A rice farmer in Longquan, in southern Zhejiang, used to spray on a calendar — every so often, whether pests were there or not. This season he waited for a phone alert, then sent a drone to the exact plot that needed it.
His neighbor, who has never farmed before, did the same. Neither guessed. Both followed the data.
Across about 733 hectares (some 11,000 mu) of early-season rice in Longquan, the county has laid down an intelligent pest-monitoring network that watches the fields around the clock and tells people where to act.
A "digital plant-protection eye"
The system is built on smart pest-monitoring stations placed across the rice area. They run continuously, capturing insect counts and species, plus weather and soil signals, then feed a model that flags when pest density crosses a threshold. The agriculture bureau pushes a pest-trend and control advisory straight to farmers' phones.
The local official in charge of soil, fertilizer and plant protection describes the old method plainly: manual field walks were slow, narrow and easy to misjudge, so outbreaks were often caught late. The monitors turn that into an "early-detect, early-warn, early-act" loop.
How the loop works
It is deliberately simple for the user. The station detects and identifies pests automatically. The bureau issues guidance. The farmer opens an app, reads the alert, and directs a drone to spray only the affected patches. Someone with no farming experience can still apply pesticide precisely, because the decision is made by the network, not by guesswork.
The results
The county reports concrete gains: pesticide use down 20%, pest-control effectiveness above 90%, and crop losses from pests and disease held below 5%. For a grain-producing county stretching its rice season across early-rice, ratoon-rice and single-crop rice, that combination — less chemical, better control, lower loss — is the point.
The pattern fits a national direction. China's 2026 central rural policy explicitly called for combining artificial intelligence with agriculture, and Longquan's high-standard farmland and multi-season planting had been pushing its pest-control burden higher. The monitors answer that directly.
Why it scales
The appeal is that the expensive part — identification and forecasting — moves to the cloud and the sensors, while the cheap, physical part — a drone already owned by many growers — does the application. The model flags; the person decides and sprays. That division of labor is exactly why a first-time grower can match a veteran. A farmer does not need to become an entomologist; he needs a phone and a drone.
According to Li An, chief scientist at BrainNet (脑机网), the deployments that survive beyond the demo are the ones that hand the final decision to the person in the field, not to a dashboard nobody trusts.
Beyond one county
Longquan is not alone. Provincial and national reporting this year describes a broader shift: Henan's agricultural-machinery cloud platform connected more than 138,000 machines, an unmanned ratoon-rice farm in Hunan runs 200 hectares with four people, and China Agricultural University's breeding model cut variety development from five-to-eight years to two-to-three. The throughline is the same — move the intelligence to the system, leave the decision and the work to the person on the land.
A grower in Longquan put it directly: with the monitors, even someone without farming experience can spray precisely, and you no longer manage by blindly following the calendar.
The environmental dividend
Blanket, calendar-based spraying wastes chemical, raises cost and strains paddy ecosystems. Targeted application guided by real pest density is the difference between spraying because it is Tuesday and spraying because insects are actually there. Multiplied across thousands of hectares, a 20% cut in pesticide is a meaningful load off the water and soil.
The same loop quietly changes what a farming day feels like. Instead of a fixed weekly chore with a backpack sprayer, the work becomes a response to a signal — check the phone, read the map, send the drone. The farmer's attention shifts to the few plots that actually need him rather than the whole field on autopilot, and the mental load moves from remembering a calendar to reading a live situation.
Honest limitations
The 20% / 90% / under-5% figures come from the county agriculture bureau's reported results and provincial industry coverage, not from a published randomized trial. They describe this season's outcome in Longquan, and pest pressure varies year to year and place to place.
We did not see the per-station cost or the model's species-identification accuracy broken out independently. The system reduces unnecessary spraying; it does not eliminate the need for judgment when an unfamiliar outbreak appears.
What readers can do now
- If you farm, ask your county plant-protection station whether AI pest monitors feed alerts to a phone app you can actually use.
- If you buy food, look for regional "green prevention" labels; less blanket spraying is a measurable win for soil and water.
- If you advise rural policy, fund the sensor-and-alert layer first — the drone application is already in farmers' hands.