Element 4.15
Smart farming: sensors, data, satellites and AI
- Sheet
- 4.15
- Chapter
- 4 Food and farming
- State
- Drafted
- Release
- 16 November 2026
- Editor
- Contributed by Praecise as open research
- Reviewers
- To be named at review
- Safety
- S1, as the chapter
- Licence
- CC BY-SA 4.0
- Resources
- 14
The element
- Requirement
Any sensor, satellite product, model or AI assistant the group uses answers a question its records cannot, is checked against field observation before it is trusted, and keeps the group's data in a form it owns.
- How to tell
- Each digital tool has a stated question (when to irrigate, which disease is this, how is the field changing) and a check against the ground.
- Data lives in an open record the group controls (farmOS, LiteFarm) and is backed up.
- Image or AI advice is tested on local cases before decisions rest on it: a disease model scoring 99.35 per cent on its own test images fell to just above 31 per cent on images from elsewhere (Using deep learning for image-based plant disease detection (Mohanty, Hughes and Salathé, 2016)).
- Options
- Farm records: farmOS and LiteFarm (farmOS, LiteFarm). A shared vocabulary such as FAO's AGROVOC, published as linked open data, helps records from different tools and groups line up (AGROVOC).
- Low-cost sensors: microcontroller boards read soil moisture, temperature and humidity (Arduino Documentation); long-range radio links carry readings across fields (About LoRaWAN); open controller software logs readings and switches pumps or vents (Mycodo).
- Satellites: open imagery (Copernicus Data Space Ecosystem), annual 10 m embeddings for crop and land maps (AlphaEarth Foundations helps map our planet in unprecedented detail) and satellite water productivity (WaPOR: FAO portal to monitor water productivity through open access of remotely sensed derived data).
- Drones: images stitched into maps and surface models with open software (ODM (OpenDroneMap)).
- Disease identification by phone: a cassava model trained on field photographs in Tanzania reached 93 per cent on unseen data and runs on Android phones (Deep learning for image-based cassava disease detection (Ramcharan and others, 2017)).
- AI advisory assistants: Farmer.Chat, a generative-AI advisory service, states deployment in four countries, over 15,000 farmers and over 300,000 queries answered (Farmer.Chat: scaling AI-powered agricultural services for smallholder farmers (Singh and others, 2024)). Answers should be checked with local extension staff.
- Robots: open hardware can plant, water and weed small beds (FarmBot). This research opened no primary source on the maturity of field-scale autonomous machines; the machines topic covers them.
- Yield prediction: no open, peer-reviewed source was opened in this work; the gap is listed.
- Settled, open, contested
Settled: validate models locally; keep data in a form the group owns (farmOS). Open: how well AI advice holds up in a given place; the Farmer.Chat paper reports gains in practices, trust and engagement without giving figures in its abstract (Farmer.Chat: scaling AI-powered agricultural services for smallholder farmers (Singh and others, 2024)). Contested: sharing farm data with platforms and buyers. Pooled data improves models; it can also be used against the group, as 4.8 notes.
The four stages
- Learn: what each tool can and cannot see; the field-versus-lab accuracy gap (Using deep learning for image-based plant disease detection (Mohanty, Hughes and Salathé, 2016)).
- Design: one question per tool; an open record as the base (farmOS); a vocabulary for shared data (AGROVOC).
- Build: sensors and radio links (Arduino Documentation, About LoRaWAN), a controller (Mycodo), drone mapping (ODM (OpenDroneMap)).
- Deploy: check readings against hand measurements each season; keep raw data; review whether each tool earns its keep.
Resources at each stage
1 Learn
Understand
- AlphaEarth Foundations helps map our planet in unprecedented detail Case study, S0
- Arduino Documentation Guide, S0
- Deep learning for image-based cassava disease detection (Ramcharan and others, 2017) Paper, S0
- Farmer.Chat: scaling AI-powered agricultural services for smallholder farmers (Singh and others, 2024) Paper, S1
- About LoRaWAN Standard, S1
- Using deep learning for image-based plant disease detection (Mohanty, Hughes and Salathé, 2016) Paper, S0
2 Design
Size and choose
- AGROVOC Standard, S0
- Copernicus Data Space Ecosystem Dataset, S0
- About LoRaWAN Standard, S1
- ODM (OpenDroneMap) Tool, S0
- Using deep learning for image-based plant disease detection (Mohanty, Hughes and Salathé, 2016) Paper, S0
- WaPOR: FAO portal to monitor water productivity through open access of remotely sensed derived data Dataset, S0
4 Deploy
Install, run and measure
- AGROVOC Standard, S0
- Deep learning for image-based cassava disease detection (Ramcharan and others, 2017) Paper, S0
- Copernicus Data Space Ecosystem Dataset, S0
- Farmer.Chat: scaling AI-powered agricultural services for smallholder farmers (Singh and others, 2024) Paper, S1
- farmOS Tool, S0
- LiteFarm Tool, S0
- Mycodo Tool, S1
- ODM (OpenDroneMap) Tool, S0
- WaPOR: FAO portal to monitor water productivity through open access of remotely sensed derived data Dataset, S0
What changes between 30 adults and 300 people
At 30 adults, a notebook, a weather station and a phone app may be enough; sensors pay off mainly in greenhouses. At 300 people farming several hectares, a sensor network, satellite or drone maps and a shared farm record become worthwhile, with a named data keeper and a backup routine.
Reference designs
-
Reference design
An open-source machine that plants, waters and weeds a raised bed. Its maker sells kits and publishes the designs, the software and the documentation.
External resource. Not made or verified by Public ProtocolMaturity: Not assessedSafety S1Open, with paid products or servicesListed 16 November 2026
Parts list
Every resource that supports this element. Those marked External carry the label “External resource. Not made or verified by Public Protocol”.
| Resource | Type | Origin | Maturity | Safety | Licence, as stated at its source |
|---|---|---|---|---|---|
| AGROVOC | Standard | External | Established | S0 | Not stated on the page opened |
| AlphaEarth Foundations helps map our planet in unprecedented detail | Case study | External | Not assessed | S0 | Not stated on the page opened |
| Arduino Documentation | Guide | External | Established | S0 | Not stated on the page opened |
| Deep learning for image-based cassava disease detection (Ramcharan and others, 2017) | Paper | External | Working | S0 | Creative Commons Attribution (CC BY 4.0) |
| Copernicus Data Space Ecosystem | Dataset | External | Established | S0 | Not stated on the page opened |
| FarmBot | Reference design | External | Not assessed | S1 | As stated for each component: hardware to the open-source hardware definition, software under permissive open-source licences, open documentation |
| Farmer.Chat: scaling AI-powered agricultural services for smallholder farmers (Singh and others, 2024) | Paper | External | Working | S1 | CC BY-NC-SA 4.0 |
| farmOS | Tool | External | Not assessed | S0 | Software: GNU GPL. Site: CC BY-NC-SA 4.0. |
| LiteFarm | Tool | External | Not assessed | S0 | Described as "open source" at the source; the licence is not named on the page opened |
| About LoRaWAN | Standard | External | Established | S1 | Not stated on the page opened |
| Mycodo | Tool | External | Working | S1 | GNU GPL v3 or later |
| ODM (OpenDroneMap) | Tool | External | Established | S0 | AGPL-3.0 license |
| Using deep learning for image-based plant disease detection (Mohanty, Hughes and Salathé, 2016) | Paper | External | Replicated | S0 | Creative Commons Attribution (CC BY 4.0) |
| WaPOR: FAO portal to monitor water productivity through open access of remotely sensed derived data | Dataset | External | Established | S0 | Not stated on the page opened |
Depends on, and completes
It is one of 27 elements that complete Food and farming.
The standing questions
Who can afford this: sensors are cheap, but time to maintain them is not. Who does the work: a data keeper. Who decides: the group decides what leaves the farm.
Environments
the cassava model was built on field photographs in Tanzania (Deep learning for image-based cassava disease detection (Ramcharan and others, 2017)).
- Tropical humid forest, arctic cold, hot deserts, islands, temperate forest, high altitude, flood plains: no source found on sensor performance by setting in this research; the machines and environment profiles topics cover hardware in harsh settings.
Environment profiles for these settings: Tropical rainforest and humid tropics. All environments
Safety
State and evidence
Drafted
An editor has written the element's text from the resources listed under it, citing them. It is checked when it is reviewed.
Evidence from a real place or system is recorded here, with what did not work. What the states mean
History
- Drafted in the release of 16 November 2026.