Water & energy efficiency
VFD pumps, real-time ETc and per-block irrigation policies.
Irrigating with precision when water is scarce
High-value fruit crops —avocado, grape, citrus— are water-dependent and operate in a region that has already crossed the mega-drought threshold. Seasons with reduced water allocations stopped being the exception and became the rule. Every liter applied in excess is a liter missing elsewhere in the season, or in the next one.
The problem isn't only how much water is available: it's that most "tech-enabled" farms don't precisely know what happens underground. They irrigate by time-based rules, with no continuous moisture reading per block, no contrast between theoretical and effective application rate, and no closed loop between what the soil needs and what the pump delivers.
This project redesigns that architecture: it replaces traditional commercial controllers —limited, expensive and blind to the soil— with an open, modular, densely sensorized platform that gives you back fine control over water, fertilizer and energy use.
The invisible ceiling of conventional tech adoption
Farms considered tech-enabled today run on commercial controllers carrying a structural limitation: most support a maximum of 4 or 5 sensor or actuator inputs per control unit. Faced with high CAPEX per unit and a low channel ceiling, the design prioritizes actuation over sensing. The few available channels go to opening valves and starting pumps; what happens in the soil, block by block, goes uninstrumented.
The problem isn't a lack of technology in the field. It's that the installed technology was never designed to sense densely: it was designed to actuate cheaply. The consequence is irrigating with eyes closed, and paying the bill in water, fertilizer and fruit.
An open, modular and dense architecture
The proposal replaces the "one expensive controller with few inputs" logic with a distributed network of low-cost nodes, linked by long-range low-power radio (LoRaWAN) and orchestrated from a cloud platform the grower owns. A single central gateway receives the signal from every node in the field; each node reads multiple sensors or controls multiple actuators. The 5-channel controller stops being a bottleneck.
Field nodes
ESP32 + LoRaWAN. They sense moisture, EC, flow, pressure and control valves, powered by solar panel and LiFePO4 battery.
Central gateway
A single elevated point listens to all nodes over radio and reaches the internet via 4G or Ethernet. No spectrum license.
Cloud platform
Over MQTT, data lands on a VPS with a Supabase database under the grower's name. This is where decisions, alerts and the closed loop happen.
The system speaks the industrial protocols needed to integrate with existing infrastructure: Modbus RTU over RS485 for ABB, Siemens, Delta or Schneider drives; Modbus TCP for equipment on Ethernet; and dry contacts for direct start-stop where there's no communication. The field's heterogeneity stops being an obstacle.
| Dimension | Traditional system | Proposed architecture |
|---|---|---|
| Sensors per controller | Max. 4-5 physical inputs | No practical limit, nodes added on demand |
| RF coverage | Wiring or short-range proprietary radio | LoRaWAN, 2-10 km line of sight, license-free |
| Node autonomy | Requires grid power or dedicated install | Solar + LiFePO4 battery, 3-5 years untouched |
| Integration protocol | Closed, single-vendor ecosystem | Open: Modbus RTU/TCP, MQTT, dry contact |
| Data ownership | Vendor-hosted, monthly subscription | Own database, no recurring license |
| Cost per measurement point | High for each added channel | A fraction, per multi-sensor node |
| Installing a new node | Wiring, trench, config: days | Plug-and-play: hours |
| Total CAPEX for full sensing | Reference: 100% | ≈ 50% of the equivalent CAPEX |
What each piece does
The system is made of seven functional blocks. Each performs a specific role and talks to the rest over standard protocols. These are their reference specifications:
| Component | Role | Key specs |
|---|---|---|
| Central LoRaWAN gateway | Communications core | 8 channels · 2-10 km LoS · 1,000+ nodes · 4G + Ethernet · IP65 |
| Soil sensor nodes | Soil sensing | ESP32 + LoRa · moisture, temp and EC · capacitive RS485 IP68 · solar + LiFePO4 · 3-5 years |
| Weather station | Climate sensing | Temp, RH, rain, radiation, wind · Penman-Monteith ETo calculation · 1 per farm |
| Sectoral flow meters | Hydraulic measurement | Hall / electromagnetic · DN50 to DN200 · pulses or 4-20 mA · 1 per block + 1 main |
| Valve control nodes | Actuator control | ESP32 + LoRa Class C · 24VAC (Hunter / Rain Bird) and DC latching · up to 8 valves |
| Universal pump controller | Pump room | Modbus RTU + Modbus TCP + 8 dry-contact relays · pressure, flow and current reading · local safe mode |
| Software platform | Intelligence and operation | ChirpStack + MQTT · PostgreSQL + TimescaleDB · FAO-56 + Kc model · web dashboard · push, email and WhatsApp |
Each component has a reading for the grower, beyond its datasheet:
Central gateway
The system's radio booth. A single one, on a high point of the field, listens to and talks with every installed sensor and actuator.
Soil nodes
The plant's glucose monitor. It tells you, hour by hour, whether the soil holds the moisture the tree needs or whether it's over- or under-watered.
Weather station
It computes how much water the climate will ask of the crop each day, to anticipate rather than react.
Sectoral flow meters
They let you know how much water entered each block, not just how much left the pump. They're the utility meter of every sector.
Valve nodes
The remote switches of the valves. They replace the trip to the block with a tap on the dashboard.
Pump controller
The brain of the pump room. It talks to the drives already installed, regardless of brand, and keeps irrigating even if the internet drops.
Software platform
The field's control panel. A single screen shows everything, proposes what to do, and lets the operator approve, without removing human control.
Catching the failure before it becomes one
Dense sensing doesn't only improve irrigation decisions: it transforms the operation. Every field sensor is also a sensor of its own infrastructure. Anomalous flow betrays breaks, low pressure signals worn pumps, erratic readings indicate batteries at end of life, and EC drift suggests sensors needing recalibration. The platform includes a condition-based preventive maintenance module.
Automatic detection
The system watches trends and thresholds across every variable: flow, pressure, motor current, service hours, battery voltage and sensor drift. A deviation fires an event.
Task generation
The platform creates a task with context: affected equipment, observed variable, historical values and failure probability. The technician gets a half-solved case, not a raw alarm.
Assignment and notification
The task goes to the responsible role —electrical, hydraulic or agronomic— by the nature of the event. Push, email and WhatsApp notification with a direct link to the detail.
Closure and learning
The technician documents the intervention from the phone in the field. The record links to the equipment and feeds the maintenance history. Every solved failure trains the system.
The result is a regime change: from corrective, reactive maintenance —which pays the full cost of failure, including the irrigation stoppage— to preventive maintenance based on real data. Main-line breaks are caught in hours, pumps are serviced before burning the motor, and sensors are recalibrated before poisoning agronomic decisions.
Where the savings become visible
The benefits don't concentrate on a single front: they spread across the operational chain and compound on each other. The ranges below are indicative and subject to pilot validation, but they correspond to orders of magnitude documented in agronomic literature and industry.
Two second-order effects add to those four axes: recovery of the soil profile —without chronic washing, microbial activity and soil retention stabilize— and improved fruit vigor and size, because better-controlled water stress in critical phenological windows yields greater size uniformity and less culling, straight to the sale price.
The system doesn't compete on a single axis. It competes on water, fertilizer, energy, maintenance and fruit quality at once. The sum of marginal gains on each front is what makes the return on investment typically materialize within one to two seasons of operation in high-value fruit.
How to start: a pilot before scaling
A solution of this scale is validated before being scaled. The right way to start is a bounded pilot, measure results, and only then decide the roll-out to the rest of the farm.
Pilot on 1 hectare · 4-6 weeks
One gateway, two sensor nodes, one valve control node and a sectoral flow meter. RF coverage, sensor calibration and the basic agronomic model are validated.
Scaling to the farm · 8-10 weeks
If the LoRa link budget holds, full deployment of sensors and valve nodes. The recommendation engine, automated ETo and the team's operational dashboard go live.
Pump room integration · 4-6 weeks
Last, for the larger blast radius: integration with VFDs over Modbus, activation of the universal controller and closing the irrigation-monitoring-control loop end-to-end.