Estimating Storage Tank Emissions — From API MPMS and AP-42 to Better Decisions

1. The main calculation frameworks

In the United States, EPA AP-42 Chapter 7.1 provides the principal public procedures for estimating emissions from organic liquid storage tanks. EPA finalised a major revision in October 2024. EPA’s browser-based TANKS application implements the AP-42 approach; TANKS 5.3 was released in July 2026.

API MPMS Chapter 19 is the industry framework from which many of the storage-loss methods originate. Chapter 19.2 addresses floating-roof tanks, while related parts cover fixed-roof tanks, reference information, transfer, roof landings and test methods for loss factors.

2. First classify the tank correctly

The calculation must represent the real configuration: fixed roof, external floating roof, internal floating roof, domed external floating roof or another defined arrangement. Misclassification changes which loss mechanisms are calculated and which factors are applied.

For internal floating roofs, deck construction, seam length, rim-seal type and fitting population can materially affect the estimate. Default values are useful for screening but should not replace known project data.

3. Product properties matter

Vapour pressure is one of the strongest drivers of evaporative loss. Use representative liquid properties at the relevant storage temperature and follow the property method required by the calculation framework. For mixtures, a generic product label can hide significant variability.

Where a product changes seasonally or by batch, a sensitivity analysis is more useful than one annual average with no uncertainty range.

4. Operating data matter just as much

Throughput, turnovers, liquid level, roof landings, tank temperature and operating periods influence calculated loss. A tank that is almost static and a tank with rapid cycling should not share the same annual operating assumptions.

Non-routine losses deserve explicit treatment. Roof landing, cleaning and gas-freeing events can create large short-duration emissions that disappear when only routine annual losses are considered.

5. Defaults are assumptions, not facts

Default fitting counts, seal factors, seam factors, paint absorptance or meteorological data make a calculation possible when site data are missing. They also create uncertainty. Every default that can materially affect the result should be visible in the calculation record.

A useful discipline is to label each key input as measured, documented, calculated or assumed. The emission total can then be accompanied by a data-confidence rating.

6. Modern equipment can expose limitations in historical factors

Some floating-roof loss factors are based on test programmes and equipment designs developed decades ago. Modern seam, fitting and seal designs may have supplier-specific test data obtained under recognised API test methods. Where permitted by the calculation framework, such data can improve representation.

That does not mean every vendor certificate should be accepted at face value. The test method, boundary conditions, laboratory competence, tested configuration and ability to reproduce the tested installation in the field all need to be checked.

7. Use scenarios instead of one number

• Baseline: current configuration and best available current data.
• Verified baseline: current configuration updated with inspection or measurement results.
• Control scenario: expected configuration after each proposed ECM.
• Sensitivity: high and low cases for uncertain product or operating inputs.
• Post-project: as-built configuration and actual operating data after commissioning.

8. Calculation should lead to a field question

If the model shows that one fitting group or rim-seal term dominates the loss, inspect that source. If the result is highly sensitive to vapour pressure, improve the product data. If the estimate changes dramatically when the roof-landing count changes, verify operating records. The calculation becomes more valuable when it directs the next measurement or inspection.

That is the point where emission estimation becomes engineering rather than reporting.