Industries / Food, farming and forestry / Agriculture and farming / AgTech
Sensors and imagery
This work measures crops, soil and water with satellites, aircraft, drones and probes in the ground, so that a farmer can see a problem before walking to it.
7
Milestones achieved, from 1973 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1973
A ratio that reads green
At Texas A&M University's Remote Sensing Center, J. W. Rouse Jr., R. H. Haas, J. A. Schell and D. W. Deering took two bands from the new ERTS-1 satellite (later renamed Landsat 1) and divided their difference by their sum. On Great Plains rangeland the result tracked how much green plant matter was growing; at one Texas site green biomass alone explained 89% of its variation. NASA calls this normalized difference vegetation index (NDVI), calculated from near-infrared and visible light, the most common satellite measure of plant growth.
Credit
Rouse, Haas, Schell and Deering, Texas A&M University, under NASA's ERTS program
1978
Satellites forecast Soviet wheat
The Large Area Crop Inventory Experiment (LACIE), run from 1974 to 1978 by NASA, USDA and the National Oceanic and Atmospheric Administration (NOAA), estimated wheat harvests by measuring planted area on Landsat images of sample plots and predicting yield from weather data. Its final summary says it met USDA's goal, production within 10% of the true figure 90% of the time, in the US Great Plains and the USSR for two years running, and in August 1977 gave an accurate early sign of the Soviet spring wheat shortfall, well before firmer figures came from the USSR. USDA then began running the system itself.
LessonIt did worse in Canada, where fields were close in size to the smallest area the satellite could see, so image sharpness limits which farms can be measured.
Credit
NASA, USDA and NOAA
1980
Soil water read electrically
G. C. Topp, J. L. Davis and A. P. Annan used time-domain reflectometry (TDR), which times an electrical pulse along a probe, to measure in the laboratory how the electrical properties of four mineral soils, from sandy loam to clay, change with their water content. They found one equation that held regardless of soil type, density, temperature and salt content, so an electrical reading could stand for the share of water in the soil. Crossref counts more than 4,200 later works citing the paper.
Credit
G. C. Topp, J. L. Davis and A. P. Annan
2018
Satellites check subsidy claims
The European Union pays most of its farm support by area, almost 80% of its agriculture and rural development funding, and its paying agencies used to send inspectors to about 5% of farmers each year to check what was grown. The EU's Sentinel-2 satellites, launched in June 2015 and March 2017, photograph all land every five days at 10 m detail, free of charge, and from 2018 agencies could use their images to check every claim instead of a sample of visits; an agency in Foggia, Italy, was first in May 2018, and 15 agencies used the method in 2019.
LessonFree images were not enough on their own: by 2019 only 15 of 66 paying agencies had switched, because it meant new computer systems, staff skills and clearer rules, the European Court of Auditors found.
Credit
European Commission; European Space Agency (ESA); the paying agencies of Italy, Belgium, Denmark, Malta and Spain
2019
Farm imagery stays rare
By USDA ERS's survey count, images from drones, aircraft or satellites were used on 7.0% of US corn acres in 2016, 9.8% of soybean acres in 2018, 3.5% of winter wheat acres in 2017, and 2.8% of cotton and 4.6% of sorghum acres in 2019, far below guidance or variable-rate use. ERS suggests cost and complexity as reasons, along with farmers' view that the images have few uses on the farm.
LessonOffering farmers a picture of the field was not enough; ERS points to farmers seeing few uses for it.
Credit
measured by USDA ERS with NASS
2023
Irrigators still judge by eye
In USDA's 2023 Irrigation and Water Management Survey, 212,714 US farms reported how they decide when to irrigate. 164,779 (77.5%) went by the look of the crop, 85,093 (40.0%) by the feel of the soil, 27,831 (13.1%) used a soil moisture sensor, 1,627 (0.8%) used a computer model, and 10,228 (4.8%) started when their neighbors did; farms could give more than one method, and the shares are this file's arithmetic from NASS counts.
Credit
USDA NASS
2025
Cheap nutrient probes fail
Low-cost soil probes, sold for about $80 to $266, claim to read nitrogen, phosphorus and potassium (NPK) directly in the field. A team from the Alliance of Bioversity International and CIAT, part of the CGIAR research network, tested three models against laboratory results on 10 soil samples in Guatemala and 63 in Colombia. Agreement was poor (adjusted R², the share of variation a model explains, mostly 0.05 to 0.48; best 0.54, for potassium), and readings changed with soil moisture, so the team does not recommend them for deciding fertilizer rates.
LessonA reading that moves with soil moisture cannot set a fertilizer rate, whatever accuracy the seller lists.
Credit
Oscar Estrada, Gresia Ramos, Miguel Lizarazo and Daniel Jimenez, Alliance of Bioversity International and CIAT
Watch for
Daily calibrated images from orbit
EarthDaily says its constellation, with eight satellites in orbit by July 2026, each carrying 16 imagers across 22 spectral bands, will deliver "globally consistent, calibrated daily measurement" matched to the Sentinel and Landsat archives, with commercial service expected later in 2026; agriculture is one of its named markets.
Achieved whenan outside party, such as a space agency calibration team or a university, publishes a comparison of its images with Sentinel-2 and Landsat over farmland and confirms daily coverage.
Claimed by
EarthDaily, July 2026
Open
01
Score satellite alerts against scouting
A one-season test of the crop-stress alerts that satellite field-monitoring apps send to farmers. For every alert, a scout visits the flagged spot within three days and records whether there is a real problem and what caused it; scouts also walk every field on a fixed schedule, to catch problems the alerts missed.
The briefCredit · unclaimed02
Scorecard for cheap soil probes
A named, public scorecard of the low-cost NPK soil probes sold online. Test the ten most-sold models on 100 soil samples against laboratory measurements of nitrate and ammonium nitrogen, phosphorus and potassium, at fixed soil moistures; then test whether each model's three nutrient numbers can be predicted from its own electrical conductivity and moisture readings alone, which would mean they carry no separate nutrient information.
The briefCredit · unclaimed
Open 01 · the brief
Score satellite alerts against scouting
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A one-season test of the crop-stress alerts that satellite field-monitoring apps send to farmers. For every alert, a scout visits the flagged spot within three days and records whether there is a real problem and what caused it; scouts also walk every field on a fixed schedule, to catch problems the alerts missed.
Why it is still open
The closest independent work compares satellite vegetation indices with USDA's survey-based crop condition ratings at state and national level, 1986 to 2023 (Bundy and colleagues, Crop Science, 2026), not alerts on single fields. A 2022 study (Buchaillot and colleagues, Remote Sensing) tested Sentinel-2 signs of fall armyworm damage to maize against field surveys in Zimbabwe, Kenya and Tanzania and got mixed results, because clouds, intercropping, weeds and other pests confused the signal; it scored a research method, not a commercial service. No published hit rate, false-alarm rate or miss rate for a commercial field alert service was found.
Built with what exists
Free Sentinel-2 images through Copernicus Browser or Google Earth Engine; the free tiers of commercial field-monitoring apps; a phone that tags photos with their location; an open vision-language model, run locally, to sort scouting photos by cause (see the Note "A first stable number for local AI").
First step
Enroll ten fields in two free monitoring apps this month, and agree a weekly scouting walk with photos with each field's farmer.
Done looks like
40 fields of one crop through one season, every alert and every scouting visit logged, and a published table of alerts confirmed, false alarms and problems missed, split by cause (weeds, disease, insects, water, nitrogen), with the photos released.
Open 02 · the brief
Scorecard for cheap soil probes
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A named, public scorecard of the low-cost NPK soil probes sold online. Test the ten most-sold models on 100 soil samples against laboratory measurements of nitrate and ammonium nitrogen, phosphorus and potassium, at fixed soil moistures; then test whether each model's three nutrient numbers can be predicted from its own electrical conductivity and moisture readings alone, which would mean they carry no separate nutrient information.
Why it is still open
The 2025 CGIAR test covered three models, did not name them in its summary, and compared nitrogen readings with organic matter because the reference laboratory did not measure nitrogen in all samples. Sellers still list accuracy such as "±2% F.S." (full scale) for these probes, though at least one seller says they should not be used alone for fertilizer prescriptions.
Built with what exists
Probes bought from online marketplaces; an RS485-to-USB adapter and a short Python script to read them over the Modbus protocol; a commercial soil-testing laboratory; a kitchen oven and scale to set sample moisture.
First step
Buy the three best-selling probes this month, connect them to a laptop, and read 20 soil samples at dry, medium and wet moisture.
Done looks like
Ten named models, each read three times on 100 samples at two moisture levels, with R² against the laboratory for each nutrient, and the share of each model's nutrient readings explained by its own conductivity and moisture readings, all data released.