Industries / Food, farming and forestry / Agriculture and farming / Crop dusting
Safety
This work keeps pilots, loaders and people near treated fields from harm: crashes into wires and towers, poisoning by the chemicals carried, and spray that lands where it should not.
7
Milestones achieved, from 1978 to 2023.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1978
Pilots poisoned by their loads
The FAA's aviation medicine office compared toxicology samples from 174 pilots killed in aerial application crashes with those of 2,449 other general aviation pilots. Over half the ag pilots had below-normal cholinesterase, a nerve enzyme that organophosphate insecticides block, which the authors read as "a continuing problem of acute and/or chronic toxicity from the pesticides being applied".
LessonThe chemical in the hopper could disable the pilot flying it.
Credit
D. J. Lacefield, P. A. Roberts and C. W. Blossom, FAA (report DOT/FAA/AM-78/31)
2011
Tower marking left voluntary
Meteorological evaluation towers (METs), thin masts put up to measure wind before a wind farm is built, needed no FAA notice or marking if under 200 feet (61 m). After unmarked METs killed ag pilots near Ralls, Texas, in 2005 (a tower put up 15 days earlier) and near Oakley, California, in January 2011, the NTSB issued a safety alert and the FAA published only a voluntary marking policy, on 24 June 2011. Congress ordered marking and a database in 2016, narrowed the order in 2018 and repeated it in 2024; the FAA's November 2024 proposal covers METs but no other short towers, and no final rule had been published by 20 November 2026.
LessonA voluntary policy and three acts of Congress have not yet put the short towers in a public database.
Credit
NTSB; FAA; NAAA, which pressed for the rule
2011
Drift illness counted
Drift is spray, vapor or dust that moves off the field being treated. Soo-Jeong Lee and colleagues pooled reports from NIOSH's SENSOR-Pesticides program and California's Department of Pesticide Regulation and found 2,945 acute illnesses from agricultural pesticide drift in 11 US states from 1998 to 2006. Aerial applications caused 39% of the 643 drift events and 24% of the cases, and 14% of all those made ill were children under 15.
Credit
Soo-Jeong Lee and colleagues, NIOSH and California DPR, in Environmental Health Perspectives
2013
Spray falls on a school
On 3 May 2013 a crop plane spraying insecticide on a corn field beside the São José do Pontal rural school in Rio Verde, Goiás, Brazil, made 29 children and 8 adults ill during their break, according to the Brazilian Senate's news agency; federal prosecutors later counted 92 people affected. Prosecutors said Ibama, Brazil's environment agency, had barred aerial use of the product on corn since July 2012, and in 2016 they sued the operator, Aerotex, and the maker, Syngenta, for at least R$10 million.
LessonA limit on how a product may be applied protects no one unless it is checked before the flight.
Credit
Brazil's Federal Public Prosecutor's Office (MPF) in Goiás; the Senate human rights committee, which heard the families in September 2013
2014
A year of crashes reviewed
The NTSB reviewed all 78 accidents of 2013 linked to agricultural flying; nine were fatal and killed ten people. The top defining events were hitting an obstacle (power lines, guy wires, trees and a MET), loss of control, and failed parts, and the board found no fatigue guidance for a trade where seasonal days run long.
Credit
National Transportation Safety Board (report SIR-14/01)
2021
New Zealand's rate falls
New Zealand's Civil Aviation Authority reported that its fixed-wing agricultural sector fell from 17 accidents per 100,000 flying hours in 2014 to 6 in 2021, a drop of 65%. The sector had 270 accidents and 25 deaths in the 1970s, and 51 accidents and four deaths in the decade to 2020. The manager of the CAA's analysts credits safety management systems; an operator credits turbine aircraft, loaders that send the load's weight to the pilot, and tighter maintenance.
Credit
Civil Aviation Authority of New Zealand; New Zealand operators
2023
Pilots knew the wire
FAA researchers held focus groups with 22 ag pilots who had survived hitting a wire, at the 2022 Ag Aviation Expo. In 20 of the 31 wire strikes they described, the pilot already knew the wire was there; one had pulled up over it 39 times and hit it on the 40th. An earlier review of 2020 to 2022 accident reports by the lead author found about half of pilots aware of the obstacle.
LessonA map of wires prevents only some strikes; many happen to pilots who know where the wire is.
Credit
Hannah Baumgartner and colleagues, FAA Civil Aerospace Medical Institute (report DOT/FAA/AM-23/20)
Watch for
Fewest crashes on record
NAAA says US agricultural aviation had 43 accidents in 2025, the fewest on record, and 4 fatal accidents, tied with 2012 for the fewest; it counted 54 accidents, 13 of them fatal, in 2024. It credits its PAASS safety courses, though its own figures show the fatal-accident rate in 1999 to 2024 only 3.1% below 1993 to 1998, against 27.9% for all accidents.
Achieved whenthe NTSB's final records for 2025 confirm 43 or fewer Part 137 accidents and 4 fatal ones.
Claimed by
National Agricultural Aviation Association, February 2026
Tower rule due July 2026
The FAA's regulatory agenda moved its rule for meteorological towers of 50 to 200 feet (15 to 61 m) to the final-rule stage and gave July 2026 as the date for the final rule. A search of the Federal Register on 20 November 2026 found only the 2011 policy and the 2024 proposal.
Achieved whenthe final rule is published and existing METs of 50 to 200 feet appear in the FAA's public obstacle database.
Claimed by
Federal Aviation Administration, Unified Agenda (RIN 2120-AK77), edition after spring 2025
Open
01
Towers the obstacle file misses
For one heavily sprayed farm county, a list of every standing structure 50 to 200 feet (15 to 61 m) tall, found in public lidar (laser scans flown for elevation maps) and checked in aerial photos, compared one by one with the FAA's Digital Obstacle File (DOF) and the FCC's antenna structure registry. Publish the ones missing from both, with coordinates and heights.
The briefCredit · unclaimed02
Twenty-five years of wire strikes
A public table of every NTSB report on a Part 137 wire strike from 2000 to 2025, each coded for whether the pilot knew of the wire, the kind of wire (power line, guy wire, MET), the pass it happened on (a swath run, a trim pass along a field edge, a turn), whether the wire was marked, and whether anyone died. Two people code a sample of 200 reports separately, and AI models code all of them and are scored against the two.
The briefCredit · unclaimed
Open 01 · the brief
Towers the obstacle file misses
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
For one heavily sprayed farm county, a list of every standing structure 50 to 200 feet (15 to 61 m) tall, found in public lidar (laser scans flown for elevation maps) and checked in aerial photos, compared one by one with the FAA's Digital Obstacle File (DOF) and the FCC's antenna structure registry. Publish the ones missing from both, with coordinates and heights.
Why it is still open
The DOF does not have to hold towers under 200 feet, and the FAA's 2024 proposal would log only METs; from 2021 to 2023 sponsors in 39 states filed just 391 voluntary notices for new METs of that height. In NAAA's 2019 survey, 59% of operators had met an unmarked communications tower while spraying and 52% an unmarked MET or wind turbine. The nearest published work, a 2024 University of Michigan method, corrected the positions of Delaware's 1,063 DOF entries and noted that a government-only obstacle file holds over 15,000 Delaware entries; it published no count of low structures the public file misses.
Built with what exists
USGS 3D Elevation Program (3DEP) lidar point clouds, free to download; PDAL to read them and keep points standing well above the ground; USDA NAIP aerial photos to confirm each find; the DOF, reissued every 56 days; the FCC Antenna Structure Registration download; QGIS; an image model to flag masts in the photos.
First step
Download the current DOF for one state and the 3DEP tiles for one of its farm counties this month, and extract every point cluster standing 15 m or more above the ground.
Done looks like
Every lidar find 15 m or taller in the county sorted into "in the DOF", "in the FCC registry only" or "in neither", with a second person checking 100 random finds against photos and the false-alarm share published.
Open 02 · the brief
Twenty-five years of wire strikes
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table of every NTSB report on a Part 137 wire strike from 2000 to 2025, each coded for whether the pilot knew of the wire, the kind of wire (power line, guy wire, MET), the pass it happened on (a swath run, a trim pass along a field edge, a turn), whether the wire was marked, and whether anyone died. Two people code a sample of 200 reports separately, and AI models code all of them and are scored against the two.
Why it is still open
The only awareness counts found are the lead FAA author's review of 2020 to 2022 reports (about half aware) and 31 strikes described by 22 pilots in focus groups (20 aware), and those authors name "the lack of accurate base rate information pertaining to wire-strike events" as a limit on what is known.
Built with what exists
The NTSB's CAROL query tool and its downloadable accident data; a written coding guide; a spreadsheet; AI models that answer fixed-choice questions about a text, such as Gemini Decision 1 (see the Note "A thousand questions, one pass") or a Claude-based reviewer of long records (see the Note "Seven more flagged").
First step
Pull the 2020 to 2022 Part 137 reports from CAROL this month, code 30 by hand with a draft guide, and compare the share of pilots who knew of the wire with the FAA's "about half".
Done looks like
All Part 137 wire-strike reports from 2000 to 2025 coded and published with the guide; the two people's agreement on each field over the 200-report sample, as Cohen's kappa (a standard agreement score); and each AI model's agreement with them, field by field.