BRENNEN

Industries / Food, farming and forestry / Agriculture and farming / Crop dusting

Application and drift

This work decides how seed, fertilizer or pesticide leaves the aircraft, how evenly it lands across each swath (the strip one pass covers), and how much of it drifts beyond the field.

6

Milestones achieved, from 1947 to 2025.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1947

    Dust drifts onto cotton

    Planes dusting the new weedkiller 2,4-D on rice farms on the Texas coastal plain damaged about 10,000 acres (about 4,000 ha) of cotton in Wharton and Matagorda counties, and the dust drifted as far as 15 miles (about 24 km). Plant pathologist A. A. Dunlap of Texas A&M College traced the damage to 2,4-D, and the advice Time reported was to spray it rather than dust it, or to mix the dust with oil.

    LessonA fine dry dust can travel miles downwind, so the form in which a product leaves the aircraft decides where it lands.

    time.com

    Credit

    A. A. Dunlap, Texas A&M College, for the diagnosis

  2. 1949

    Fertilizer spread from the air

    In May 1949 John Brazier of Airwork (NZ) Ltd flew a de Havilland Tiger Moth, ZK-ASO, to spread superphosphate on Sir Heaton Rhodes's property at Tai Tapu, Canterbury, which MOTAT, New Zealand's Museum of Transport and Technology, gives as the country's first commercial aerial topdressing. Before this, workers hauled bags of fertilizer across hill country by horse and spread it by hand; air force and earlier experiments had tried sowing and spreading from the air since the 1930s. MOTAT notes that hill-country sheep farmers gained most, and that sheep numbers rose from 35 million in 1950 to over 70 million in the early 1980s.

    medium.com

    Credit

    John Brazier and Airwork (NZ) Ltd; Royal New Zealand Air Force trials before them

  3. 1950

    Planes sow flooded rice

    A USDA bulletin issued in September 1950 reported that California's rice land, 290,000 acres (about 117,000 ha) in 1949, was flooded first and the seed "sown broadcast on the water, largely by airplanes". Seed sown on water was seldom buried under clods, and fertilizer could also be put on from airplanes up to 65 days after seeding.

    agronomy-rice.ucdavis.edu

    Credit

    Jenkin W. Jones and colleagues, US Department of Agriculture

  4. 1981

    Spray patterns get tested

    The National Agricultural Aviation Association (NAAA) began Operation S.A.F.E. (Self-regulating Application and Flight Efficiency) in 1981. At its fly-in clinics a pilot makes test passes over a line of collectors, and an analyst, often a USDA or university scientist, measures how evenly the spray lands across the swath, the effective swath width and the droplet sizes, then suggests changes. In February 2026 NAAA's research foundation began classing analysts by airframe: fixed-wing, single-rotor and multi-rotor drones.

    agaviation.org

    Credit

    NAAA and the National Agricultural Aviation Research and Education Foundation (NAAREF)

  5. 1993

    Drift traced to three factors

    The Spray Drift Task Force, about 40 pesticide makers that joined in 1990, ran 180 aerial field tests in Texas in 1992 and 1993. Reviewers for the US Environmental Protection Agency's (EPA) Scientific Advisory Panel found that droplet size, spray release height and wind speed were the main factors in drift, and that drift did not depend on which non-volatile chemical was sprayed. The data fed AgDRIFT, a model built by EPA, the US Forest Service and the task force from the Forest Service's AGDISP model, which EPA still uses for aerial drift.

    LessonBecause drift follows the physics of droplets rather than the chemistry, one shared set of trials could cut the number of drift studies each product needed.

    archive.epa.govepa.gov

    Credit

    Spray Drift Task Force; USDA Forest Service; US EPA

  6. 2025

    EPA updates its aerial model

    In April 2025 EPA changed the aerial settings it uses in AgDRIFT to size buffer zones: the Tier III module, the Air Tractor AT-802A in place of the smaller AT-401, medium droplets as the default, and new values for swath displacement (how far upwind the pilot offsets each pass), atmospheric stability and wind height. NAAA had told EPA that the old defaults did not match common practice and over-predicted drift. EPA counted AT-802 registrations rising from 488 to 583 between 2020 and 2023.

    LessonA model's default inputs age as the fleet changes, and stale ones overstated off-field deposit.

    epa.gov

    Credit

    EPA Office of Pesticide Programs, after comments from NAAA

Watch for

  1. Drone drift in the middle

    The Unmanned Aerial Pesticide Application System Task Force (UAPASTF), formed by pesticide makers in December 2021, has filed nine Good Laboratory Practice drift trials of spray drones (five from 2023, four from 2024) with regulators in the US, Canada, Australia and the UK. It says averaged drone drift was generally lower than crewed aerial spraying, higher than ground booms and overlapping airblast orchard sprayers, judged against EPA's Tier 1 AgDRIFT curves. NAAA objected in July 2026 that the comparison used the Tier 1 aerial curves rather than the Tier III settings EPA adopted in 2025, and that the trials used one drone, the DJI T30, making three or four passes.

    Achieved whena regulator such as EPA publishes its own review of the trials, or drone-specific drift curves appear in an EPA risk assessment.

    uapastf.comagaviation.org

    Claimed by

    UAPASTF, April 2026

Open

  1. 01

    Finish the open AGDISP

    A Python version of AGDISP, the Forest Service's aerial drift model, that anyone can script and run in batches, with a public test suite showing it gives the same downwind deposit as the official program.

    The brief
    Credit · unclaimed
  2. 02

    One table of drift trials

    One open file of published field measurements of downwind deposit from crewed aircraft and spray drones: distance, deposit, droplet size, release height, wind, temperature, humidity, aircraft and nozzles, each row linked to its paper. Then run AGDISP on each trial and publish how far its predictions fall from the measurements.

    The brief
    Credit · unclaimed

Open 01 · the brief

Finish the open AGDISP

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A Python version of AGDISP, the Forest Service's aerial drift model, that anyone can script and run in batches, with a public test suite showing it gives the same downwind deposit as the official program.

Why it is still open

The official AGDISP 8.26 and AgDRIFT 2.1.1 are downloads from EPA's website and run as Windows programs. A public port, pyAgDisp on GitHub, holds the AGDISP 8.29 source code; its notes say its droplet paths match AGDISP for two fixed-wing airplanes, that input handling was still being written, and its last change was on 30 August 2022.

github.comepa.gov

Built with what exists

The AGDISP 8.29 source in that repository; Python and NumPy; AI coding agents to translate the code and write the tests, as in AWS's cloud emulator that agents wrote and tests alone checked (see the Note "Code no person read"); the official Windows AGDISP to make the reference outputs.

  1. First step

    Run the official AGDISP on ten cases this month, save the outputs, and run the port's droplet-path code on the same ten.

  2. Done looks like

    500 scenarios (aircraft, nozzles, droplet sizes, wind, release height, stability) run in both programs, with the port's deposit within 1% of AGDISP's at every downwind distance out to 300 m (about 1,000 ft) in at least 95% of scenarios, and every mismatch listed.

Open 02 · the brief

One table of drift trials

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

One open file of published field measurements of downwind deposit from crewed aircraft and spray drones: distance, deposit, droplet size, release height, wind, temperature, humidity, aircraft and nozzles, each row linked to its paper. Then run AGDISP on each trial and publish how far its predictions fall from the measurements.

Why it is still open

France's CAPRIV project has published open drift datasets for vineyard and orchard ground sprayers (2024 and 2025), and Germany's JKI has published the data from its own drone drift tests in vineyards, one study's worth. Searches found no table that pools aerial trials across studies. The drone task force's trials went to regulators, and its public notice gave no details of how they were run. A USDA ARS agreement at College Station, Texas, from July 2026 to June 2028, will draft drone drift protocols and a database for drones only.

pmc.ncbi.nlm.nih.govojs.openagrar.deagaviation.orgars.usda.gov

Built with what exists

Papers and their supplemental tables; an AI model that reads long PDFs to draft each row, checked by a person (see the Note "Seven more flagged"); WebPlotDigitizer to read values off published graphs; a CSV file with a written schema; AGDISP.

  1. First step

    List every paper this month that reports downwind deposit from an aircraft or drone at three or more distances.

  2. Done looks like

    At least 40 aerial trials from at least 15 papers, crewed and drone, entered by two people working separately, with disagreements resolved, released under an open license, with a chart of model error by distance.