Gas Testing Before Anyone Opens a Tank

Purpose

This SOP produces a documented, instrument-based decision about whether a structure is safe to work OVER, made before any lid or riser cap is lifted, and made the same way regardless of who is on the truck.

Without it a shop runs on smell, and smell is the sense hydrogen sulfide takes away first. Low concentrations stink; as concentration climbs the olfactory nerve fatigues, so the tank that stops smelling bad is usually the tank that got worse. The recurring fatality in onsite wastewater is not one person, it is two or three: the tech goes down and the people who reach in after him go down on top of him.

Scope

Covers non-entry work at any onsite structure with a removable cover: septic tanks, pump and dosing chambers, aerobic treatment unit compartments, grease interceptors, lift station wet wells, and distribution boxes deep enough to hold a headspace.

Does NOT cover entry. The moment any part of a person breaks the plane of the opening this SOP has ended and the shop's permit-required confined space program under 29 CFR 1910.146 takes over, with a permit, an entry supervisor, an attendant who does nothing else, and a retrieval system (1910.146(k)(3) requires a mechanical retrieval device for a vertical space more than 5 feet deep unless it would increase the risk). Crews installing rather than servicing may fall under 29 CFR 1926 Subpart AA instead; the office decides which at booking. Pumping, electrical isolation and lid excavation are owned by their own SOPs.

Roles and handoffs

Procedure

1. Confirm the job is non-entry before the truck leaves. Anything that cannot be done from outside the opening plane, such as retrieving a pump off the floor, is an entry job. Acceptance: an explicit non-entry flag and a named second person on the ticket. Wrong is "service tank" on a one-man dispatch; do not roll until dispatch writes the scope down. A tech who learns this is an entry job while standing over an open tank improvises, and that is how the second person ends up in the hole.

2. Bump test the four-gas meter in clean air on the tailgate. Apply challenge gas, confirm every sensor responds and every alarm latches audible, visual and vibrating, then let it clear to a 20.9 percent oxygen baseline. Acceptance: all four sensors respond inside the manufacturer's stated response time. Wrong is a sensor that lags or alarms visually only; tag it out and use the spare. A bump test is not a calibration, so check the due date and honor the interval in that instrument's manual.

3. Kill the ignition sources before the lid moves. Park so truck exhaust and blower discharge sit downwind of the opening, and set a zone the customer, children and pets stay outside of. Acceptance: discharge downwind and a vehicle length off the opening, no smoking, no torch, no grinder, no unrated drop light. Wrong is the truck idling upwind with the blower engaged, pulling its own exhaust across the manway; reposition before the lid is touched. Headspace can sit above methane's 5 percent lower flammable limit, and an idling engine is a continuous ignition source at ground level.

4. Crack the cover from upwind with a long-handled tool. Break the seal with a lid lifter, key or pry bar while standing upwind and to the side, never straddling the opening, and open only far enough to admit the sample line. Acceptance: no part of the tech has crossed the opening plane. Wrong is a tech on both knees with his face over the gap because the lid is heavy; stop and bring a mechanical lifter. The first slug out of a freshly cracked tank is concentrated headspace arriving at face height for a kneeling person.

5. Take the breathing-zone reading first, at the opening plane. Hold the instrument at the height a working tech's face will occupy for a full 30 seconds; this is the exposure figure and it never shares a record slot with the depth readings. Acceptance: oxygen 19.5 to 23.5 percent, LEL under 10 percent, hydrogen sulfide under the general-industry ceiling. Wrong is a reading taken at chest height or on an instrument still clearing from the last site; a failed gate here means back away upwind and go to step 8. This reading decides whether respiratory protection is required, and that call is made under a written program at 29 CFR 1910.134, not by grabbing a cartridge mask off the truck.

6. Sample three depths in the standard's order, with purge time at each. Lower the draw-pump line to a foot below the cover, then mid-headspace, then 6 inches above the liquid, reading oxygen first, then flammable gases, then toxic contaminants, the order 29 CFR 1910.146 sets because an oxygen-starved sensor misreports combustibles. Acceptance: a stable display held for the manufacturer's purge and response time, which grows with sample-line length. Wrong is three readings inside 20 seconds, which is the top layer read three times. Methane runs about 0.55 the density of air and collects at the top, hydrogen sulfide about 1.19 and pools at the bottom, so a clean read at the cover says nothing about the layer a vacuum hose stirs up.

7. Compare every reading to a stated gate and write the call down. Oxygen below 19.5 or above 23.5 percent is a hazardous atmosphere under 29 CFR 1910.146(b); flammable gas above 10 percent of its lower flammable limit is hazardous under the same definition; hydrogen sulfide carries a 20 ppm general-industry ceiling at 29 CFR 1910.1000 Table Z-2, with construction under its own table at 29 CFR 1926.55. Acceptance: every value recorded beside its gate with a pass or fail. Wrong is a record reading "tested, OK"; one failed gate at any depth means no work over the opening until a re-test clears it. NIOSH puts the hydrogen sulfide IDLH at 100 ppm, so a bottom reading near that hurts someone the moment the crust breaks even when the opening plane reads clean.

8. Ventilate mechanically, discharge to open air, never with oxygen. Duct a blower to the failed depth, aimed to sweep rather than straight down at the liquid, and run it while nothing else happens at the opening. Acceptance: at least the run time the blower manufacturer states for that volume, treating 15 to 20 minutes as a working minimum for a residential tank rather than a proven number for yours. Wrong is a shop fan laid across the lid, moving air over the hole instead of through it. Two cycles that fail to clear a gate is a site problem: close the lid and call the service manager. Never ventilate with oxygen, since above 23.5 percent the space becomes an ignition hazard on its own.

9. Shut the blower off, wait, then re-test all three depths. Kill the blower and give the space at least 2 minutes to re-equilibrate, longer where a heavy crust is holding gas underneath it. Acceptance: all three depths pass every gate from step 7 with the blower OFF. Wrong is a re-test taken while the blower runs, which reads dilution air at the sample point rather than the space. A re-test that passes only with the blower running is a fail: either continuous ventilation becomes a written condition of the work or the job stops. The crust is the reservoir, so on a crusted tank plan a further reading once it is disturbed.

10. Monitor continuously while the cover is off, and log the peak. Clip the instrument at the opening in continuous mode with alarms audible over the blower and the vacuum pump. Acceptance: no alarm during the work, and the peak opening-plane value recorded at the end. Wrong is the meter back on the tailgate because it kept alarming; any alarm means everyone steps back upwind and the space is re-tested from step 5. The peak arrives when the hose breaks the crust or the liquid drops and exposes wet sidewall, the moment the tech is least focused on the display.

When the site does not match

A lid grouted, tarred or cast in that will not crack with a hand tool is an excavation job; reschedule it. A structure surcharged to the underside of the lid has no headspace, so the gas test is meaningless: treat it as failed and work it as a backup call. A customer who wants the test skipped because the last contractor just opened it gets one sentence and no debate. A meter that alarms in fresh air is a dead instrument, not a dirty site.

The record this produces

One atmospheric test record on the work order: date and time, structure identifier, technician and second person, instrument make and serial, last bump and last calibration dates, the opening-plane values for all four gases, the three depth readings for all four gases, each gate marked pass or fail with its value beside it, ventilation start and stop times, the re-test values, the peak opening-plane value, and the final call.

The service manager reads the peak column monthly and finds the sites that need standing ventilation before a tech does. Whoever prices the next visit reads the ventilation time and stops bidding that account as a straight pump-out. If an exposure claim lands, this record is the only thing showing what was known and what was done.

Worked pass: 1,000 gallon two-compartment tank, riser to grade

Ticket flagged non-entry, two techs. Bump test passed all four sensors with all three alarm modes latching. Cover cracked from upwind with a lid key. Opening-plane read at face height for 30 seconds: oxygen 20.8 percent, 0 percent LEL, 3 ppm hydrogen sulfide, 0 ppm carbon monoxide, all inside their gates. Depth reads followed, oxygen then combustible then toxic at each stop:

Step 7 failed. The bottom depth read 18.9 percent oxygen against the 19.5 percent floor at 1910.146(b), so that gate failed outright. The 9 percent LEL sat under the 10 percent gate with no margin, and 96 ppm hydrogen sulfide is 4 percent below the 100 ppm NIOSH IDLH. Nobody was entering, but the vacuum hose was about to break the crust and lift that bottom layer to the opening, so the step 7 stop rule applied and work stopped.

Blower ducted to mid-headspace and swept rather than aimed at the liquid, run 18 minutes. Blower off, 3 minutes to settle, re-test all three depths: oxygen 20.4 / 20.2 / 20.1 percent, LEL 2 / 2 / 3 percent, hydrogen sulfide 6 / 14 / 21 ppm, top to bottom. Every gate cleared with the blower off, so step 9 passed on the first re-test.

Work proceeded with the instrument clipped at the opening in continuous mode. The peak came where it usually does: 11 ppm hydrogen sulfide at the opening plane when the hose broke the crust, under the 20 ppm Table Z-2 ceiling, so no respiratory protection was triggered, and the account moved to a longer standard visit length.

Checking the run against the rules above: the opening-plane figures, 3 ppm rising to a peak of 11 ppm, are the exposure quantity compared against the 20 ppm ceiling, and the depth figures of 27, 61 and 96 ppm are the hazard inventory and were never compared to it. Ventilation ran 18 minutes against the stated 15 to 20 minute minimum, and the re-test was taken with the blower off.

References

• 29 CFR 1910.146, permit-required confined spaces (general industry): hazardous atmosphere definition at (b), testing order, retrieval at (k)(3); 29 CFR 1926 Subpart AA for construction work.
• 29 CFR 1910.1000 Table Z-2 for the general-industry hydrogen sulfide ceiling, 29 CFR 1926.55 for construction limits, 29 CFR 1910.134 for any respirator decision.
• NIOSH Pocket Guide to Chemical Hazards for immediately dangerous to life or health values.
• See related: the septic tank pumping SOP, the lift station and pump service SOP, the tank locating and lid excavation SOP.