Methane emissions: choosing the right climate metric and time horizon & Impacts CRITICAL REVIEW

ab Methane is a more potent greenhouse gas (GHG) than CO 2 , but it has a shorter atmospheric lifespan, thus its relative climate impact reduces signi ﬁ cantly over time. Di ﬀ erent GHGs are often con ﬂ ated into a single metric to compare technologies and supply chains, such as the global warming potential (GWP). However, the use of GWP is criticised, regarding: (1) the need to select a timeframe; (2) its physical basis on radiative forcing; and (3) the fact that it measures the average forcing of a pulse over time rather than a sustained emission at a speci ﬁ c end-point in time. Many alternative metrics have been proposed which tackle di ﬀ erent aspects of these limitations and this paper assesses them by their key attributes and limitations, with respect to methane emissions. A case study application of various metrics is produced and recommendations are made for the use of climate metrics for di ﬀ erent categories of applications. Across metrics, CO 2 equivalences for methane range from 4 – 199 gCO 2eq. /gCH 4 , although most estimates fall between 20 and 80 gCO 2eq. /gCH 4 . Therefore the selection of metric and time horizon for technology evaluations is likely to change the rank order of preference, as demonstrated herein with the use of natural gas as a shipping fuel versus alternatives. It is not advisable or conservative to use only a short time horizon, e.g. 20 years, which disregards the long-term impacts of CO 2 emissions and is thus detrimental to achieving eventual climate stabilisation. Recommendations are made for the use of metrics in 3 categories of applications. Short-term emissions estimates of facilities or regions should be transparent and use a single metric and include the separated contribution from each GHG. Multi-year technology assessments should use both short and long term static metrics ( e.g. GWP) to test robustness of results. Longer term energy assessments or decarbonisation pathways must use both short and long-term metrics and where this has a large impact on results, climate models should be incorporated. Dynamic metrics o ﬀ er insight into the timing of emissions, but may be of only marginal bene ﬁ t given uncertainties in methodological assumptions. Methane emissions are a key contributor to climate change but have a substantially di ﬀ erent impact on global warming than carbon dioxide: methane has a much high radiative e ﬃ ciency but is relatively short-lived. Consequently, the use of Global Warming Potentials over a single 100 year time frame has been frequently called into question as it hides the substantial variation in impact over time. This study compares a comprehensive range of di ﬀ erent climate metrics and their key qualities to provide an insight on which metric and time horizon is most appropriate for use in di ﬀ erent applications.


Introduction
Methane emissions are the second largest contributor to climate change next to carbon dioxide, with its direct impact representing around 20% of additional climate forcing since 1750 according to the Saunois et al. 1 Further, the estimated direct and indirect forcing effects of methane (including oxidation to CO 2 and impact on ozone creation) is estimated to be 58% of the value of CO 2 (0.97 W m À2 for methane compared to 1.68 W m À2 for CO 2 ). 2 Annual emissions are only 3% w/w of those associated with CO 2 (0.56 GtCH 4 /year vs. 14.5 GtCO 2 /year for methane and CO 2 respectively), 1,3 but methane has a radiative forcing approximately 120 times more than CO 2 immediately aer it is emitted. On the other hand, methane has a perturbation life of only 12.4 years, 2 whereas CO 2 lasts in the atmosphere for much longer: 50% of an emission is removed from the atmosphere within 37 years, whilst 22% of the emission effectively remains indenitely. 4 Consequently, the relative impact of methane compared to CO 2 changes over time.
Global warming potentials (GWP) are used to compare the relative impact of different greenhouse gases (GHGs) on climate forcing, by converting emissions into 'CO 2 equivalents'. It is dened as the average (time-integrated) radiative forcing of a pulse emission over a dened time horizon, compared to CO 2 . GWP is used widely across industrial, regulatory and academic applications to compare the effect of a change in product or process. The 100 year time horizon is most common, giving a CO 2 equivalent value of 28-36 for methane (depending on whether various indirect climate effects are included). 2 However, there is much criticism about the use of GWP, because: The selected time horizon has a large impact on the value of the metric; Despite its name, it does not compare gases against their effect on global temperature; Measures an average climate forcing effect of a single pulse emission over time but gives no indication of the climate impact at an end-point in time, or that of a sustained emission.
Increasingly there are calls for the use of different time horizons (e.g. 20 years) or even different metrics that better reect climate change or align with climate targets (e.g. the global temperature change potential as described in the IPPC AR5 2 ). But which metric is most appropriate for different applications and over what time horizon?
Previous studies have assessed the impacts of a small selection of alternative metrics on natural gas versus coal for electricity 5 and the climate impacts of transportation. 6 Deuber et al. 7 and Johansson 8 examine the physical basis and relationship between some metrics, whilst others assess the cost of emissions mitigation using different metrics. 9, 10 Mallapragada and Mignone 11 classify a selection of metrics based on some key characteristics and apply metrics to a case study of natural gas versus gasoline-fuelled vehicles. This paper goes further by assessing a large suite of climate metrics regarding their key differentiating characteristics and applies a case study technology assessment to demonstrate the impact of metric selection on technology preference. The study makes recommendations for which metrics and time horizons are most appropriate for different applications, including short term regional emissions estimates, life cycle technology assessments and energy systems pathways.
The contribution this paper makes is to provide insight for industry, policy makers and academics to ensure the appropriate use of metrics. A range of metric values and methods are presented and synthesised, and clear guidelines are given for the use of metrics across different applications.
First, the report describes the procedure for assessment for the climate metrics. Section 3 gives a summary of the climate impact of GHGs and methane in the atmosphere. Section 4 describes the global warming potential metric, including its history and limitations. Alternative metrics are dened in the following Section 5 and key differences and factors that affect the choice of metrics are outlined in Section 6. Evidence around the impact of using the various metrics are described in Section 7, before recommendations and conclusions are made.

Assessment methods
Given the purpose of this study is to assess the impact of using different climate metrics and to make recommendations for their use in different applications, the following stages of assessment are undertaken: Contextualising the climate cause-effect chain. Assessing climate metrics and key characterising factors. Applying a case study. To place the analysis of different climate metrics in context, the study rst describes the climate cause-effect chain, against which metrics will be categorised and assessed. Methane is the focus of this study and is explained in this context, but it should be noted that the assessment is applicable for the study of other emissions and environmental impacts.
A review of a full suite of proposed climate change metrics is then carried out. Firstly, the standard GWP metric is dened and characterised relating to its physical basis, methodological construction and associated uncertainty. Alternative metrics are synthesised from a wide body of literature and compared against GWP and each other, relating to their 'CO 2 equivalent' quantities as well as their basis for construction, intuitiveness and associated uncertainty. Key characteristics are developed and analysed against typical applications of each metric. Characteristics considered are: The time horizon or associated discount rates; The physical/economic basis of the metric; Static versus dynamic metrics; The level of uncertainty versus tangibility; and The suitability of metrics for different applications. To demonstrate the impact of the broad range of metrics and CO 2 equivalent values, a case study is given: a climate assessment of the use of LNG as a shipping fuel, against alternative fuels. The case study is based on the outputs of a full environmental assessment, but focuses on the change in rank preference of fuel based on different CO 2 equivalents, as well as the use of dynamic versus static metrics.
Different applications of metrics from industry, policy and academic are characterised in terms of factors such as their required simplicity and their time-frames of consideration. From this, a series of recommendations for the use of metrics are made, which may serve as guidelines for further discussion.

Greenhouse gases and the climate cause-effect chain
The link between GHG emissions, climate change and damage to human health and ecosystems is multifaceted. Fig. 1 illustrates a simplied cause-effect chain linking emissions with climate change-related damage, and later in this report the metrics will be placed in this context. Firstly, a GHG is emitted, which increases the concentration of this GHG in the atmosphere. Each GHG has a radiative efficiency, which is the capacity of an atmospheric concentration of gas to trap and reradiate heat downwards, measured in W m À2 ppb À1 . 2 When multiplied by the atmospheric concentration, this gives the total radiative forcing attributed to the GHG. Thus, radiative forcing is the total change in heat balance in the atmosphere from the increase in concentration of a greenhouse gas, 5 measured in W m À1 . 12 An increase in radiative forcing results in a temperature increase, where the degree of temperature rise is governed by the magnitude of emission and radiative efficiency, as well as the existing atmospheric concentration of the GHG and the concentrations of other gases in the atmosphere. The increase in global average temperature causes damage via increased extreme weather events, sea level rise, oceanic circulation changes, species extinction and more. This damage is likely to increase faster than the rate of change in global temperature. 13 Two important points require emphasis. First, increased radiative forcing is not the same as temperature increase. Temperature change is a result of increased forcing, but the value of temperature change is governed by other factors as well. There is also a lag between radiative forcing and temperature change of approximately 15-20 years, 14 as shown in Fig. 2. Second, global average temperature change is not the only indicator that may describe climate change. Other important factors describe climate change, including the rate of temperature rise and the cumulative temperature rise. Each of these climate change attributes are interrelated but cause damage to health and ecosystems in different ways, examples of which are described in Table 1. The global average temperature rise increases the variation and volatility of temperatures and results in more extreme weather events. The rate of temperature increase governs how much time species may take to adapt to new conditions and so a fast rate will cause more species extinction. The cumulative temperature rise (i.e. prolonged increases) strongly affects longer term changes such as glacial melt and seal level rise. Emissions of GHGs affect each of these climate attributes differently, depending on: emission quantity; existing concentration of pollutant in the atmosphere; residence time of emission in the atmosphere; and the concentration of other molecules in atmosphere (e.g. OH À and O 3 ).
For methane, an emission has a much larger radiative forcing effect than CO 2 given the difference in radiative efficiency and indirect impacts. 4 However, methane is a short-lived climate pollutant (SLCP) and has an atmospheric lifetime of 8.4 years, dened as the atmospheric burden divided by the sink strength. 15 Methane comes out of the atmosphere and troposphere by typically reacting with hydroxyl radicals, oxidising to form CO 2 and water (which are also both greenhouse gases). 88% of the methane reacts this way, meaning that one gram of methane will form 2.4 grams of CO 2 . 13 The other 12% of the methane forms molecules such as methanal (formaldehyde) and methyl hydroperoxide. The increasing concentration of methane in the atmosphere reduces the availability of the hydroxyl radicals for further reactions which in turn would increase the lifespan of methane. Thus, the perturbation lifetime of methane, which allows for the gases inuence on other atmospheric species during its life, is 12.4 years. 2 In comparison, the lifespan of CO 2 is more complicated due to the different mechanisms that take CO 2 out of the atmosphere, but 50% of a pulse emission is removed from the atmosphere within 37 years, whilst 22% of the emission effectively remains indenitely. 4 Thus, whilst the initial radiative forcing is low compared to methane, the lasting and cumulative effects are large. The change in radiative forcing over time is shown in Fig. 3 for methane and CO 2 .
The effect of GHG emissions on the climate is multifaceted and detailed climate models are required to understand the effects of changing emissions and the environment over time. Such models as MAGICC6 17 are used in integrated assessment projects to estimate the impacts. However, these are detailed Fig. 1 The cause-effect chain linking greenhouse gas emissions to climate change-related damage. Fig. 2 The relative impact of a pulse emission of methane on radiative forcing and subsequent impact on temperature change. Source: ref. 14.

Global warming potential
Global warming potential (GWP) is the standard metric used to compare GHGs emitted from different products and services. The metric was developed for use following the Kyoto Protocol and adapted by the Intergovernmental Panel on Climate Change 18 to help in the design of emissions strategies, accounting for the trade-offs between different types of GHG. 19 It is dened as the time-integrated radiative forcing of an emission pulse of a gas, relative to that of CO 2 , over a dened time horizon. For a 100 year time horizon, methane GWP is 36 gCO 2eq. / gCH 4 , meaning that the average radiative forcing of a methane emission over 100 years aer the emission is 36 times that of an equivalent mass of CO 2 . The IPCC have typically given estimates of GWP for time horizons of 20, 100 and 500 years (although the most recent 5 th assessment report excluded 500 years) and the 100 year GWP (GWP100) remains the most common metric used.
With a high radiative efficiency and short lifetime compared to CO 2 , methane has a much higher GWP over short timescales: GWP20 is 87 gCO 2eq. /gCH 4 . Fig. 4 shows the GWP of methane over different timescales, but not including the effect of climatecarbon feedback (CCFB), resulting in slightly lower numbers than those expressed within this paragraph (e.g. a GWP100 of 30 rather than 36).
The values of GWP for each GHG have been developed over each IPCC assessment report, to account for better understanding of radiative forcing and the various indirect radiative forcing effects, such as cloud albedo and CCFB. 2,21 CCFB is a broad term that encompasses both negative and positive feedback effects associated with increased forcing or temperature. For example, a positive feedback is an increase in temperature causing greater concentrations of water vapour, which itself results in further radiative forcing. The cloud albedo effect is the impact of clouds reecting radiation and contributing to climate cooling. The concentration of GHGs in the atmosphere and troposphere has an impact on cloud formation and consequently the cloud albedo effect. Additionally, most atmospheric methane eventually oxidises into CO 2 , which raises the total GWP values by 1 and 2 for 20 and 100 year time horizons, respectively. This is summarised in Table 2, presenting the change in GWP for methane across IPCC publications.  Additionally, indirect effects have been inconsistently included in historical IPCC publications. In the second and third assessment reports calculations of GWP did not include CCFB. In the fourth assessment report, CCFB were included in the calculation of CO 2 absolute global warming potential (AGWP), the baseline against which the GWP for other gases is based. However, while CCFB also impacts on the radiative forcing of other gasses, these impacts were not included in the GWP calculations until AR5, which results in a large increase, especially for the 100 year horizon GWP, as shown in Table 2.

Criticism of GWP
There are a number of criticisms levelled at the use of GWPs relating to the three key aspects of this metric: a time horizon must be set; it is modelled on a single pulse emission; and it measures time-integrated radiative forcing.
First, the need to select a time horizon requires the metric user to decide a timeframe that is important. This is a particular issue for methane given that the GWP values change so signicantly over time. The selection of a single time horizon is arbitrary and means that other timeframes are disregarded: selection of a short timeframe for methane will ignore the long-term impacts of CO 2 , whereas selection of a long timeframe for methane will largely ignore the short term forcing of methane. Indeed, the fact that any time horizon is set means that longer term impacts are systematically underrepresented.
Second, the GWP was designed to equate pulse emissions, i.e. one-off emissions, rather than sustained or developing emissions, such as those modelled using life cycle assessment methods. This does not generally reect the consequences of real-world investment or policy decisions. 12 Last, the physical basis of the GWP is the integrated radiative forcing and does not represent the temperature (or other climate) impact. As described in Section 3, radiative forcing is a precursor to temperature change, but they are not synonymous. Additionally, the fact that GWP is based on an integrated measure means that the GWP indicates the average impact over a time horizon rather than the impact at the end-point of the time horizon (both are useful in estimating the impacts of climate change).
The limitations associated with GWP have given rise to the creation of alternative climate metrics over the last 20 years. These metrics are dened in the following section, aer which their key differentiating factors are discussed in Section 6, including time horizons and physical basis.

Alternative metrics
The many climate metrics that have been proposed in the last few decades can be categorised in a number of ways, which are summarised in Table 3. Table 3 lists the most cited metrics and categorises them based on key factors: CO 2 equivalency value, their physical basis, whether they are static or dynamic metrics, cumulative or end-point estimates, and their level of uncertainty. The following section rstly describes the most used alternative, GTP, before outlining the characteristics of each other metric in order that they appear in the table.

GTPglobal temperature change potential
Global temperature change potential (GTP) is the most popular and most researched alternative climate metric to GWP. 2 It was developed by Shine et al. 24,32 and is included in the IPCC Assessment Reports. It is dened as the change in mean surface temperature aer a specied time due to a pulse emission, relative to the effect from an equivalent pulse emission of CO 2 . The key differences compared to the GWP are: It is an end-point metric, 11 measuring the impact at the end of a time period, rather than a cumulative effect within a time period; and It estimates the effect on temperature, rather than radiative forcing (which gives rise to temperature but the relationship is not linear).
Values of GTP for methane are currently estimated as 13 gCO 2eq. /gCH 4 (GTP100) and 71 (GTP20) including an allowance for CCFB and the eventual oxidation of methane into CO 2 . Whilst the GTP20 is around 20% lower than the equivalent GWP20 (87), the 100 year time horizon differs greatly, over 60% lower than GWP, as shown in Fig. 5. This is because the GTP gure measures at the end-point and does not account for the strong forcing prior to this time. At 100 years the proportion of the pulse emission remaining in the atmosphere is relatively small. Indeed, at this time aer the emission, the dominant force is from only the indirect effects such as CCFB and methane oxidation (without which the GTP100 would be only 4). The GTP goes one step further down the cause-effect chain (see Fig. 8) than GWP by estimating the relative temperature change resulting from the increased radiative forcing. This brings more clarity when using the metric for temperature-based analyses (e.g. keeping global temperatures below 2 C). However, the estimation of GTP incorporates additional assumptions about physical processes, such as climate sensitivity and the exchange of heat between the atmosphere and the ocean. 2,24 This consequently brings more uncertainty compared to GWP. 4 The IPCC estimate an uncertainty of GTP100 of AE75% (with a 90% condence), compared to AE30% and AE40% for GWP20 and GWP100, respectively. 2 The 500 year value is not given in the most recent IPCC assessment report, so the gure presented is from the 4th assessment report. c The IGTP metric values are estimated to be 12% higher than equivalence GWP values and are thus calculated. The original estimation was based on the 4th assessment report values of the GWP.

SGWP À sustained-ux global warming potential
The sustained-ux global warming potential (SGWP) has been previously called the step-change global warming potential 4,34 and is designed to eliminate the dependence of the GWP metric on the single 'pulse' emission. This metric measures the relative radiative forcing of a sustained emission of a GHG relative to that of CO 2 . This metric is otherwise the same as GWP, but the sustained emission measurement results in a larger CO 2 equivalence and is 40% higher than GWP for the 100 year horizon. 4

ICI and CCIinstantaneous and cumulative climate impact
Edwards and Trancik 16 developed a new set of metrics in 2014, intended to be a simplied dynamic method to account for changing emissions proles over time, in order to assist with development of effective emissions pathways. Instantaneous climate impact (ICI) measures the radiative forcing associated with emissions at a specic time point, similar to an instantaneous version of GWP. It is dynamic in that the time horizon end-point is xed, rather than the time period aer an emission (further explained in Section 6). Consequently, in a multi-year emissions assessment (e.g. a life cycle assessment), as the year of emission increases, the time period decreases until the end time point is reached. The result is that any methane emissions incurred at the start of the time frame contributes relatively little, but the values increase signicantly as the emissions approach the end-point. The second of the set of impacts developed by Edwards and Trancik 16 is a cumulative version of the ICI, the CCI. As such, it measures the cumulative radiative forcing of an emission or emission prole. It is similar to the GWP in that it measures cumulative radiative forcing, but whereas the time horizon is xed with GWP (e.g. 100 years), the end point is xed with CCI (e.g. 2080). In other words, the CCI is a dynamic version of GWP. 11

TWPtechnology warming potential
Technology warming potential (TWP) is designed specically for comparing technologies or products over variable time and is classed as a dynamic metric. 12 TWP does not produce a CO 2 equivalency metric as such, but produces a 'technology equivalency', as it gives relative improvements (or otherwise) associated with technology switching over a time frame. It is dened as the relative proportional change in cumulative radiative forcing over different timescales and may be as a result of a pulse or sustained emission. 5 The effect is broadly similar to the ratio of GWPs associated with two different technologies, but the initial set-up of TWP did not allow for climate carbon feedbacks, suggesting that the methane impact may be underestimated in this metric. 5

IGTPintegrated global temperature change potential
The integrated global temperature change potential (IGTP) is a cumulative version of the GTP. Unlike the GTP which estimates the temperature impact of a pulse emission at a specic time, the IGTP estimates the cumulative temperature impact from the time of a pulse emission to a specic time horizon, relative to CO 2 . 6 In this respect, it is a temperature equivalent of the global warming potential. This means that IGTP values are higher than GTP, as the initial high radiative (and temperature) forcing is effectively 'remembered' in the cumulative time horizon estimates. 26,28 Values are approximately 12% higher than the GWP for the 20, 50, 100 and 500 year time horizons.

TEMPtemperature proxy index
The temperature proxy index (TEMP) was developed by Tanaka et al. 29 in 2009 to provide a temperature based equivalency metric similar to the GTP but integrated over a specic time horizon (similar to the IGTP). Instead of a projected impact metric derivation such as the GWP, TEMP values are numerically estimated based on the historical contribution of different GHGs over the post-industrial time period. 30 The TEMP metrics and analysis suggest that GWP100 underestimates the contribution from methane and that a value of 39 would be most appropriate (which is not dissimilar to the current GWP100 value of 36 including carbon climate feedbacks and oxidation to CO 2 ).

CCIPclimate change impact potential
The climate change impact potential (CCIP) metric was created by Kirschbaum 14 in 2014 and is the only mid-point type metric that combines the effects of temperature rise with cumulative warming as well as rate of warming. Key assumptions associated with this metric are that each impact (temperature, cumulative temperature and rate of rise) are weighted equally in importance and the values are only available for 100 year time horizon, which is similar to the GWP100 at 32 gCO 2eq. /gCH 4 . This is a unique metric in its attempt to incorporate the different types of climate impact. If there were a specic calculator that allowed the selection of weighting and time horizon to generate the appropriate CO 2 equivalence, this would be a useful bridge between simple static metrics and more complicated climate models.

GSP and IGSPglobal sea level rise potential
The global sea level rise potential was developed in 2014 and goes a step further than the temperature impacts of emission by estimating the specic impact on sea level rise. 28 It is a static metric based on a set time horizon, estimating the relative change in sea level at the end of the time horizon. The values for 20, 100 and 500 year time horizons lie between those associated with GWP and GTP for methane, at 78, 18, 3.8 gCO 2eq. /gCH 4 respectively. 28 The relative uncertainty associated with GSP is likely to be higher than GWP or GTP as it is further in the line of damage estimation (see Fig. 8). However, this is still a physical metric with no required socio-economic evaluation, unlike the GDP and GCP.
The IGSP is a cumulative version of the GSP, similar to the GWP but estimating average sea level impacts. The metric values for IGSP are slightly higher than those of GWP at 95, 39 and 11 gCO 2eq. /gCH 4 for 20, 100 and 500 year horizons respectively.

GPPglobal precipitation change potential
Global precipitation change potential is a static equivalency metric created in 2015 that compares GHGs against their effect on global average change in precipitation, due to a pulse or a sustained emission. 30 The precipitation estimate over time uses both a radiative forcing element (GWP) and a temperature change element (GTP) and their relative impact changes over time. 26 Similar to the sea level rise metric, this metric goes further along the cause and effect chain, whilst still being physically based (rather than socio-economic). The metric values are higher than GWP and GTP values for the 20 year horizon (120) and slightly lower for the 100 year (8.1). This indicates that the effect of methane on global precipitation change is large in the short term, much larger than the temperature change impact.

GDPglobal damage potential
Global damage potential (GDP) goes beyond mid-point physical impacts to estimate the end-point damages caused by climate change, relating to human health, increased rates of mortality and ecosystem losses, which are aggregated using an economic value. 7 It is still an equivalency metric in that it estimates the relative damage impact of an emission compared to CO 2 and is based on the cumulative impact over time. The end-point economics-based metric removes the requirement to specify a timeframe by setting an innite horizon and setting a discount rate at which future emissions are discounted against near term emissions. Recently estimated GDP equivalences for methane are between 19 and 100 with a base case of 50 (with an additional outlier of 420, associated with high discount rate). 35 The estimation of an economic value on damage represents signicantly higher uncertainty than other mid-point metrics, owing to the additional assumptions that must be made to estimate: The damage caused by an increase in concentration (e.g. number of extreme weather events, sea level rise, extinction events); and The economic value placed on such damage. The GDP is an intuitively useful method to determine the least-cost mitigation strategy. 25 However, the move from a physical to economic basis and the high uncertainty reduces the transparency and useability of such a metric for many applications and it is typically utilised within an integrated climate-cost model framework. 2

GCPglobal cost potential
Global cost potential (GCP) is also an end-point economic metric and denes price ratios between GHGs and CO 2 that deliver the least-cost mitigation solutions to meet a specic climate target at a specic time. 2,27 Similar to the GDP, this metric is typically an output from a climate-economic model generating price ratios for different GHG mitigation options using an optimisation model 36 and are not normally used in carbon equivalency-related studies due to their complexity and dependence on system assumptions. Tanaka et al. 36 recently estimated GCP values that t with a 2 C climate target, resulting in a range of values from 5 to 65 gCO 2eq. /gCH 4 , with a peak at the time of stabilisation around 2060.

SCMsocial cost of methane
The social cost of methane (SCM) is another estimator of the economic costs of damage associated with methane. As indicated by the name, the damages focus on methane rather than the climate effect, as it includes damages associated with air quality and tropospheric ozone creation which has a large impact on crop yield and premature deaths. 13 Impacts are monetised and levelized per tonne of emission, and subsequently compared to the social cost of carbon. Instead of using specic time horizons, the time horizon is innite and a discount rate is set. Thus, instead of varying values over time horizons, they vary signicantly over discount rate: 10% discount rate equates to a CO 2 equivalency of 199; 5-102%; 4-76%; 2.5-42%; 1.4-26%. These values are higher than most other equivalency metrics, partly due to the incorporation of the damage effect of ozone creation.

The key factors that differentiate climate metrics
There are many important differentiating factors associated with the climate metrics, which are analysed below to inform recommendations for metric selection. The following section assesses metric in relation to: selecting the timeframe; static vs. dynamic metrics; the physical basis; level of uncertainty; simplicity vs. tangibility; and suitability for the application.

Selecting the timeframe
The need to select an appropriate timeframe is the most common criticism of the GWP and has the largest impact on metric value. This variation is shown in Fig. 6, giving equivalencies for different metrics for methane over different time horizons.
There is no single correct time horizon to use: it depends on the perspective and reason for which the estimation is being carried out. 11,26,[37][38][39] The IPCC typically uses a 100 year time horizon (GWP100), being commensurate with the scenario timescales used in its modelling work. However, 20 year time horizons are increasingly used, which can signicantly alter results, oen leading to disagreement and conicting conclusions in the literature. 12,40 Using a short-term metric inherently ignores the impact of long term, long-lived forcers (CO 2 ) and on a systems scale this means prolonging the point at which the globe reaches climate stabilisation. Conversely, a long-term metric inherently ignores the large impact of short-lived forcers (methane), which may cause more rapid temperature increases require more drastic emission reduction measures earlier to meet temperature targets.
Using a GWP100 gives the average radiative forcing occurring over the 100 years aer an emission. But why is the average effect over the next 100 years important and are there other important time horizons? The selection of time horizon is a policy decision: are there concerns about short-term or longterm global temperatures? Many countries have committed to reducing GHG emissions by 2030 or 2050, but these are interim targets with the aim of long term decarbonisation. There is an argument to suggest that an appropriate time horizon should be in accordance with 1.5 or 2 C decarbonisation pathways that require stabilisation of GHG concentrations by 2050-2100 : 30-80 years. [41][42][43] However, the GWP metric does not measure the impact at a specic time, but the average effect over a period. When concerned with a specic time for stabilisation, an instantaneous metric (such as GTP) may be more appropriate.
As the time of required climate stabilisation grows closer, the importance of methane mitigation grows stronger. Conversely, in 2100, an emission of methane from 2015 will be seen as relatively unimportant. The timeframe aer a stabilisation year will also be extremely important in maintaining a stabilised climate, whilst the application of a short time horizon effectively reduces the importance of longer term emissions to zero, which may be inappropriate.
Alvarez et al. 12 suggest that for technological environmental analyses, it is most appropriate and transparent to plot estimated GHG emissions over different time horizons. Other studies suggest that a comparison should span a exible range of time horizons, e.g. 12, 16 Ocko et al. 65 suggest simply presenting GWP from both a 20 and 100 year time horizon. For larger-scale integrated assessment models which project emissions up-to, and beyond, climate stabilisation periods, the use of a single GWP value such as the GWP100 would signicantly undervalue the impact of methane emissions. Thus the inclusion of both short and long-term metrics is imperative to assess the robustness of any projections, especially where the contribution of methane emissions is signicant.
From the development of metrics that analyse impacts on sea level and precipitation, 28,30 it is clear that potent short lived pollutants like methane may play a strong role in climate change in both the shorter (20 years) and longer (100+ years) time horizons. Both the short term and longer term effects of emissions must be understood and thus the inclusion of multiple time horizons help to prevent any unintended consequences associated with a technology or product switch.
As described in Section 5, there are three metrics described here that do not require the setting of a time horizon, but instead use a discount rate to estimate impacts over an innite time: the GDP, GCP and SCM metrics. Whilst the avoidance of a time horizon is benecial, the need to apply a discount rate represents a similar arbitrary weighting of preference for shorter (or longer) time horizons and so there is little advantage from this perspective. The numerical values are even more wide ranging as shown in Fig. 7, perhaps due to the compounding of assumptions relating to discount rates and the cost of damages.

Physical basis of the metric
The various metrics differ with respect to their physical or socioeconomic basis, and are primarily categorised as: radiative forcing; temperature; economic; or a mix of the aforementioned. They can also be categorised in relation to their position along the climate cause-effect chain as shown in Fig. 8. Metrics sitting closer to the end-point effects are more intuitively useful and understandable. As described, GWP is based on radiative forcing, but there is suggestion that a switch from GWP to a temperature-based metric such as GTP is more appropriate given that our climate targets revolve around global mean temperature changes. 2 However, at the point in the cause-effect chain where metrics estimate end-point damage, they convert from a physical basis to socio-economic and this carries additional uncertainty. These damage indicators may be extremely useful for broader studies into decarbonisation pathways, but typically require energy/climate/economic system models and are a step away from a simple metric design. The use of simpler physical metrics is preferable for such uses as annual emission inventories from a company or national perspective, or for simpler technological evaluations.
More recent metrics estimating contribution to sea level rise, the GSP, and to precipitation change, GPP, are very useful in improving our understanding of the physical effects of emissions across different timeframes and will help to inform the appropriate CO 2 equivalencies. It is notable that these metrics are broadly within ranges bounded by the GWP and GTP for equivalent time horizons.

Static vs. dynamic metrics
The way that GWP (and GTP) is used in most abatement studies does not take into account the timing of emissions. Typically, one metric (e.g. GWP100) is used to estimate emissions, for example from a natural gas well, over the lifetime of the well. However, as a well may be active and emitting for 30 years or more, this means that the end-point of the time horizon is not xed. For example, if a well emits within the rst year of operation, say 2015, the GWP100 would consider the impact up to 2115. If the well still operates and emits at 2045, the GWP100 estimation would consider the impact up to 2145.
Static metrics like the GWP and the GTP use xed time horizons. This means that the time horizon (e.g. 100 years) stays the same length, even when emissions studies may span multiple years (e.g. life cycle assessments). However, these metrics may also be used dynamically instead, using a xed end-point in time rather than a xed time horizon. This means that for multiple year studies, the end-point (e.g. the year 2100) stays the same and the horizon reduces as the year of emission advances. For example, a GWP100 may be used with an emission in 2015, a GWP99 in 2016 and GWP98 in 2017 etc. 44 Fig. 9 shows the difference between static (GWP and GTP) and dynamic (ICI and CCI) metrics by dening the CO 2 equivalency value over time.
To use a dynamic approach in a technology assessment, rst an end-point must be selected (e.g. 100 years from the start of the assessment time). Estimations of emissions must be made for each year of the assessment period (e.g. over a 30 year lifetime of a technology). Additionally, a different metric value for each year must be estimated. For example, emissions at year zero will be multiplied by the 100 year metric value, whilst emissions at year one will be multiplied by the 99 year metric value, and so on until the end of the assessment period (e.g. emissions at year 30 multiplied by the 70 year metric value). Thus, the use of dynamic metrics adds signicant complexity to   44 and Edwards and Trancik. 16 The use of static metrics must be carried out with care for emissions scenarios over long timeframes, for example with life cycle assessments. When doing so, the denition of the metric changes from its original meaning, for instance with GWP, which is intended to measure the average effect of a single pulse emission over a specic time horizon. Both the pulse and specic time horizon aspects are no longer applicable as there may be sustained emissions over many years.
The use of a dynamic metric may result in signicantly different results compared to the use of static metrics. 16 Using the example above, the methane emissions during the rst year would have a signicantly lower impact on global warming than equivalent methane emissions during the 30 th year. Such metrics are the ICI 16 or a dynamic version of the GTP. 2 Whilst the use of dynamic metrics may be preferable when comparing technologies over long timescales, static metrics are most appropriate for emissions estimates based on shorter timescales, for example annual emissions estimates. Additionally, the projection of a specic stabilisation year for use with a dynamic metric is an assumption, with atmospheric GHG concentration stabilisation years spanning 40 years or more across different emission pathways, as mentioned in Section 6.1. Thus, the use of a simpler static GWP for an LCA that spans 30 years would fall within this uncertainty range. Thus, there may be only marginal benet in applying a dynamic metric methodology, which may be outweighed by the relative increase in complexity of calculation.

Simplicity vs. tangibility
As metrics move along the cause-effect chain, they become more policy relevant 2 and relatable as an output. For example, temperature change may be a more tangible measure than radiative forcing, whereas damage estimates as a result of climate change are even more so. However, with greater tangibility comes more assumptions, uncertainty and complexity. For example, moving from a physical temperature change to estimating the socio-economic damage caused by that temperature change requires the modelling of climate impacts, population and demand projections, as well as technological resilience and innovation. Thus, there is a trade-off between simplicity, uncertainty and tangibility.
Myhre et al. 2 show that uncertainty is higher for GTP than for GWP for example: AE40% for GWP100 compared to AE75% for GTP100 (with a 90% condence interval). However, the impact of different time horizons gives even more variation in results than this uncertainty. Further, the uncertainty in estimates of methane emissions in the rst place have relatively high uncertainties in some cases e.g., 51 which are likely to be of similar order of magnitude to those from GWP or GTP. Some uncertainty is to be expected, which is why sensitivity analyses should be carried out wherever an investment or policy decision is marginal or at risk. It is the authors' opinion that for technology assessments and annual emission inventory estimates, physical climate metrics that enable CO 2 equivalency over a broad range of values best serve the purpose of understanding the range of potential climate impacts.

Suitability for application
Perhaps most importantly, the chosen metric must be appropriate for the application. Different applications require different levels of complexity and span different time scales as shown in Table 4. Typical uses of climate metrics are: Emissions inventories from industry operations. National/regional emissions contributions. Technology assessments e.g. LCA for policy planning. Energy system mitigation pathways. When the result will inform a long-term investment decision or policy, it is imperative that the impacts of using different metrics and time horizons on the result are explored.
Broadly, estimates of emissions over a short timeframe, e.g. annual emissions estimated from a company or national perspective, are likely to require a simple and static metric, given the lack of time variation and the requirement for fast and repeated estimation. For a technology assessment or a life cycle assessment that spans multiple years, a suitable metric may be:  a dynamic metric which accounts for the longer time frame considered; and a simple metric, given that the scope boundary is small and does not consider wider global implications. Estimates of emissions pathways to meet climate targets over longer time scales and multiple technologies may require metrics that: estimate the effects of climate change, either physical or economic damage; and may utilise more complex approaches such as climate models or end-point metrics.

The impact of different metrics on emissions results
As seen in the summary Table 3, the CO 2 equivalency values of methane range from 4 to 120 across metrics and time horizons. Additionally, the end-point metrics SCM and GDP have even higher values associated with the highest discount rates (for example the SCM estimates an equivalency of 199 at 10% discount rate 13 ). It is clear that the time horizon (or discount rate) has the largest impact on variation, more so than the metric type. Given that these are static multipliers in emission estimates, the impact of using different static values is large and linear.
To determine the impact of using different static and dynamic metrics and time horizons, this study applies the various metrics and equivalency values to an emissions case study: an estimate of greenhouse gas emissions associated with the production and consumption of various shipping fuels, including liqueed natural gas (LNG), heavy fuel oil (HFO) and methanol. Multi-year technology or fuel assessments typically use a single metric (e.g. the GWP100), but this assessment shows that the use of a singly metric inappropriately ignores the importance of timing of emissions and of the differences between short-term and long-term climate impact.
LNG exhibits 25-30% lower CO 2 emissions than liquid fossil fuels such as HFO upon combustion on an energy output basis, but typically has greater methane emissions. [45][46][47][48] Total methane emissions are governed by both the upstream supply chain and the engine type: this study investigates the use of a lean-burn spark ignition (LBSI) and a high-pressure dual fuel (HPDF) engine. 45 HFO and methanol are both used within diesel engines, where methanol also has lower CO 2 emissions due to its relatively higher H-C ratio. [48][49][50] A full environmental assessment has been conducted and is presented in a parallel paper to this, but a summary of the life cycle CO 2 and methane emissions are given in Fig. 10.
For the natural gas supply chain, upstream methane emissions arise from extraction, gathering and processing, liquefaction, storage and bunkering. Median estimates from Balcombe et al. 51 were used for production, gathering and processing. Liquefaction gures were estimated based on mean values derived from 6 studies 52-57 and synthesised in Balcombe et al. 58 For LNG storage the study uses assumptions made in Lowell et al., 53 whereas for bunkering, it is assumed that 0.22% of LNG is boiled off or displaced as vapour during fuelling, with a 50% capture resulting in 0.11% emission. 53,59 For methanol, the production and processing of natural gas is the same as included for the LNG supply chain. The inventory for gas reforming and methanol synthesis is derived from the NREL database, 60 using the Ecoinvent 3.3 database for the ancillary impacts. 61 The upstream allocated impacts to heavy fuel oil and marine diesel oil are taken from the Ecoinvent 3.3 database. For HFO, bunker oil with an average sulphur content of 3.5% w/w is assumed. For diesel, the production of low sulphur light fuel oil is used, with a sulphur content of 0.005% w/w. For upstream carbon dioxide emissions, 440 gCO 2 /kg HFO and 524 gCO 2 /kg diesel is associated with the production up to point of use. 61 Engine efficiencies, total methane emissions and total CO 2 emissions are given for each fuel/engine option in Table 5. For engine efficiencies, average values from various sources: ref. [45][46][47][48]53, 62 and 63 were taken and emissions are expressed per kWh of power output considering the average efficiency.
As can be seen in Fig. 10, large differences exist across the options in methane emissions both upstream and at end-use, as well as some moderate variation in CO 2 emissions. Combined life cycle GHG emissions are represented in Fig. 11 for different CO 2 equivalency values assumed. Given the different emission proles, there exist some crossover points where the rank order of fuels change. Under low equivalency values of less than 20 gCO 2eq. /gCH 4 , both LNG fuelled engines exhibit the lowest GHG emissions. Putting this in context, CO 2 equivalence values of less than 20 are those associated with longer time horizons and end-point metrics which do not account for the high initial forcing impacts. Such metrics with less than 20 gCO 2eq. /gCH 4 are the GTP at timeframes greater than 45 years, the ICI at timeframes greater than 30 years and the global sea-level rise potential (GSP) and global precipitation change potential (GPP) at 100 year time horizon.
As CO 2 equivalency value increases, the higher methane emissions associated with LBSI LNG engine result in this fuel/ engine option exhibiting the highest GHG emissions. Conversely, the LNG fuelled HPDF engine exhibits the lowest impacts across all equivalency values beside the highest at 120 gCO 2eq. /gCH 4 , due to its signicantly lower methane slip rates. It should be noted that methanol fuelled engines exhibit higher GHG emissions than HFO across all time horizons due to the high CO 2 emissions associated with methanol production from natural gas, as well as the moderate upstream methane emissions.
To understand the time dependence of emissions, we employ dynamic versions of the GTP and GWP for the above case study. The climate impact of the different fuels varies over time signicantly, as shown in Fig. 12. When long time horizons are considered, LNG engines perform favourably, especially in the case of GTP. For GTP and time horizons greater than 40 years, LNG presents a reduced climate impact by 10-20%. However, the LBSI engine with high levels of methane slip performs very poorly with respect to short term climate forcing. With respect to GWP, the integrated nature of the metric means that the initial high climate forcing of LNG engines maintains its impact for the LBSI engine across all timeframes considered, resulting in a higher climate impact than HFO. The HPDF with lower methane slip and low CO 2 emissions has the lowest climate impact across all time horizons.
Two implications arise from this assessment. Firstly, shortterm impacts are substantially different to long-term impacts across different technologies and the selection of timeframe may change the rank order of preference. It is imperative that both short and long-term climate impacts are accounted for when considering industrial investment or policy decisions. Secondly, for LNG fuelled engines to reduce GHG emissions compared to HFO, both upstream and end-use methane emissions must be constrained. Engines which inherently exhibit high methane slip are inappropriate for reduction of climate impacts. It should be noted however that LNG offers other benets than just climate impact, including reduced NO x , SO x , particulates as well as cost improvements.
The effect of changing equivalency value on the climate impact of other technology groups is also noticeable. For example, Edwards and Trancik 16 compare the operation of a CNG passenger vehicle versus one fuelled with petrol. Using a GWP100 results in the CNG vehicle improving GHG emissions by 10-15%, but with a GWP20 the CNG vehicle exhibits 20% higher emissions than for petrol. Producing a dynamic assessment using ICI and CCI Fig. 11 Estimates of total CO 2 equivalent GHG emissions for different shipping fuels and engines. metrics shows that CNG passenger vehicles offer a climate benet only over timeframes longer than 20 years. The comparison of natural gas against coal for power generation is robust in favour of natural gas and shows preference in all but the most conservative of assumptions about GWP values and methane emissions. 64 However, for estimates where carbon capture and storage is used to reduce combustion emissions by up to 90%, the impact of methane emissions proportionally increases. In this case, the choice of metric and time horizon is likely to have a large impact on the relative benet.
Thus, the selection of metric, and more importantly, time horizon, has a large impact on the ranking of these fuels and technologies, as well as the magnitude of estimates. Investment or policy decisions that trade-off different greenhouse gases like above must ensure that both short-term and long-term climate impacts are taken into consideration.

Conclusions and recommendations
This report has investigated the use of various climate metrics and analysed their key attributes and limitations, with respect to methane emissions. There is no single metric or time horizon that is appropriate for all applications and situations. One key point is that methane emissions for the most part are transitory, 33 whereas CO 2 emissions are persistent. Consequently, when considering time horizons the emphasis must not be lost on eliminating CO 2 emissions as, if they are not largely eliminated, the climate will not stabilise. Therefore, any adoption of a shorter time horizon should be tempered with a comparatively longer one.
Given the requirement to stabilise GHG concentrations and to ensure there is no long-term climate change beyond a 2 C limit, it is inadvisable to use only a 20 year time horizon. A 20 year horizon effectively disregards the impact of emissions aer this point, which in the context of comparing methane to CO 2 emissions, dangerously undervalues the long term impact of CO 2 . A two-value approach, which indicates the effect over two different time horizons, is suggested by a number of studies. 65 In selecting an appropriate metric, there is a trade-off between simplicity and transparency. 66 The most appropriate metric depends on the application and which aspect of climate change is most pertinent to the study. 2 Using a single value equivalency such as the GWP100 or GTP100, is the simplest option but hides much information which may be needed to make an investment decision or a policy recommendation. For example, a GHG with a short life but strong radiative forcing may have the same GWP value over a set time horizon as a GHG with a long life but weak forcing effect: the impact of each GHG on climate change may be signicantly different but this is lost with such a simplication. 32 A temperature-based metric such as GTP ts well with a temperature based climate target, but it is suggested that the damage caused by climate change will increase faster than the temperature increase. 13 Consequently, reducing our CO 2 equivalencies from GWP values to GTP values may cause an underestimation of the impact of methane. Even the use of GWP100 may cause an underestimation of the contribution of methane, 16 for example to impacts relating to sea level rise. 28 The overarching recommendation from this study is to present emissions results with transparency. It is prudent to report methane and CO 2 emissions separately and where climate metrics are used, a summary of the magnitude and type of metric should be given. If the equivalency value has a large impact on results, both low and high values should be used to assess the impact.
Broadly, metric applications can be placed into three categories: short-term (e.g. annual) emissions estimates of processes, facilities or regions; multi-year technology assessments or life cycle assessments; and long-term modelling of energy systems and decarbonisation pathways. Recommendations are made for each category.
Estimates of emissions on a short timescale in the order of 1 year typically involve aggregating estimates for a facility or region and require simple static metrics such as GWP or GTP. Two recommendation options are to: present emissions using a single GWP or GTP metric (50 or 100 year), and include the separated contribution from both methane and CO 2 ; present two time horizons, a short term (e.g. 20 or 50) and a longer term (e.g. 100 or more), such that any comparative arguments for technology change holds in both the short term or the long term, or at least that a detriment to either short or long term has been considered.
For technology assessments or life cycle assessments that span 20 or 30 years, suitable metrics could be static (GWP or GTP) or dynamic (e.g. ICI or TWP) to account for the emissions timing. However, given the uncertainty associated with a projected stabilisation year, this report considers dynamic metrics to be of only marginal benet. Additionally, given the increase in complexity associated with using a dynamic metric, the selection of a static metric and incorporating two (or more) time horizons would be appropriate.
For longer term analyses of multiple energy systems over long timeframes, higher levels of complexity are acceptable and application of climate models is most suitable. Where this is not feasible, the application of dynamic metrics or the assessment of both short and long-term time horizons is imperative, especially under scenarios where methane emissions are signicant.
In summary, the use of climate metrics in GHG estimation must be carried out with great care and the standard usage of a single global warming potential is not acceptable as it may hide key trade-offs between short and long-term climate impacts. To counter this, transparent reporting of methane and CO 2 emissions is required. It is vital to test any GHG estimates with high and low equivalency values to ensure that we are not simply replacing long-term climate forcing with short-term, or vice versa.

Conflicts of interest
There are no conicts to declare.