Crawl Space Entry and Atmosphere Check

Purpose

Nobody from this shop goes under a house until the air under it has been measured with an instrument, an attendant is posted at the opening, and the way out is clear.

Without that, the failure is quiet. Crawl spaces under the houses we pier and drain hold oxygen displaced by standing water and rotting organics, sewer gas from a cracked cast-iron drain, carbon monoxide from a flue that has separated under the floor, and vapor off whatever the homeowner stored down there. None of it announces itself at the hatch. The tech who has crawled two hundred without incident is the most likely person to go in on the two hundred and first without a meter, because the pattern he has learned is that crawl spaces are fine, and the one that is not fine looks exactly like the rest from the access door.

Scope

Covers every entry under a suspended floor through a hatch, an exterior door, or an opened rim: classification, atmosphere test, attendant, retrieval plan and abort. It covers inspection visits as well as production work, because the atmosphere does not know which one you came for. It does not cover the structural work once you are inside (see the pier and beam crawlspace underpinning sequence) or entry into an open excavation, which belongs to the exterior waterproofing and backfill SOP.

Which standard binds turns on the activity, not the building. Under-floor work that alters the structure - piers, beams, drain tile, encapsulation - is construction, so 29 CFR 1926 Subpart AA governs and 1926.1203 puts the identification and classification duty on a competent person. A pure inspection call by a general-industry employer falls under 29 CFR 1910.146 instead. They differ on paperwork, not on the test you run or the numbers below.

Roles and handoffs

Procedure

1. Classify the space before the truck rolls. The competent person states in writing on the ticket whether this crawl is permit-required and why. Acceptance: a classification, a name, and a date within 12 months or later than the last recorded change to the space. Wrong looks like a blank field, or one copied from a neighboring address. Stop rule: no entry, call for a fresh classification, and if nobody answers treat the space as permit-required. The hazard is silent: a non-permit crawl becomes permit-required the day a plumber runs a gas line through it or the homeowner starts storing fuel down there, and neither generates a call to you.

2. Shut down the carbon monoxide sources you control, and find the ones you do not. Before the hatch is touched, turn the truck off rather than leaving it idling near the access, and walk the perimeter for a running generator, pressure washer, kerosene heater, or an attached garage with a vehicle in it. Acceptance: no combustion engine running within 20 ft of any crawl opening or foundation vent, and the homeowner asked directly. Wrong looks like an extension cord snaking out of a garage side door. Stop rule: the source goes off before the meter goes in, and if the homeowner will not shut it down the entry reschedules. Exhaust reaches an under-floor space through vents and rim penetrations far more readily than anyone expects, and it is odorless, so nobody's nose confirms anything.

3. Bump-test the meter at the truck. Use a four-gas instrument covering oxygen, combustible gas as percent of lower explosive limit, carbon monoxide and hydrogen sulfide, bumped against known gas per the manufacturer's instructions before the day's first entry. Acceptance: every sensor responds and returns to baseline, calibration date inside the stated interval. Wrong looks like a sluggish sensor or one reading low on bump gas. Stop rule: tag it out and take the spare, because a meter that failed a bump test never gets used with a mental allowance for the error. A dead sensor reads a clean, believable zero.

4. Test from outside the opening, with the pump and hose. A crawl space is not a tank. In a vault you sample top, middle and bottom; in a 24 in crawl the whole space is the bottom, so heavier-than-air gases pool exactly where your chest will be. Lay the probe hose along the ground an arm's length past the opening and run the pump for the manufacturer's stated purge time for that hose length. Acceptance, all four together: oxygen 19.5 to 23.5 percent, the deficient and enriched definitions at 29 CFR 1910.146; combustible below 10 percent LEL, that standard's flammable threshold; carbon monoxide below 35 ppm, the NIOSH recommended 8-hour limit, which sits under OSHA's 50 ppm permissible limit at 1910.1000 Table Z-1 and so alarms with margin; hydrogen sulfide below 10 ppm, the NIOSH ceiling, against OSHA's 20 ppm acceptable ceiling at Table Z-2. Stop rule: nobody enters on three good readings and one bad one.

5. Ventilate, then retest, and confirm the source is dead before trusting the second number. Set an electric blower at one opening and give the space a path out at another, computing the volume rather than guessing: length by width by clear height. Acceptance: all four back in band on a fresh test, with the source of the original reading identified. Wrong looks like the trap this step exists for, a number that falls and then stalls, which means you are diluting a live source rather than clearing stale air. Stop rule: an improving reading that holds short of the limit means the source is still producing, so stop ventilating, kill the source, then retest. A gasoline or propane blower at the intake feeds its own exhaust into the space, so site any engine well clear of every vent.

6. Post the attendant and establish egress before anyone goes prone. The attendant stays at the opening for the whole entry and does not leave to fetch tools. Clear stored material so the access is a full opening, and confirm the entrant can reverse out of the deepest planned position without turning around. Acceptance: voice contact demonstrated by a real call and response from that deepest point, with a second person on site who is not the entrant. Wrong looks like a one-man call where the attendant is the homeowner. Stop rule: no attendant, no entry, reschedule rather than improvise. Outside responders cannot perform a fast rescue through an 18 by 24 in hatch, which is why the retrieval line is set rather than why there is no plan: name the evaluated rescue service on the permit rather than "911", record that it was contacted and can reach this address inside a stated time, and put the entrant in a full-body harness with a retrieval line back to the opening, per 29 CFR 1910.146(k) for general industry and 1926.1204(i) for construction. Where no service can be evaluated for that address, the entry does not happen.

7. Enter with the meter in the breathing zone and monitor continuously. Clip the instrument at collar height rather than leaving it in the bag behind you, and call readings at a fixed interval set before entry. Where rodent droppings, nesting material or a dead animal are present, do not sweep, dry-brush or vacuum any of it: wet it with disinfectant per CDC hantavirus guidance and wear at least a half-face respirator with P100 cartridges under a written program per 29 CFR 1910.134, because that is an inhalation route and gloves do nothing for it. Acceptance: all four readings hold in band throughout. Wrong looks like carbon monoxide rising as you near the furnace or water heater, usually a separated flue rather than an instrument fault. Stop rule on any alarm: exit immediately, leave the tools, retest from outside once clear.

8. Close the permit at the hatch, not at the shop. Record exit time, the readings taken during the entry, and anything that changes the classification for the next crew. Acceptance: every field carries a value, including the uneventful ones. Wrong looks like a permit filled from memory in the truck at day's end, which reliably loses the intermediate readings. Stop rule: where a field cannot be filled honestly, write what happened, because a blank reads later as a skipped step. The hazard managed here is the next crew's, not yours.

The record this produces

One entry permit per entry, per opening, per day. Fields: date, address and ticket number; competent person, classification and reason; instrument serial with its last bump-test and calibration dates; the four pre-entry readings with times; every retest with its time and the ventilation run time between them; blower type and placement; entrant and attendant names; entry and exit times; conditions found that change the classification; both signatures.

It attaches to the job ticket and copies into the site file for the address. The next competent person reads it before dispatching anyone here, which is the only reason step 1 can ever be answered from a file. The safety lead reads the month's permits for the pattern no single job shows, such as three aborts on one street after a storm. And after an incident it is the dated record of what was known and when; whether it carries further legal weight is a question for the shop's own attorney and insurer.

Worked pass: pier and beam, sticking doors, day after a regional outage

Elevation survey under a 38 by 26 ft crawl with about 22 in of clearance. Site file classifies it permit-required from 8 months ago, reason recorded as gas line and sewer lateral both under the floor, so step 1 passes. Step 2: truck off, perimeter walked, and the homeowner confirms she ran a portable generator in the attached garage overnight and shut it off about an hour before arrival. No live source found. Step 3: bump test clean, calibration current.

Step 4 fails. Hose laid 6 ft in at ground level reads oxygen 20.6 percent, combustible 0 percent LEL, hydrogen sulfide 0 ppm, carbon monoxide 62 ppm. Three of four in band, one out, so nobody enters.

Step 5: volume is 38 by 26 ft, which is 988 sq ft, at 1.83 ft of clear height, so 1,808 cu ft. The blower is rated 1,200 cfm, one nominal air change every 1.5 minutes and five changes in about 7.5 minutes. That nominal figure assumes perfect mixing, and a crawl with one blower opening and scattered vents short-circuits badly, so the crew treats it as a floor rather than a schedule and runs 20 minutes.

Retest at 20 minutes: carbon monoxide 41 ppm. It fell, so the tempting move is another 20 minutes and then entry. It stalled instead, so the stop rule says the source is still live. They find the generator inside the closed garage with a warm tank and the door down. It was off, but a hot engine in a small closed garage was still off-gassing into a space that communicates with the crawl rim. They roll it to the driveway and open the garage. Twenty-five more minutes: carbon monoxide 3 ppm, oxygen 20.9 percent, 0 percent LEL, 0 ppm hydrogen sulfide. Step 5 passes.

Step 6: attendant at the hatch, call and response confirmed from the far corner. Step 7: meter at the collar, readings called every 10 minutes and holding; rodent droppings along the east rim, so the survey is worked in a P100 half-face and nothing is brushed or swept. Step 8: permit closed at the hatch with both retests and their times, 45 minutes of blower run time, and a new classification note that this address has an attached garage communicating with the crawl.

What the crew banked is not that generators make carbon monoxide. It is that the first retest moved in the right direction and was still wrong.

References

• 29 CFR 1926 Subpart AA, Confined Spaces in Construction, including the competent-person identification duty at 1926.1203
• 29 CFR 1910.146, Permit-Required Confined Spaces, for general-industry inspection work, including the oxygen and 10 percent LEL definitions
• 29 CFR 1910.1000 Table Z-1 and Table Z-2, and the NIOSH Pocket Guide to Chemical Hazards, for the carbon monoxide and hydrogen sulfide alarm points
• 29 CFR 1910.134, for the written respiratory protection program required before any respirator is worn
• CDC guidance on hantavirus and cleanup of rodent-contaminated spaces
• See related: pier and beam crawlspace underpinning sequence; crawlspace encapsulation and vent screening