Fields you can read, season after season

Fields you can read, season after season

What growers, agronomists, crop consultants and research teams fly drones for, and how Omniris turns those flights into calibrated crop maps, per-zone statistics across the season, crop water stress and drainage layers, with a sealed custody record behind every file.

Summary

In agriculture a drone is mostly a camera that sees what the eye cannot: near-infrared and red-edge light that shows crop vigor and chlorophyll, and heat that shows a crop running short of water. The flights are flown again and again through a season, over the same fields, and their value is in the comparison — which zone is falling behind, whether a block recovered after irrigation, how much of a field a storm took out.

Omniris covers the imaging and analysis side of that work end to end, from a camera's raw files to a ranked list of zones to walk, and keeps every flight of a field lined up by date. Index maps and zone boundaries go on to farm software in open formats, ready for the prescription.

How agriculture uses drones today

Drones in agriculture do two quite different jobs. Imaging drones carry colour, multispectral or thermal cameras and tell people where to look and what changed. Application drones carry a tank and spray or spread. This paper is about the first kind. Everything in this section describes the industry, with a numbered source for each claim.

Who flies, and who buys

• Adoption on farms is real but limited. The USDA Economic Research Service reports that drone adoption "plateau[ed] at 12 percent of large-scale, crop-producing family farms and 13 percent of nonfamily farms" in its 2023 survey data.²⁸ Aerial imagery of any kind (aircraft, drone or satellite) was used on 7.0% of US corn acres (2016), 9.8% of soybean acres (2018), 3.5% of winter wheat (2017), 2.8% of cotton and 4.6% of sorghum (2019).²⁹
• Retailers and agronomists are the main channel. In the CropLife/Purdue survey of US agricultural dealers, the share offering drone imagery rose from 19% to 55% in the eight years to 2022 that the question had been asked, and dealers estimated drone imagery was used on 17% of their trade-area acres, up from 6% in 2017.³⁰
• Imagery is a hard business. The same survey found imagery offerings "much less profitable" than fertilizer services,³⁰ in 2025 dealers reported that most farmers are not using imagery,³¹ and in 2026 a quarter of dealers who offer drone imagery reported losing money on it.²³ Dealers who use AI use it most for imagery analysis.²³
• Spraying is where the growth is. More than 90% of dealers in the 2026 survey know of drone input application in their area and half offer it, in-house or through a contractor; dealers put the cost of setting up a crew at about $145,000 and running one at about $36,000 a month.²³

In practice the people who fly imaging missions are growers with their own drone,⁴ agronomists and crop consultants, often working for a retailer,³⁰ independent drone service providers, and university extension and research teams.⁷¹⁸

The missions

Cadence across the season

Imaging is repeated, and its value depends on timing. In corn, the strongest link between vegetation index and yield in Virginia trials came from flights in the late vegetative stages, from about V11–V12, before tassels appear;¹ in a Cornell study of twelve weekly flights from emergence to R5, yield models were best from sunny flights between R1 and R4, and cloud during a flight degraded the results.³ A typical row-crop season therefore looks like this:

Sensors

Good capture practice is consistent across vendors and extension: photograph a calibrated reflectance panel immediately before and after every flight, with no shadow on it; avoid rolling partial cloud;²⁷ fly near solar noon (10 a.m. to 2 p.m., or within 2.5 hours of solar noon) in wind under 10 mph and not right after rain, with the same settings every flight;¹²⁷ and plan about 75% front and side overlap, with terrain following on slopes such as vineyards.²⁷

Rules and standards that govern the work

Pain points

• Data volume. One 40-acre weed-scouting flight with a five-band camera produced about 2,900 images and 6.8 GB.⁵
• Turnaround. Full mapping missions take hours of preparation, flying and photo handling, and their stand data typically comes in too late to replant.⁴
• Consistency between dates. Calibration depends on panel and sunlight-sensor practice,²⁷ DJI states NDVI "cannot be obtained" when the sunlight sensor is blocked,²⁵ and clouds during a flight degrade the result.³
• From map to decision. A stand assessment still needs ground-truthing before a replant decision,⁵ and a prescription needs a format the display understands.¹³
• Economics. Imagery services are less profitable for retailers than other services, and some lose money on them.²³³⁰

Use case by use case: what Omniris does

The table maps each drone use described above to what Omniris delivers for it. The sections after the table describe each workflow in the order a season runs.

Crop scouting and vigor mapping

A vigor map is only useful if it gets someone to the right corner of the field. Omniris takes a multispectral flight from the camera's files to a list of places to walk.

Which index, when

The product's own guidance follows the agronomy described above: NDVI early in the season, NDRE once the canopy closes, soil-adjusted indices (SAVI, OSAVI, MSAVI2) where bare soil shows between the rows, and PSRI for senescence and ripening. Each standard index states its valid range and what kind of limit that is — mathematical, set by the band scale, or only a plausibility limit — so a ratio index that blows up over shadow or water is never mistaken for a measurement.

Orion picks the index from the wording of the question and can combine them: "stressed, but not just ripening" looks for low NDVI that high PSRI does not explain. On a calibrated flight it uses published class limits; otherwise it marks the weakest quarter of the field, or the share you ask for ("the worst 10%").

The full multispectral workflow is in the Multispectral paper.

Water stress and irrigation

A crop short of water closes its pores and warms up before it shows stress to the eye. Omniris reads that from two directions: canopy temperature from a thermal camera, and water and vigor indices from a multispectral one.

Temperatures, not a heat picture

Thermal photos from DJI radiometric cameras (Mavic 3 Thermal, Matrice 4 Thermal, Matrice 30T, Zenmuse H20T / H20N / H30T and others), FLIR radiometric cameras, and the thermal band of the MicaSense Altum family are converted to temperature and stitched into a map where every pixel is a surface temperature. For DJI and FLIR photos, air temperature and humidity for the conversion are prefilled from the weather reading nearest the capture time on the capture day. A map is labelled THERMAL only after its values are checked to be temperatures, and every thermal tool reads only those maps. Readings show °F first with °C beside them.

On the map you can click for a temperature, outline a block for its minimum, mean, maximum and 2nd and 98th percentiles, and highlight everything above, below or between temperatures you choose.

Crop Water Stress Index

The Crop Water Stress Index (CWSI) turns canopy temperature into a score from 0 (well watered) to 1 (severely stressed). You must choose the crop; none is assumed:

Only canopy is scored: plants are separated from soil with NDVI from the same flight when multispectral bands exist, at the site's vegetation threshold (the one zonal analytics uses; 0.20 when the site has none), otherwise with an index from the colour map, and the result says which it used. Without one of those maps, CWSI is refused rather than averaged over plants and soil together. The result is a saved map layer in five classes, from well watered to severe, with its own custody record. Zonal analytics trends it per zone across flights and two dates can be compared, the baseline and relative versions as separate indices. A baseline CWSI flight counts in the zone trend line even without multispectral calibration, because it is worked out from temperature against the crop's published limits; the relative version, rescaled per flight, is never in the zone trend line, on any flight, calibrated or not; it appears only in the zone rankings. On a flight without multispectral bands, CWSI zone values are measured over the whole zone (the layer is already limited to the canopy), and the view says it is showing all pixels.

Irrigation blocks and sets

Draw each irrigation block, set or pivot span as a zone of the field. Zonal analytics then measures each one on every flight — NDVI, NDRE, the water index NDWI (measured by default over vegetation only, so it reads crop water content, not puddles), any other standard index or a formula of your own — and the uniformity measure flags a zone that is unusually uneven against the rest of the field, which is often where a block is under-watered or a line is plugged. For an orchard or vineyard, the zones can be the blocks the irrigation is actually run by.

Ask Orion "is this field water-stressed?" and it reads the saved CWSI layer, with the crop and method used, or works out the relative version on the spot and says that nothing was saved. Thermal flights are covered in full in the Thermal Imaging paper.

Nitrogen and in-season management

In-season nitrogen decisions rest on comparing the crop with a well-fertilised reference and acting on the difference. Omniris supplies the measurement side of that workflow.

• Red-edge indices. NDRE is made with every flight from a camera with a red-edge band; CI-RedEdge, MTCI and CI-Green are a click away in Band Analytics, and any published variant can be written as your own formula and saved.
• The reference strip as a zone. Draw the nitrogen-rich or reference strip as its own polygon. Zonal analytics gives each zone's mean on every flight and, for the latest flight, its standing against the field in standard deviations ("vs field"), and the strip can be read beside the zones around it; those zone and strip means are the inputs to a sufficiency index.
• Crop only. The default vegetation mask measures only pixels at or above an NDVI of 0.20 (adjustable for the site, with a preview, or for any one zone), so soil between young rows does not drag the average down. A thin strip just inside each zone's edge, about 6 in (0.15 m) and at least one pixel wide, is also left out, except on zones too small to spare it.
• Calibration respected. Trends, season totals and green-up dates use only flights corrected with a sunlight sensor or a reflectance panel (at least two of them), so a change in the light is not read as a change in the crop.

From map to prescription

Omniris hands the measurement on to the software that writes the prescription, in open formats it reads:

Research plots and trials

Trial work asks the same question as zonal analytics, at a smaller scale and with more bookkeeping: what did every plot do on every flight, and which treatment was it?

Plots as zones

• Bring the plot map in. Import plot boundaries from a shapefile (any projection stated in its .prj file is reprojected), GeoJSON, KML or KMZ, adding to the site's polygons or replacing them. Each plot gets a label, a colour and its area in acres, so the plot grid you already have becomes the trial's zones.
• Keep the plot data on the plot. Custom fields on each polygon hold text, whole numbers or decimals — entry, treatment, replicate, hybrid, rating — either one value for the whole season or a separate value per flight date, so a visual score or a hand measurement can sit beside the drone numbers for that date. A set of fields saves as a template for the next trial.
• Every plot, every flight. Zonal analytics measures each plot on each flight and shows its mean, spread, vegetation cover, standing against the whole trial on the latest flight, trend per week, season total (the area under the index curve, iNDVI for NDVI), green-up and peak dates. Freak values far outside a plot's own spread are set aside first, and boundary pixels are left out, so the alley between plots does not leak in.
• Season against season. Compare years overlays each year's curve on one January-to-December axis, for the trial or any single plot.

What a trial manager should know

• Pixel size matters. The planner shows the ground detail (GSD) for the chosen camera and height, so the flight can be planned to put enough pixels inside the narrowest plot.
• Position matters. Comparing plots across dates relies on the maps lining up. Ground control points and RTK are supported in the stitch, checkpoints measure the result, and a re-flight can be aligned to an earlier one (see Accuracy and calibration).
• Getting the table out. The plot table is read in Omniris — sorted by its main columns, with a history chart per plot — and Orion can answer questions about it in the chat. For analysis in your own statistics software, the index maps and plot boundaries download in open formats.

Drainage, ponding and terrain

The same colour flight that makes an orthomosaic also measures the shape of the ground: a surface model (DSM, the top of everything) and a bare-earth terrain model (DTM). On an elevation layer in the Site Navigator, one button runs terrain analysis on the terrain model (the surface model can be chosen instead). The run is free and makes any of five products, each a map layer filed under the flight date:

Every setting can be changed before the run. A probe reads values from the slope, aspect and ponding layers. The products are filed with a custody record and file fingerprint, and the ponds and drainage paths are GeoJSON files you can download.

• Vertical reference is labelled. Heights are labelled with their vertical datum when the ground control says what it is; when it does not, they are labelled "heights relative — vertical datum unknown".
• For the drainage designer. Contours and the surface export as DXF (contour polylines or a surface of 3D faces) and as a LandXML surface, the formats CAD and design packages take in.
• Seen against the crop. Ponding and drainage layers sit in the same Site Navigator as the season's NDVI and thermal maps, so a drowned-out area on the vigor map can be checked against where water collects.

Damage assessment and insurance claims

After hail, wind, flood or drought, the questions are how much of the field was affected, where, and whether the imagery can be relied on later. Omniris answers the first two with measurement and the third with its custody record.

Before and after

The Site Navigator's Compare tool takes any index and two flight dates of the field — the last flight before the event and the first after — and works out the change at every spot:

• a change map, saved as a GeoTIFF beside the later flight, red where the index fell and green where it rose;
• the average and middle change, the spread, and the area that fell, rose or stayed the same past a change threshold (0.05 index units unless you set another); for NDVI and NDRE, also the area where vegetation cover disappeared or appeared;
• the whole flight, one zone, or an area you draw, taken away as a report or a CSV file.

If the two flights were not calibrated on a common footing, the comparison is marked as an estimate. Two colour photos are compared by eye — fade, swipe and blink between dates — because changing light would swamp a computed difference. A situation report puts the two dates side by side at the same scale in one PDF, with each image's source, capture date, custody history and fingerprint, ready for someone who will never open Omniris.

Imagery that stands up later

• Every upload and every map made from it gets a custody record and a SHA-256 file fingerprint as it is created, sealed and timestamped.
• A legal hold on the field's site freezes its imagery, so nothing can be deleted or replaced while a claim is open.
• A signed custody certificate lets an adjuster, insurer or court check the file against its record, online or offline, without relying on Omniris.
• Photo locations can be pinned on the map, so each original photo can be opened from where it was taken.

The custody model is described in the Evidence Integrity paper.

Farms, fields and seasons

A consultant's work is organised by grower and field, and a season's value is in seeing the same field again and again. Omniris is organised the same way.

A customer is the grower, farm or estate; a site is a field, ranch, orchard or vineyard; its polygons are the field boundaries, management zones, blocks, irrigation sets or trial plots; and every flight of the site is filed under its flight date.

• Agricultural is the default site type. It adds zonal analytics to the site. Solar, roof and other types hide the farm tools.
• Flights file themselves. Give the site and customer name when you start a map; when it finishes, every result is labelled with the site, the customer, the flight date and the layer type, and the site is created if it is new.
• Boundaries once. Draw polygons on the map or import them from KML, KMZ, GeoJSON or a shapefile, and export them in the same formats. The same boundaries are measured on every flight; edit one and past flights are re-measured to match. A polygon is a single outline: holes in an imported shape are not kept as holes (a shapefile's holes come in as separate polygons), and a multi-part shape is split into separate polygons.
• Your own data per block. Custom fields hold crop, variety, rootstock, planting year, or anything else, for the season or per flight date, and are filled in by clicking the block on the map.
• The season on one map. The Site Navigator lists every flight date of the field with its layers; tick dates and layers on and off, fade one over another, and click a zone for its index values on each date.
• A field cut out of a bigger flight. A map can be cropped to an area you draw or have measured, and the Site Navigator can show the cropped map in place of the original for that date.
• Imagery from elsewhere. A satellite or aerial map you already hold — a picture or one of ten vegetation and water index types — can be added to the field's timeline with its real capture date and a source label, and compared with the drone flights.
• Told when it changes. A change alert on an index comparison emails you when a new flight's mean change, or the share of the area that changed (10% unless you set another), passes your limit.

Seasons follow the calendar year: by default you see the latest year with flights, and every trend is worked out within that year. Sites, boundaries and the Site Navigator are covered in the Site Management paper.

Getting the flights flown

Planning the mission

Every job and every site has a mission planner; on a marketplace job the pilot builds the plan and the client can see it. Pick the field polygon, the aircraft from the pilot's own equipment and the camera — the catalogue includes common DJI, Autel, Skydio and MicaSense cameras, with the Mavic 3 Multispectral, MicaSense RedEdge-P and Altum-PT as its multispectral cameras — and the planner draws the flight lines and gives the photo count, flight time and battery count, updating with every change.

• Mapping grid. Parallel lines with the camera straight down, starting at 75% front and 70% side overlap (both adjustable). Enter the ground detail you need or the height you want to fly; the other is worked out for the camera.
• Camera and blur. The default speed of 18 mph (8 m/s) is lowered when the camera cannot fire fast enough for the photo spacing, and motion blur above one pixel is warned about.
• Terrain following. On rolling ground the planner can hold a steady height above the ground from a global terrain model or your own earlier survey, keeping a minimum clearance you set (33 ft (10 m) by default). It recommends this when the ground across the field varies by 49 ft (15 m) or more.
• 400 ft warning. A planned height above 400 ft (121.9 m) is warned about as above the Part 107 ceiling.³²
• Into the aircraft. The planner writes a mission file for DJI Pilot 2, DJI Fly, Litchi or QGroundControl, and the pilot flies it from the aircraft's own app.

Airspace and weather

• Airspace. The whole route is checked against FAA airspace, including the floors of controlled airspace; where authorization is needed, Omniris works out what the pilot needs to file through LAANC, and records the authorization the pilot enters after filing it in a LAANC app. Rural fields near an airport are where this matters.
• Flyability. An hour-by-hour score from 0 to 100 for up to 16 days, against the aircraft's limits — wind and gusts, temperature, precipitation, cloud cover, visibility and daylight among the factors — with the best window long enough for the job. A flyability alert watches the days you picked and emails when their forecast score gets better or worse by at least the amount you set.
• Pre-flight checklist. A Part 107 checklist that cannot skip the regulatory items is saved with the job and frozen once complete.

Planning, airspace and weather are covered in the Mission Planning paper.

Hiring the pilot

In the cloud and data center editions, a grower, consultant or co-op can post the work as a job in the Omniris marketplace: draw the fields on the map, tick the deliverables (orthomosaic, multispectral, thermal and others), set the dates, and choose a pilot or let approved pilots nearby apply. A job can repeat weekly, every two weeks, monthly, quarterly or on a custom cadence, each occurrence with its own flight window, which fits a season of scouting flights. Pilots verify their FAA Part 107 certificate in their profile. Results arrive in the client's vault with their custody history, and the client accepts or rejects them. Commerce — the marketplace, payments and payouts — is not part of the on-premises edition.

Asking Orion about a field

Orion, the Omniris assistant, answers questions in plain English from your own flights and records. It only reads, and it sees only what your account may see. While the Site Navigator is open it knows which field, dates and layers you are looking at, so "this field" is enough.

Orion's highlights are images in the chat, not drawings on the live map. A count of objects comes only from an AI detection run; Orion does not count from a glance. It is described in full in the Orion paper.

Accuracy and calibration

Two different kinds of accuracy matter in agriculture, and Omniris reports both separately.

Radiometric: can two flights be compared?

A camera-only flight supports comparisons within itself (zone against zone) and is left out of trends, season totals and green-up dates. A sunlight-sensor flight is corrected for changing light and suitable for relative comparison across flights, though it is not absolute reflectance. A reflectance-panel flight gives reflectance that is comparable across dates, sites and aircraft, the best footing for multi-flight and multi-season comparison.

Sunlight-sensor and panel-calibrated maps carry their calibration label; a camera-only map carries none. Where a site has more than one map for a date, zone statistics and site trends use the best-calibrated one. Reflectance values far outside what a surface can reflect (below −0.5 or above 2.0) are marked as no data before any index is made. A flight comparison between flights on different calibration footings is marked as an estimate, and no error margin is given for index change, because there is no published error model for it.

Positional: does the map sit where the field is?

• Relative accuracy — distances and areas within one map — comes from the photogrammetry and is enough for most scouting.
• Absolute accuracy — the map sitting where a surveyor would put it — needs ground control points or RTK. Omniris supports both, measures checkpoints (surveyed points left out of the fit) to give horizontal, vertical and 3D RMSE, and can align a re-flight to an earlier flight of the same field.
• When a flight has both a colour map and a multispectral map, the offset between them is measured and shown beside the date: grey under 1.6 ft (0.5 m), amber to 3.3 ft (1 m), red above. Neither map is moved.
• Without ground control, the results say plainly that absolute accuracy cannot be stated from the flight alone.

How maps are made and how their accuracy is stated is covered in the Stitching paper.

Evidence and records

Most crop maps are working documents. Some end up in an insurance claim, a dispute with an applicator or a neighbour over drift, a lease or crop-share settlement, or a research record that has to be reproducible. Every file in Omniris carries the same record either way:

• A chain of custody per file — upload, the maps made from it, views, downloads, shares, moves — each event sealed into a tamper-evident record that is signed by Omniris and timestamped by an independent authority.
• Lineage. Index maps, CWSI layers, change maps and terrain products are linked to the stitch they came from; a saved index records the formula that made it; the thermal conversion settings are saved with the job.
• Legal hold and retention. Freeze a file, folder, job or a whole site for as long as a claim or dispute runs; or set a retention period on a trial's folder or site, after which its data is kept or deleted, with a signed certificate of what was deleted and when.
• Certificates. A signed custody certificate for a file can be checked online, or offline with free standard tools.

The record is described in the Evidence Integrity paper, and protection of the data itself in the Security paper.

Integrations and delivery

Results reach growers and farm software through open files and links.

Sharing and delivery are covered in the Sharing and Delivery paper.

Cloud, data center and on-premises

Omniris runs in four ways, with the same multispectral, thermal, zonal, terrain and Orion tools in each:

A research group or agribusiness that must keep trial data on its own hardware can run on-premises. A self-hosted installation can send a processing job to the Omniris cloud with Hybrid — a choice made per job when it is started — and the result lands in the local vault like any other run.

What the results are, and what they are not

Crop maps feed real decisions, so Omniris states what each result is where it applies:

• An index is a measurement of light, not a diagnosis. It shows where a part of the field reflects differently; whether the cause is nitrogen, water, disease or soil is found on the ground.
• AI detection is a detection aid, not a census. Its counts of trees or other objects are reviewed and corrected before they are relied on.
• Relative crop water stress is relative. Published CWSI baselines cover corn, soybean, cotton and wheat before heading. Every other crop, orchards and vineyards included, gets the relative index, which compares a plant only with the rest of its own flight and is labelled that way everywhere it appears.
• Seasons follow the calendar year. Trends and green-up dates are worked out per calendar year, not from a planting date; a season that spans the new year shows in one line in the All years view.
• Accuracy is stated, not certified. Omniris reports the calibration, ground control, checkpoint error and weather it has. Thermal accuracy is the camera's. Without ground control, heights are relative, and under a tall crop the terrain model follows the canopy.

Sources

Industry claims in this paper come from the sources below. Numbers that only a manufacturer states are marked "vendor-reported" in the text.

1. Khulal, U. and Walsh, O. "When to Fly a Drone for Accurate Corn Yield Potential Prediction." Virginia Cooperative Extension, SPES-752NP. https://pubs.ext.vt.edu/SPES/spes-752.html
2. Campbell, S. (Cottonwood Extension District), with information from D. Joshi, Precision Ag Extension Specialist. "Crop monitoring with drone-based sensing technology." Great Bend Tribune (reprint). https://anvil.gbtribune.com/news/business/crop-monitoring-with-drone-based-sensing-technology
3. Sunoj, S., Cho, J., Guinness, J., van Aardt, J., Czymmek, K. J. and Ketterings, Q. M. "Corn grain yield estimation with drones: timing is key." Cornell University, What's Cropping Up?, February 14, 2022. https://blogs.cornell.edu/whatscroppingup/2022/02/14/corn-grain-yield-estimation-with-drones-timing-is-key
4. Brooks, K. "Using your farm drone to conduct field stand counts in corn and soybeans." University of Illinois Extension, June 30, 2025. https://extension.illinois.edu/news-releases/using-your-farm-drone-conduct-field-stand-counts-corn-and-soybeans
5. Thompson, L. "Using drones for early-season stand assessment and weed scouting." University of Nebraska–Lincoln CropWatch, April 26, 2018. https://cropwatch.unl.edu/2018/using-drones-early-season-stand-assessment-weed-scouting
6. Vong, C. N., Conway, L. S., Zhou, J., Kitchen, N. R. and Sudduth, K. A. "Early corn stand count of different cropping systems using UAV-imagery and deep learning." Computers and Electronics in Agriculture 186 (2021) 106214. https://digitalcommons.unl.edu/usdaarsfacpub/2535
7. Westfall, A. "Green snap damage in corn." Purdue Extension UAV program. https://extension.purdue.edu/uav/corn-soybean-wheat/green-snap-damage-in-corn.html
8. Gonzalez-Dugo, V. et al. "Transpiration from canopy temperature: Implications for the assessment of crop yield in almond orchards." European Journal of Agronomy 105 (2019) 78–85; cites Idso et al. (1981), Agricultural Meteorology 24:45–55, and Jackson et al. (1981), Water Resources Research 17:1133–1138. https://quantalab.ias.csic.es/pdf/CWSIalmond_gonzalez2019.pdf
9. Atencia Payares, L. K. et al. "Thermal imaging from UAS for estimating crop water status in a Merlot vineyard in semi-arid conditions." Irrigation Science 43:87–103 (2025). https://link.springer.com/article/10.1007/s00271-024-00955-1
10. Kelley, L. and Anderson, E. "Camera or drone video aid greatly in identifying needed center pivot sprinkler repairs." Purdue Pest&Crop Newsletter, September 19, 2019. https://extension.entm.purdue.edu/newsletters/pestandcrop/article/camera-or-drone-video-aid-greatly-in-identifying-needed-center-pivot-sprinkler-repairs/
11. Scott, J. "Irrigation management." Purdue Extension UAV program. https://extension.purdue.edu/uav/structural/irrigation-management.html
12. Thompson, L. J. and Puntel, L. A. "Transforming Unmanned Aerial Vehicle (UAV) and Multispectral Sensor into a Practical Decision Support System for Precision Nitrogen Management in Corn" (University of Nebraska–Lincoln). Remote Sensing 2020, 12, 1597. https://digitalcommons.unl.edu/agronomyfacpub/1340
13. ISO 11783-10:2015, Tractors and machinery for agriculture and forestry — Serial control and communications data network — Part 10: Task controller and management information system data interchange. International Organization for Standardization. https://www.iso.org/standard/61581.html
14. Allred, B. et al. "Overall results and key findings on the use of UAV visible-color, multispectral, and thermal infrared imagery to map agricultural drainage pipes." Agricultural Water Management, 2020, doi:10.1016/j.agwat.2020.106036 (U.S. Geological Survey / USDA-ARS). https://www.usgs.gov/publications/overall-results-and-key-findings-use-uav-visible-color-multispectral-and-thermal
15. Adair, A. "Tile drain observation: installation to harvest." Purdue Extension UAV program. https://extension.purdue.edu/uav/corn-soybean-wheat/tile-drain-observation-installation-to-harvest.html
16. Bruner, B. "Interaction of land features with developing crops." Purdue Extension UAV program. https://extension.purdue.edu/uav/corn-soybean-wheat/interaction-of-land-features-with-developing-crops.html
17. USDA NRCS Minnesota. Conservation Practice Standard, Subsurface Drain (Code 606), March 2017 (marked archived), Field Office Technical Guide. https://efotg.sc.egov.usda.gov/api/CPSFile/281/606_MN_CPS_Subsurface_Drain_2017
18. Han, X. et al. "Measurement and Calibration of Plant-Height from Fixed-Wing UAV Images" (Texas A&M). Sensors 2018, 18(12):4092. https://pmc.ncbi.nlm.nih.gov/articles/PMC6308534/
19. Matias, F. I., Caraza-Harter, M. V. and Endelman, J. B. "FIELDimageR: An R package to analyze orthomosaic images from agricultural field trials." The Plant Phenome Journal, 2020. https://github.com/filipematias23/FIELDimageR
20. Federal Aviation Administration. "Dispensing Chemicals and Agricultural Products (Part 137) with UAS." https://www.faa.gov/uas/advanced_operations/dispensing_chemicals
21. Federal Aviation Administration. Report to Congress on UAS chemical aerial application, July 16, 2020. https://www.faa.gov/sites/faa.gov/files/2021-11/Report_Congress_UAS_Chemical_Aerial_Application.pdf
22. Liu, W. and Ampatzidis, Y. "Regulatory Requirements to Operate Spraying Drones." EDIS publication AE608. University of Florida IFAS Extension, February 2025. https://ask.ifas.ufl.edu/publication/AE608
23. CropLife. "2026 CropLife/Purdue survey reveals shifting priorities in precision agriculture," July 1, 2026. https://www.croplife.com/smart-tech/2026-croplife-purdue-survey-reveals-shifting-priorities-in-precision-agriculture/
24. MicaSense (AgEagle). Knowledge base: band centre wavelengths and bandwidths. https://support.micasense.com/hc/en-us/articles/214878778
25. DJI. Mavic 3 Multispectral specifications and FAQ. https://enterprise.dji.com/mavic-3-m/specs · https://enterprise.dji.com/mavic-3-m/faq
26. Sentera. 6X multispectral sensor. https://senterasensors.com/hardware/sensors/6x/
27. MicaSense. "Best practices: collecting data with MicaSense sensors." https://support.micasense.com/hc/en-us/articles/224893167-Best-practices-Collecting-Data-with-MicaSense-Sensors
28. USDA Economic Research Service. America's Farms and Ranches at a Glance: 2024 Edition, EIB-283, p. 25. https://ers.usda.gov/media/9120/eib-283.pdf?v=70077
29. McFadden, J. and Njuki, E. "Aerial imagery remains mostly grounded on U.S. farms." Chart of Note, USDA Economic Research Service, August 24, 2023. https://ers.usda.gov/data-products/charts-of-note/107207
30. CropLife. "2022 precision survey: as ag economy shifts, so does technology use by retailers," July 25, 2022. https://www.croplife.com/management/2022-precision-survey-as-ag-economy-shifts-so-does-technology-use-by-retailers/
31. CropLife. "The 2025 CropLife/Purdue Precision Adoption Survey: Making Way for New Ag Tech," July 8, 2025. https://www.croplife.com/smart-tech/2025-croplife-purdue-precision-adoption-survey/
32. 14 CFR §107.51, Operating limitations for small unmanned aircraft. https://www.law.cornell.edu/cfr/text/14/107.51
33. 14 CFR §107.31, Visual line of sight aircraft operation. https://www.law.cornell.edu/cfr/text/14/107.31
34. Martinez-Casasnovas, J. A. et al. "Analysis of vineyard differential management zones and relation to vine development, grape maturity and quality" (satellite NDVI at veraison). Spanish Journal of Agricultural Research, 2012, doi:10.5424/sjar/2012102-370-11. https://sjar.revistas.csic.es/index.php/sjar/article/view/1998
35. MicaSense. Knowledge base: Altum-PT thermal specifications. https://support.micasense.com/hc/en-us/articles/360036377834
36. MicaSense. "Comparison of MicaSense cameras." https://support.micasense.com/hc/en-us/articles/1500007828482-Comparison-of-MicaSense-Cameras
37. DJI. P4 Multispectral support and specifications. https://www.dji.com/support/product/p4-multispectral

Glossary

CWSI (Crop Water Stress Index)

A 0–1 score of water stress from canopy temperature: absolute with a published crop baseline, relative within one flight without one.

DSM / DTM

Digital surface model (the top of everything, crop and trees included) and digital terrain model (the bare ground), both measured from the same flight.

GSD (ground sample distance)

How much ground one pixel covers. Smaller is finer.

LAANC

Low Altitude Authorization and Notification Capability: the FAA's system for near-real-time authorization to fly in controlled airspace.

NDRE

Normalized Difference Red Edge: a vegetation index using the red-edge band, which keeps responding in dense canopy where NDVI flattens.

NDVI

Normalized Difference Vegetation Index: the classic vigor index, comparing near-infrared with red.

Part 107 / Part 137

The FAA rules for small drone operations, and for agricultural aircraft operations such as spraying.

Prescription (Rx) map

A map of application rates that a variable-rate controller follows.

Radiometric calibration

Correcting images for the camera and the light on the day, so values mean the same thing from flight to flight.

Reflectance panel

A target of known reflectance photographed on the ground before or after a flight, for absolute calibration.

RTK

Real-time kinematic positioning: GPS corrected from a base station or network to centimetre-level precision.

Vegetation threshold

The NDVI value (0.20 unless you change it) at or above which a pixel counts as plant for zone statistics.

Zonal analytics

Per-polygon index statistics, flight by flight, with trends, alerts and season totals.