Designing field telemetry for smallholder farms with intermittent connectivity and practical alerting.
The connectivity assumption problem
Most IoT tutorials assume reliable Wi-Fi or 4G. Rural Bangladesh has neither in many farming areas.
LoRaWAN gives us a workable path: low-power, long-range, and tolerant of intermittent gateway uptime. But designing for LoRaWAN introduces its own constraints — message rate limits, small payload windows, and ACK reliability trade-offs that shape every part of the firmware and backend architecture.
The key difference from a city deployment: we cannot assume the gateway is always reachable. Nodes must buffer readings locally, prioritise critical alerts, and retry efficiently when the network window opens.
Soil moisture sensing in the field
Capacitive soil moisture sensors are cheap and widely available. They are also highly sensitive to soil composition, salinity, and burial depth — factors that vary enormously across a single farm, let alone across the country.
Our calibration approach:
- Two-point calibration per sensor — we calibrate each node against air (dry reference) and saturated soil (wet reference) at the installation site.
- Salinity correction — saline soils read artificially high on capacitive sensors. We use a correction factor derived from a conductivity measurement.
- Depth consistency — burial depth variation of ±2 cm measurably shifts readings. We fix depth with a mechanical guide and verify with a gauge before sealing the access point.
The calibration procedure, correction formula, and tolerance range are under agronomist review and will be published with the deployment guide.
Solar harvesting for node autonomy
A node that a farmer has to charge manually is a node that stops working within a month.
Our design target is full autonomy through consecutive overcast days during monsoon season, with zero manual charging.
Key decisions:
- Panel sizing — matched to the node daily energy budget and its mounting geometry.
- Battery capacity — a lithium pack sized to survive consecutive overcast days without sunlight.
- MPPT firmware — tuned on the node microcontroller to maximise harvest at low irradiance.
- Low-power mode — node enters deep sleep between telemetry windows, waking only for scheduled readings and alert checks.
Final specifications and the full power budget will be published once hardware validation completes.
Alerts, not dashboards
Farmers do not want to read sensor graphs. They want to know when to irrigate and when not to.
We built a threshold-rule system for the first version:
- Low-moisture alert: triggered when soil moisture stays below a calibrated threshold for consecutive readings.
- Waterlogging alert: triggered when moisture stays above a calibrated threshold for a sustained period.
- Delivery path: SMS via a regional gateway for areas with limited smartphone access. App notification for users with smartphones and data.
The lesson: alert fatigue is as dangerous as no alerting. Our thresholds are conservative on the first deployment and tightened based on operator feedback after two weeks of observation.
Threshold values and delivery integrations are being tuned during pilot review and will be documented with each deployment.
ARIOT Engineering Team
[Author bio pending — will appear once engineering team profiles are approved.]