SIP4
AI-driven irrigation scheduling that delivers
more yield with less water
Sector
Sugar Beet & Potatoes
Country
Poland
SIP Leader
PSNC
Participants
L-PIT, DIH WODR
The challenge
Water scarcity and unpredictable rainfall are increasing the pressure on Polish arable farmers, particularly those growing water-intensive crops like potatoes and sugar beets. Without real-time soil and weather data, irrigation decisions are made on fixed schedules, wasting water, raising energy costs, and leaving yield potential unrealised.
SIP4 set out to replace calendar-based irrigation with a data-driven controller that responds to what is actually happening in the soil and what is forecast in the weather, automating decisions that previously required daily manual measurement.
How the system works
Soil moisture sensors placed across the field transmit readings continuously via LoRaWAN. The controller combines this data with real-time weather forecasts and evapotranspiration models to calculate exactly how much water the crop needs and when.
When irrigation is needed, the controller activates valves and pumps automatically, delivering water only to the zones that need it. If heavy rainfall is forecast, the system reduces or cancels a scheduled irrigation event, avoiding unnecessary water use. Farmers can monitor and override the system at any time via a mobile dashboard.
The system was tested in two modes: a Cloud mode that connects sensor data to a remote platform for centralised optimisation; and an Edge mode that processes all decisions locally without internet dependency.
Soil moisture sensors transmit real-time hydration readings via LoRaWAN.
Controller integrates soil data with weather forecasts and evapotranspiration models.
Irrigation scheduled automatically. Valves and pump activated only when needed.
Farmer reviews dashboard remotely. All decisions logged for traceability.
Two deployment modes tested
CLOUD MODEL
Sensor data transmitted to a cloud platform via LoRaWAN and internet. Integrates weather forecasts, evapotranspiration data, and remote soil readings to dynamically optimise each irrigation dose. Achieved 21% higher yields with 40% less water than conventional irrigation.
Best for: farms with internet connectivity, farmers who want remote control and weather-integrated optimisation
All processing on a local mini-PC on-site. No internet required. Decisions based on soil sensor readings only, without weather forecast integration. Achieved 6% higher yields with 20% water savings.
Best for: farms without reliable connectivity, long-term autonomous operation as local models mature
What farmers said
After Year 1, participating farmers and advisors shared their experience with the smart irrigation controller in both Cloud and Edge modes.
Strong support for wider adoption - led by Cloud
Both Cloud and Edge modes produced consistent user satisfaction. Farmers strongly support broader adoption, particularly with the Cloud mode. Organisational resistance was low for Cloud and moderate for Edge, reflecting the greater maturity of the Cloud system at this stage.
Cloud requires higher digital skills - and contributes to closing the digital divide
The Cloud mode requires a higher level of digital skills, indicating a need for training when scaling. However, it is also seen as contributing to reducing the digital divide by bringing advanced data-driven tools to farming communities.
Edge is a long-term opportunity as local data matures
As more soil and weather data accumulate over successive seasons, local models can be trained and improved, enabling greater autonomy and lower connectivity dependence. Farmers see Edge as a future direction rather than an immediate first choice.
“More yield, less water. The data does the deciding.”
Year 1 headline results
The smart irrigation controller delivers significant advantages to farmers, including:
higher yield
Cloud mode achieved 48.4 t/ha versus 38.2 t/ha with conventional irrigation. Edge mode achieved 6% higher yield.
less water used
Targeted spraying versus conventional blanket application across all vineyard rows.
less manual effort
Cloud mode reduced manual labour by 84% by eliminating daily field measurements and manual scheduling. Edge reduced manual effort by 67%.
Year 1 results: full breakdown
Results are organised across 4 dimensions. Each compares the Cloud and Edge deployment mode side by side. Dimension 4 is represented above in the section What farmers said.
| Metric | Cloud | Edge | What this means |
|---|---|---|---|
| Irrigation efficiency | Cloud: 48.4 t/ha (+21% vs conventional). 5 doses of 15mm. Dose reduced 3x based on rainfall forecast. | Edge: 42.5 t/ha (+6% vs conventional). No dynamic adjustment — no weather data access. | Cloud weather integration is the decisive advantage. Dynamic dose adjustment prevented overwatering and saved significant water. |
| ML algorithm performance | Dynamically reduced irrigation doses from 25mm to 15mm on 3 occasions based on heavy rainfall forecast. | No dynamic adjustment. Decisions based on soil moisture only. No forecast data available locally. | Cloud algorithm responds to future conditions, not just current soil state. This is the core performance differentiator. |
| Reduction in manual labour | Approximately 84% reduction. Eliminates daily manual measurements, soil sampling, and irrigation scheduling. | Approximately 67% reduction. Local automation handles scheduling but requires more manual oversight. | Both modes significantly reduce manual effort. Cloud enables full remote management. |
| Metric | Cloud | Edge | What this means |
|---|---|---|---|
| Network utilisation (full season) | Cloud: ~48.4 MB total (soil sensor data + weather queries + evapotranspiration + forecast). | Edge: ~4.8 MB total (soil sensor data only). No external queries. | Cloud uses 10x more network data but this enables weather-integrated optimisation that Edge cannot achieve. |
| Computing resources | Cloud: 10W MiniPC (proxy) + PSNC cloud servers. Total season: ~86.7 kWh. | Edge: 15W MiniPC (local compute). Total season: ~55 kWh. | Edge uses 36.5% less computing energy. Cloud offloads to PSNC server infrastructure. |
| Total farming operation energy | Same pump energy for both modes. ADS computing adds 55-87 kWh depending on mode. Submersible pump dominates total energy for both modes (~553 kWh for irrigation). Computing and transmission are minor by comparison. Optimising irrigation scheduling has far greater impact than computing efficiency. | ||
| Metric | Cloud | Edge | What this means |
|---|---|---|---|
| Data formats | Same for both modes. Redis, PostgreSQL, InfluxDB — identical stack for both modes. REST API with Swagger. OCSM semantics. Validated against OpenAgri Farm Calendar. | ||
| Networking standards | LoRaWAN for sensor to gateway. GSM/LTE for gateway to cloud. | LoRaWAN for sensor to local gateway. No internet required. | LoRaWAN ensures open interoperable sensor communication. Choice of internet depends on farm infrastructure. |
| Software switchability | Same modular open-source architecture for both modes. Docker containerised. Open-source stack. Modular architecture — management and optimisation modules updated independently. No vendor lock-in. | ||
| Farm data switchability | Same data access model. No vendor lock-in. Full REST API with Swagger. Cloud API online. Edge API on local network. Data downloadable in both cases. | ||
| Hardware switchability | Same flexibility for both modes. No dependency on specific sensor vendor. LoRaWAN-compatible sensors from multiple manufacturers. IRRTEC bridge sprinkler integrated. Raspberry Pi compatible. Any LoRaWAN actuator. | ||
OpenAgri OS-based Services used
The following OpenAgri OS-based services are integrated in SIP4:
OPENAGRI OS-BASED SERVICES USED
The following OpenAgri OS-based services are integrated in SIP4:
- Weather Data Service (climate forecasts and historical data for predictive irrigation planning)
- Soil Moisture Monitoring Module (aggregates field sensor data)
- Irrigation Management Platform (centralised control and scheduling)
PILOT DETAILS
- SIP Leader: PSNC (Poznan Supercomputing and Networking Center)
- Partners: L-PIT (ADS provider), DIH WODR (advisor)
- Sector: Arable / Sugar beet and potatoes
- Country: Poland
- Target group: Farmers seeking to increase irrigation efficiency