I live in the countryside. The sound of the birds. The smell of the grass after the rain. Drinking my coffee surrounded by my chickens.
The countryside provides me with a lovely life. But it also gives me some interesting challenges when I want to add a bit of technology to it.
One of them appeared when I asked myself: how can I automate watering my garden? In most cases, it doesn't require much effort. You can buy a watering timer, set it to run at a certain time and… that's it. But I wanted something a bit smarter. A garden that gets the water it needs, without wasting any.
So how can I know if I need to water my vegetables and fruit trees? And how much water do they actually need?
It was the starting point of an interesting journey!
The first step was very similar to my other projects: brainstorming with my pal ChatGPT.
My original idea was to have an n8n workflow that takes, at least, the current weather into account to determine if I need to water my garden. I discovered the free API from Open-Meteo. It answered pretty much all my needs. And even more.
I also (re)discovered the concept of evapotranspiration. It's a cool name for something quite simple: water evaporates from the soil, and plants release water through their leaves. Open-Meteo provides an estimate of this process, called reference evapotranspiration (ET₀).
Basically, it helps estimate how much water is being lost.
Another important requirement was to isolate my different garden areas. I have two vegetable gardens and an area dedicated to my fruit trees. Their needs are totally different.
I did some research (in fact, I asked ChatGPT), and determined an initial estimate of the water needs for each area. Each zone now has its own configuration: number of plants or surface area, water requirements, minimum interval between watering events, and a few other parameters.
Nothing too complicated. But enough to make better decisions.
The workflow is quite simple.
It starts every morning with a scheduled trigger. It fetches weather data from Open-Meteo, normalizes it, retrieves the latest watering events, does some maths, and determines if watering is needed or not.
The interesting part is the water deficit.
Think of it as a balance. Every day, plants consume water, and rain brings some back. The workflow estimates this balance for each zone, using rainfall, evapotranspiration, and the specific needs of the plants. If the deficit reaches a certain threshold, it's probably time to water.
But there's another question: what if it's going to rain tomorrow?
That's why the workflow also looks at the forecast. There's no point in watering my garden today if nature is planning to do it for me tomorrow. At least, in theory. Weather forecasts aren't always right.
Finally, the workflow updates the state of each zone and decides whether to send me a notification.
Everything runs on n8n, hosted on my Raspberry Pi. And the best part? I don't have to think about it every morning.
The event system is divided into two parts: the push notifications and the watering events I manually record.
First, the notifications.
Every morning, the workflow checks which zones need water.
If watering is recommended, I receive a push notification on my phone through Pushover.
Second, the watering events.
Knowing when to water is useful. But the system also needs to know when I actually do it. So I created a simple form using n8n.
Every time I water a zone, I open the form on my phone, select the garden area, and enter how much water I used. This event is then stored in a database and taken into account during the next calculation.
The important thing is that a notification doesn't mean the garden has been watered. The workflow needs confirmation. It's a small detail, but without it, the whole system would quickly become unreliable.
That's a good question.
I could probably connect some valves, add a few sensors, and let the system water everything automatically.
And to be fair, that sounds pretty cool.
But I don't think I need it. At least, not yet.
First, my calculations are still estimates. I'm using weather data and theoretical water requirements, not real-time soil moisture measurements. Some plants might need more water. Others might need less. And there's something else.
I actually enjoy taking care of my garden.
Walking outside, checking the vegetables, looking at the fruit trees. It's part of why I love living in the countryside.
I don't want technology to replace that. I want it to help me make better decisions.
For now, a simple notification telling me when and how much to water is already a big improvement. Maybe I'll add soil moisture sensors in the future. Maybe I'll automate watering for some areas.
But I'll do it when there's a real need for it.
This little project reminded me of something I really enjoy about software development.
You don't always need to build a big application to solve a problem.
Sometimes, a free API, a few calculations, a small database, and a couple of n8n workflows are more than enough.
Is my system perfect? Absolutely not.
It still needs some calibration. The weather isn't always predictable, and every garden is different. But it's a good starting point. And more importantly, it's already useful.
I also learned a few things about plants, water, and the weather along the way. Not bad for a project that started with a simple question.
And now, if you'll excuse me, I have some chickens to feed.