Compliance10 min readFebruary 18, 2026

Annualizing vs. Rolling Average: EPA Leak Rate Methods Explained

The EPA allows two methods for calculating refrigerant leak rates under 40 CFR Part 84, Subpart C: Annualizing and Rolling Average. Both are fully compliant, but they can produce significantly different results from the same data. Choosing the right method — and understanding what the numbers mean — is essential for accurate compliance tracking and avoiding unnecessary repair triggers.

Key Takeaways

  • Annualizing projects a single event over 365 days; Rolling Average sums the additions in the previous 365 days, or over the period since the last successful follow-up verification test showing all identified leaks were repaired, if that period is less than one year
  • Annualizing is more sensitive to short intervals between additions
  • Rolling Average is more sensitive to cumulative additions over time
  • You must choose one method and apply it consistently — no switching per system
  • For infrequently serviced systems, Annualizing is usually simpler
  • For frequently serviced systems, Rolling Average provides more stability
  • Under 40 CFR 84.106(b)(1)–(2), an appliance's first calculation after January 1, 2026 substitutes 365 days as the number of days since the last refrigerant addition (Annualizing) or the pounds added since January 1, 2026 (Rolling Average) — which EPA allows only when previous records are absent (89 FR 82724)

The Annualizing Method

The Annualizing method projects a single refrigerant addition over a full 365-day period. It answers the question: "If this system kept leaking at this rate all year, what would the annual leak rate be?" The formula is: Leak Rate = (lbs added ÷ full charge) × (365 ÷ D) × 100, where D is the number of days since the previous refrigerant addition, capped at 365 — § 84.102 Step 2 takes "the shorter of the number of days that have passed since the last day refrigerant was added or 365 days". For an appliance's first calculation after January 1, 2026, § 84.106(b)(1) substitutes 365 days for D — so the first addition is not annualized upward — but only when records of earlier additions are absent: EPA's preamble says the substitution "is only allowed when previous records are absent" (89 FR 82724). If you have a record of the last addition before 2026, D counts from it.

Note

The annualization factor (365 ÷ D) amplifies small additions over short periods. Adding 2 lbs to a 50-lb system after 30 days produces a 48.7% leak rate — much higher than the 4% raw ratio suggests.

The Rolling Average Method

The Rolling Average method sums the refrigerant added over the previous 365 days — or since the last successful follow-up verification test showing all identified leaks were repaired, if that is less than a year (§ 84.102, rolling average Step 1) — and compares the total to the system's full charge. The formula is: Leak Rate = (sum of lbs added in the window ÷ full charge) × 100. For an appliance's first calculation after January 1, 2026, § 84.106(b)(2) substitutes the pounds added since January 1, 2026 — only when records of earlier additions are absent (89 FR 82724); with those records, the calculation counts them. This method is cumulative — every counted addition in the window contributes to the rate, regardless of timing. Purged refrigerant destroyed at a verifiable destruction efficiency of 98 percent or greater is not counted (§ 84.106(k)), and an addition that qualifies as a seasonal variance is excluded (§ 84.106(b)). The rolling window moves forward with each new service event, and older events naturally drop off as they fall outside the window.

When Annualizing produces higher results

The Annualizing method tends to produce higher leak rates when service events are close together in time. The shorter the interval between additions, the higher the annualization factor. For example, if you add refrigerant twice in a 2-week span, the second event will be annualized over just 14 days — amplifying the rate by a factor of 26. This makes Annualizing more sensitive to burst patterns and can trigger threshold exceedances faster than Rolling Average would for the same data. Systems that require frequent top-offs in a short period are more likely to show elevated rates under Annualizing.

When Rolling Average produces higher results

Rolling Average tends to produce higher rates when sizeable additions sit far apart within a year. Since every counted addition in the window is added up in full, the total can exceed what any single annualized event would show. Adding 4 lbs to a 40-lb system twice, 364 days apart, with nothing added in the year before, reads 10.0% under Annualizing each time and 20.0% under Rolling Average at the second addition. For systems with chronic, low-level leaks that get topped off monthly, the two methods converge: Annualizing shows the yearly rate from the second top-off, while the Rolling Average climbs to about the same rate as the year's additions accumulate.

Which method should you choose?

The EPA requires that the same method be used for every covered appliance at an operating facility (40 CFR 84.102), and § 84.106(b)(3) treats an owner's other facilities as sharing one method — a change is allowed only after acquiring a facility that used the other one. In practice: choose one method and apply it consistently. You cannot switch methods per system or per event to cherry-pick the lower number. That said, here are practical guidelines. Annualizing is often better for systems serviced infrequently (once or twice a year), because it gives a meaningful snapshot from each individual event. Rolling Average is often better for systems serviced frequently (monthly or quarterly), because it smooths out individual variations and provides a more stable compliance picture. For most small HVAC/R contractors whose covered systems are a handful of walk-ins, racks, and chillers, Annualizing is the simpler choice — it needs only the date of the previous addition, and each event produces an immediate result.

Pro Tip

Pick your method once and stick with it. Ref LeakLog lets you set your organization's calculation method in Settings and applies it automatically to every service event.

Practical example: same data, different results

Consider a 50-lb comfort-cooling chiller (10% threshold) with no additions on record before 2026 and two service events in 2026: 3 lbs added on March 15, and 2 lbs added on June 15 (92 days later). Under Annualizing, the March 15 event calculates as (3 ÷ 50) × (365 ÷ 365) × 100 = 6.0% (compliant). The June 15 event calculates as (2 ÷ 50) × (365 ÷ 92) × 100 = 15.9% (exceeds 10% threshold — repair required). Under Rolling Average, the June 15 calculation sums both events: (5 ÷ 50) × 100 = 10.0% (right at the threshold). Same system, same service history — but Annualizing triggers a repair action while Rolling Average shows the system at exactly the limit. This is why method selection matters.

Frequently Asked Questions

Can I use Annualizing for some systems and Rolling Average for others?

No. The same method must be used for all covered appliances at an operating facility (40 CFR 84.102), and the rule allows a change only in the acquisition case in § 84.106(b)(3). You cannot switch methods per system, per event, or per refrigerant to produce a more favorable result.

Which EPA leak rate method produces lower results?

Neither method is inherently lower. Annualizing tends to produce higher rates when service events are close together, while Rolling Average tends to produce higher rates when sizeable additions sit far apart within a year. The better choice depends on your service patterns.

What happens on the first service event after January 1, 2026?

If there are no records of earlier additions, the first calculation substitutes D = 365 for the Annualizing method (40 CFR 84.106(b)(1)), so the rate is not amplified, and counts only the pounds added since January 1, 2026 for Rolling Average (84.106(b)(2)) — on the first event, just that event's pounds: (lbs added ÷ full charge) × 100. EPA's preamble says this "is only allowed when previous records are absent" (89 FR 82724): with a record of an earlier addition, Annualizing counts D from it (capped at 365), and Rolling Average counts it if it falls inside the window.

Can I change my calculation method later?

Only in one narrow case. Under 40 CFR 84.106(b)(3) an owner or operator may switch leak rate calculation methods only when all three conditions are met: it has purchased or otherwise acquired an operating facility whose appliances were using a different method than its other facilities; it has determined that no appliance at any facility whose method would change is exceeding the applicable leak rate under either method; and it keeps the § 84.106(l)(3) record of the change — basic identification (owner or operator, facility name, facility address and appliance ID or description), the acquisition date, both methods' leak rates for every affected appliance, the date the new method is adopted, and the method in use afterward. Outside that acquisition scenario the rule gives no path to switch, so pick a method and commit to it.

Do both methods count the same service events?

They draw on the same service records. For an appliance's first calculation after January 1, 2026, Annualizing substitutes 365 days as the number of days since the last refrigerant addition and Rolling Average substitutes the pounds added since January 1, 2026 (40 CFR 84.106(b)(1)–(2)), which EPA allows only when previous records are absent (89 FR 82724). The difference is how they use the data: Annualizing looks at individual events in relation to the previous event, while Rolling Average looks at the cumulative total within its window — the previous 365 days, or the period since the last successful follow-up verification test showing all identified leaks were repaired, if that period is less than one year.

Ready to Automate Compliance?

Stop worrying about calculations, deadlines, and audits. Ref LeakLog handles EPA compliance tracking for your entire fleet — automatically.

No credit card required · Full access for 14 days