Industries / Food, farming and forestry / Forestry and logging
Worker safety
Keeping the people who fell, move and haul timber alive and unhurt.
4
Milestones achieved, from 1983 to 2014.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1983
Damped saws, fewer numb fingers
Chainsaw vibration causes vibration white finger, in which the fingers turn pale and numb. Among Finnish forest workers its rate fell from 40% in 1972 to 5% in 1983, which the researchers put down mainly to lower chainsaw vibration after damping measures began in 1970.
LessonTingling and numbness in the hands and arms fell from 78% to 37% by 1976 and then rose to 51%, so one fix did not end all hand damage.
Credit
I. Pyykkö and co-authors, Scandinavian Journal of Work, Environment & Health (1986)
1994
A US logging standard
OSHA published its logging operations standard (29 CFR 1910.266) on 12 October 1994, and most of it took effect on 9 February 1995. It sets safety practices "for all types of logging, regardless of the end use of the wood", including protective equipment, first aid, machines and felling.
Credit
Occupational Safety and Health Administration
2014
A review after deaths
An independent review of New Zealand forestry reported in October 2014, after 32 deaths in the sector between 2008 and 2013. It found a weak safety culture, said mandatory standards were "not consistently being met across the sector", and called for contractor certification, clear rules on matters such as fatigue, and better data on near misses, injuries and deaths.
Credit
Independent Forestry Safety Review panel, chaired by George Adams
2014
Winches hold machines on slopes
In winch-assist harvesting, a cable from an anchor machine helps a harvester or feller-buncher keep traction on steep ground. The first winch-assist machine in British Columbia began work in December 2014, and FPInnovations counted 22 in use there by January 2017, with 27 more planned or on order.
LessonIn New Zealand in 2016 an anchor machine slid downhill and pinned a feller-buncher operator, who was left paraplegic after a suspected winch failure, so the assist system brings hazards of its own.
Credit
FPInnovations' count; BC contractors and machine makers
Watch for
Smartwatches that know the task
Researchers at the University of Idaho and the PNASH Center are building smartwatch models that tell in real time whether a logger is felling by hand, setting chokers (looping cables around logs for a cable yarder to pull) or running a machine, combined with location sharing, under NIOSH funding for 2022–2027. Fieldwork is planned on active operations in Idaho, Oregon and Washington, and the project page reports no accuracy figures.
Achieved whena field trial on working logging jobs, measured by people outside the project, shows the system tells crews where a co-worker is and what they are doing quickly and correctly enough to act on.
Claimed by
Robert Keefe and Eloise Zimbelman, PNASH Center, project page current in 2026
Open
01
Hazards overhead, flagged
A model that reviews video from a helmet camera as a feller walks up to a tree and flags hazards that OSHA tells fellers to look for, such as broken limbs hanging in the crown (widowmakers), standing dead trees and trees lodged against others, before any cut is made.
The briefCredit · unclaimed02
Near misses reported by voice
A voice app in which crew members describe a near miss in their own words at the end of a shift; speech-to-text and an AI model turn each account into a coded report for the safety lead.
The briefCredit · unclaimed
Open 01 · the brief
Hazards overhead, flagged
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A model that reviews video from a helmet camera as a feller walks up to a tree and flags hazards that OSHA tells fellers to look for, such as broken limbs hanging in the crown (widowmakers), standing dead trees and trees lodged against others, before any cut is made.
Why it is still open
OSHA says "More people are killed while felling trees than during any other logging activity". A search found drone and aerial work on dead trees, and a 2025 dataset paper saying data and methods for forestry vision "are still too scarce", but no tested system that checks felling hazards from the ground.
Built with what exists
a helmet camera, an AI model that reads images, such as Gemini or Claude, to make first labels, a detector such as YOLO fine-tuned on those labels, and a local model runtime such as ggml so it runs on a phone without a signal (see the note "A first stable number for local AI").
First step
Film 50 standing trees in a woodlot with an arborist's ratings recorded, with no felling involved.
Done looks like
On 300 filmed trees, each rated by a certified arborist, the system catches at least 90% of the trees the arborist rates as hazardous, with false alarms reported.
Open 02 · the brief
Near misses reported by voice
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A voice app in which crew members describe a near miss in their own words at the end of a shift; speech-to-text and an AI model turn each account into a coded report for the safety lead.
Why it is still open
New Zealand's 2014 forestry safety review called for action on "the lack of data on near-misses, injuries and deaths".
Built with what exists
a speech-to-text model such as ElevenLabs Scribe v3 (see the note "Fewer word errors"), a language model that codes each report by activity and hazard, using OSHA's list of felling hazards as categories, and offline recording for sites without a signal.
First step
Write a ten-category coding scheme from OSHA's felling hazards and test it on 20 near-miss accounts collected by interview.
Done looks like
Over one season at least three crews file reports, and a safety adviser checks a random 100 of them and agrees with the coding on at least 90%.