Deaths and injuries in US tree work, from federal records.

Every figure on this page is quoted from a federal dataset, and those datasets have a vocabulary. None of it is obvious, and most of it changes what a number means.

Tree work does not map to one clean code, and the answer moves depending on which is chosen. That makes the choice the most important thing on this page, so here it is with the alternatives that were rejected and why.

What was used

Fatality rates come from the BLS Census of Fatal Occupational Injuries for grounds maintenance workers, SOC 37-3010, labeled throughout as a proxy. Death counts come from the same census, Table A-5, for tree trimmers and pruners, SOC 37-3013. Enforcement comes from OSHA’s own enforcement database for landscaping services, NAICS 561730, federal jurisdiction. Non-fatal rates come from OSHA’s Injury Tracking Application for the same industry, and firm sizes from the BLS Quarterly Census of Employment and Wages. Each of those is defined in Terms.

Figure notes

How the 99 died

A count, not a rate. It carries no information about danger per hour worked, and BLS publishes no denominator for this occupation, so it must never be read as one or drawn on the fatal injury rate chart.

Compared with other jobs

The green row is a proxy, not tree work. BLS publishes no fatal injury rate for tree trimmers and pruners in any of the last four data years, because it cannot estimate the hours they work. Grounds maintenance is the narrowest group it does publish, and it contains landscaping and groundskeeping workers — 131 of the group’s 239 deaths — which is the safer half. The rate almost certainly understates tree work. The other six jobs are familiar comparisons, not a ranking: nine occupations rate higher than grounds maintenance, logging and roofing among them, and the full ranking is printed in this section. The chart rounds each published rate to whole people. Grounds maintenance sits above each job the headline names by more than both published margins of error.

Severe injuries by year

Federal jurisdiction only, and under-reporting by small employers is likely. A single report can carry more than one category, so the outcomes do not sum to the total. It is a count with no worker denominator, so it is never turned into an injury rate. The pace in the headline is the same count spread over time: the 3,982 days from 1 January 2015 to the file’s last report on 2025-11-25, divided by the 1,270 reports. Federal OSHA covers about half the country, so the real pace is faster. This is the whole industry code, tree care alongside lawn care — the three figures below it cut tree work out of it, and say how.

What a severe injury does

Tree work is our classification rather than an OSHA category — the rule is that the coded source or the narrative names tree work and the coded source does not name lawn care equipment, and it is in scripts/incident-data/classify.ts. A report counts as an amputation when OSHA flagged one, whether or not the worker was also admitted to hospital. 45 of the 55 amputations in this subset were a hand, a finger or a thumb. Exposure, transportation and everything else together produced none. Federal jurisdiction only, and under-reporting by small employers is likely.

Which equipment takes a body part

The rows are OSHA’s own coded source and secondary source fields, so which equipment a report names is not our classification — only the tree-work subset it is drawn from is. A report can name two sources and counts toward both, so the rows do not sum to the subset. The chart rounds each share to whole people out of 100. Read each one against the reports behind it, because a share out of 17 reports is a much softer number than a share out of 238: wood chipper 35, trees 238, bucket or boom truck 38, limbs and branches 128, chainsaw 68, skid steer or loader 17. One group is narrower than its label suggests — OSHA’s title "Saws-powered, except chainsaws" contains the word chainsaws, and those reports are excluded from the chainsaw row rather than folded into it.

When it happens

Eleven calendar years pooled, not a single year: the subset averages 65 reports a year, and one year of it split twelve ways would be noise rather than a season. Months are not equal in length, and February is not corrected for being shorter — at this scale the correction is smaller than the gap it would explain. Tree work is our classification and excludes lawn care, which matters here more than elsewhere: mowing is summer work, and leaving it in would flatten the fall peak into one long summer.

What OSHA fines for

The chart rounds penalties to the thousand. The exact amounts, the standard numbers and the citation counts are in the table in this section. Federal OSHA only. This excludes every State Plan jurisdiction — roughly half of US workers, California and Washington among them — so it is not a national picture. California has its own tree-work standard and issued 220 citations to this industry with only 5 General Duty Clause citations, which is the opposite shape. NAICS 561730 is Landscaping Services, which contains tree care but also lawn care and grounds maintenance. Penalties are current post-settlement amounts, not proposed.

What an injury costs in time

Establishments with 20 or more employees, which is the population OSHA requires to file a 300A summary electronically. That misses most of this trade — the average firm in it has seven people — and skews to the large employers that do file, utility line clearance among them. The tell is that this population’s own recordable rate runs below the BLS survey’s for the same industry, which is the opposite of what under-reporting by small firms would produce. Three filing years are pooled, because one year of one industry code moves on who filed as much as on who was hurt. Days away are counted the way OSHA counts them, in calendar days and capped at 180 per case, so a worker who never returns is recorded as 180 rather than as the career it was.

The ten most-cited standards, exactly

Federal OSHA, fiscal year 2025, landscaping services. Penalties are current post-settlement amounts. The names are ours, in the words a crew uses; the standard number is what makes each row checkable against OSHA’s own table.

The full fatal injury ranking

Every occupation BLS rated above grounds maintenance in 2024, in deaths per 100,000 full-time equivalent workers, with the margin of error it publishes. Rows 4 to 8 are separated by less than their margins of error, so their order carries no information.

Where the tree-work boundary was drawn#

NAICS 561730 is Landscaping Services. It files tree crews alongside lawn crews and has no tree-care sub-code, so the three figures above count a subset we defined: the coded source or the narrative names tree work, and the coded source does not name lawn care equipment. That gives 716 of the 1,270 reports. The rule is not described anywhere else — it is committed as code, in scripts/incident-data/classify.ts, and every number above is regenerated from OSHA’s raw file by that script rather than counted by hand.

Excluding mowing removes the more dangerous half, not the safer one. Lawn care is 172 reports in this industry code — about a seventh of them — and 82 amputations, which is more than a third of every amputation in the code. Its reports end in an amputation 48 percent of the time against 8 percent for tree work. The two trades have opposite injury profiles and an average of them describes neither, which is the whole reason for the cut — but it is a cut that makes our own trade look better, so the numbers it removes are printed here.

One more caution, because it caught us. OSHA changed its source coding part way through this dataset: it reuses a code across the break with a different meaning, and it reformatted every title. Counts keyed on either alone are wrong in opposite directions. Four figures in our own sourcing record were wrong for exactly that reason before the script existed, and none of them had reached this page.

What was rejected

SOC 37-3013 · Tree trimmers and prunersnot published

The occupation this page is actually about. BLS publishes no rate for it in 2021, 2022, 2023 or 2024: a rate needs 5+ fatalities and 40 million+ hours worked, and while the deaths clear the bar, the hours estimate does not support one. Used for counts only.

NAICS 561730 · Landscaping services18.5

An industry, not an occupation. Mixing an industry row into a chart of occupations would compare two different things on one axis. Used to corroborate the proxy — 18.5 against 20.9 is mild evidence the proxy is not wildly off.

SOC 37-0000 · Building and grounds cleaning and maintenance7.2

The major group. Includes janitors. Far too broad, and would understate badly.

SOC 37-1012 · First-line supervisors12.3

Supervisors only. Wrong population.

Line-clearance tree trimmingno code

No separate SOC code exists. It falls variously under tree trimmers and under electrical power-line installers, so it cannot be isolated.

Known weaknesses

• No fatality rate exists for tree trimmers specifically. The rate on this page is the broader group that contains them, and that group’s safer half is larger than its tree-work half — so it likely understates.
• In the full ranking, five occupations in the middle are separated by less than their published margins of error. Their order carries no information, and the table says so rather than implying a ranking the data cannot support.
• Two of the six event categories behind the 99 deaths are suppressed by BLS to prevent disclosure. They are drawn as a single gray slice rather than dropped, because dropping them would rescale every other slice against the wrong total.
• Every population on this page misses part of this trade, and the reason is its shape: the average landscaping business employs about 7 people. Electronic 300A filing starts at 20 employees, so the time-off figure is drawn from the large employers rather than the typical ones — its own recordable rate runs below the rate SOII reports for the same industry, which is the opposite of what under-reporting by small firms would produce. The self-employed are outside both.
• Every OSHA figure is federal jurisdiction only, excluding State Plan states — roughly half of US workers. California, which has its own tree-work standard, shows a different enforcement shape entirely.
• Three figures on this page count tree work as a subset of an industry code that has no tree-care sub-code. Where that boundary was drawn is our judgment, not OSHA’s, and it is printed in full below rather than described. The equipment rows inside it are OSHA’s own coded fields.
• No rate was derived by dividing a death count by an employment estimate. CFOI rates are hours-based; a headcount rate is a different measure and cannot share an axis with them.
• One widely repeated trade figure — 110 deaths per 100,000 tree trimmers — was traced and rejected. It is a magazine estimate built on a worker count its own author calls a significant underestimate, not a BLS rate.

Publishing a reference is a recurring obligation rather than a launch. These are the dates this page next becomes wrong, so you can judge how much to trust it and we can be held to fixing it.

BLS CFOI, 2025 dataExpected December 2026 to February 2027

Every row of the rate chart and the 6.3× multiple change.

OSHA FY2026 citation dataAfter September 2026, published with a lag

The citation, penalty and General Duty Clause figures all move. This is why the fiscal year is printed beside them.

OSHA tree care rulemakingProjected October 2026; re-check monthly from September 2026

High. A proposed federal rule for tree care operations is projected under RIN 1218-AD04. If it publishes, any statement that no OSHA standard is written for tree care needs rewording immediately — even though a proposal is not a rule.

OSHA Severe Injury ReportsPeriodically; the current file covers through 30 November 2025

The severe-injury counts move. The download filename encodes its own range.

OSHA ITA 300A summaries, 2026 filing yearFiled by 2 March 2027, published with a lag

The days-away figure moves, and the pooled window rolls forward a year. Each year is a separate file whose name encodes its own range.

BLS QCEW, 2026 annualExpected mid-2027

The firm-count and average-firm-size panel changes. Nothing else depends on it, but the small-operator argument in Method does.