## Introduction

Molecular hydrogen (H2) is increasingly presented as a key element of the worldwide energy transformation required to limit global warming and meet the Paris Agreement1 climate objectives. H2 is recognised as an important future energy vector for applications ranging from power generation, transportation, industry, building heating and energy storage2,3,4,5,6,7,8. The use of hydrogen produced by a renewable source of electricity enables the conversion and the storage of energy, and may provide a way to decarbonize sectors of the economy difficult to decarbonize such as long-distance transportation by trucks and airplane, heavy industries, or for domestic uses blended with natural gas4.

H2 is a symmetric molecule with no direct impact on infrared radiation at temperature and pressure conditions prevailing in the Earth’s atmosphere9. However, hydrogen is recognized as an indirect greenhouse gas through its involvement in chemical reactions which affect the lifetime and concentration of other greenhouse gases10,11. In particular, tropospheric oxidation of H2 depletes the hydroxyl radical (OH). Since OH is the main sink for methane, a potent greenhouse gas, this results in a lengthening of the methane atmospheric lifetime. H2 and methane are also precurors of tropospheric ozone and a photochemical source of water vapour in the dry stratosphere12,13, both of which also act as greenhouse gases.

Firn air measurements from South Pole indicate that the concentration of hydrogen has increased from about 350 parts per billion (ppbv) in 1910 to about 540 ppbv in 200014. Today, the sources of H2 are oxidation of methane and other hydrocarbons in the atmosphere (56%), fossil fuel combustion (23%), biomass burning (12%), and nitrogen fixation in the soils and ocean (12%)15. Hydrogen is removed from the atmosphere by oxidation by OH (about 25%) and by soil uptake due to bacterial activity (about 75%)15. The burden of H2 in the troposphere (136–162 Tg)15 and its production and destruction terms are however uncertain in both magnitude and distribution12,15,16,17. In particular, the critical role played by biological H2 soil consumption is subject to considerable differences among the various estimates (55–88 Tg/year)15,16,17,18,19. This is reflected in the large range of estimates for H2 tropospheric lifetime (1.4–2.3 years)12,13,14,15,16,17,18,19,20.

The development of a future large-scale hydrogen economy has the potential to increase the atmospheric source of H2 through leakage during production, transport, storage and use. Based on previous reports2,6,21, leakage rates are likely to range from 1 to 10%. Hereafter, we will refer to these leakage rate as low and high leakage rates, respectively. These fugitive hydrogen emissions were found to have minor implications for both air quality2,21,22,23,24 and stratospheric ozone depletion24,25, unless unrealistic leakage rates of 20%8,26, are considered27. The climate impact of hydrogen is assessed in terms of the Global Warming Potential (GWP) climate metric3,10,11,28 or more recently in terms of radiative forcing and equilibrium temperature change15,29. With calculated GWP100 of 5.0 ± 1.028 and 3.3 ± 1.43, earlier assessments indicate that the climate impact of hydrogen is relatively minor compared to other potent greenhouse gases with more important anthropogenic emissions such as carbon dioxide (CO2), nitrous oxide (N2O) or methane (CH4). However, recently, the climate benefit of a hydrogen economy has come into question. A first reason is that the climate impact of a hydrogen economy depends on how hydrogen is produced. In particular the production of “decarbonized” hydrogen from oil and gas depends on capture and storage of the CO2 produced4,30, and on the associated upstream methane leakage4,31,32,33. Secondly, since the indirect hydrogen impact on climate involves chemical reactions in the atmosphere, the uncertainties on key processes affecting the H2 distribution and budget hamper a precise quantification of the associated radiative forcings13,34. As an illustration of this uncertainty, a recent study revised the hydrogen GWP100 value to more than twice its previously estimated value35. Furthermore, GWP100, the climate metric used by the United Nations Framework Convention on Climate Change (UNFCCC) when reporting emissions of different greenhouse gases, is often criticized for over- or understating the importance of Short-Lived Climate Forcers (SLCFs), like H24,29,36,37,38,39.

Here, we use recent global H2 perturbation simulations and associated radiative forcings of climate15 to derive hydrogen climate metrics. The metrics are decomposed into their various contributions from methane, tropospheric and stratospheric ozone (O3) and stratospheric water vapour (H2O). Based on these metrics and on future roadmaps for transition to an hydrogen economy in Europe and worldwide, we quantify the climate benefit of such hydrogen decarbonization pathways. In particular, we illustrate the impact of a hydrogen economy on the cumulative CO2 emissions abatement over the Paris Agreement long-term climate goal.

## Results

### Hydrogen climate metrics

Future hydrogen demand levels reaching up 3000 TgH2/year were recently applied to derive an equilibrium global surface temperature increase29. However, the climate steady-state to a perturbation in hydrogen emissions will be reached over centuries, providing that all other parameters are constant. In order to quantify the climate impact of H2, it is essential to consider the transient behaviour of the perturbation. In particular, other climate metrics are required to effectively estimate the climate impact of a transition to a hydrogen economy during the next 75 years. The two most commonly used climate metrics are the GWP and the Global Temperature Potential (GTP)42,43,44. The GWP is the time-integrated radiative forcing due to a pulse of emission, over a given time-horizon relative to an equal mass pulse of CO2. In order to better represent the actual impact of a greenhouse gas on the global-mean surface temperature, the GWP concept has been extended to the GTP metric which provides the temperature effect of a gas emission at a given time-horizon42. The GWP and GTP of H2 are calculated as a function of the time-horizon, based on the radiative efficiencies15. An uncertainty analysis for these different metrics based on the IPCC methodology41 has been performed (details are available in the “Methods” section).

Figure 1a shows the Absolute GWP (AGWP) calculated for H2 and decomposed into its different contributions. Rather than focusing on a particular time-horizon, we estimate and show the AGWP as a function of the time-horizon. The total AGWP increases with the time-horizon, reaching a maximum value of 1.11 $${{{{{\rm{\pm }}}}}}\,0.34$$ × 10−12 W/m2/kg year about 20 years after emission. The indirect AGWP associated with ozone and stratospheric H2O reach a maximum after 8 years. The indirect AGWP of methane shows a different time-scale and increases over a period of about 30 years. This shows the importance of decomposing the H2 indirect forcing into a short-term component characterized by the H2 lifetime and a longer-term component characterized by the methane perturbation lifetime of 12.4 years43. As shown on this figure, even after 300 years, the CO2 AGWP is still increasing, stressing the importance of considering the difference in lifetime between the short-lived H2 forcing and the long-term CO2 forcing. Figure 1b shows the resulting total hydrogen GWP and its individual components as a function of the considered time-horizon. The evolution and decrease of the GWP with the time-horizon after 3 years is controlled by the increase of the CO2 AGWP as seen in Fig. 1a. The H2 GWP decreases from a value of 106 ± 38 for a very short-time horizon (GWP1) to a value of 40.1 ± 24.1 for a 20-year time-horizon (GWP20) and 12.8 ± 5.2 for the 100-year time-horizon (GWP100) (Table 1). The impact of H2 on the methane lifetime and the associated GWP component shows a different temporal profile than the other short-term contributions from tropospheric and stratospheric ozone and stratospheric H2O, and peaks at a time-horizon of 10 years. The methane contribution represents 48 and 53% of the H2 GWP20 and GWP100, respectively (Supplementary Fig. S1a, b). The H2 GWP100 is larger than the earlier estimate of 5.8 with the STOCHEM three-dimensional model10,11, the value of 5.0 ± 1.0 recalculated with the same model28, and with the value of 3.3 ± 1.4 calculated with the TROPOS two-dimensional model3. This discrepancy can be attributed to a higher methane radiative efficiency45, to the indirect effects of stratospheric H2O and stratospheric ozone not included in the previous estimates, and to the longer H2 tropospheric lifetime (2.1 years)15 compared to these earlier estimates (1.6 years)28. This emphasizes the sensitivity of the results to assumptions made for H2 soil uptake. Our results are consistent with a more recent estimate35 which derived a H2 GWP100 of 11 ± 5 using a state-of-the-art global climate-chemistry model.

The calculated Absolute GTP (AGTP) of H2 is illustrated as a function of the time-horizon in Fig. 1d. The AGTP shows a short-term increase and peaks after 4 years. The AGTP then falls down by several orders of magnitude, in line with the short timescale of the temperature change response function. This sharp decrease is then followed by a slower decrease for time-horizons >60 years where the longer time-scale of the temperature response becomes more important. The AGTP of CO2 increases rapidly with the time-horizon before peaking after about 10 years and decreasing slowly thereafter. The resulting H2 GTP is illustrated in Fig. 1e. The GTP decreases rapidly from a value of 106 ± 139 for a very short time-horizon to a value of 17.8 ± 23.3 for a 20-year time horizon (GTP20) and 2.3 ± 1.5 for a 100-year time horizon (GTP100). The longer-term methane indirect effect increases to a maximum value reached about 10 years after emission and decreases thereafter. The methane response contributes 82% to the hydrogen GTP20 and 54% to the GTP100 (Supplementary Fig. S1c, d).

Due to the larger fraction of land in the northern hemisphere than in the southern hemisphere, this results in a larger sink by soil uptake and hence a shorter atmospheric residence time of H2 in the northern hemisphere. This contrast explains the observed H2 enhancement in the southern hemisphere16,20. The H2 global lifetime in the troposphere derived based on the GFDL-AM4.1 model simulations is 2.1 years15. Over the whole atmosphere this corresponds to a global atmospheric lifetime of 2.5 years. This global atmospheric lifetime is used in this study for the GWP and GTP calculations. Averaged over the northern hemisphere solely, the calculated H2 atmospheric lifetime is 2.0 years. The shorter lifetime of H2 in the northern hemisphere reduces the climate impact of H2 emitted in this region (GWP100 = 10.7 ± 4.4 and GTP100 = 1.9 ± 1.2) (Table 1). These northern hemisphere values will be used hereafter to quantify the climate impact of European H2 leakages.

The GWP and GTP climate metrics are defined in terms of a pulse emission of a species. For a SLCF such as H2, several authors also suggested the use of metrics defined in terms of a sustained change in emissions to express equivalence between cumulative climate pollutants and SLCFs31,42,46,47,48. The ‘s’ subscripts denote metrics based on sustained emissions. The use of a sustained emission increases the calculated GWP of H2 after a few years and the 100-year time-horizon GWPs is 20.8 ± 8.5, 1.6 times larger than the GWP. Sustained metrics are barely used and are essentially calculated here in order to derive the combined climate metrics (see below).

Figure 1b, e and Table 1 show that both GWP and GTP metrics vary considerably with the considered time-horizon. For a SLCF such as hydrogen, there is a scalar factor of 3.1 for the GWP and 7.7 for the GTP in the value of the metric for a 20-year versus a 100-year time-horizon. In order to reconcile the impact of cumulative climate pollutants and SLCFs, combined metrics44 have been defined as the ratio of a sustained response of the considered SLCF to a pulse response of CO2 (details available in the “Methods” section). These combined metrics, based on the end-point metrics GFP (Global Forcing Potential) and GTP, measure the progress towards a steady-state temperature. The combined metrics are no longer dimensionless, but have units of time, reflecting the need to compare rates (kg/year) with pulses (kg). Figure 1c show the calculated Combined GWP (CGWP) and Combined GTP (CGTP) of H2. As shown also in Table 1, CGWP and CGTP vary less with the time-horizon and there is a scalar factor of 1.7 for the CGWP and 1.4 for the CGTP in the value of the metric for a 20-year versus a 100-year time-horizon. After about 20 years, the time variation of these metrics for a short-lived species like H2 is essentially due to the decrease of CO2 following the pulse emission.

### Climate consequences of a future hydrogen economy

To deliver on the sustainability promise it is important to maximize the hydrogen decarbonization potential. The most common way of producing hydrogen today is by reforming natural gas into hydrogen and CO2, which is referred to as grey hydrogen if the CO2 is emitted into the atmosphere, or blue hydrogen if the CO2 is captured and permanently sequestered. If renewable sources are used for water electrolysis, the hydrogen produced is referred to as green. The climate impact of a future hydrogen economy depends on the intensity of the deployment of the hydrogen energy supply in the different sectors of the economy but also on the hydrogen production pathways.

The climate impact of an hydrogen economy also depends on the fraction of H2 lost to the atmosphere during production, transport, storage and use through venting, fugitive emissions and incomplete combustion5,49. The future leakage rate of hydrogen into the atmosphere is a major uncertainty and our assessment of the climate impact of an hydrogen economy transition is performed assuming different leakage rates. In order to investigate the climate benefit of a future hydrogen powering economy, several scenarios for worldwide and European energy supply by hydrogen are considered in this study (details available in the “Methods” section). In the future, the global hydrogen production portfolio will contain a mix of both green, blue and grey hydrogen production pathways. For green hydrogen, the equivalent CO2 emission is calculated by multiplying the total hydrogen production by a leakage fraction and by the considered climate metric for different time-horizons. The total hydrogen production is itself increased by the leakage rate in order to meet the given supply. For grey and blue hydrogen, in addition to this CO2 equivalent emission associated with the H2 leakage rate, we add carbon dioxide and methane emissions (details available in the “Methods” section).

We first evaluate the carbon-saving potential of an installed 2050 worldwide hydrogen economy based on scenario HC2017 for global decarbonization and hydrogen supply (See “Methods” for a brief description of the HC2017 scenario proposed by the Hydrogen Council in 201750,51). The climate benefit in 2050 is determined by dividing the hydrogen CO2 equivalent emissions, calculated based on the chosen climate metric and time-horizon, by the avoided CO2 annual emission. According to scenario HC2017, in 2050, hydrogen provides worldwide some 21,667 TWh (78 EJ), an almost eight-fold increase of hydrogen consumption over the 2020–2050 period. Such an energy production from hydrogen is estimated to abate the CO2 emissions by 6 GtCO2/year50. This abatement corresponds to 10.9 kgCO2 avoided per kgH2 across all energy supply sectors. Figure 2 shows the percentage of remaining carbon equivalent emissions considering an hydrogen economy as a function of the metric time-horizon and assumed leakage rate. We illustrate the climate impact of an hydrogen economy assuming an optimistic case of 100% green hydrogen supply, or a mix of blue + green hydrogen. Figure 2a provides the climate benefit assuming a green hydrogen economy and based on the GWP climate metric. By definition, no leakage of green hydrogen would provide 0% of CO2 equivalent emission. When a very low leakage rate of 0.1% is assumed, the H2 emission to the atmosphere represents 0.1% of the abated CO2 emission based on the GWP100. In this case, 99,9% of the fossil fuel based CO2 emissions are avoided. Using a shorter time-horizon (GWP20), increases the CO2 equivalent emissions to 0.4% (a 99.6% benefit in terms of avoided CO2 emissions). In contrast, in the case of a high leakage rate of 10%, the CO2 equivalent emissions represent about 13% of the abated CO2 emissions for a 100 year time-horizon (about 40% at a 20 year time-horizon). It is only for leakage rates of 10 and 15%, and at a very short time-horizon of 1 year (not realistic from a climate pespective) that the hydrogen economy results in a climate penalty, reaching 167% (CO2 equivalent emissions higher than the abated emissions by 67%). In the case of a mix of blue + green hydrogen (assuming, according to the 2050 HC2017 scenario, 30% of blue hydrogen and 70% of green hydrogen), the CO2 equivalent emissions increase and represent 20.1% (resp. 28.3%) of the abated emissions for a 100 year (resp. 20 year) time-horizon and a 1% leakage rate, and represent 64% (resp. 153%) for a 10% leakage rate (Fig. 2b). In a mix of blue + green hydrogen supply, equivalent CO2 emissions increase strongly with the hydrogen (and methane) leakage rate and largely offset the climate benefit of the hydrogen economy. At a high leakage rate of 10%, the equivalent CO2 emissions represent 64.5% of the abated CO2 emissions at a 100 year time-horizon but 153.3% of the abated emissions at a 20-year time-horizon. This moderate climate benefit obtained in the case of a mix of blue + green hydrogen strongly depends on the assumed blue hydrogen supply. For a 1% leakage rate, the equivalent CO2 emissions associated with an hydrogen economy increase to 26.5%, 32.8%, 39.1% and 51.7% at a 100 year time-horizon for a 40%, 50%, 60 and 80% fraction of blue hydrogen (Fig. 3). It shows that, based on the GWP100, a portfolio composed of >60% of blue hydrogen has basically no climate benefit if the leakage rate is >7%. Based on the GWP20, the climate benefit is lost if the blue hydrogen fraction is >30% and if the leakage rate exceeds 3%. In the case of a green hydrogen supply, the use of GWPs increases the CO2 equivalent emissions by a factor of 1.3 and 1.6 for a 20 year and 100 year time-horizon, respectively, independently of the leakage rate (Supplementary Fig. S2). In the case of a blue + green supply, this ratio between GWPs and GWP based equivalent emissions depends on the leakage rate and is reduced compared to the green hydrogen case since a fraction of the equivalent CO2 emission is associated with the methane leakage. With a lifetime longer than H2, the CH4 GWPs is less affected than H2 by the use of the sustained metric at longer time-horizons (Supplementary Fig. S3).

In order to illustrate the sensitivity of the results to the hydrogen technologies, the calculations were repeated using the hydrogen Lower Heating Value (LHV) instead of the Higher Heating Value (HHV) used in previous studies3,4,11,50 (details available in the “Methods” section). Assuming the LHV implies that about 18% more hydrogen needs to be produced in order to supply the same energy. In the case of a green hydrogen economy, when a low leakage rate of 1% is assumed, the hydrogen CO2 equivalent emissions increase from 1.2% with the HHV to 1.4% with the LHV based on the GWP100. In the case of a high leakage rate of 10%, the CO2 equivalent emissions increase from 13% with the HHV to 15% with the LHV for a 100 year time-horizon (Supplementary Fig. S4a). In the case of a mix of blue + green hydrogen, the CO2 equivalent emissions increase from 20% with the HHV to 24% with the LHV for a 1% leakage rate and from 64% with the HHV to 76% with the LHV for a 10% leakage rate (Supplementary Fig. S4b).

Unlike the GWP, the GTP quantifies the climate effect at the end of a timeperiod rather than averaging over the whole timeperiod. The GTP therefore gives more weight to the radiative forcing that comes later in the timeperiod. Whereas, using a GWP to trade a short-lived species like H2 against the long-lived like CO2 tends to overestimate the importance of H2 for reaching long-term climate goals42,52. If Article 2 of the Paris Agreement is interpreted as a long-term climate goal, the GTP appears as an appropriate metric to use. When the GTP is used, the CO2 equivalent emissions associated with H2 are reduced and represent <5% of the abated emissions for a time-horizon in the range 20–100 year and a 1–3% leakage rate in the case of a green hydrogen economy (Fig. 2c). More generally, in the case of a green hydrogen supply, the use of the GWP instead of the GTP increases the CO2 equivalent emissions by up to a factor of 6.2 for a 75 year time-horizon (Supplementary Fig. S5). In the case of a blue + green hydrogen supply, this ratio between GWP and GTP emissions depends on the leakage rate and is reduced to less than a factor of 4 reflecting the role of the leaks on the longer-lived methane (Supplementary Fig. S6).

We now calculate the cumulative CO2 abatement resulting from the transition to an hydrogen economy over the entire 2030–2100 period. We use the HC2017 scenario for the 2030, 2040 and 2050 decades and assume 2050 values for decades thereafter. We add the CO2 or H2 emissions over the entire 2030–2100 period in order to calculate the cumulative emissions needed to supply the same energy over the full period. For this given energy supply scenario, we compare the CO2 emissions abated due to a greater use of hydrogen in the economy (i.e., the emissions that would have resulted from a business-as-usual fossil-fuel based economy without hydrogen) and the hydrogen equivalent CO2 (CO2e) emissions released into the atmosphere to supply the same energy. For the full period, this scenario results in a total abatement of CO2 emissions worldwide by 331 GtCO2. The use of green hydrogen over the full period with 0% leakage rate would then results in a reduction of 331 GtCO2 emitted into the atmosphere. The CO2e emissions resulting from hydrogen leakage into the atmosphere and, in the case of grey and blue hydrogen, its production and transport, will reduce this avoided cumulative CO2. Figure 4a shows that assuming a green hydrogen economy would result, assuming a 1% leakage rate and a 100-year time-horizon (GWP100), in 327 GtCO2e avoided. This abatement decreases to 289 GtCO2e for a 10% leakage rate. For a mix of blue + green hydrogen, 259 GtCO2e are avoided for a 1% leakage rate at a 100 time-horizon (Fig. 4b). In 2030, a 33% fraction of grey hydrogen is still assumed in this scenario. When a of mix of grey + blue + green hydrogen is considered, according to the HC2017 scenario, the cumulative CO2 emissions avoided total 258 GtCO2e for a 1% leakage rate and using the GWP100 metric (Fig. 4c). This abatement decreases to 103 GtCO2e for a 10% leakage rate. The results show that in the case of blue and grey hydrogen, the climate benefit rapidly decreases when the leakage rate is >3%. A similar conclusion arises for the IEA2021 scenario (see “Methods” for a brief description of the IEA2021 scenario proposed by the International Energy Agency in 202153) characterized by a higher contribution of blue hydrogen in 2050 compared to HC2021 (Supplementary Fig. S7a). Over the full 2030–2100 period, this scenario provides a cumulative CO2 emission abatement of 353 GtCO2. When the climate impact of the hydrogen leaks and production is considered, for a very low leakage rate of 0.1%, the CO2 abatement is 270 GtCO2e based on the GWP100 climate metric. For a leakage rate of 1%, we calculate an abatement of 252 GtCO2e. This abatement is reduced to 60 GtCO2e for a high leakage rate of 10%.

An alternative, more ambitious, roadmap for a transition to a hydrogen energy supply has also been considered. According to this HCMK2021 scenario (see “Methods” for a brief description of the HCMK2021 scenario proposed by the Hydrogen Council/McKinsey & Company in 202154), over the full 2030–2100 period, the resulting cumulative CO2 emission abatement is 417 GtCO2. When the climate impact of the hydrogen leaks and production is considered, the cumulative equivalent CO2 emissions avoided is affected as depicted in Supplementary Fig. S7b. For a very low leakage rate of 0.1%, the CO2 abatement is 395 GtCO2e based on the GWP100 climate metric. For a leakage rate of 1%, we calculate an abatement of 388 GtCO2e. This abatement is reduced to 310 GtCO2e for a high leakage rate of 10%. This scenario, characterized by a decrease of grey and blue hydrogen supply in order to reach 100% green hydrogen in 2050, clearly shows a larger climate benefit compared to HC2017 stressing the need for a rapid transition to a 100% green hydrogen economy.

Achieving an energy transition of the European Union (EU) compliant with the Paris Agreement and with the European ambitions for achieving climate neutrality by 2050 will require hydrogen deployement at large scale55. According to the TYNDP2022 scenario (see “Methods” for a brief description of the TYNDP2022 scenario proposed by ENTSOG and ENTSO-E in 202256), 2496 TWh (9 EJ) are produced from hydrogen in 2050. Based on the energy content of gaseous hydrogen3,7, this corresponds to a supply of 63 Mt H2/year in 2050 with a 6% supply of blue hydrogen (3.8 Mt/year). The total energy production from hydrogen provides a total avoided CO2 emissions of 620 Mt CO2/year in 2050, or a total cumulative CO2 abatement of 36.1 GtCO2 over the full period 2030–2100. For this scenario related to European emissions, we use the hydrogen metrics (i.e., GWP in Fig. 5 and CGTP in Fig. 6) calculated based on the hydrogen northern hemisphere lifetime (see Table 1). Figure 5 shows the cumulative CO2 emissions abatement calculated based on the GWP as a function of the leakage rate and time-horizon. In the case of a green hydrogen supply (Fig. 5a), the cumulative CO2 abatement accounting for the impact of hydrogen leaks is 35.7 GtCO2e for a 1% leakage rate and a 100 year time-horizon. This abatement decreases to 31.8 GtCO2e for a 10% leakage rate. For a mix of blue + green hydrogen (Fig. 5b), the abatement is reduced to 33.6 GtCO2e and 26.1 GtCO2e for a 1 and 10% leakage rate, respectively, based on the GWP100. For a 20 year time-horizon metric, an abatement of 32.2 GtCO2e is calculated for a 1% leakage rate.

## Discussion

The GWP100 is the reference metric for reporting emissions under UNFCCC to determine CO2 equivalent emissions for a basket of greenhouse gases. Shorter time-horizons have also been suggested for SLCFs4,36,37,38. Reducing SLCFs can slow the projected global warming over the next 25 years with many synergies towards achieveving Sustainable Developments Goals (SDGs), in particular in terms of human and ecosytems health37,57. In the case of H2, limited air quality or ecosystem impacts are however expected21,58 and these synergies are probably minor. For methane fugitive emissions associated with grey and blue hydrogen production, the situation is different and a shorter time-horizon appears appropriate to account for synergies with air quality impacts. The time-horizon of the metric is therefore variable in our calculations in order to illustrate the sensitivity to the chosen metric and time-horizon. Based on our calculations, a clear benefit for climate arises from a transition to a green hydrogen economy. For a leakage rate of 1–3% and based on the GWP climate metric at 20–100 year time-horizon, the green hydrogen CO2 equivalent emissions represent 1.2–11.3% of the CO2 emissions avoided (0.2–5.0% based on the GTP). In the scenario considering a mix of blue (30%) and green (70%) hydrogen, we calculate for a leakage rate of 1–3%, a significantly lower climate benefit with hydrogen CO2 equivalent emissions representing 20–54% of the abated CO2 emissions based on the GWP20 and GWP100. At leakage rates >10% (resp. 5%) and for a time-horizon of 100 year (resp. 20 year) the mix of green + blue hydrogen economy leads to a climate penalty with CO2 equivalent emissions larger than the avoided CO2 emissions. This moderate climate benefit (or even climate penalty) obtained in the case of a mix of blue + green hydrogen strongly decreases as the assumed blue hydrogen contribution to the energy portfolio increases.

Shorter time-horizons are less adequate for estimating the importance of a short-lived species such as H2 for mitigating long-term climate change, as envisaged within the Paris Agreement44,48,52. The endpoint metric GTP appears as a more appropriate climate metric to estimate the effect of a short-lived climate forcer like H2 on end of century climate goals and evaluate the mitigation potential within this framework. The CGTP metric combining sustained and pulse emission responses, also appears as a particularly appropriate instrument to convert emissions of the short-lived H2 to equivalent cumulative carbon emissions44. The CGTP provides the cumulative emissions associated with a structural change such as the transition to a hydrogen energy supply. As a summary, and as an illustration of the use of CGTP to measure the mitigation progress towards a climate change objective at the end of the century, Fig. 6 shows the cumulative carbon emissions saved by 2100 when the hydrogen economy is introduced for the different scenarios considered in this study over the 2030–2100 period. In the case of a green hydrogen economy, a cumulative worldwide CO2 emission abatement of 324 GtCO2e is calculated for a 1% leakage rate for the HC2017 scenario, 346 GtCO2e for the IEA2021 scenario, and of 408 GtCO2e for the more ambitious HCMK2021 scenario. A cumulative carbon emission abatement of 35 GtCO2e is calculated for the TYNDP2022 European green hydrogen scenario. For a 10% leakage rate, under a green hydrogen economy assumption, the abatement is reduced to 250, 273, 317 and 28 GtCO2e for the HC2017, IEA2021, HCMK2021, and TYNDP2022 scenarios, respectively. Interestingly, the comparison with Figs. 4a and 5a shows that these results are very close to the emission abatements calculated based on the GWP40. This suggests, in agreement with previous work59, that 40 years appears as an appropriate time-horizon to estimate the mitigation of CO2 emissions due to a hydrogen economy and the resulting benefit on end of century climate. The cumulative CO2 emission abatement is significantly reduced in the case of a mix blue + green hydrogen and decreases to 247 GtCO2e for a 1% leakage rate for the HC2017 scenario considering 30% of blue hydrogen after 2050. A climate penalty of −31 and −205 GtCO2e is even calculated for higher leakage rates of 10 and 15%, respectively. For the IEA2021 scenario characterized by a higher blue hydrogen fraction, a climate penalty of −86 and −292 GtCO2e is also calculated for leakage rates of 10 and 15%, respectively. The comparison with scenario HCMK2021 (0% blue hydrogen after 2050) providing CO2 emissions abated by 389 and 278 GtCO2e for 1 and 10% leakage rates, respectively, clearly shows the importance of a rapid replacement of the blue hydrogen supply by green hydrogen in order to preserve a clear climate benefit from a hydrogen economy.

All considered scenarios for a future hydrogen transition worldwide and in Europe, clearly show that a green hydrogen economy is beneficial in terms of mitigated CO2 emissions for all policy-relevant time-horizons. In contrast, our results suggest that the CO2 and CH4 emissions associated with blue (and grey) hydrogen production and transport significantly reduce the benefit of a transition to a hydrogen economy and even introduce a climate penalty at high leakage rate or blue hydrogen market penetration. Reducing the leakage rate of H2 (and of CH4 in the case of blue hydrogen production) and increasing the green hydrogen production pathway appear as the key leverages towards a maximum mitigation of equivalent CO2 emissions from a large-scale structural transition to a hydrogen economy.

It should be kept in mind, that the H2 climate impact is indirect and is therefore subject to larger uncertainties than direct impacts from other climate pollutants. We estimate uncertainties on the H2 metrics of about 40% on the GWP100, 60% on the GTP100 and 70% on the GTP100. These typical uncertainties on the metrics directly translate into a similar uncertainty on the carbon equivalent emissions. Therefore, we stress the need for better constrained atmospheric models based on a reliable observational network in order to derive even more robust estimates of the hydrogen climate benefit.

## Methods

### Climate metrics calculations

We use analytical Impulse Response Functions (IRFs) to calculate the hydrogen GWP and GTP52. The mass of a species S evolves in time in the atmosphere according to:

$${\Delta {{{{{\rm{M}}}}}}}_{S}\left({{\mbox{t}}}\right){{\mbox{=}}}{\int }_{{{{\mbox{t}}}}_{0}}^{{{\mbox{t}}}}{{{\mbox{E}}}}_{{{\mbox{S}}}}\left(t^{\prime} \right){{{\mbox{IRF}}}}_{{{\mbox{S}}}}\left(t-t^{\prime} \right)\,{{\mbox{dt}}}^{\prime}$$
(1)

where MS(t) is the species atmosphere mass (kg) at time t, $${{{{{{\rm{E}}}}}}}_{S}({{{{{\rm{t}}}}}})$$ the emission of the species (kg/year), and t0 is the initial year of integration (we assume t0 = 0). The atmospheric mass MS is converted to mixing ratio qS (ppbv) with $${q}_{S}$$ = 5.625 × 10−9 $${{{{{{\rm{M}}}}}}}_{S}$$/mS, where mS is the molecular mass.

The H2 IRF is a simple exponential decay function with a characteristic atmospheric global lifetime $${\tau }_{{{{{{{\rm{H}}}}}}}_{2}}$$ of 2.5 years (corresponding to a global tropospheric lifetime of 2.1 years)15:

$${{{{{{\rm{IRF}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}(x)={{\exp }}\left(-\frac{x}{{\tau }_{{{{{{{\rm{H}}}}}}}_{2}}}\right)$$
(2)

A similar IRF is used for methane with a perturbation lifetime $${\tau }_{{{{{{{\rm{CH}}}}}}}_{4}}$$ of 12.4 years43. The IRF for CO2 is calculated with refs. 52, 60:

$${{{{{{\rm{IRF}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}\left(x\right)={a}_{0}+\mathop{\sum }\limits_{i=1}^{3}{a}_{i}\,{{\exp }}\left(-\frac{x}{{b}_{i}}\right)$$
(3)

with a0 = 0.217, a1 = 0.259, a2 = 0.338, a3 = 0.186, b1 = 172.9 year, b2 = 18.51 year, and b3 = 1.186 year. These parameters used in the IRF were determined from a 400 GtC impulse emission in the Bern carbon cycle model52,60.

The CO2 radiative forcing (RF, or ERF) is then calculated based on the calculated $$\triangle {q}_{{{{{{\rm{CO}}}}}}2}$$(t) and the radiative forcing efficiency41 of 1.33 × 10−5 W m−2 ppbv−1. For H2 indirect forcings, a similar procedure is used based on the GFDL-AM4.1 radiative forcings efficiencies15. The H2 forcing is further decomposed into its various components corresponding to the stratospheric H2O, tropospheric and stratospheric ozone components and for these forcings the H2 lifetime is used.

A special treatment is used for the methane perturbation associated with the change in OH resulting from H2 emissions in order to account for the different lifetime of the perturbation. The methane lifetime associated with OH oxidation 9.7 years in the GFDL-AM4 model40 is ajusted to account for the stratospheric loss of methane with a lifetime of 120 years, and for the methane soil uptake with a lifetime of 160 years. Based on the change in OH of +8% for a 200 Tg H2 emission increase15, the steady state methane concentration due to the H2 perturbation is calculated with61,62;

$$\triangle {q}_{{{{{{\rm{CH}}}}}}4}^{{SS}}=f\,{q}_{{{{{{\rm{CH}}}}}}4}^{{ref}}\triangle {{{{{{\rm{\tau }}}}}}}_{{{{{{\rm{CH}}}}}}4}$$
(4)

where $$\Delta {q}_{{{{{{\rm{CH}}}}}}4}^{{SS}}$$ is the steady-state change in methane mixing ratio due to H2 increase in the atmosphere per TgH2, $${q}_{{{{{{\rm{CH}}}}}}4}^{{ref}}$$ (1808 ppbv) is the reference methane mixing ratio, f = 1.315 denotes the methane feedback on its own lifetime and ΔτCH4 (%/TgH2) is the change in methane lifetime per TgH2. We approximate the change in methane due to the H2 impact on OH with63:

$${\Delta {{{{{\rm{M}}}}}}}_{{{{{{\rm{CH}}}}}}4}\left({{{{{\rm{t}}}}}}\right)\,=\, \Delta {q}_{{{{{{\rm{CH}}}}}}4}^{{SS}}\,{\int }_{{t}_{0}}^{t}{\triangle {{{{{\rm{M}}}}}}}_{{{{{{\rm{H}}}}}}2}\left({t}^{{\prime} }\right)\,\left[(1-{e}^{-\frac{1}{{{{{{{\rm{\tau }}}}}}}_{{{{{{\rm{CH}}}}}}4}}})\delta \left({t}^{{\prime} }-{t}_{0}\right)\right.\\ +\left.(1-{e}^{-\frac{1}{{{{{{{\rm{\tau }}}}}}}_{{{{{{\rm{CH}}}}}}4}}}){e}^{-\frac{(t-{t}^{{\prime} })}{{{{{{{\rm{\tau }}}}}}}_{{{{{{\rm{CH}}}}}}4}}}(1-\delta \left({t}^{{\prime} }-{t}_{0}\right))\right]{dt}^{\prime}$$
(5)

where $$\delta$$(x) is the Dirac function. Each year following the hydrogen pulse emission, due to the change in H2, CH4 increases towards the steady-state perturbation (first term of equation) and decreases exponentially from this value the following years (second term). Supplementary Fig. S8a shows the response of the atmospheric mass of H2, CH4, and CO2 after a pulse emission of H2 (and CO2) of 1 Tg. H2 decreases exponentially after emission with a characteristic lifetime $${\tau }_{{{{{{{\rm{H}}}}}}}_{2}}$$and CO2 decreases with a much longer lifetime $${\tau }_{{{{{{{\rm{CO}}}}}}}_{2}}.$$ Methane resulting from the H2 increase in the atmosphere and the subsequent perturbation in OH, increases from 0 to a peak of 60 pptv 4 years after the initial 1 TgH2 emission. This value is in agreement with the 80 pptv calculated with the STOCHEM three-dimensional model28 for 1.67 TgH2 (i.e. 48 pptv/TgH2). The peak CH4 increase is however larger in our estimate, essentially due a shorter H2 tropospheric lifetime in STOCHEM (1.6 year) compared to GFDL-AM4.1 (2.1 year). Based on the calculated change in methane mixing ratio, the radiative forcing efficiency of 44.3 × 10−5 W m−2 ppbv−1 is used45 to derive the associated radiative forcing. The direct methane forcing is increased to account for the indirect tropospheric ozone (0.116 × 10−3 W m−2 ppbv−1) and stratospheric H2O (0.027 × 10−3 W m−2 ppbv−1) forcings64. The indirect CO2 produced from CH4 oxidation64,65 is not accounted for in this work. The tropospheric ozone and stratospheric water vapour forcings reported with GFDL-AM4.1 are associated with both H2 and CH4 changes. In our calculations, the methane forcing also accounts for the indirect tropospheric O3 and stratospheric H2O indirect contributions. In order to avoid double counting, we calculate the methane forcing at steady-state for a 1 Tg H2 perturbation. This forcing is higher that the methane forcing calculated with GFDL-AM4.1 due to these indirect ozone and stratropheric H2O contributions included. The tropospheric ozone and stratospheric H2O radiative forcing efficiencies from GFDL-AM4.1 are then rescaled in order to account for this difference and hence separate the H2 and CH4 impact on these forcings.

The climate response, in terms of global surface temperature change, is also estimated based on an IRF52:

$$\Delta T\left({{{{{\rm{t}}}}}}\right)={\int }_{{t}_{0}}^{t}{{{{{\rm{RF}}}}}}\left({{{{{\rm{t}}}}}}^{\prime} \right)\,{{{{{{\rm{IRF}}}}}}}_{{{{{{\rm{T}}}}}}}\left(t-t^{\prime} \right)\,{dt}^{\prime}$$
(6)

The climate response IRFT is provided by:

$${{{{{{\rm{IRF}}}}}}}_{{{{{{\rm{T}}}}}}}\left(x\right)=\frac{1}{\lambda }\mathop{\sum}\limits_{i=\{s,f\}}\frac{{c}_{i}}{{d}_{i}}{{\exp }}\left(-\frac{x}{{d}_{i}}\right)$$
(7)

where $$\lambda$$ is the climate sensitivity parameter, and subscripts f and s refer to the fast and slow climate responses, respectively. The updated coefficients66,67 of IRFT are: cf = 0.587, cs= 0.413, df = 4.1 year, and ds = 249 year. The sum of the cs and cf coefficients is normalized and the climate sensitivity parameter, $$\lambda$$ = 1.04 W/m2/K, is introduced in the IRF, assuming a 3.78 K warming for a CO2 doubling radiative forcing (3.93 Wm−2)41. Supplementary Fig. S8b, c show the resulting radiative forcing and temperature change after a 1Tg pulse emission of H2 (or CO2). The additional carbon cycle responses to temperature are excluded in our calculations68.

The Absolute GWP (AGWP) of H2 is calculated based on:

$${{{{{{\rm{AGWP}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}\left({t}_{h}\right)={\int }_{{t}_{0}}^{{t}_{h}}{{{{{{\rm{RF}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}\left(t\right)\,{dt}$$
(8)

The GWP of H2 is then calculated by definition as the ratio between the AGWP of H2 (or individual components) relative to the AGWP for CO2:

$${{{{{{\rm{GWP}}}}}}}_{{{{{{\rm{H}}}}}}2}\left({t}_{h}\right)=\frac{{{{{{{\rm{AGWP}}}}}}}_{{{{{{\rm{H}}}}}}2}\left({t}_{h}\right)}{{{{{{{\rm{AGWP}}}}}}}_{{{{{{\rm{CO}}}}}}2}\left({t}_{h}\right)}$$
(9)

This methodology is repeated in order to decompose GWPH2 into its various individual components (i.e., GWPCH4, GWPH2O, GWPO3t, GWPO3s), substituting $${{{{{{\rm{RF}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}$$by the individual radiative forcings of methane, stratospheric water vapour, tropospheric ozone, and stratospheric ozone.

The Absolute GTP (AGTP) of H2 is calculated based on:

$${{{{{{\rm{AGTP}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}\left({t}_{h}\right)={\int }_{{t}_{0}}^{{t}_{h}}{{{{{{\rm{RF}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}\left(t\right)\,{{{{{{\rm{IRF}}}}}}}_{{{{{{\rm{T}}}}}}}\left({t}_{h}-t\right)\,{dt}=\,\Delta {T}_{{{{{{{\rm{H}}}}}}}_{2}}\left({t}_{h}\right)$$
(10)

The GTP is calculated by definition as the ratio between the AGTP of H2 (or of its individual componenets) and the AGTP for CO2:

$${{{{{{\rm{GTP}}}}}}}_{{{{{{\rm{H}}}}}}{2}}\left({t}_{h}\right)=\frac{{{{{{{\rm{AGTP}}}}}}}_{{{{{{\rm{H}}}}}}2}\left({t}_{h}\right)}{{{{{{{\rm{AGTP}}}}}}}_{{{{{{\rm{CO}}}}}}2}\left({t}_{h}\right)}$$
(11)

The emission profile $${{{{{{\rm{E}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}({{{{{\rm{t}}}}}})$$ (or $${{{{{{\rm{E}}}}}}}_{{{{{{{\rm{CO}}}}}}}_{2}}$$) is taken to be a pulse emission of 1 Tg for the calculation of the pulse metrics GWPp and GTPp or a step sustained emission of 1 Tg for the sustained metrics GWPs and GTPs. It should be noted, that the exact definition of the GWP refers to an instantaneous pulse emission. In our calculations, a time-step of 1 year is considered, implying a 1-year pulse emission even for GWPp. This limitation induces a 10% difference in the GWPp,100 calculation3.

In order to derive the combined metrics44, we first define the Absolute Global Forcing Potential (GFP):

$${{{{{{\rm{AGFP}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}\left({t}_{h}\right)={\int }_{{t}_{0}}^{{t}_{h}}{{{{{{\rm{RF}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}\left({t}_{h}-t\right)\delta \left(t-{t}_{0}\right)\,{dt}={{{{{{\rm{RF}}}}}}}_{{{{{{{\rm{H}}}}}}}_{2}}\left({t}_{h}\right)$$
(12)

Similar to the AGTP which provides the temperature at a considered time-horizon, the AGFP is an end-point metric providing the radiative forcing at a considered time-horizon. The H2 Combined GWP (CGWP) is then calculated with:

$${{{{{{\rm{CGWP}}}}}}}_{{{{{{\rm{H}}}}}}2}\left({t}_{h}\right)=\frac{{{{{{{\rm{AGFP}}}}}}}_{{{{{{\rm{H}}}}}}2,{{{{{\rm{s}}}}}}}\left({t}_{h}\right)}{{{{{{{\rm{AGFP}}}}}}}_{{{{{{\rm{CO}}}}}}2}\left({t}_{h}\right)}$$
(13)

and the H2 Combined GTP (CGTP) is calculated with:

$${{{{{{\rm{CGTP}}}}}}}_{{{{{{\rm{H}}}}}}2}\left({t}_{h}\right)=\frac{{{{{{{\rm{AGTP}}}}}}}_{{{{{{\rm{H}}}}}}2,{{{{{\rm{s}}}}}}}\left({t}_{h}\right)}{{{{{{{\rm{AGTP}}}}}}}_{{{{{{\rm{CO}}}}}}2}\left({t}_{h}\right)}$$
(14)

where the subscript ‘s’ refers to the sustained emission metrics.

The uncertainties in the H2 and CH4 metrics are estimated following the exact same methodology as developed and presented by IPCC41. Since a large fraction of the H2 indirect climate impact is associated with methane, several uncertainties are similar for these two species. Supplementary Tables S1 and S2 provide, for respectively CH4 and H2, the total uncertainty calculated in various metrics, as a percentage of the best estimate and expressed as 90% confidence interval. The uncertainties are also provided by component of the total emission metric calculation (radiative efficiency, chemical response and feedbacks, atmospheric lifetime, CO2 combined uncertainty in radiative efficiency and CO2 impulse response function, carbon cycle response, fate of oxidized fossil methane, and impulse-response function for temperature increase). The uncertainties in individual terms are taken from Section 7.6 of IPCC AR6 report41, except for the CO2 impulse response which comes from Joos and Bruno60. The uncertainties are however modified for H2 when needed, in particular the lifetime uncertainty is estimated assuming a H2 atmospheric lifetime uncertainty of 0.5 year around our best estimate of 2.5 year. The uncertainties on the temperature impulse-response function were recalculated by taking 1.645 x standard deviation of the GTPs generated from 600 ensemble members of the impulse response derived from two climate emulators (see Forster et al.41 for more details). The resulting estimated total uncertainty in the H2 GWP is 31–36%. A value in agreement with previous estimates35. The uncertainty on the H2 GTP ranges from 131 to 64% and is largely dominated by the climate response function uncertainty. The uncertainty on the GTP is in the range 47–73%. These uncertainties on the metrics directly translate into similar uncertainties on the calculated CO2 equivalent emissions.

### Future hydrogen emission scenarios

The estimated hydrogen production in 2020 is 8753–9054 Mt H2/year. It is beyond the scope of this study to propose future hydrogen energy supply scenarios. Instead, we base our analysis on the scenarios developed by the Hydrogen Council for the worldwide hydrogen energy transition and on the European Union (EU) roadmap for a future hydrogen economy in Europe. These scenarios are succinctly summarized below and we refer the reader to the full reports for a more in-depth overview of the underlying economic and energy assumptions. The worldwide energy transition scenario used in this study is based on the scenario developed by the Hydrogen Council as a roadmap for long-term hydrogen deployment50. This scenario (referred to as HC2017) analyses the possible deployment of hydrogen in the various sectors: transportation, industrial energy, building heating and power, industry feedstock and power generation, and assumes an almost eight-fold increase in hydrogen demand between 2020 and 2050. According to this scenario the global energy supply of hydrogen was 10 EJ (2778 TWh) in 2020, and is projected to increase to 14 EJ (3889 TWh) in 2030; 28 EJ (7778 TWh) in 2040, and 78 EJ (21,667 TWh) in 2050. In this study, we use a hydrogen energy density of 142 MJ/kg as used in previous work3,4,11,50. This value is the Higher Heating Value (HHV) of the hydrogen fuel defined as the amount of heat released once it is combusted and the products have returned to a temperature of 25 °C, considering the latent heat of vaporization of water in the combustion products. Currently, the hydrogen technologies do not condense the water to use the full heating value of hydrogen. The hydrogen Lower Heating Value (LHV) of 120 MJ/kg, defined as the amount of heat released by combusting hydrogen and returning the temperature of the combustion products to 150 °C, and hence assuming that the latent heat of vaporization of water in the reaction products is not recovered, should be used. For consistency with previous work but also assuming an improvement of the hydrogen technologies over the next 30 years, we apply the HHV in our analysis. Based on this value we derive a production of 98 Mt H2/year in 2030, 197 Mt H2/year in 2040 and 549 Mt H2/year in 2050. Using the LHV instead of the HHV would increase the hydrogen production for a given energy demand by 18% and provide 650 Mt H2/year in 2050. The sensitivity of the results to the use of the LHV instead of the HHV will be illustrated. Please note that using the LHV (more consistent with current hydrogen technologies), we derive a hydrogen production in 2020 (estimated in 201750) of 83 Mt H2/year, a value lower by 4–8% compared to the actual production reported for 202053,54. In addition to this total energy supply, this scenario further assumes a phase-out of grey hydrogen by 2040 and an increased supply of blue and green hydrogen to about 33% grey, 33% blue, and 33% green in 2030; 50% blue in 2030, and 30% blue in 205051. For the year 2050, the HC2017 scenario results in an annual 6 GtCO2/year abatement. We assume the 2050 values over the period 2050–2100. Over the entire 2030–2100 period, the HC2017 scenario results in a cumulative CO2 emission abatement of 331 GtCO2.

There is a continued acceleration in hydrogen deployment and a strong momentum for future transition to an hydrogen economy. A more ambitious hydrogen scenario with hydrogen reaching 22% of the global energy demand in 2050 has been developed by the Hydrogen Council and McKinsey & Company54. According to this scenario (referred to as HCMK2021), the hydrogen worldwide end-use demand increases from 20 EJ (5555 TWh) in 2030; 55 EJ (15,278 TWh) in 2040; to 94 EJ (26,111 TWh) in 2050. This corresponds to 140, 385, and 660 Mt H2/year in 2030, 2040 and 2050, respectively. This scenario further assumes 50%, 30 and 20% of grey, blue, and green hydrogen, respectively, in 2030; 5%, 40%, and 55%, respectively in 2040; and 100% green hydrogen in 2050. The HCMK2021 scenario results in an annual CO2 emission abatement of 7 GtCO2/year in 2050. If we further assume the 2050 values constant for the following decades, over the entire 2030–2100 period, the HCMK2021 scenario results in a cumulative CO2 emission abatement of 417 GtCO2. Other scenarios for future hydrogen production have been developed. In particular, in its “Net Zero by 2050” roadmap, the International Energy Agency (referred to as IEA2021) assumes a future hydrogen supply of 212 Mt H2/year in 2030, 390 Mt H2/year in 2040, and 528 Mt H2/year in 205053. This scenario further assumes 30%, 32 and 38% of grey, blue, and green hydrogen, respectively, in 2030, and 9%, 38%, and 53%, respectively in 2040. In 2050, this scenario still assumes a significant fraction of blue hydrogen with 1% of grey, 37% of blue, and 62% of green hydrogen. The IEA2021 scenario results in an annual CO2 emission abatement of 5.7 GtCO2/year in 2050. If we further assume the 2050 values constant for the following decades, over the entire 2030–2100 period, the IEA2021 scenario results in a cumulative CO2 emission abatement of 353 GtCO2. Another set of three future scenarios for global scale transition to an hydrogen economy has also been proposed by Bloomberg NEF69. These scenarios estimate a total energy demand for hydrogen in 2050 of 27 EJ (190 Mt H2/year), 99 EJ (697 Mt H2/year), and 195 EJ (1373 Mt H2/year) for the “Weak Policy”, “Strong Policy” and “Theoretical Max” scenarios, respectively. We note that the “Strong Policy” scenario is close to the HCMK2021 scenario. In addition, the “Theoretical Max” scenario, which assumes that all unlikely-to-electrify sectors in the economy use hydrogen, suggests that there is an enormous potential for hydrogen use in the economy, resulting in a yearly CO2 abatement of almost 15 GtCO2/year in 2050. In this study we focus on the HC2017 and further illustrate the HCMK2021 and IEA2021 scenarios which cover a reasonable range allowing to assess the sensitivity of the results to the development of a future hydrogen economy, but we note that other scenarios are also available.

The European Union (EU) has issued an hydrogen roadmap aiming at generating 8.1 EJ (2250 TWh) of hydrogen in 2050, representing roughly a quarter of the EU’s total energy demand53. Achieving this objective would reduce EU annual CO2 emissions by about 560 Mt CO2 in 2050. Instead of using this scenario, we rather use a more recent and slightly more ambitious scenario developed for European future energy transition by the European Network of Transmission System Operators for Gas and Electricity (ENTSOG/ENTSO-E)55. The “Global Ambition” scenario55 (referred to here as TYNDP2022) pictures a EU pathway to achieving carbon neutrality by 2050 with a hydrogen demand developing as of 2030 and becoming the main gas energy carrier in 2050. According to this scenario 1.7 EJ (466 TWh) are produced from hydrogen in 2030; 5.7 EJ (1575 TWh) in 2040; and 9.0 EJ (2496 TWh) in 2050; corresponding to a supply of 12 Mt H2/year in 2030, 40 Mt H2/year in 2040 and 63 Mt H2/year in 2050. The TYNDP2022 scenario further assumes a phase-out grey hydrogen before 2030, and 51%, 13.5 and 6% of blue hydrogen in 2030, 2040, and 2050, respectively. This scenario reduces the annual CO2 emissions by 620 MtCO2/year in 2050. Assuming 2050 values for the following decades, the TYNDP2022 scenario leads to a mitigation of cumulative CO2 emissions of 36 GtCO2 over the full period 2030–2100.

### Carbon dioxide and methane emissions from blue and grey hydrogen

At present, 95% of the hydrogen is produced via the Steam Methane Reforming (SMR) process using fossil natural gas as a feedstock70. The CO2 emissions from grey hydrogen production were recently calculated4 considering both the CO2 emissions during the SMR process itself (38.5 gCO2/MJ), during the heat and electricity generation needed to drive the SMR process (31.8 gCO2/MJ), and the upstream emission from the energy used to produce, process and transport natural gas and hydrogen (5.3 gCO2/MJ). The sum of these three terms (75.6 gCO2/MJ), combined with the energy content of gaseous hydrogen (7 gH2/MJ) provides an emission of 10.8 kgCO2/kgH2. We also account for the methane needed to produce hydrogen4. The methane needed for the SMR process itself (14.04 gCH4/MJ) plus the amount burned to generate the heat and pressure needed for SMR (11.6 gCH4/MJ) provides a total of 25.6 gCH4/MJ, or 3.65 kgCH4/kgH2. Based on this total methane amount, we account for the leakage during production and use of unburned methane to the atmosphere with a leakage fraction flCH4 (see below). The total methane amount needed is slighly larger than the previous estimates of 3.0 kgCH4/kgH271 and 3.2 kgCH4/kgH25.

Blue hydrogen differs from grey hydrogen because some of the carbon dioxide released by the SMR process or/and heat and pressure generation is captured. We use capture rates4 of 85% efficiency for SMR and 65% efficiency for driving the SMR. The total CO2 emissions remaining, assuming CO2 capture for both processes, is therefore 15% × 38.5 gCO2/MJ + 35% × 31.8 gCO2/MJ = 16.9 gCO2/MJ. In addition, we consider the energy required to capture carbon dioxide4. The CO2 emissions associated with carbon sequestration during both SMR and energy production to drive SMR is 16.3 gCO2/MJ4. An additional emission of 6.5 gCO2/MJ is added for indirect upstream emissions, providing a total CO2 emission from blue hydrogen production and transport of 39.7 gCO2/MJ, or 5.7 kgCO2/kgH2. This value, which accounts in particular for the energy needed to ensure the carbon sequestration, is significantly higher than previous estimate51 of 1.5 kgCO2/kgH2.The methane amount needed for blue hydrogen is similar to grey hydrogen except for the amount associated with the increased energy needed to drive the carbon sequestration process4. This amount is estimated at 6.0 gCH4/MJ, providing a total ammout of CH4 of 31.6 gCH4/MJ, or 4.5 kgCH4/kgH2. Based on this total methane amount, we apply a leakage fraction flCH4 (see below) to account for unburned methane emission to the atmosphere. The equivalent CO2 emissions, at a given time-horizon, from methane leakage during the grey and blue hydrogen production are calculated by multiplying the methane fugitive emissions by the corresponding recalculated methane climate metric (see Supplementary Table S3).

The hydrogen leakage rate flH2 is varied in our calculations from 0.1 to 15% in order to calculate the hydrogen climate footprint. The overall methane leakage rate flCH4 was recently estimated at 3.5%4. In this study, we derive flCH4 based on the considered flH2. Due to its lower volumetric density compared to natural gas, it is estimated that H2 will be transported at three times the pressure of natural gas5. At these higher pressures and since the gas viscosity of H2 is lower than CH4, the leakage rate of H2 is calculated 3.7–4.5 times higher than CH4 (assuming that most of the flow is laminar). Based on these values, we assume a volumetric leakage ratio between flH2 and flCH4 of 4. For emissions, we are interested by the mass of leakage and this ratio needs to be divided by the density ratio of CH4 to H2 (0.72 kg m−3/0.09 kg m−3 = 8), providing flH2 / flCH4 = 0.55. The methane fugitive emission needs to be corrected in order to account for the fact that hydrogen is deployed in order to replace natural gas use29. A fraction of the CH4 fugitive emissions would hence also be released to the atmosphere under the use of fossil fuel natural gas use. We have assumed a methane amount of 25.6 gCH4/MJ needed for hydrogen production4. For methane (natural gas), we assume an energy content of 53.6 MJ/kg72 or 18.6 gCH4/MJ. The difference (25.6–18.6) 7.0 gCH4/MJ is the additional methane amount needed for hydrogen production. The calculated fugitive methane emission is hence corrected and multiplied by a 0.27 ratio (7.0/25.6). In the case of blue hydrogen, accounting for the additional methane needed for carbon sequestration provides a difference between hydrogen production and natural gas use of (31.6–18.6) 13.0 gCH4/MJ, and a correction factor of 0.41 (13.0/31.6). The methane leakage rate flCH4 is then equal to 0.54 flH2 in the case of grey hydrogen and 0.82 flH2 for blue hydrogen.