Coffee farming is full of tiny decisions: When to fertilize, when to prune, when to spray, and when to do nothing. A proposed “digital coffee farmer” aims to turn daily field observations into simple recommendations that support growers.
The idea mixes low-cost sensors, a wireless network, and prediction models. Fieldwork in Quindio helped map what expert farmers watch most closely, then translate that knowledge into data that a model can learn from.
From gut feeling to data
Many farms do not log weather or soil conditions, even when growers spot patterns. At Las Acacias in Salento, sharp day-night temperature swings can burn leaves and delay regrowth by two months.
A digital approach does not replace experience. It tries to capture key signals, then show them on a simple dashboard, so a grower can see what matters without drowning in numbers.
This proposal treats the farm like a living system, where soil, rain, and temperature push the crop in different directions. Better records make it easier to compare seasons, test changes, and avoid expensive mistakes.
It also accepts reality; budgets are tight. That is why the plan includes low-cost sensors plus public weather data and shared knowledge between nearby growers.
The IoT network on the farm
Visits to Las Acacias and La Morelia in Quindio, plus interviews with expert coffee farmers in May 2023, helped pick the first variables to monitor. The goal was to track what farmers already use.
For Quindio, the scale is not small. The coffee area was described as 5,662 farms covering 18,051 hectares, so a tool that works for small producers has to be practical and easy to maintain.
On Las Acacias, the proposal lists a LoRaWAN (device linking low-power sensors to the internet) gateway and sensors for air temperature and humidity, plus soil moisture and electrical conductivity. The goal is steady measurements, sent wirelessly, without constant manual work by using Internet of Things (IoT) concepts.
The key variables are simple and familiar: air temperature and humidity, rainfall, and soil moisture. These match what farmers already use, only now they can be logged, graphed, and compared.
Turning measurements into predictions
The “digital coffee farmer” is an AI model meant to copy how an expert chooses field actions. Training needs inputs from sensors and soil tests, plus labeled outputs, for the real decisions taken in the plot.
The output classes include actions such as applying agrochemicals, using integrated pest and disease prevention practices (MIPE), pruning, or taking no action. That last option matters because sometimes the best move is to wait.
A first training attempt used a test dataset and a multilayer perceptron (MLP) network, a simple “stack” of connected layers of artificial neurons that learns patterns from examples and then makes a prediction, but results were limited because there was not yet a large volume of real sensor data. Some values were simulated.
Prediction improves as seasons pass because each harvest adds more “stories” to the dataset, including what was done and what the crop produced.
Soil, climate, and coffee quality
Soil tests remain central. The approach highlights pH, phosphorus, and potassium, plus climate variables, because nutrient balance and weather swings can show up later as yield losses or quality problems.
Las Acacias is described with 20,000 productive trees on 4.5 hectares, with 90% mountainous topography, and varieties Castillo and Cenicafé. Those details matter because terrain and variety change how water and nutrients behave.
Coffee also faces pests and diseases. The proposal links climate shifts with issues like coffee borer and rust, and it points to research using vision-based tools and models for early disease detection.
A simple dashboard can help connect the dots. When humidity spikes and rain keeps coming, a grower can plan field work better, avoid wasted inputs, and protect quality before defects show up in the cup.
Brewing the future of coffee farming
The “digital coffee farmer” is not a magic robot. It is a plan to capture the best part of expert intuition, then back it with sensors and models that keep learning across harvests.
If the IoT network is deployed and actions are logged carefully, small producers can gain clearer signals about their farms. That can mean fewer surprises, smarter timing, and better coffee with less guesswork.