A Bay Area Air Quality Management District study tested what happens when a Method 21 probe is held one centimeter off a valve’s component interface instead of right at it, and the results should make any LDAR program manager uncomfortable. At that distance, inspectors caught only 79 percent of valves leaking between 100 and 500 ppm, and just 43 percent of valves leaking above 500 ppm.
One centimeter. About the width of a pencil. More than half the bigger leaks, missed.
It’s worth pointing out where the confusion comes from, because Method 21 actually has two different placement rules depending on what you’re monitoring. For pumps and compressors with rotating shafts, the rule is “within 1 cm of the shaft seal interface,” and that allowance exists because pressing a probe onto a spinning shaft is a safety hazard. For valves, flanges, connectors, and other static components, there is no 1 cm allowance at all. The probe goes at the surface of the interface. What the Bay Area study captured is what happens when crews carry the rotating shaft habit over to valves, and the data shows exactly what that costs.
EPA cites this study in the LDAR Best Practices Guide for good reason. When EPA’s NEIC audited 17 refineries, the facilities reported an average leak rate of 1.3 percent, while NEIC’s own monitoring at the same plants came back at 5.0 percent. That isn’t fraud. It’s technique.
Why it happens in the field
Method 21 is pretty specific about what good monitoring looks like: probe at the interface for valves, find the maximum reading, and hold for about twice the instrument’s response time. On paper that’s straightforward. On a hot day with 600 valves left before the truck rolls, here’s what actually goes wrong.
The probe gets swept along too fast for a TVA’s 2 to 4 second response time to catch up, so the reading never has a chance to register. Grease, paint, insulation, or even rain droplets on the component block the probe from reaching the seal in the first place. The “maximum reading” gets eyeballed instead of methodically worked for. And underneath all of it, pay per component contracts push everyone toward speed rather than accuracy.
The numbers tell the story. EPA has pulled data logger timestamps showing valves logged at one per second, and at one facility a single tech reportedly hit 8,000 components in a shift. By contrast, a well trained two person team typically covers 500 to 700 valves a day. You can do that math.
What the good programs do differently
The programs that hold up under audit share a handful of habits, none of them glamorous.
The first is end of shift drift checks, not just morning calibrations. If you come back more than 10 percent negative at the end of the day, you re-monitor everything above 100 ppm for valves or 500 ppm for pumps since the last good calibration. It’s a pain, but it keeps false negatives off your reports before they become enforcement findings.
The second is internal leak thresholds set tighter than the rule. EPA recommends 500 ppm for valves and 2,000 ppm for pumps as an internal trigger, even when the applicable subpart allows higher. The field is never as clean as the spreadsheet, and that buffer is where you absorb the noise.
The third is spot checks of the actual monitoring, not just the records. Walk the unit with a tech once in a while. Run a second analyzer side by side on a sample of valves. Records can look immaculate while the probe technique is wrong, and that’s exactly what an inspector will find when they show up.
The last one is QA/QC before the data leaves the tech’s hands at all. That means daily sign off, components per shift sanity checks, and somebody other than the monitoring tech reviewing flagged readings.
A clean monitoring report and a compliant LDAR program are not the same thing, and inspectors have caught on. The reason recent consent decrees keep getting more specific about task cycle definitions, training records, drift thresholds, and contractor oversight is because regulators stopped trusting the paperwork a while ago.
If your last internal audit looked at records but didn’t watch how the probe gets held, that’s a blind spot worth closing before someone else closes it for you.