Combined Effects of Refrigerant Substitutions of Residential Air Conditioners and Improvement in Lifecycle Refrigerant Management on Reduction of Global Greenhouse Gas Emissions

Abstract

This study analyzes the effects on global greenhouse gas (GHG) emissions of various combinations of lifecycle refrigerant management (LRM) practices and refrigerant substitutions in residential air conditioners (ACs) until the year 2070. Six scenarios involving three refrigerant types with different levels of global warming potential (GWP)—high, medium, and ultralow—and three levels of LRM involving leakage reduction during operation and end-of-life refrigerant recovery are examined. The findings reveal that combining medium GWP refrigerants (e.g., R32) with the highest level of LRM could achieve as much as a 95% reduction in emissions (933 MtCO₂eq) by 2050 and a 97% reduction (1660 MtCO₂eq) by 2070, when compared to using high GWP refrigerants (e.g., R410A). The substitution of ultralow GWP refrigerants (e.g., R290) is projected to achieve up to a 97% emissions reduction (954 MtCO₂eq) by 2050 and a 100% (1709 MtCO₂eq) reduction by 2070. Global mean temperature decreases in 2070 are nearly identical under scenarios in which either medium GWP refrigerants or ultralow GWP refrigerants are combined with the highest level of LRM (0.067 °C versus 0.069 °C). The implication is that combining medium GWP refrigerants, already underway in some regions, with the highest level of LRM offering an effective and pragmatic strategy for mitigating the climate impacts of refrigerant emission from residential ACs.

1. Introduction

1.1. Background

Global cooling demand is expected to grow significantly in the coming years, particularly in developing countries such as India, China, and Indonesia [1]. UNEP [1] projects that by 2050, if current trends continue, the installed capacities of cooling equipment will triple, electricity consumption will more than double, and greenhouse gas (GHG) emissions from cooling will reach 6.1 billion tons of CO₂ equivalent, raising serious environmental concerns. Cooling technologies such as air conditioning, refrigeration, and freezing will be crucial for purposes such as protecting people from high temperatures, maintaining the quality of foods and medicines, and enhancing labor productivity through improved workplace comfort. Given this situation, UNEP [1] has identified three key action areas to pursue in transitioning to sustainable cooling: (1) passive cooling such as insulation and ventilation; (2) higher energy efficiency; and (3) an accelerated phasing down of climate-warming refrigerants.

In 2023, the Global Cooling Pledge was announced at the 28th Conference of the Parties (COP28) to the United Nations Framework Convention on Climate Change (UNFCCC) as part of the Cool Coalition initiative of UNEP [2]. The pledge seeks to reduce cooling-related GHG emissions by 68% by 2050 relative to 2022 levels. Additionally, at the Group of Seven (G7) Climate, Energy, and Environment Ministers Meeting in 2024, the following statement was included in the group’s final communiqué: “We will simultaneously address the refrigerant choice and energy efficiency of equipment, while maintaining the safety of the equipment, in order to further reduce HFC emissions and maximize the benefits of newly introduced energy-efficient heating and cooling systems.” [3].

GHG emissions from air conditioners (ACs) and refrigeration equipment are classified into three categories in order of emissions volume: (1) CO₂ emissions associated with electricity consumption during operation, (2) refrigerant emissions due to leakage during operation and atmospheric release at the end of life (EOL), and (3) CO₂ and refrigerant emissions during manufacturing. Most of the GHG emissions are CO₂ emissions associated with electricity consumption during operation, depending on the CO₂ intensity of the electricity [4,5].

The largest portion of global anthropogenic greenhouse gas (GHG) emissions is CO₂, followed by CH₄ and NO₂. Fluorinated greenhouse gases (F-gases), which include HFCs (hydrofluorocarbons), PFCs (Perfluorocarbons), SF₆ (sulfur hexafluoride), and NF₃ (Nitrogen trifluoride), are the fourth largest contributor, accounting for approximately 2% of the anthropogenic GHG emissions. HFC emissions account for approximately 90% of F-gas, with the majority used as refrigerants in refrigeration and air conditioning equipment [6].

Refrigerants utilized in refrigeration and air conditioning equipment have traditionally included chlorofluorocarbons (CFCs) and hydrochlorofluorocarbons (HCFCs). Due to their chlorine content and ozone-depleting properties, the Montreal Protocol was adopted in 1987 to protect the ozone layer. The production and consumption of ozone-depleting substances (ODS) were to be phased out, and the transition to HFCs with an ozone depletion potential of zero was promoted as an alternative to CFCs and HCFCs. While HFCs do not contribute to ozone layer depletion, it was later discovered that they have a significant impact on the greenhouse effect. The Kigali Amendment (KA) to the Montreal Protocol, which was adopted in 2016, established a schedule for gradually reducing HFC production and consumption (HFC consumption includes production HFCs plus import minus export HFCs). The schedule for HFC reductions was set for two classifications of countries: developed countries and developing countries. Developed countries are required to reduce HFCs by 85% by 2036. Developing countries, which are major consumers of ACs, are further disaggregated into two groups based on varying national circumstances and capacities. The first group (Gr1) includes most developing countries (e.g., China), while the second group (Gr2) includes countries such as India, Pakistan, and the countries of the Middle East. Gr1 countries are expected to reduce HFCs by 80% by 2045, while Gr2 countries are expected to reduce HFCs by 85% by 2047.

For residential ACs, R22 (HCFC) and R410A (classified as an HFC) have been widely utilized as refrigerants. R410A is an HCFC alternative with zero ozone depletion potential (ODP) [7]. However, because both refrigerants have high global warming potential (GWP), the substitution of R32, which has a GWP value approximately one-third that of R410A, is underway in many countries. The adoption of ACs utilizing R32 was largely completed in Japan around 2016 and is now spreading globally. However, further efforts to substitute refrigerants with much lower GWP will be required to contribute to achieving net-zero GHG emissions. So far, there is still no practical ultralow GWP alternative to R32 and R454B for residential ACs. Any acceptable alternative would need to satisfy the essential requirements of environmental performance, energy efficiency, cost-effectiveness, and safety [8,9].

While the KA encourages all parties to phase down their HFC emissions (expressed in CO₂ equivalents) in their consumption and production, it does not explicitly deal with emissions reduction over the entire lifecycle of the product, which would include both operation and EOL. More recently, discussions dealing with the transition to sustainable cooling have begun to include lifecycle refrigerant management (LRM), which focuses not only on the production process but also on refrigerant management practices aimed at minimizing emissions throughout the product lifecycle. For example, LRM for fluorocarbons, including HFCs, CFCs and HCFCs, was highlighted as one of the commitments of the Global Cooling Pledge. Additionally, a workshop on LRM was held at the 36th Meeting of the Parties to the Montreal Protocol (MOP36) in 2024, indicating growing international interest in LRM [10,11].

Table 1. GWP classification of refrigerant.

Classification	GWP-100 Year	Representative Refrigerants
High GWP refrigerants	1000–3000	R22, R410A, and R407C
Medium GWP refrigerants	300–1000	R32, and R454B
Ultralow GWP refrigerants	<30	R290, R474A, R1234yf, and R744

Source: Table 3-1 of UNEP/TEAP [12].

shows the GWP classifications used throughout this study, following UNEP/TEAP [12]. In this study, refrigerants are classified into three groups based on their GWP values.

1.2. Literature Review

Several prior studies featuring scenario analyses of HFC emissions reductions have been conducted. Some of the more recent studies have revealed that if current trends continue and no additional measures are taken, HFC emissions related to refrigerants will increase significantly [13,14,15]. Akanle [16] discussed the various interconnections between efforts to protect the ozone layer and mitigating climate change. Hurwitz [17] discussed the impact of HFCs on climate change, particularly focusing on their role as refrigerants in residential ACs. His study used simulations based on a chemistry-climate model to show that early actions on HFC mitigation can be highly effective and that more than 90% of the climate change impacts of HFCs are avoidable.

Studies have also been conducted to compare alternative refrigerants. Wang et al. [18] evaluated R22, R410A, and R290 emissions from residential ACs in China. Their analyses demonstrated that R290 brought significantly greater environmental benefits, both directly by reducing greenhouse gas emissions due to its ultralow GWP and indirectly because of its higher energy efficiency than either R22 or R410A. The authors concluded that if the safety issue with R290 (R290 is highly flammable) was overcome, the refrigerant could become an important substitute in residential ACs. Li et al. [19] conducted scenario analyses of GHG emissions in China through 2050 and found that the emissions reduction potential of residential ACs is the largest among all sectors. The reduction combines three key factors: replacing high/medium GWP refrigerants (R410A/R32) with ultralow GWP R290, making significant energy efficiency improvements in cooling systems, and decarbonizing the electricity supply. The most stringent mitigation scenario assumes complete market penetration of R290 by 2045, alongside enhanced energy performance standards and power sector transformation, creating a comprehensive approach beyond mere refrigerant substitution. The U.S. EPA [14] performed a marginal abatement cost analysis of non-CO₂ GHG emissions by country, evaluating R32/R452B with microchannel heat exchangers and R290 as reduction options for residential ACs. According to the data annex of the EPA [13], the global reductions in 2050 due to R32/R452B and R290 in residential ACs could reach approximately 760 MtCO₂eq relative to the business-as-usual scenario.

Recent modeling approaches have examined HFC emission impacts that include those of ACs. Velder et al. [20] conducted a model analysis focused on HFC emissions that included stationary ACs until the year 2050. It was shown that stationary ACs have a large HFC emissions volume, as do industrial and commercial refrigerators. The study’s findings indicated that, without KA compliance, total HFC emissions would result in a global surface warming of 0.28−0.44 °C by 2100, as compared to 0.14−0.31 °C with compliance.

Purohit et al. [15] evaluated the GHG emission reduction effects of replacing R410A with R32 and R290 refrigerants in residential ACs using the Greenhouse Gas and Air Pollution Interactions and Synergies (GAINS) model. The study assumed that refrigerant alternatives with the ultralow GWP of R290 would be dominant after 2060 and found significant reductions in GHG emissions due to the use of these refrigerants. In a later study, Purohit et al. [21] discussed the significance of LRM and used the GAINS model to analyze the reduction in HFC/HCFC emissions for all sectors due to LRM. It was estimated that the use of an enhanced LRM that includes leakage prevention and high recovery at EOL would reduce HFC/HCFC emissions by approximately 39 GtCO₂eq between 2025 and 2050.

The concept of a circular economy has emerged as an important framework for addressing refrigerant emissions. Castro et al. [22] pointed out the importance of strategies to manage F-gas, including handling and operation, disposal, destruction, and recycling, in addition to production and consumption, as a means to mitigate its impact on climate change. The study’s particular focus is on EOL, arguing that F-gas is a valuable material that is well-suited for recycling due to its high stability. The authors also emphasize the need to intensify efforts to separate and recycle value-added F-gas in order to promote cost-effective adaptation to its phase-down based on circular economy principles.

1.3. Focus of This Study

Previous studies analyzing the climate change impacts of ACs and their refrigerants have generally focused on four primary aspects: (1) CO₂ emissions from increased electricity consumption during operation; (2) emissions reductions through improved energy efficiency; (3) GHG emissions from increased use of HFCs as refrigerants; and (4) emissions reductions through refrigerant substitutions. UNEP [1] reported that emissions related to the electricity consumption of ACs were higher than those related to refrigerants. However, with the exception of Purohit et al. [21], such studies have failed to provide quantitative analyses of the effects of measures directly aimed at suppressing refrigerant emissions, such as reducing refrigerant leakage during operation and recovering refrigerants at EOL.

To fill this gap, this study seeks to quantitatively clarify the GHG emissions and reduction effects of refrigerant substitutions by combining them with LRM considerations (including leakage reduction during operation and refrigerant recovery at EOL) for residential ACs. In contrast to Purohit et al. [21], our study explicitly shows the effects of LRM for residential ACs, taking into account differences in LRM enhancement by country and by process.

The remainder of the paper is organized as follows: Section 2 describes the methodology and data used for this study. Section 3 shows the GHG emissions results for the various mitigation scenarios involving residential AC refrigerants and their contributions to global surface temperature reduction. The effects of the production and consumption of HFCs are also estimated from the perspective of the KA. Section 4 discusses the results and their implications. Section 5 summarizes the paper.

2. Methods

2.1. Estimation Formula

This study estimates refrigerant emissions produced by residential ACs from three related processes—production, operation, and EOL—as applied to the UNFCCC GHG inventories, where the diagram of refrigerant life cycle showing the various stages were provided in EPA [23]. The estimation method for the three processes follows the GIO [24] approach to estimating the GHG emissions of residential ACs, consistent with the 2006 IPCC guidelines. Residential AC stocks during operation are first estimated based on cooling degree days (CDDs). Next, the number of units in the production and EOL processes are estimated. Finally, the GHG emissions corresponding to the refrigerants in the individual processes are estimated.

Based on the applicable guidelines, the refrigerant emissions for the three AC-related processes are formulated as follows:

$$EM=\sum (NP\ast MM\ast XM)$$

(1)

$$EO=\sum (NO\ast MO\ast XO)$$

(2)

$$ED=\sum \left(ND\ast MD\right)-RE$$

(3)

EM, EO, and ED are the refrigerant emissions at the manufacturing stage, during operation, and at the EOL stage, respectively. NP, NO and ND are the number of units produced, operated, and disposed of, respectively. MM, MO, and MD are the average amounts of refrigerant in each of the three processes. XM and XO are the refrigerant leakage rates during each of the processes. RE is the amount collected, based on the relevant IPCC guidelines. This study does not assume reuse of equipment collected at EOL because of focusing on only residential ACs.

The stocks of residential ACs are obtained using the formulation described by Gi et al. [25]. This study also adopts the parameters assumed by Gi et al. [25], as indicated in Figure S1 of the Supplementary Information, except for the future per capita GDP and temperature change scenarios. Residential cooling demand is estimated from a household scenario by country/region. Cooling demand per household assumes sigmoid functions that incorporate both per capita GDP and CDDs as independent variables, following the methodology established by Gi et al. [23] for a temperature pathway meeting the below 2 °C target (compared with the preindustrial level) by 2100. The future projections of population, households, and GDP provided in the SSP2 “middle of the road” scenario in the SSPs (Shared Socioeconomic Pathways) [26,27] are used. The stocks of ACs are estimated using operating hours and capacity per unit. Operating time is assumed to be a linear function proportional to 0.25 times the CDD increment relative to 2015. The estimates were conducted for 54 regions/countries which cover the whole world, and they were aggregated into seven regions to represent the regional results.

The number of AC units disposed of is estimated using the function provided by Tasaki et al. [28]. We assume that the proportion Wt(y) of units disposed of by the end of year y follows the Weibull distribution probability density function.

$$Wt\left(y\right)=1-\mathrm{e}\mathrm{x}\mathrm{p}[-{\left\{{\displaystyle \frac{y}{\overline{y_t}}}\right\}}^b{·\left\{\mathsf{\Gamma }\left(1+{\displaystyle \frac{1}{b}}\right)\right\}}^b\}],$$

(4)

where $\mathsf{\Gamma }$ represents the gamma function and parameter b represents the range of operating years and is assumed to be 3.0 based on the approximation by Tasaki et al. [28]. The parameter $\overline{y_t}$ represents the average lifetime of an AC: the average lifetime in developed countries is assumed to be 15 years, based on a survey of Japan conducted by METI [29]. Reflected here is the trend of an increasing number of AC operating years. The average lifetime of ACs in developing countries is assumed to be 10 years, based on a survey of China reported by Guo et al. [30].

Consider $F_t\left(i\right)$, the percentage of ACs newly installed in year t − i that are disposed of by the end of year t. Since ACs newly installed at the beginning of year t − i are operating for i + 1 years and ACs newly installed at the end of the year t − i are operating for i years, the average percentage disposed of within i + 0.5 years is $F_t\left(i\right)=W_t(i+0.5)$, which is the fraction of products disposed of within i + 0.5 years on average, where $F_t\left(i\right)=0$ and $W_t\left(0.5\right)=0$ are assumed. For simplicity, the disposals with very small probabilities (less 0.2%) beyond 30 years are accounted for collectively in the 30th year. Accordingly, if we let $f_t\left(i\right)$ represent the fraction of ACs newly installed in year t − i that are disposed of in year t, ${NI}_{t-i}$ represent the number of newly installed ACs in year t − i, and ${ND}_t$ represent the number of ACs disposed of in year t, then the following equations can be specified:

$${ND}_t={\sum }_{i=1}^{30}{ND}_{t,t-i}={\sum }_{i=1}^{30}\left\{{NI}_{t-i}·f_t\left(i\right)\right\}$$

(5)

$${NI}_t={ND}_t+{NO}_t{-NO}_{t-1}$$

(6)

Given the values of ${NO}_t, {NO}_{t-1}$ during operation and the value of ${ND}_t$ at EOL, the number of newly installed ACs, ${NI}_t$, can be estimated. ${NI}_t$ and ${NO}_t$ can then be used to determine ${ND}_{t+1}$. The historical number of newly installed ACs in Japan each year through 2023 can be found in JRAIA [31]. For countries other than Japan, we calculated the number of newly installed ACs using the assumed stock during operation and the number of ACs that are disposed of.

2.2. Assumed Refrigerants

Table 2. Assumed scenarios.

Scenario	Refrigerant Type	Ultimately Dominant Refrigerant (Representatives)	Leakage Reductions	Recovery Improvements
S 1	High GWP (Reference)	R410A	Normal	Normal
S 2-1	Medium GWP and Flammable Gas	R32	Normal	Normal
S 2-2	Medium GWP and Flammable Gas	R32	Enhanced	Normal
S 2-3	Medium GWP and Flammable Gas	R32	Enhanced	Enhanced
S 2-4	Medium GWP and Flammable Gas	R32	Enhanced	Strongly Enhanced
S 3	Ultralow GWP and Extremely Flammable Gas	R290	Enhanced	Strongly Enhanced

shows the six GHG emission scenarios and the associated refrigerants assumed in this study. Since representing the actual refrigerant replacement situation in each country would result in a highly complex model, the refrigerant substitutions were modeled using four refrigerant classifications: (1) ODS and high GWP refrigerants (e.g., R22), (2) high GWP refrigerants (e.g., R410A, R407C), (3) medium GWP refrigerants (e.g., R32, and R454B), and (4) ultralow GWP refrigerants (e.g., R290, R474A, R1234yf, and R744). The starting periods for refrigerant substitution were set in the order of (1), (2), (3), and (4).

We selected R22, R410A, and R32 as representative refrigerants for categories (1), (2), and (3), respectively, taking into account the history of widespread adoption. The flammability classification used in the study is based on ISO [32]. As for the representative refrigerant in category (4), since an ultralow refrigerant that meets all the requirements for environmental performance, efficiency, economy, and safety has yet to be identified, we selected R290 as an example of ultralow GWP refrigerants despite the safety concerns. So far, all of the potential refrigerants for category (4) have some issues to be overcome, and other potential refrigerants will be required to develop the emissions reduction effects in future research. Since the difference in GWP values is negligible, the results can also be applied to R474A, R1234yf, and R744.

This study employs 100-year GWP values based on the fifth assessment report of the IPCC (AR5). The GWP-100 values for R22, R410A, and R32 are 1760, 1920, and 677, respectively. The one exception is that we used 0.02 as the GWP-100 for R290 according to the sixth IPCC assessment report (AR6), since it is not provided in AR5. While the Montreal Protocol employs GWP values reported in the fourth assessment report of the IPCC (AR4) and latest values are also available in the AR6, the GWP values by the AR5, which are used in the current inventory of GHG emissions by the UNFCCC, were basically employed. The main factors determining the amount of refrigerant charged are the density of condensed refrigerant and the pressure loss in the heat exchangers and pipes. In fact, the amount of refrigerant charged will also depend on product design specifications such as equipment efficiency. Considering the density of the liquid refrigerant at a condensation temperature of 45 °C, pressure loss characteristics and some design factors such as subcooling and pipe diameter, we set the relative ratios of the amount of charge for each refrigerant as 1.0, 1.0, 0.7, and 0.5, for R22, R410A, R32, and R290, respectively. We assumed that the average refrigerant charge in all regions is 1000 g/unit based on GIO [24] and the unit capacity (kW/unit) by country is based on IEA [33].

Scenario S1, which assumes that only R410A is used during the evaluation period, serves as the reference scenario. In scenario S2-1, R32 is the dominant refrigerant. Thus, comparing S1 and S2-1 allows us to evaluate the effects of substituting R32 for R410A. In both S1 and S2-1, the leakage and recovery rates remain unchanged from their initial values; that is, no improvement was made to the rates.

S2-2 features enhanced leakage control (relative to S2-1) during the operation process. Scenario S2-3 adds enhanced recovery control at EOL to the enhanced leakage control in S2-2. In S2-4, the enhanced recovery control is stronger than in S2-3. The effects of these improved recovery measures can be evaluated by comparing S2-2, S2-3, and S2-4. Scenario S-3 features an ultralow GWP refrigerant (R290).

This study simplified the substitution pattern for refrigerants, assuming a sequential transition from R22 to R410A and then to R32 in all the S2 scenarios. Thus, for example, the direct substitution from R22 to R32 (not thorough R410A), for example, was not considered. In the case shifting from R22 to R410A through R410A, cumulative GHG emissions could be higher compared to those direct shift to R32. In addition, GHG emissions could be reduced by approximately 25% compared with those estimated in the simplified scenarios, which assumed R32 only as a medium GWP refrigerant, if R454B, which is a medium GWP refrigerant (GWP: 466) and is widely used in the Unites States, was completely employed instead of R32.

For our analysis, we divided the world into seven regions or classifications in Figure 1. The developed country classification includes (1) Japan, (2) North America, (3) Europe and Oceania, and (4) Russia and other Former Soviet Union (FSU) countries. Based on the KA, the developing country classification includes two subgroups: Gr 1 ((5) China and (6) other Gr1 countries) and (7) Gr 2 (India, Pakistan, and the Middle East).

2.3. Assumed Emission Leakages, Recovery, and Schedule of Refrigerant Substitutions Considering Differences Across Countries

The leakage rates for refrigerants, their recovery rates, and the schedule of refrigerant substitutions were considered in the various scenarios.

The assumed leakage rates for refrigerants during operation are shown in Figure 2. As indicated, S-1 and S2-1 feature a constant leakage rate until the year 2070. The leakage rate for Japan in S-1 and S2-1 is 2% (the rate that appeared in the government’s 2008 report). Although this 2% value has not been revised since 2008, scenarios S2-2, S2-3, S2-4, and S3 assume that improvements in design and manufacturing quality, construction technology, and service quality will reduce the leakage rate to 1% by 2030. The assumptions were based on the trend that the average leakage rate of all equipment will decrease over time due to decreases in leakage rates of newly installed equipment [34]. For the other countries, the leakage rate was set according to the IPCC’s GHG emission calculation guidelines. It should be noted that these assumptions, considering economic factors and insufficient qualification of personnel, are not supported by empirical verification. Other countries are expected to promote LRM in order to achieve their KA targets: S2-2, S2-3, S2-4, and S3 assume that the leakage rates in all countries will improve.

The refrigerant recovery rates at EOL are shown in Figure 3. Scenarios S1, S2-1, and S2-2 assume that the current situation in which refrigerant recovery is not being implemented in any of the developing countries will remain unchanged until 2070. (It is assumed that refrigerant recovery in most developing countries is not being implemented because prevailing legal systems, insofar as these apply to the reclamation of home appliances after disposal, are inadequate or poorly functioning and because there are no real incentives for refrigerant recovery.)

The recovery of EOL ACs in Japan is based on the Home Appliance Recycling Law (Act for the Recycling of Specified Home Appliances). The recovery rate has been steadily improving, reaching 41.2% in 2022 [35]. Despite the inherent uncertainties, it is assumed that the current upward trend will continue through 2030, at which time the recovery rate will reach 45%, and that this 45% rate will remain in effect through 2070.

S2-3 represents a world in which environmental concerns are given greater importance and recovery rates are significantly improved based on the incremental targets of the KA. S2-3 assumes recovery rates up to 90% for developed countries and up to 50% for developing countries. In Japan, the 2022 recovery rates for televisions, washing machines, and refrigerators—the three items covered by the Home Appliance Recycling Law—were 93.4%, 92.5%, and 87.7%, respectively [36]. Given the high recovery rates for these three items, a rate of 90% for ACs is considered to be perfectly feasible for developed countries. Until now, relatively little attention has been paid to the recovery of refrigerants when cooling equipment is disposed of. However, the number of countries promoting legislation related to electronic waste, including managing EOL electrical equipment through regulated collection and material recovery processes, is gradually increasing [37]. Given the growing interest in creating a circular economy, the movement to reduce GHG emissions and to enhance LRM is expected to advance rapidly. In this study, it is assumed that a recovery rate of 50% is attainable in the near future in most countries and that, ultimately, a recovery rate of 90% will be reachable. The regional differences in rates shown in Figure 3 reflect regional differences in the reduction targets set by the KA. Scenario S2-4 features significantly greater improvements than are assumed in S2-3, with recovery rates of 90% for all countries after 2045. We assumed, conservatively, the complete absence of recovery until 2030 in some countries in all scenarios. Future studies will require exploring sensitivity analysis to recovery rates, such as gradual increases in recovery rates.

Finally,

Table 3. Assumed time for refrigerant substitutions.

Region Category	Country/ Region	R410A (S1)		R32 (S2-1–S2-4 and S3)		R290 (S3)
		A1	A2	B1	B2	C1	C2
Developed countries	Japan	1998	2001	2013	2016	2035	2040
	US and Canada	2008	2011	2024	2028	2040	2045
	Europe and Oceania	1999	2002	2016	2020	2040	2045
	Russia and other FSU	1999	2002	2025	2030	2040	2045
Developing countries Gr1	China	2013	2017	2017	2026	2040	2050
	Other countries	2013	2014	2014	2026	2040	2050
Developing countries Gr2	India and Pakistan	2010	2013	2013	2024	2045	2055
	Middle Eastern countries	2013	2025	2025	2030	2045	2055

Note: A1 and A2 represent the year when R410A refrigerant exceeds 20% and 80% shares of new installations, respectively. B1 and B2 represent the period when R32 refrigerant exceeds 20% and 80% shares of new installations, respectively, in the scenarios except for S1. C1 and C2 represent the period when R290 refrigerant exceeds 20% and 80% of new installations, respectively, in the scenarios, except for S1.

shows the calendar years in which the various substitute refrigerants are introduced in newly installed ACs. We assumed that replacement times resulted from cooling performances and costs, mainly based on the historical Japanese experiences. For example, substituting from R410A to R32 improves performance [38], and reduces costs when comparing products with the same performance. The cost reductions resulted in substitution. Because we did not gain the detailed datasets of residential ACs for countries sufficiently, we assumed replacement periods from the limited recognition of refrigerants. In scenario S3, the substitution of ultralow GWP refrigerants is assumed to completely eliminate the global warming impact of refrigerants. Because some ultralow GWP refrigerants have safety concerns, minimizing leakage and ensuring safe handling at EOL will be essential to the widespread diffusion of such refrigerants. In this study, the installation years were set assuming the highest level of leakage reduction and maximum recovery rates in order to meet the GHG emissions reduction schedule specified in the KA.

As noted earlier, under the KA, developed countries are required to achieve an 85% reduction in GHG emissions (compared to 2011–2013 levels) by 2036. For these countries, we assume that the installation of new ACs using ultralow GWP refrigerants begins in 2035. On the other hand, the KA expects the developing countries in Gr1 and Gr2 to achieve an 80% reduction by 2045 (compared to 2020–2022 levels) and an 85% reduction by 2047 (compared to 2024–2026 levels), respectively. Accordingly, we assume that the installation of ACs using ultralow GWP refrigerants will begin in 2045.

Based on the historically observed average adoption rates for R410A and R32, the adoption rates for newly installed AC units using the substitute refrigerants are assumed to be five years in developed countries and ten years in developing countries. The refrigerant substitution from R410A to R32 in new installations of residential ACs in Japan occurred within approximately four years from 2013 to 2017 [39]. Based on this Japanese experience, we assumed five years in developed countries.

3. Results

Our analysis of global refrigerant emissions from residential ACs reveals significant patterns across the different scenarios, regions, and processes. Figure 4 shows the temporal evolution of these emissions under various mitigation scenarios from 2010 to 2070. Figure 5 provides a breakdown of the emissions by process and refrigerant type.

Table 4. Global refrigerant emissions in 2050 and 2070 and emission reduction ratios.

Year	Emissions/Reduction	S1	S2-1	S2-2	S2-3	S2-4	S3
2050	Emission (MtCO2eq)	982	277	138	71	49	28
	Emission reduction ratio (%, relative to S1)	0%	72%	86%	93%	95%	97%
	Emission reduction ratio (%, relative to S2-1)	-	0%	50%	74%	82%	90%
2070	Emission (MtCO2eq)	1710	425	177	93	51	1
	Emission reduction ratio (%, relative to S1)	0%	75%	90%	95%	97%	100%
	Emission reduction ratio (%, relative to S2-1)	-	0%	58%	78%	88%	100%

shows the global refrigerant emissions in 2050 and 2070, along with the emission reduction ratios relative to S1 and S2 for the various scenarios. As described in Figure 4 and Figure 5, global refrigerant emissions from residential ACs in the reference scenario (S1), which assumes that the KA will not be implemented and that R410A will continue to be used, increase markedly from 2010 to 2070, reaching approximately 1.7 GtCO₂eq annually in 2070. The results highlight that leakage reduction during operation offers the greatest potential for emissions mitigation.

Figure 6 shows refrigerant emissions from residential ACs by country/region, revealing significant geographical variations in both emissions and mitigation potential. Developing countries experienced the fastest growth in emissions under S1, reflecting increasing AC adoption. Their share of global emissions is projected to be approximately 64% and 78% in 2050 and 2070, respectively.

Meanwhile, the analysis of the alternative scenarios demonstrates the potential for significant emission reduction. In S2-1, where medium GWP R32 replaces R410A, emissions decrease by approximately 72% by 2050 (compared to S1) and by approximately 75% by 2070. Thus, a substantial reduction effect from substituting R32 can be expected. Since a large number of ACs are likely to be installed over this period, particularly in developing countries, the level of refrigerant emissions in 2070 under scenario S2-1 is expected to be nearly the same as in 2025.

While the reduction effect of refrigerant substitution is significant, the effect of enhanced LRM is also shown to have a major effect on emissions reduction. In S2-2, which features enhanced LRM to reduce leakage rates, emissions are reduced by approximately 86% and 90% by 2050 and 2070, respectively, compared to scenario S1. The refrigerant emissions reduction in S2-2 compared to that in S2-1 is also notable. The emissions of S2-2 in 2070 are 40% compared to that of S2-1 in 2025.

In S2-3 and S2-4, which assume enhancements in refrigerant recovery, emissions in 2070 are reduced by approximately 95% and 97%, respectively, compared to S1, indicating the significant effectiveness of an enhanced LRM that includes refrigerant recovery at EOL. The most aggressive mitigation scenario, S3, which involves both a fully enhanced LRM and the substitution of the ultralow GWP refrigerant R290, indicates the possibility of a nearly 100% emissions reduction by 2070 relative to S1 in 2020.

The estimated mitigation effect of enhanced LRM in developing countries under S2-4 is a 97% reduction in emissions in 2070 relative to S1. This represents a significantly larger effect than that when substituting alternative refrigerants due to the long lag time behind developed countries. Even in terms of cumulative emissions for the period 2020−2070, an 87% reduction relative to S1 is possible under S2-4. Moreover, our analysis shows that an 89% reduction is achievable under S3, the most ambitious scenario.

More moderate emissions growth in S1 and earlier emissions reductions in the mitigation scenarios are expected in the developed countries compared to those in the developing countries. The reason for this is that residential ACs have already been widely diffused in the developed countries and refrigerant substitutions have been implemented early, as specified in the KA. This suggests that mitigation strategies that combine refrigerant substitution with the most enhanced LRM can achieve the near-complete elimination of refrigerant emissions from residential ACs by 2070, underlining the importance of technological advancements in refrigerant substitutions and operational improvements in enhanced LRM in addressing the climate impact of cooling systems.

Figure 7 shows the changes in global mean surface temperature (relative to S1) under the assumed scenarios. The MAGICC6 model [40], a simplified climate model, was used for these estimates. Various parameters of the model were tuned to meet the representative projections made by multiple AOGCMs (Atmosphere-Ocean General Circulation Models), as well as past climates. The simplified climate model has been used to calculate global mean surface temperatures for the long-term emissions scenarios in the IPCC’s AR5. Refrigerant kg equivalents of HFC32 and HFC125 for residential ACs were input into the model. This conversion reflects the fact that R410A is a mixture of HFC32 and HFC125, while R32 consists of HFC32. Using these as the inputs for MAGICC, the difference in global mean temperature between scenarios was estimated. Assuming that the other GHG emissions except for those from residential ACs are the same as those used for estimating CDDs of the below 2 °C target by 2100 described in Section 2, the global temperatures for 2050 and 2070 under S1 are approximately 1.79 and 1.84 °C (compared with the preindustrial level), respectively.

In 2070, the mitigation of global temperature increases under the S2-1, S2-4, and S3 scenarios is estimated to be 0.051, 0.067, and 0.069 °C, respectively, compared to S1. Although S3 shows the greatest potential for mitigating global temperature increases, the difference between its effect and that of S2-4 is small, indicating that even with the use of R32, a medium GWP refrigerant, the deleterious environmental impact can be reduced to nearly the same level as in S3 (where R290, which has a near-zero GWP, is used) by improving the leakage and recovery rates. These findings imply the importance of a multiple pathway approach that focuses not only on lower GWP refrigerant substitutions but also on LRM practices that minimize leakage and enhance recovery rates, even (or especially) when only a medium GWP refrigerant such as R32 is being used.

Figure 8 shows the global reductions in HFC production and consumption achieved through enhanced LRM. These global reductions consist of leakage reductions during operation as well as increases in the amount of refrigerant recovered and reclaimed at EOL. The contribution of leakage reductions during operation is calculated as the difference in HFC emissions between S2-1 and S2-2. The contribution of recovery and reclamation at EOL is calculated as 90% of the recovered emissions estimated in S2-4 in all the regions, where the 90% assumption approximately corresponds to the reclamation ratio at facilities for Japan observed in 2023, as reported in [41]. For years when the estimated amount of reclaimed refrigerant exceeds refrigerant demand, the maximum amount of reclaimed refrigerant is constrained by the level of demand; otherwise, the estimated amount of reclaimed refrigerant is applied. The refrigerant demand is assumed to be the sum of the refrigerant put into newly installed ACs and the amount of refrigerant leakage during the operation of existing ACs.

Global HFC refrigerant production and consumption will be reduced by 119, 224, and 384 MtCO₂eq in 2030, 2050, and 2070, respective ly. The contribution of leakage reduction is more significant than the contribution of recovery and reclamation across all periods, accounting for 83, 138, and 248 MtCO₂eq in 2030, 2050, and 2070, respectively. In contrast, recovery and reclamation contribute 36, 85, and 136 MtCO₂eq for the same years. R32 contributions exceed those of R410A after around 2035, reflecting ongoing refrigerant substitution. This substitution is partially affected by the relatively short 10-year average lifetime of ACs in developing countries. The global cumulative reduction will be achieved by approximately 10,800 MtCO₂eq between 2020 and 2070.

4. Discussion

Two key strategies are essential for reducing the GHG emissions from residential AC refrigerants: substituting refrigerants with lower GWP and implementing measures to reduce emissions through improvements in leakage reduction and refrigerant recovery. In terms of substitutions, the substitution of R32 for R410A is effective due to its medium GWP; the substitution of R290, which has an ultralow GWP, is highly effective only when focusing GWP. Leakage reductions during operation have the greatest potential for emission reductions. When R32 is subjected to the significantly enhanced measures on leakage rates and recovery rates, the emissions approach those of ultralow GWP. This demonstrates that the combination of refrigerant medium GWP and LRM effectiveness would be considered, as well as focusing solely on ultralow GWP values.

Well-managed measures to promote leakage reduction and refrigerant recovery have not been broadly implemented in most developing countries, meaning that there is significant potential for emissions reduction. To meet the 1% leakage and 50−90% recovery rates assumed in S2-3 and S2-4, there need to be significant changes from the current situation. Meanwhile, in Japan, the current recovery rate for ACs is only around 40%, while the rate for refrigerators and washing machines is approximately 90%. This lower recovery rate for ACs primarily results from their high resource value, which leads to a diversion to informal disposal patterns whereby contractors bypass official recycling routes [42]. Further improvements in recovery at EOL are expected. The promotion of waste management practices for electronic devices, including the implementation of legislation related to electronic waste in some countries, is already underway and is expected to expand significantly in the near future, meaning that the leakage and recovery rates assumed in this study would appear to be highly feasible and ultimately attainable. In order to achieve these goals, however, technical and financial support from international organizations such as UNEP and the Multilateral Fund for the Montreal Protocol (MLF), including capacity-building programs, technology transfer, and funding assistance for developing countries, will be needed.

Implementing enhanced LRM not only reduces HFC refrigerant emissions; it also reduces the consumption and production of refrigerants. In the medium GWP refrigerant scenario defined in our study, the most enhanced LRM reduces production and consumption by 224 MtCO₂eq in 2050 and 384 MtCO₂eq in 2070. These estimates assume that reducing leakage results in a decrease in refrigerant demand and that 90% of the recovered refrigerants can be reclaimed. Enhancing LRM provides a pragmatic approach to addressing the KA targets that consider different countries’ conditions. In addition, it helps reduce the cost of foreign currency outflow for refrigerant-importing countries. Moreover, it is consistent with broad circular economy strategies that promote sustainable development.

Our study showed that combination of a strongly enhanced LRM and the use of medium GWP refrigerant can substantially reduce GHG emissions from refrigerants and reach up to a 97% reduction by 2070 (compared to S1). To date, no refrigerant has been found that meets all the requirements of environmental performance, high efficiency, economy, and safety. While the ultralow GWP refrigerant R290 is one of the candidates, it is highly flammable, meaning that its use as a substitute will require the implementation of stringent safety measures well beyond those currently in place. The scope of the IEC 60335-2-40 international standard [43] addressing the safety of residential ACs has been described as follows: “As far as practical, this standard addresses common hazards presented by appliances encountered in normal use and assumes that installation, operation, decommissioning, and EOL are safely handled by competent persons and accidental release of refrigerants is avoided. Safety requirements at EOL are not specified in this standard.”

In order to ensure that residential ACs are handled safely, and to prevent the accidental release of refrigerants, it is essential that we establish appropriate standards and guidelines, as well as implement strict safety measures by installation contractors and service providers for equipment, tools, and safety protective gear for safe handling. Strict and standardized safety measures in logistics, including transportation and warehousing, and during the relocation or disposal of equipment, will also need to be established, recognizing that proper management at EOL is a particularly difficult challenge. Standardizing safe LRM practices that promote leakage reduction and proper handling at EOL can be positioned as an essential step for reducing emissions.

The following six limitations should be noted in interpreting our results: (1) not accounting for heterogeneity within the assumed seven regions, (2) assumptions based on limited literature due to insufficient data, (3) simplified refrigerant transition scenarios without considering alternative transition pathways, (4) absence of cost analysis, (5) limited selection of refrigerants, (6) not considering the application of other possible technologies for cooling demands.

Firstly, this study assumed the uniform implementation of LRM within each country group, whereas actual implementation is expected to vary significantly based on socioeconomic conditions and economic and political disparities. We assumed that, particularly in developing countries, the development of waste management systems and progress toward a circular economy, driven by economic growth, would accelerate LRM in the cases of enhanced LRM. Secondly, since we were not able to obtain sufficiently detailed datasets on residential ACs for all countries, we based our assumptions regarding replacement periods, leakage and recovery rates, and average lifetime on limited data from specific regions such as Japan and Australia. It should be noted that such assumptions may not fully reflect the situation in other countries with different economic and technological conditions. Future research will require sensitivity analysis on several assumptions and will incorporate more region-specific data if available. Thirdly, we assumed that simple refrigerant substitutions in this study, and we did not consider other potential transitions, such as direct transitions from R22 to R32. It is worth noting that the impact of substitution scenarios not considered in this study on the quantitative results requires further evaluation in future studies.

Fourthly, it should be noted that this study does not address the costs required to achieve the various GHG emission scenarios. LRM improvements will require investments related to technology and personnel training. The next step requires further research into how to effectively implement the scenarios for leakage and recovery, as well as the necessary investments. We did not consider cooling demand changes affected by energy price changes and climate change impacts under the scenarios. A comprehensive cost analysis, including energy price changes and the relevant demand effects, would be essential for policy implementation. Fifthly, this study did not analyze selecting the potential of various alternative refrigerants with ultralow GWPs. There are uncertainties for the possibility of developing unknown superior refrigerants in the future, and European F-gas regulations and the possibility of PFAS (per- and polyfluoroalkyl substances) regulations may further limit the range of refrigerant options and development possibilities [8,44]. Analyzing the feasibility of the alternative refrigerants will be an important future study. Sixthly, we excluded alternative cooling technologies such as absorption chillers and thermoelectric coolers due to their very low market penetration in current residential applications. This will be important for analyzing the impacts of such alternative cooling technologies on GHG emissions in the future study.

For other future endeavors, it would be desirable to estimate the investment required to implement enhanced LRM and ensure the safe installation of ultralow GWP ACs. In addition to employing enhanced LRM, further reductions in HFC emissions and production/consumption can be expected by taking measures in the passive cooling range advocated by UNEP and by extending the lifespan of the equipment. Promoting the development of high-efficiency equipment and ensuring its efficient operation are also important issues to be considered.

5. Conclusions

This study conducted a scenario analyses of global refrigerants emissions from residential ACs through 2070. While most existing studies have focused on refrigerant substitutions as climate change mitigation strategies, this study sought to analyze both refrigerant substitutions and the LRM effects of better leakage control and improvements in recovery and reclamation. The analysis described here showed the significant impact of a combined approach, even when using only medium GWP refrigerants such as R32 (which is already underway in some regions). When such a substitution is accompanied by enhanced LRM practices—including measures to lower leakage rates during operation and improve refrigerant recovery at EOL—global refrigerant emissions from residential ACs can be reduced by as much as 95% in 2050 and 97% in 2070 compared to the reference case. This reduction in emissions can be expected to lower the global mean surface temperature by 0.067 °C in 2070 compared to the reference case. In the S3 scenario, which involves the substitution of an ultralow GWP refrigerant, a 97% reduction in emissions can be achieved by 2050 and a 100% reduction can be achieved by 2070, with a global temperature decrease of 0.069 °C in 2070 (relative to the reference case).

Under the medium GWP scenario, enhancing LRM to reduce leakage, improve recovery rates, and reclaim recovered refrigerants reduces the global production and consumption of HFCs by approximately 10,800 MtCO₂eq over the period 2020–2070, assuming that 90% of the recovered emissions are reclaimed. This contributes to achieving the national targets of the KA while also aligning with a circular economy strategy.

Currently, no ultralow GWP refrigerants that satisfy all requirements for residential ACs have yet been identified in terms of environmental performance, economic performance, and safety. Considering this situation, a multiple approach that encompasses not only refrigerant substitution but also the enhancement of LRM for leakage reduction and recovery improvement may offer an effective strategy for reducing global HFC emissions and contributing to the achievement of the KA.