse Response (FaIR) model v1.6, which has been calibrated to match several aspects of the overall WGI assessment (Forster et al. 2021a; specifically Cross-Chapter Box 7 in [[IPCC:Wg3:Chapter:Chapter-10|Chapter 10]] therein). Additionally, its temperature response to emissions with shorter atmospheric lifetimes such as aerosols, methane or ozone has been adjusted to broadly match those presented in [[#Szopa--2021a|Szopa et al. (2021a)]] . There are some differences in actual warming compared to the GWP100 weighted emissions of each gas (Figure 2.4), in particular a greater contribution from CH 4 emissions to historical warming. This is consistent with warming from CH 4 being short-lived and hence having a more pronounced effect in the near-term during a period of rising emissions. Nonetheless, Figure 2.4 highlights that emissions weighted by GWP100 do not provide a fundamentally different information about the contribution of individual gases than modelled actual warming over the historical period, when emissions of most GHGs have been rising continuously, with CO 2 being the dominant and CH 4 being the second most important contributor to GHG-induced warming. Other metrics such as GWP* (or GWP star) ( [[#Cain--2019|Cain et al. 2019]] ) offer an even closer resemblance between cumulative CO 2 -eq emissions and temperature change. Such a metric may be more appropriate when the key objective is to track temperature change when emissions are falling, as in mitigation scenarios.

<div id="_idContainer014" class="Basic-Text-Frame"></div>

[[File:59c543168b88b39268bec2d454e47046 IPCC_AR6_WGIII_Figure_2_4.png]]

'''Figure 2.4''' '''|''' '''Contribution of different GHGs to global warming over the period 1750 to 2018.''' Top row: contributions estimated with the FaIR reduced-complexity climate model. Major GHGs and aggregates of minor gases as a timeseries in '''(a)''' and as a total warming bar chart with 90% confidence interval added in '''(b)''' . Bottom row: contribution from short-lived climate forcers as a time series in '''(c)''' and as a total warming bar chart with 90% confidence interval added in '''(d)''' . The dotted line in (c) gives the net temperature change from short-lived climate forcers other than CH 4 . F-Kyoto/Paris includes the gases covered by the Kyoto Protocol and Paris Agreement, while F-other includes the gases covered by the Montreal Protocol but excluding the HFCs. Source: [[#Minx--2021|Minx et al. (2021)]] .

<div id="cross-chapter-box-2" class="h2-container box-container"></div>

<span id="cross-chapter-box-2-ghg-emissions-metrics"></span>