Documenting clearance work so it survives an audit

A clearance record in California is written to four content standards: GO 165 Section III.C for the inspection, GO 95 Rule 18A(1)(b) for the corrective work, FAC-003-5's measures on applicable transmission, and none at all for the Public Resources Code work itself. An auditable entry names the object, the rule and subdivision, the measurement, the level, and the closure.

An audit of a vegetation program does not begin under a conductor. It begins with a records request, and the field visit that follows is a test of whether the records were true.

The Safety and Enforcement Division's electric distribution audit of Turlock Irrigation District is a public example of the shape. Two Electric Safety and Reliability Branch staff, Dmitriy Lysak and Sajjad Mansuri, spent March 18 to 22, 2024 on it.

The report, EA2024-1181, went out June 4, 2024, gave the district until July 2 to respond, and told it that both the report and the response would be posted on the CPUC website. Before anyone walked a circuit, staff had asked for eleven categories of record, most of them covering January 2019 to January 2024.

That is the reader this page is written for: someone who was not there, who arrives years later, and who is going to decide from the file alone whether the work happened.

There is no single clearance record

The phrase "the clearance record" describes a thing that does not exist. A single trimmed span in California generates entries under as many as four regimes, written by different people into different systems. Three of them set a content standard. The fourth sets none.

The inspection that found it. GO 165 Section III.C sets the content standard for a distribution inspection record, and it is five elements in one sentence:

• The circuit, area, facility or equipment inspected.

• The inspector.

• The date of the inspection.

• Any problems or items requiring corrective action.

• The scheduled date of corrective action.

What that standard does and does not reach is its own subject.

The work that closed it. GO 95 Rule 18A(1)(b) governs the other half, and it names three different facts:

Upon completion of the corrective action, the company's records shall show, with sufficient detail, the nature of the work, the date, and the identity of persons performing the work. These records shall be preserved by the company for at least ten (10) years and shall be made available to Commission staff upon 30 days notice.

Five elements on one side, three on the other, and no field in common except the date. Nothing in either standard requires the two entries to reference each other, which is why so many programs cannot demonstrate that a particular trim closed a particular finding.

The transmission line above it. On an applicable line, FAC-003-5's measures set out what evidence each requirement is proved by, and they are drafted as lists of acceptable artifacts rather than as content standards. Which lines the standard reaches decides whether this layer exists at all on a given span.

The statute the same conductor also sits under. Nothing. Sections 1250 through 1258 of Title 14 impose no record obligation on clearance or firebreak work at all; the article's only affirmative documentation duty attaches to a claimed exemption, not to the work. So the record that proves PRC 4292 and 4293 compliance is built to the Commission's rules or to the utility's own program, and on a non jurisdictional line it is built to nothing but the program.

One more layer sits underneath all of them and is easy to miss. A dedicated vegetation patrol is not a GO 165 inspection, because neither GO 165 inspection type is a vegetation inspection. Its record is written to the utility's own procedure and to the cycle committed in its wildfire mitigation plan under Public Utilities Code section 8386(d).

The correction record, though, does not care how the finding was made. A Rule 35 nonconformance is a GO 95 nonconformance, and Rule 18A governs the closure whether it was spotted from a truck, from a helicopter, or by a customer on the phone.

What the records request actually looks like

Staff asked Turlock Irrigation District for these:

TID Overhead/Underground Preventative Maintenance Program. Overhead and underground facilities statistics. Completed work orders with notifications, canceled work orders with notifications, and open work orders with notifications from January 2019 to January 2024. Patrol and detailed inspection records from January 2019 to January 2024. Reliability metrics and sustained outages from January 2019 to January 2024. TID system map. New Construction projects (both overhead and underground) from January 2023 to January 2024. Pole loading and safety factor calculations completed from January 2023 to January 2024. Inspector list from January 2019 to January 2024 and inspector qualifications. Equipment test records from January 2019 to January 2024. Intrusive inspection records from January 2019 to January 2024.

Three features of that list are worth transferring to any program's expectations.

The program document comes first, and it is read as a requirement. The maintenance program a company writes under Rule 18A(2)(a) is the document the rest of the file is measured against, which means an internal standard that is more demanding than the rule becomes the standard the audit applies.

Canceled and open work orders are requested by name, alongside completed ones. A file that can produce closures and not cancellations has answered a third of the question. The interesting entries in any maintenance system are the ones that were opened and then went somewhere other than done.

And the longest window in the list is five years, against a ten year retention obligation. Staff asked for half of what the utility is required to hold, and for new construction and pole loading calculations only one year. There is no rule fixing the length of a request, and the retention period is a separate question with a separate answer for each artifact, but a program built to survive exactly the minimum has no margin for the audit that reaches further back.

The identifier is the spine, and two poles at 1125 Paulson Rd are one row

Table 3 of that audit lists the locations staff inspected in the field. The first five are street addresses:

• 921 S Central Ave.

• 980 S Central Ave.

• 1033 S Central Ave.

• 1125 Paulson Rd, location 4.

• 1125 Paulson Rd, location 5.

Both Paulson Rd entries are recorded as poles in Turlock. Locations 6 through 10 are structure IDs, Y08 022 through Y08 026, in Lake Don Pedro.

Two different naming schemes in one week of fieldwork, and in the first of them two distinct structures carrying the same name. Whatever those two poles at 1125 Paulson Rd were, the file that closes a finding on one of them has to say which.

Every regime in this stack leaves the unit of record to the utility, so this is not a problem any rule will solve. GO 165 Section III.C accepts "the circuit, area, facility or equipment inspected," four levels of granularity in one clause.

Note 1 to the Section III.D annual report template says outright that each utility will define its reporting unit basis, whether circuit, grid or facility. FAC-003-5 R6 requires inspection of 100 percent of applicable lines "measured in units of choice" and offers circuit, pole line, line miles and kilometers as examples, and R7 repeats the phrase for the annual work plan.

The constraint that survives all that freedom is narrow and absolute: the finding and its closure have to land on the same object. A Rule 18A correction attaches to a nonconformance on a facility. A vegetation clearance is a property of a span, which is to say of an ordered pair of structures and a phase.

A GO 165 patrol record may legitimately be written at circuit level. If the program records findings by span and closures by circuit, no amount of subsequent work will let anyone prove that span 1892096E to 1892097E was the one that got cut.

That is also why the audit's own location numbers matter less than they look. Table 7 of the Turlock report records the Rule 35 findings against locations 17, 47, 51, 115 and 117. Those numbers belong to the audit, not to the district's work management system. Somebody has to map them, and the mapping is easier where the audit was able to write down a structure ID than where it wrote down an address.

The finding as written, and the number it does not contain

Here is the whole of Table 7:

Location	Finding
17	Service drop strain
47	Overgrown vegetation near guy wire
51	Overgrown vegetation
115	Service drop strain
117	Vegetation near primary lines

Five vegetation violations of GO 95 Rule 35, and not one distance among them.

Two of the five use the rule's own defined term. Rule 35 says strain is present "when vegetation contact significantly compromises the structural integrity of supply or communication facilities," and it exists as a defined term precisely because contact by itself is not a nonconformance. A finding of "service drop strain" is therefore a conclusion under a definition, and closing it means either correcting the condition by one of the four remedies the rule names or showing that the definition was not met.

The other three name a condition and leave the measurement open. "Overgrown vegetation" is not testable against Table 1 by anyone. To close it the utility has to supply what the finding does not: which case and column applied, what was measured, at what point on the span, and under what conditions. The case and column and the annual weather variations condition footnote (ddd) puts on those two cases are what turn a description into a distance somebody else can check.

This is the ordinary condition of an inbound finding, not a defect in it. A fire agency notice arrives the same way on CAL FIRE's form LE-38a, and what that notice obliges is set out elsewhere. The point for the file is that the burden of the number falls on the utility's record in both cases, and a closure entry that mirrors the vague wording of the finding has closed nothing.

The due date in the file is the utility's own number

Rule 18A(2)(a) is the only place in the vegetation stack where a date is created by a person rather than by a table. The Level 2 correction period, it says, is

to be determined at the time of identification by a qualified company representative

within ceilings of six, 12 and 59 months that turn on the condition and where it sits.

Four facts are created in that moment and all four belong in the record:

• The priority level.

• The fire risk characterization and the tier it was made against.

• The representative who made the call.

• The period they set.

A file that carries the level and not the person has recorded a decision with no decider, and "determined at the time of identification" cannot be demonstrated after the fact.

The ceiling is not the deadline. Turlock's own maintenance program defines Priority 2 as a condition to be repaired within 30 calendar days, Priority 3 within six months, Priority 4 within two years. Staff reviewed 60 months of work orders against those self set intervals and found 296 that were not addressed by their assigned due date, 143 of them still pending.

Work order 22000078, opened February 10, 2022 to relocate a down guy with a required end date of February 11, 2022, was still open when staff rechecked on April 22, 2024, 809 days past due. Work order 19000937, opened October 11, 2019 to replace a transformer with a required end date ten days later, was 1,653 days past due on the same check.

None of those numbers came from a general order. They came from subtracting one field in the district's own system from another. That arithmetic is available to anyone who is handed the file, which is the single most useful thing to understand about a due date: it is not a target, it is a claim, and it is checkable by subtraction forever.

Rule 18A(2)(b) provides the release valve, and it is a list rather than a discretion. Correction times may be extended under reasonable circumstances

such as: Third party refusal; Customer issue; No access; Permits required; System emergencies (e.g. fires, severe weather conditions).

Nothing in the rule says to record which one applied. So an extension entered without a stated basis, an author and a new date is indistinguishable in the file from a missed date, and it will be read as one. Where the basis is a refusal, Exception 2's two artifacts are a separate file answering a separate question, and the extension record should point at it.

A record that cannot be produced did not happen

The most instructive violation in the Turlock report is not about a pole. Staff wrote it up under GO 165 Section III.B, the inspection standard:

TID's underground inspection tracking software MIMS has been inaccessible since 2021. As a result, TID was not able to provide any records of underground inspection in the requested time period of January 2019 to January 2024.

The inspections may well have been performed. The finding is written against the rule that requires them anyway, because from outside the file there is no difference. Three years of a system being unavailable produced the same result as three years of nobody going out.

That is what the 30 day clause is really specifying. GO 165 Section III.C says records "shall be made available to parties or pursuant to Commission rules upon 30 days notice," and Rule 18A(1)(b) repeats the period for corrective action records.

Thirty days is a requirement on the system, not on the staff who answer the letter. Any archive, any retired platform, any contractor's field application has to be able to yield a span level history inside a month, for the full retention period, after the vendor contract has ended and the people who ran it have left.

The federal side adds a wrinkle in the other direction. FAC-003-5's evidence retention clause holds three calendar years of evidence for R1, R3, R5, R6 and R7. It then says that where the retention period is shorter than the time since the last audit, the Compliance Enforcement Authority may ask for other evidence covering the whole interval. A program that deletes on the three year boundary can be asked for the years it deleted.

Attestation and measurement are different kinds of evidence

FAC-003-5's measures are unusually explicit about what an auditor will accept, and reading them next to the California rules shows two regimes with different theories of proof. M1, on the core requirement not to let vegetation encroach into the MVCD:

Each applicable Transmission Owner and applicable Generator Owner has evidence that it managed vegetation to prevent encroachment into the MVCD as described in R1. Examples of acceptable forms of evidence may include dated attestations, dated reports containing no Sustained Outages associated with encroachment types 2 through 4 above, or records confirming no Real-time observations of any MVCD encroachments.

M5 lists initially planned work orders, documentation of constraints from landowners, court orders, inspection records of increased monitoring, documentation of the de-rating of lines, revised work orders, invoices, or evidence that the line was de-energized. M6 lists completed and dated work orders, dated invoices, or dated inspection records. M7 lists a copy of the completed annual work plan as finally modified, plus the same three.

The word dated appears eight times across M1, M6 and M7, attached to attestations, reports, work orders, invoices and inspection records. That is the standard's real content requirement and it is worth taking literally: an undated work order proves nothing under FAC-003 whatever else is on it.

An attestation is available for R1 because R1 is a negative. Proving that nothing encroached is not the same kind of act as proving that a crew cut a limb, and the standard accommodates it. California offers no equivalent. Rule 18A(1)(b) asks for the nature of the work, the date, and the identity of the persons who did it, which are three facts about an event, and GO 165 Section III.C asks for five facts about an inspection. Neither will take a signature certifying a conclusion in place of them.

The difference matters most two years later. An attestation that a span was clear can only be attacked with contrary evidence, and there usually is none. A measurement is a datum: it can be compared against the specification it was cut to, against the same span's previous reading, and against the growth rate the trim cycle was built on. One certifies; the other can be re-derived by a stranger. Only the second one improves the program.

What a photograph settles, and what it does not

Photographs are the most common thing in a modern vegetation file and the most commonly overestimated.

Start with what a photograph can decide against Rule 35, because the rule is specific enough to give a clean answer. Abrasion is "damage to the insulation resulting from the friction between the vegetation and conductor," and the rule then says that scuffing or polishing of the insulation or covering is not abrasion.

An image showing a bright polished stripe on a triplex jacket, with no cut into it, is a photograph of a condition the rule expressly excludes. It is useful evidence and it is evidence in the utility's favor, which is exactly why it should be in the file.

Now what it cannot decide. A photograph carries no scale, and it collapses depth, which is frequently the axis the clearance runs along. An image of a limb near a conductor establishes that the two were near each other from one viewpoint on one day. It cannot establish 30 inches, and it cannot establish the conditions the measurement has to hold under, whether that is Table 1 footnote (ddd)'s normal annual weather variations or PRC 4293's 120 degree position. A photograph corroborates a measurement and never substitutes for one.

What raises an image from illustration to evidence is mostly metadata and discipline. The device timestamp and coordinate, retained rather than stripped by the upload. A stated bearing or a landmark, so the viewpoint can be reoccupied.

Something of known dimension in frame, an insulator, a crossarm, a marked pole tag. And the pair: the same view before and after, taken from the same position, which is the only version of a photographic record that shows what changed rather than what exists. A single after image proves that somebody stood there once.

The entry that says nothing found

GO 165 Section III.C opens its content standard with three words that get skimmed: "For all inspections." Not for all findings. The fourth element, any problems or items requiring corrective action, is allowed to be empty. The row is not.

This is the field most often lost in the move from paper to tablets, because a blank row and an absent row look identical in a summary view and are opposite facts.

An auditor's report can legitimately carry findings alone. Table 7 of the Turlock report names five locations and is silent on the rest, because its job is to record violations. The utility's file has the opposite job. A circuit walked in one pass with five findings on it proves that five spans were looked at, if the file holds five rows. It proves the circuit was covered only if it holds a row for every span, and the rows carrying nothing are the entire difference between the two claims.

FAC-003-5 builds the same idea into its evidence list from the other direction, and more explicitly than any California rule does: one of M1's three named forms of acceptable evidence is "records confirming no Real-time observations of any MVCD encroachments." A record of an absence, produced deliberately, is what proves the requirement was met.

The clear entry also does something the finding cannot. It is the only thing that dates the last time a span was known to be good, and on a span that later fails, the interval between that date and the failure is the measurement the program learns from.

Amending a record without destroying it

Every long lived file needs corrections, and the question is whether the correction is visible.

FAC-003 handles this well enough to borrow from. R7 requires completion of 100 percent of the annual vegetation work plan. It then says that modifications in response to changing conditions or to inspection findings

may be made (provided they do not allow encroachment of vegetation into the MVCD) and must be documented

and then defines the score:

The percent completed calculation is based on the number of units actually completed divided by the number of units in the final amended plan.

The denominator is allowed to move, and the amendment is priced into the arithmetic. The standard then offers nine examples of reasons for modification, from change in expected growth rate at 7.1 through permitting delays at 7.7 and land ownership changes at 7.8.

That is the shape to copy: an amendment is an event with a date, an author and a reason drawn from a named list, sitting alongside the original rather than on top of it. A system that overwrites in place cannot show a stranger which version an earlier decision was made against.

On the state side the stakes are higher because the annual report is sworn. GO 165 Section III.D requires each subject utility to file by July 1 for the previous year "under penalty of perjury," which means an amended report is a new sworn statement rather than a data fix. What that looked like when one utility found its inspection system and its field results disagreed is worth reading before the situation arises, because the sequence, reclassify, reassign, self report, amend, is the same one any program will need.

What the record has to name

Every page in this set closes by naming the fields its own rule makes checkable. This is the list underneath all of them, the fields a clearance entry needs regardless of which rule produced the finding.

The object, in the identifier the closing system uses. Circuit, structure, span as an ordered pair of structures, and phase. Where the finding arrived with somebody else's identifier, an audit location number or a street address, the mapping to the internal identifier is part of the record and not a lookup somebody does later.

The rule and the subdivision, not the rule alone. GO 95 Rule 35 with the Table 1 case and column, PRC 4293 with its subdivision, FAC-003-5 R1 with the Table 2 row. Two regimes on the same span produce two findings in two units of measure, and an entry that names only a distance cannot be sorted into either.

The observation, with its number and the conditions it was taken under. The measured distance, the point on the span it was taken at, ambient temperature, wind, and the date. A defect described as an observation rather than as a conclusion, so a later reader who disagrees has something to re-examine.

The level, the person, and the date of identification. Rule 18A(2)(a) creates all three at once. The correction period the representative set, and the resulting due date as a date rather than a duration.

The closure, in Rule 18A(1)(b)'s three facts, plus the number. Nature of the work, date, identity of the persons performing it, and the clearance actually achieved. The rule asks for three; the fourth is what makes the entry useful to the program rather than merely compliant.

The link between the finding and the closure. A finding identifier carried onto the work record. This is the one field no regulation requires and the one whose absence most often defeats an otherwise complete file.

The rows with nothing in them. Every span inspected, including the ones that were fine.

The artifact behind any exemption, refusal, extension or notice. The subsection claimed and the document that proves it:

Claimed	The document that proves it
An Exempt Tree	the confirmation and database coordinate for an Exempt Tree
A firebreak exemption	the hardware or cultivation evidence
A refusal	the two artifacts Exception 2 requires
A hazard on another company's facility	the Rule 18B notice, which the rule requires all parties to preserve for at least ten years

The version of every external reference relied on. The fire threat map edition, the responsibility area layer and its effective date, the internal procedure and its revision. A tier claim and a responsibility area claim are assertions about a location as of a date, and the boundaries move.

Assembled, one span reads as a single line that a stranger can test end to end:

Sonoma 1104, span 1892096E to 1892097E, west phase, 12 kV. GO 95 Rule 35, Table 1 Case 14 column E, 48 inches required. Measured 30 inches at closest approach, 96 degrees, light wind, August 14, 2026, CPUC HFTD v.3 Tier 2. Finding VM-2026-04417, Level 2, fire risk, 90 days set by [representative] on August 14, 2026, due November 12, 2026. Trimmed to 141 inches on October 2, 2026 by [crew], [foreman]. Paired photographs from the same station on both dates.

Every fact in that line is checkable against a published table, a named rule or another entry in the same file, and there is nothing in it a crew could not have written down on the day.

Where the record stops

A record is evidence about work, and it cannot repair the work. A span trimmed short is trimmed short whatever the file says, and the entry's real job in that case is to schedule the return visit off the shortfall rather than off the cycle. Documentation that dresses up an inadequate trim is worse than a thin record, because it is checkable and it will be checked.

Retention is a separate subject with a different answer for almost every artifact in the list above. It runs from ten years under GO 95 Rule 18A(1)(b) to the life of the pole for intrusive inspection records under GO 165 Section III.C to three calendar years under FAC-003-5. Which clock each piece of the file runs on is worth settling before a system is designed rather than after.

And the file has more than one reader, arriving through more than one door. Commission staff inspect records consistent with Public Utilities Code section 314(a), on 30 days notice. A Compliance Enforcement Authority works from the Reliability Standard Audit Worksheet and samples right of way across the biomes in the footprint rather than averaging over them. CAL FIRE writes its findings on form LE-38a and sends the utility a copy. Which body can do what with what it finds is a question worth answering in advance.

The exposure behind all of it runs on dates. The Commission's penalty is set per offense, with each day of a continuing violation counted separately, which makes the identification date and the correction date in the file the multiplier rather than merely two compliance fields. On Public Resources Code land, section 4021 makes the negligent or willful omission of an act required by chapters 2 through 6 of part 2 of division 4 a misdemeanor.

Turlock Irrigation District had until July 2, 2024 to describe its corrective and preventive actions, and was told at the outset that its answer would be posted publicly alongside the findings. That is the last practical fact about a compliance file. The entry that gets quoted is not the average one.

Further reading

• CPUC GO 95 Rule 35: the vegetation clearance rule, explained: Rule 18A's record standard, and the rule the record is written against.

• GO 165 inspection cycles, and the one thing they do not cover: The inspection record's own five element standard, and the sworn annual report it feeds.

• Recordkeeping and retention periods: How long each artifact in the file has to survive, which is not one number.

Frequently asked questions

What does GO 95 require a vegetation clearance record to contain?

Rule 18A(1)(b) requires that upon completion of the corrective action, the company's records shall show, with sufficient detail, the nature of the work, the date, and the identity of persons performing the work, preserved for at least ten years and made available to Commission staff upon 30 days notice. Notice what it does not name: the measured clearance, the case in Table 1, or the conditions of the reading. Those come from the finding side of the file, under GO 165 Section III.C.

Is a photograph enough to document a vegetation nonconformance?

Not by itself. Rule 35 defines abrasion as damage to the insulation resulting from friction and states that scuffing or polishing is not abrasion, so a photograph of a polished service drop jacket documents a condition the rule expressly excludes. A photograph also carries no scale along the axis of view, which is the axis a radial clearance is measured on. It corroborates a measurement; it does not replace one.

Who sets the correction due date that appears in the record?

The utility does. GO 95 Rule 18A(2)(a) says the time period for correction is to be determined at the time of identification by a qualified company representative, subject to ceilings of six months, 12 months and 59 months depending on the condition and its location. Where a company's own maintenance program sets a tighter number than the rule's ceiling, an audit measures against the tighter number.

How long does a utility have to produce records in a CPUC audit?

Thirty days. GO 165 Section III.C says records shall be made available to parties or pursuant to Commission rules upon 30 days notice, and GO 95 Rule 18A(1)(b) uses the same period for corrective action records. Public Utilities Code section 314(a) is the authority behind Commission staff inspecting them. Thirty days is a specification for the system holding the records, not just for the people answering the letter.