USE CASES · IOT
IoT in agriculture: observing in time to make better decisions
A moisture reading can help inform an irrigation decision. Interpreting it requires knowing the soil, depth, crop and location of the measurement. The data becomes useful when it represents a relevant condition and reaches someone who can act.
In agriculture, IoT connects observations from the field, equipment and work records. We are interested in using it to coordinate water, inputs, livestock and machinery with actual production conditions. Each application needs a defined decision and a way to check the effect of an intervention.
Soil and crops: measuring a representative condition
Sensors can record moisture and temperature at selected points. Their location, depth and calibration need to reflect field variability and the crop. A single reading may describe its measurement point well while representing another part of the field poorly.
Other indicators require specific methods. Interpreting nutrient availability, for example, involves relating appropriate measurements and analyses to soil conditions and management. Choosing a microcontroller does not, by itself, provide that observation capability.
Records need to retain the plot, measurement point, unit and time. Field visits and observations by the people doing the work help check whether the measurement reflects the condition we need to understand.
Irrigation: relating need, command and water delivered
Scheduling can use moisture, rainfall, crop condition and water availability. The University of Minnesota’s irrigation scheduling resources show why observations and water balance need to be considered together.
In a design example, a reading that deserves attention is checked against the schedule. The controller applies the authorized rule, and a flow measurement helps verify whether water was delivered. A valve-opening command and water actually applied are different events.
Operation needs local limits and a defined response when readings or communication are missing. The trial compares water applied, crop conditions and operating effort, without assuming a saving in advance.
Livestock: a signal that helps guide a check
A connected device can provide location, activity or other measurements supported by its design. A change from usual behavior may justify observing an animal. That signal needs to be related to handling, environmental conditions and the state of the device.
The notification should identify the animal or group, the period and the reason for checking. Health interpretation requires a competent assessment; the sensor provides evidence. We need to establish whether the signal supports timely checks and how many notifications demand attention without informing a useful intervention.
Weather: observing the site and understanding the forecast’s scope
A station can record temperature, humidity, wind and rainfall according to its equipment and installation. The local time series helps describe conditions relevant to a task, such as applying a treatment or preparing irrigation.
Observing current weather and forecasting its development require different capabilities. External forecasts can be combined with local records while retaining their source, issue time and horizon. Installing a station does not demonstrate that an in-house model predicts events accurately enough for a particular decision.
Machinery: relating use, condition and maintenance
Machines can provide operating hours, consumption, location and condition signals. These data help reconstruct routes and waiting periods, prepare maintenance or investigate a performance difference.
Comparison requires knowing the task, equipment and working conditions. A change in soil or load may explain a variation that does not indicate a fault. Technical inspection and its outcome complete the evidence needed to evaluate alerts and future predictions.
Traceability: retaining relationships after harvest
Products move through harvesting, storage and transport. Lot identification needs to survive splitting, mixing and delivery. Condition measurements provide evidence when they remain linked to the lot and the interval observed.
The GS1 EPCIS standard provides a reference for sharing product and asset events. Recording a movement supports reconstruction of the history; claiming a quality condition also requires the corresponding evidence.
Yield estimates and planning
Historical records and observations during the growing cycle can support yield estimates. A model needs representative data and evaluation against seasons not used to fit it. The result should describe its uncertainty and the conditions that could change it.
Planning inputs, storage or transport calls for comparing scenarios. An estimate prepares a decision; its usefulness is checked against what happened and the consequences of error. A calculated number does not automatically become a reliable forecast.
Building a first application that can be sustained
Data acquisition can combine field devices, gateways and local processing. Node-RED can integrate messages using MQTT. When AWS is useful, IoT Core can route messages to processing and storage configured for those functions.
Coverage, power, maintenance and recovery after disconnection are part of the design. Open software makes it possible to understand and adapt components when documentation and maintenance capability are available.
An initial trial can focus on an irrigation zone, one machine or a lot’s journey. The aim is to complete a sequence of observation, intervention and verification. That evidence helps determine what to extend and what should be corrected first.