LoRa Vineyard Sensor

Powdery and downy mildew hit vines when temperature, humidity and leaf wetness line up in particular ways, and growers usually find out after the fact. The prediction tools that exist need a stable internet connection to reach the cloud, which rural Western Cape vineyards often do not have. Our final-year project is a fault-tolerant, offline-first LoRa sensor network that runs entirely on the farm.

I co-lead the six-person team with a classmate who holds a master's in viticulture: she brings the disease science, I designed the system architecture and built the hardware. Each node is a Heltec Wireless Stick (ESP32-S3 with LoRa) with an on-board OLED, sealed in a weatherproof enclosure with a solar panel, a battery and a microSD card for local storage. It started on a breadboard on my desk and is now strapped to a fence post in the field.

Nodes report over long-range radio to a Raspberry Pi central node, which ingests the readings into SQLite and evaluates a deterministic, rule-based prediction tree built from established disease thresholds: sustained wet runs for downy mildew, repeated warm spells for powdery. No training data, and no outbound internet connection.

The dashboard plots temperature, dew point, humidity and battery charge and marks the hours when every condition for each disease was met. The central node and the locally hosted PWA that farmers will use are being developed now.