Refrigerant Leak Statistics 2026:
Rates, Costs & EPA Rules

What are the headline statistics for commercial refrigerant leak rates?

The most widely cited figures for commercial refrigeration are 14.2% (GreenChill voluntary partner rate for 2023), approximately 21.2% (derived from EPA's modeled 2025 stock-to-leak ratio for systems over 15 pounds), 25% (EPA's historical food-retail estimate developed in 2007) and 35% (the upper end of the IPCC default range, used by EPA as a conservative screening factor for medium and large systems).^{[2][5][3][6][7]} None of these figures is an observed national average drawn from a probability sample of every U.S. commercial refrigeration system.

What do current commercial refrigerant leak statistics show?

This reference organizes every figure by evidence class because the same percentage can mean entirely different things depending on where it came from. A 14.2% program average, a 21.2% model-derived ratio and a 35% screening factor cannot be treated as three measurements of the same thing, even though they all appear in the range where people expect a commercial leak rate to land.

Complete source-normalized reference — 45 records

The table below lists all 45 records in this dataset. Each record preserves the population, method, geography, period and primary limitation of the figure it describes. Records are numbered RL-001 through RL-044, with RL-020 split into two sub-records (RL-020A and RL-020B) for the two federal chronic-leak thresholds.

Download the full structured dataset: refrigerant-leak-statistics.csv

Citation information

What do these numbers show — and what do they not show?

The 14.2% GreenChill figure and the 21.2% EPA model ratio are not contradictory measurements of the same population. The gap between them reflects real differences in what each figure measures.

Physical emissions fell because installed charge also declined — from approximately 2,862 lb/store to approximately 2,033 lb/store — while the percentage rate held roughly steady.^[2]

What this reference can and cannot support

How was this refrigerant leak statistics dataset produced?

This dataset was produced by extracting figures from identified public primary sources, assigning each figure an evidence class, and recording the population, method, geography, period and key limitation that make the figure meaningful. No figures from secondary sources, industry estimates without a named origin, or anonymous expert opinion were included.^[1]

Inclusion and exclusion rules

• Include: figures that trace to an identifiable public primary source and can be verified at that source.
• Include: derived statistics that are transparent arithmetic from primary-source inputs, clearly labeled as derived.
• Include: federal regulatory thresholds, definitions and formulas from the current eCFR.
• Exclude: secondary-source summaries, statistics attributed to unidentified experts or unnamed studies.
• Exclude: figures whose source cannot be located or verified at the original publisher.
• Exclude: statistics that claim a national U.S. average where no suitable national probability sample exists.

Normalization fields

Each record includes: record ID, metric group, figure, unit, equipment scope, evidence class, geography, period, source organization, source title, source URL, interpretation, key limitations and dataset version. The CSV download contains all fields.

Calculation rules

Derived statistics use these formulas, applied to primary-source inputs:

annual leak emissions (MT) ÷ installed stock (MT) × 100

pounds physically lost × price per pound

leaked pounds × 100-year GWP ÷ 2,204.62262

[(lbs added ÷ full charge) ÷ (min(days since last addition, 365) ÷ 365)] × 100

[applicable lbs added in the defined rolling period ÷ full charge] × 100

Rounding rules

Derived percentages are shown to one decimal place. EPA's model totals are rounded in the source; derived statistics from those totals carry that rounding forward. GWP-based CO2e is rounded to one decimal place. Prices are rounded to two decimal places from EPA's published per-kilogram values.

Source-discrepancy log

Where official statements do not reconcile, this page records both without resolving them and withholds any derived statistic.

Data files

• refrigerant-leak-statistics.csv — 45 source-normalized records with all normalization fields
• refrigerant-leak-impact-per-1000-lb.csv — 6-scenario per-1,000-pound impact matrix (pounds, R-404A tCO2e, R-448A tCO2e, R-744 tCO2e)

What is the average commercial refrigeration leak rate?

There is no defensible single current average across every commercial refrigeration system in the United States. The five source-specific figures most often cited — 14.2%, 15.6%, approximately 21.2%, 25% and 35% — reflect different populations, methods and purposes.^{[2][4][5][3][6]}

GreenChill's 14.2% is the voluntary-partner rate. It covers a real population but not a random national sample: stores that participate in GreenChill are more likely to maintain equipment proactively.^[2] DNV's 15.6% is a regional assembled estimate for one state and one store-size class, with disclosed imputation where market-actor data were unavailable.^[4] The 21.2% ratio is derived arithmetic from EPA's rounded Vintaging Model totals — not a direct measurement and not EPA's own published rate.^[5] The 25% figure is a historical estimate developed by EPA in 2007, retained as a comparison baseline, not updated with new sampling data.^[3] And 35% is EPA's conservative screening input for medium and large systems, not an average.^[6]

This page does not publish an average because doing so would combine numbers that belong to different populations, methods and years. Any number that results from that combination belongs to no real population and has no traceable source.

What does the 2025 Massachusetts grocery study show?

DNV's October 2025 study for the Massachusetts EEAC reports a 15.6% assembled estimate for grocery chains operating 15 or more stores in Massachusetts, and recommends 16% as a program baseline for the state's incentive program.^[4]

What are refrigerant leak rates by equipment type?

IPCC provides default ranges by equipment class for greenhouse-gas inventory work. These are not measured U.S. appliance rates. They are intended for national inventory estimates when better country-specific data are unavailable.^[7]

Why is 35% often quoted as the commercial refrigerant leak rate?

35% is not the average. It is the upper end of the IPCC default range for medium and large commercial refrigeration, and EPA adopted it as a conservative screening factor for that class.^[6][7] It was never intended to describe typical performance.

What ratio is implied by EPA's modeled 2025 commercial-refrigeration stock and annual leak emissions?

EPA's September 2024 regulatory-impact analysis models 2025 installed stock and annual leak emissions. For the row labeled "Commercial Ref (>15 lbs)," it assigns 69,000 metric tons of installed HFC stock and 14,600 metric tons of annual leak emissions — a site-calculated aggregate derived modeled stock-to-annual-leak ratio of approximately 21.2%.^[5]

Four original comparisons from EPA's table

Within the two rows labeled "Commercial Ref":

1. The >15-pound category contains approximately 90.1% of modeled installed stock.
2. It produces approximately 97.3% of modeled annual leak mass.
3. Its installed stock is approximately 9.1 times the 5–15-pound category's.
4. Its modeled annual leak mass is 36.5 times as large, and its derived ratio is approximately 4.0 times the smaller row's.

What the 21.2% result is — and is not

14,600 MT ÷ 69,000 MT × 100 = 21.159%, or approximately 21.2%

It is a transparent quotient from EPA's rounded model totals. It is not an EPA-published appliance leak rate, a direct field census, a legal threshold or proof that a particular system will lose 21.2% of charge. EPA's appendix explains that its model distributes equipment into leak-rate quintiles around average assumptions and, for some subsectors, adjusts values so at least one group exceeds the relevant repair threshold while preserving the weighted average.^[5]

How much refrigerant does a supermarket leak per year?

EPA reports that an average GreenChill partner store emitted 298 pounds of refrigerant in 2023, down from 411 pounds in 2007. The 113-pound decline equals approximately 27.5%, but these are voluntary-program averages — not a prediction for every supermarket or every refrigeration system.^[10]

Full charge matters. A 10% physical loss equals 10 pounds on a 100-pound system, 100 pounds on a 1,000-pound system and 150 pounds on a 1,500-pound system.

How much do commercial refrigerant leaks cost?

Replacement-only cost equals the pounds physically lost multiplied by the relevant refrigerant price per pound. EPA's May 2026 economic memo supplies a dated public-buyer price layer through April 2026, and EPA separately estimates that GreenChill partners avoided $134 million in replacement-refrigerant purchases during 2023. Neither source measures the full cost of an individual leak event.^[15][10]

Replacement-only cost formula

replacement-only cost = pounds physically lost × relevant price per pound

The price input is not universal. Refrigerant, package size, seller, purchase volume, contract terms, location and timing all matter.

EPA sampled refrigerant prices through April 2026

EPA's May 2026 economic memo reports sampled publicly available buyer prices for 2024 through April 2026. The per-pound column converts EPA's dollars-per-kilogram figures using 0.45359237 kg/lb. EPA warns that general-public prices may exceed negotiated large-volume prices.^[15]

Replacement-only cost sensitivity — 1,000-pound full charge

A full leak event can also involve diagnostics, technician labor, parts, verification testing, emergency service, energy inefficiency, product loss, downtime and administrative work. No authoritative U.S. national primary dataset supports one all-in average that can be applied honestly to every commercial system.

How much climate impact does a refrigerant leak create?

Direct climate impact depends on the leaked mass and the refrigerant's 100-year global warming potential (GWP). At a 14.2% illustrative loss from a 1,000-pound charge, 142 leaked pounds equal approximately 252.6 metric tons CO2e for R-404A, 89.3 for R-448A and 0.064 for R-744 using EPA's listed GWP values.^[16]

Direct CO2e formula

metric tons CO2e = leaked pounds × 100-year GWP ÷ 2,204.62262

This covers direct refrigerant emissions. It does not include electricity-related emissions, equipment manufacturing or other lifecycle effects.

Equal leaked masses do not mean the listed refrigerants are interchangeable. The table isolates one variable: the direct GWP-based effect of an emitted mass.

EPA also reports that GreenChill Certified Stores avoided more than 2.54 million metric tons CO2e from 2008 through 2024 through preventive maintenance and leak reduction. That is a program estimate for certified stores, not a national estimate for all commercial refrigeration.^[11]

What EPA commercial-refrigeration leak thresholds apply in 2026?

Covered commercial-refrigeration appliances are subject to a 20% federal action threshold under both Clean Air Act Section 608 and the AIM Act management rules, but the regimes cover different appliances. Section 608 generally applies at 50 pounds or more of a Class I or Class II ozone-depleting refrigerant or a blend containing one; Part 84 generally applies at 15 pounds or more of an HFC or covered substitute with GWP greater than 53, subject to the rule's details and exceptions.^[9][8]

The 20% figure is an action threshold — not an operating target, an acceptable average, or permission to leak up to 20%.

What counts as commercial refrigeration?

Part 84 defines commercial refrigeration as refrigerant-containing appliances used in retail-food and cold-storage-warehouse subsectors. Examples include systems used in supermarkets, convenience stores, restaurants, other food-service establishments and facilities storing meat, produce, dairy products and other perishable goods.^[12]

What happens after a covered appliance exceeds 20%?

The applicable rule may require repair or a retrofit/retirement path, followed by verification, inspections, records and reporting. For covered Part 84 commercial systems:

• After a successful follow-up verification test, systems from 15 pounds to under 500 pounds generally require annual leak inspections until the rule's stopping condition is met.
• After a successful follow-up verification test, systems with 500 pounds or more generally require quarterly inspections until four consecutive quarters satisfy the stopping condition.
• A qualifying automatic leak-detection system can affect periodic-inspection requirements when regulatory conditions are met.^[8][13]

When is automatic refrigerant leak detection required?

Part 84 requires automatic leak detection for covered commercial-refrigeration and industrial-process-refrigeration appliances with a full charge of at least 1,500 pounds.^[13]

Direct systems must detect 10 parts per million and alert at 100 parts per million. Indirect systems must alert at a measured loss of 50 pounds or 10% of full charge, whichever is less. The system must be audited and calibrated annually.^[13]

How does EPA calculate a refrigerant leak rate?

EPA defines leak rate as the percentage of an appliance's full charge that would be lost over 12 months if the current rate of loss continued. Part 84 provides an annualizing method and a rolling-average method, and requires an operating facility to use the same selected method for all appliances subject to the leak-repair requirements.^[12]

This regulatory rate is not automatically equal to calendar-year leak mass divided by installed stock. That is why the 20% legal threshold cannot be substituted for the 21.2% derived model ratio — or vice versa.

Full charge

"Full charge" is the amount of refrigerant required for normal operation. The rule allows manufacturer information, engineering calculations, actual measurements or an established operating range whose midpoint is used, alone or in combination.^[12]

Annualizing method

[(pounds added ÷ full charge) ÷ (min(days since the last refrigerant addition, 365) ÷ 365)] × 100

Example: A 300-pound appliance receiving 20 pounds after 90 days:

(20 ÷ 300) ÷ (90 ÷ 365) × 100 = 27.0%

Rolling-average method

[applicable pounds added during the defined rolling period ÷ full charge] × 100

Example: A 300-pound appliance with 60 applicable pounds added during the preceding 365 days:

60 ÷ 300 × 100 = 20.0%

What changes for the first Part 84 calculation after January 1, 2026?

For the first Part 84 calculation after January 1, 2026, the annualizing method substitutes 365 days as the days-since-last-addition value, while the rolling-average method uses refrigerant additions made since January 1, 2026. This transition instruction prevents pre-applicability periods and additions from being handled as though Part 84 had already operated for a full year.^[8]

Common calculation errors

How do U.S. figures compare with observed international data?

VDKF reported a 1.7% average for the commercial-refrigeration category in its German 2025 database analysis. It is a useful recent observed comparator, but it cannot be substituted for a U.S. national average because geography, equipment mix, regulation, sampling and event definitions differ.^[14]

The point of the comparison is not to rank countries. It is to show how far an observed result can move when the population, maintenance regime and reporting system change.

Why do refrigerant leak statistics matter in 2026?

January 1, 2026 marked the applicability date for Part 84 leak-repair requirements covering qualifying HFC and substitute-refrigerant appliances with at least 15 pounds. That makes old summaries risky: a page can repeat the correct 20% commercial threshold while attaching it to the wrong refrigerant scope, charge threshold, calculation method, inspection timing or reporting rule.^[8]

The 15-pound Part 84 scope did not erase Section 608's separate 50-pound ODS regime. The two rules now sit beside each other, and a technically correct percentage can still be legally misleading when its scope is wrong.

For the first Part 84 calculation after January 1, 2026, annualizing-method users substitute 365 days for the elapsed-days value, while rolling-average users count applicable additions made since January 1, 2026. That date-specific instruction is easy to miss in a generic threshold summary.^[8]

What are the limitations of this refrigerant leak dataset?

This dataset improves comparability by preserving definitions and source context, but it cannot create a national observed average where no suitable national probability sample exists.

GreenChill is a voluntary program

GreenChill covers a substantial food-retail participant group, but participation does not turn the dataset into a random national sample. Its 14.2% figure must remain a partnership rate.^[2]

EPA's modeled ratios are not appliance measurements

The 21.2% result is derived from rounded model totals. It is not a direct appliance survey or the federal leak-rate calculation.^[5]

The Massachusetts estimate includes imputation

DNV's 15.6% result covers a regional 15+ store segment and uses a 20% imputation where no interview or GreenChill rate was available. Raw market-actor values were not released.^[4]

IPCC values are inventory defaults

They are intended for greenhouse-gas inventory work when better information is unavailable. They are not measured U.S. appliance performance.^[7]

EPA screening factors are intentionally conservative

The 35% figure is useful for its screening purpose. Calling it an average changes the source's meaning.^[6]

Regulatory rates have their own definition

The federal rate projects loss over 12 months using the annualizing or rolling-average method. It cannot be assumed to equal calendar-year physical loss divided by installed charge.^[12]

Cost coverage is incomplete

EPA's sampled prices are public-buyer references, not negotiated facility prices or repair invoices. The page supports replacement-only calculations and scoped program estimates, not a universal all-in event cost.^[15]

Official EPA statements do not fully reconcile on installed charge

The 829-pound visible endpoint difference and the separate 921-pound statement remain in the source notes. No derived decline percentage is published.^[2][10]

GWP values depend on the selected reference basis

The CO2e tables use EPA's current Technology Transitions reference values as verified on July 22, 2026. A different accepted GWP basis can produce a different result.^[16]

International data are not directly interchangeable

VDKF's German result comes from a different regulatory and sampling environment and includes accidents or major accidental events. It remains an international comparator, not a substitute U.S. benchmark.^[14]

Data year and verification date are different

"Last verified July 22, 2026" means the sources and calculations were rechecked on that date. It does not mean every underlying measurement was made in 2026.

Who produced this reference, and why?

KC Commercial Refrigeration produced and publishes this source-normalized reference. The site is a commercial-refrigeration service-request website, not a contractor, regulator, laboratory, university or trade association; the research page is based on public primary sources and contains no service-request routing or commercial recommendation.^[1]

By KC Commercial Refrigeration

No named engineer, regulator, reviewer or subject-matter expert is attributed because no such person reviewed the page under a documented process.

How it was produced

• Primary-source extraction from EPA, eCFR, IPCC, DNV and VDKF material.
• Current-regulation verification on July 22, 2026.
• A 45-record source-normalized public dataset.
• A six-scenario public impact matrix.
• Arithmetic checks for percentages, pounds, prices, costs and CO2e.
• An internal claim ledger distinguishing publishable, qualified, withheld and rejected claims.
• A public source-discrepancy log for official statements that do not reconcile.

Why it exists

The page exists to make commercial refrigerant leak percentages harder to misuse. A reader should be able to identify the right figure for a specific question without losing its population, method, year or limitation.

Frequently asked questions about refrigerant leak statistics

The answers below preserve the scope and evidence class of each number.

What is the average commercial refrigeration leak rate?

There is no defensible universal current average across every commercial-refrigeration system. Source-specific figures include a 14.2% GreenChill partner rate for 2023, a 15.6% DNV Massachusetts assembled estimate, an approximately 21.2% ratio derived from EPA's modeled >15-pound category, a historical EPA estimate of 25% and IPCC default ranges that vary by equipment class.^{[2][4][5][3][7]}

Is 35% the average refrigerant leak rate?

No. Thirty-five percent is the upper end of the IPCC default range for medium and large commercial refrigeration and the conservative factor EPA uses for that class in its screening method. It is not a measured universal average.^[7][6]

What is EPA's commercial-refrigeration leak threshold?

The federal commercial-refrigeration action threshold is a calculated rate that exceeds 20% for covered appliances under both Section 608 and Part 84. Their refrigerant scope and minimum full charge differ: Section 608 generally starts at 50 pounds for covered ODS appliances, while Part 84 generally starts at 15 pounds for covered HFC or substitute appliances, subject to the rules' details and exceptions.^[9][8]

Is a 20% refrigerant leak rate acceptable?

Twenty percent should not be described as an acceptable operating target or industry average. It is a legal action threshold for defined covered equipment; a lower rate is not automatically leak-free, efficient or sufficient under every federal or state requirement.^[9][8]

What is the 125% chronic-leak reporting threshold?

Covered Section 608 appliances with at least 50 pounds and covered Part 84 appliances with at least 15 pounds must be reported to EPA by March 1 of the following year when calendar-year leakage reaches at least 125% of full charge. Repeated leakage and recharging can produce cumulative losses above 100%.^[9][8]

How many pounds of refrigerant does a supermarket leak per year?

EPA reports that an average GreenChill partner store emitted 298 pounds in 2023, compared with 411 pounds in 2007. These are voluntary-program averages, not a national prediction for every supermarket; for an illustrative calculation, multiply full charge by the selected physical-loss fraction.^[10]

How much does a commercial refrigerant leak cost?

Replacement-only cost equals the pounds physically lost multiplied by the relevant price per pound. EPA's May 2026 memo supplies dated public-buyer reference prices through April 2026, while total event costs can also include diagnostics, labor, parts, energy, downtime, spoilage and administrative work.^[15]

How do you convert leaked refrigerant to CO2e?

Multiply leaked pounds by the refrigerant's 100-year GWP and divide by 2,204.62262. Using EPA's GWP of 3,922 for R-404A, a 100-pound release equals approximately 177.9 metric tons CO2e.^[16]

How does EPA calculate the refrigerant leak rate?

EPA provides an annualizing method and a rolling-average method. The annualizing method adjusts a refrigerant addition for elapsed time; the rolling-average method uses applicable additions over the preceding 365 days or the shorter defined period since a successful follow-up verification test.^[12]

When is automatic refrigerant leak detection required?

Part 84 requires automatic leak detection for covered commercial-refrigeration and industrial-process-refrigeration appliances with at least 1,500 pounds of full charge. Installation timing depends on the appliance's installation date.^[13]

Does Section 608 apply to HFC-only appliances?

The leak-repair provision in 40 CFR §82.157 applies to appliances with at least 50 pounds of a Class I or Class II ozone-depleting refrigerant or a blend containing one, and excludes appliances containing solely substitute refrigerants. Covered HFC and substitute appliances are addressed through the separate Part 84 regime.^[9][8]

Where do the data come from, and how will they be updated?

Every consequential figure traces to an issuing agency, current regulation, inventory-methodology body or original data producer. The stable dataset follows a planned maintenance schedule: source years remain attached to records, substantive changes create a new version and prior files remain archived instead of being silently overwritten.

Primary-source map

Planned maintenance schedule

Version history

Footnotes

1. [1]Editorial and Source Standards — KC Commercial Refrigeration. Source hierarchy, claim verification rules, AI-assistance policy and correction process.
2. [2]GreenChill Annual Recognition Event 2024 — U.S. EPA GreenChill. 14.2% and 13.8% partnership rates, installed-charge endpoints, 25% historical estimate.
3. [3]About the GreenChill Partnership — U.S. EPA GreenChill. GreenChill program origin, 2007 establishment, 25% historical estimate context.
4. [4]Grocery Store Refrigerant Baseline Practices — DNV / Massachusetts EEAC. 15.6% regional assembled estimate, 16% recommended baseline, Massachusetts grocery chains operating ≥15 stores. October 30, 2025.
5. [5]Regulatory Impact Analysis Addendum: Management of HFCs and Substitutes — U.S. EPA. September 2024. Table 3-3 modeled 2025 installed stock and annual leak emissions.
6. [6]Direct Fugitive Emissions — Scope 1 and Scope 2 Inventory Guidance — U.S. EPA Center for Corporate Climate Leadership. 15%, 35% and 25% conservative screening factors and screening-method limitations.
7. [7]2019 Refinement to the 2006 IPCC Guidelines, Volume 3, Chapter 7, Table 7.9 — IPCC. 1–15%, 10–35% and 7–25% default ranges by equipment class.
8. [8]40 CFR §84.106 — Leak Repair — eCFR / U.S. EPA. Part 84 commercial-refrigeration 20% action threshold, inspection cadence and chronic-leak reporting. Applicability: January 1, 2026.
9. [9]40 CFR §82.157 — Leak Repair — eCFR / U.S. EPA. Section 608 ODS commercial-refrigeration 20% action threshold and 125% chronic-leak reporting.
10. [10]GreenChill Partnership Accomplishments — U.S. EPA GreenChill. 411 lb (2007) and 298 lb (2023) per partner store; $134 million in 2023; unresolved installed-charge statement.
11. [11]GreenChill Certified Store Achievements — U.S. EPA GreenChill. More than $1,400 per store per year; more than 2.54 million metric tons CO2e avoided 2008–2024.
12. [12]40 CFR §84.102 — Definitions — eCFR / U.S. EPA. Part 84 definitions including commercial refrigeration, full charge, annualizing method and rolling-average method.
13. [13]40 CFR §84.108 — Automatic Leak Detection — eCFR / U.S. EPA. 1,500-lb threshold, installation timing, detection sensitivity, calibration requirements.
14. [14]Kältemittel-Leckagerate liegt bei nur noch 1,0 % (2025 VDKF-LEC data) — VDKF. German 1.7% commercial category, 1.0% overall, ~307,000 systems in database, includes accidents.
15. [15]Analysis of Economic and Environmental Impacts, Appendix D, Table D-1 — U.S. EPA. May 2026. Sampled public-buyer refrigerant prices through April 2026 for HFC-134a, R-404A, R-410A, R-448A, R-449A and HFO-1234yf.
16. [16]Technology Transitions GWP Reference Table — U.S. EPA. 100-year GWP values used for all CO2e calculations. Last verified July 22, 2026.