Watering new trees with water from the air deserves a proper trial

Anyone who knows me knows I am usually the first to get excited about new technology, especially when it can reduce our dependence on imports. So when the idea came up of using atmospheric water generators, machines that condense water out of the air, to water newly planted trees in the parts of Aruba where no truck or pipe can easily reach, I was hoping it would work. This summer I worked on a model to find out. The short answer is that, on the numbers I can defend today, it comes out more expensive than bringing the water in by truck. The longer answer is that most of those numbers have never been measured here, and some of the gaps are big enough that I would not rule the idea out.

Energy first

The machines most people have in mind work much like a large dehumidifier. A compressor cools a coil below the dew point, water condenses on it, and you collect it in a tank. Our warm and humid trade-wind air is actually quite good for this, but it still takes a lot of energy. Physics puts the floor at around 0.30 kWh per liter in our conditions, and real machines use more. I assumed 0.45 kWh per liter for a mid-size unit and more for small ones, since independent tests of small units have measured two to four times the energy their vendors advertise. Off-grid, that energy has to come from small wind turbines or solar panels and batteries sitting next to the machine, and that is where most of the money goes.

What the model says

I set up a planting site of 1,000 trees that need water at the roots through their first two dry seasons, roughly 310 liters per tree in total, and ran it through 200 simulated rain sequences based on the Meteorological Department’s 30-year averages so that dry years are included. I didn’t treat any water as free. Trucked water pays for the water itself, the truck hours, a tank on site that has to be moved to the next planting, and the drip lines. The machine pays for its panels, batteries, storage, upkeep and moves.

With my central assumptions, spraying trees from a water truck, which is how many plantings could be watered today, comes to about US$56 per tree over two years. If the same truck instead fills a tank on site that feeds drip lines, the cost falls to about $16. A solar-powered AWG feeding the same drip lines comes to about $84, or $58 if you add a plastic sheet to catch rain. For 1,000 trees it would take a machine drawing about 29 kW with close to 60 kWp of solar panels behind it. When I varied all the uncertain inputs together, the truck-filled tank came out cheaper.

Those are central assumptions, though, and the machine gets a lot closer at the optimistic end. If the machines perform near the physical limit, cost about $1,300 per kW, last 12 years and rarely break down, the AWG drops to around $45 per tree, which already beats spraying. I also looked at a more modest job for the machine, only keeping seedlings alive from planting until the heavy rains arrive, using low-end pricing. That costs about $12 per tree for the season against $5.70 for a truck-filled tank, and with the best-case inputs it falls to about $7.20. At a site a kilometer from the nearest road, where people would otherwise have to carry the water in by hand, the machine was the cheaper option in more than a quarter of the runs, without counting the effort of getting the machine up there in the first place.

Where I would try it first

The machine makes most sense where trucks can’t get in reliably but a vehicle can get in once to install it, and where it can stay put. In the bridging case the machine, panels and batteries are only about a third of the cost. The rest is storage, setting up, moving the kit to a new site every year, drip lines and maintenance visits. A unit that stays in one place for years, feeding several plantings around it or a permanent orchard or food forest, avoids much of that, and it’s a case I still need to model. A machine that also supplies drinking water at a ranger post or field station would share its cost with another use. Rain helps as well. Gutters on the machine’s own solar panels cost very little, and every liter of rain they catch lowers the cost of the system. With the low turbine prices I used, small wind turbines on our trade winds even made it into the cheapest design.

Too many unknowns

To me the most important result is not the cost comparison but how much of it rests on guesses. I couldn’t find reliable numbers on how much water our native seedlings would need, or how many survive under each way of watering. For the first dry season, the literature supports anything from a survival ration of 10 to 30 liters per tree, delivered through a deep pipe, to a few hundred liters for healthy growth. A tenfold range like that moves the cost of every option, machine or truck.

We also don’t know what a water truck really costs here. Taking conservative numbers for trucking into account, making water on site can look interesting. Machine efficiency here is still a question mark. In the field it could be anywhere from 0.30 to 0.90 kWh per liter, and we still don’t know how the coils will hold up in our salty coastal air or how often the machines would sit idle waiting for parts.

The rain is another question mark. Our Meteorological Department describes the rainy season as September to January, which brings almost three quarters of our annual rain, but there is no official date for when it starts. In six years of daily records, the rain since 1 September reached 100 mm, enough to leave young trees on their own, around 22 October in a typical year. In 2015 it didn’t get there, and October to December brought only 38 mm. KNMI’s climate scenarios for Aruba point to a drier island with a longer dry season, which would make a water source that doesn’t depend on rain more valuable over time.

How the water reaches the tree matters as well. In our heat and wind only about 45% of the water sprayed into a basin around a tree will stay in the root zone, compared with about 85% for drip or a deep pipe. In a trial in the Sonoran desert, seedlings watered through deep pipes had 71% survival against 23% for surface watering with the same small amount of water. An AWG feeding drip lines gets that benefit by design, but we haven’t tested whether a small, steady supply helps young trees more than a good soak every week or two. Frequent shallow watering could keep roots near the surface, so it could go either way. And of course none of this helps if goats get to the trees first. There’s an estimate that Arikok alone has around 1,500 goats, and unfenced seedlings would rarely survive a week.

What is next

Most of these questions can be answered in two planting seasons. A trial with about 200 seedlings, split between truck spraying, a tank with drip and a tank with deep pipes at two water levels, with soil moisture sensors and a monthly count of what survives, would tell us what our trees actually need. Alongside it, a mid-size AWG with coated coils and a modern refrigerant such as R-32 or R-290, run for at least three months at a coastal and an inland site, would show what a machine really produces per kWh in our air and how often it breaks down.

We have started these trials, but in Aruba, with limited resources, the research goes slowly. At the University of Aruba we have a small machine for testing, and this summer two Penn State students, Brian and Bradley Roseborough, have been working to find a good off-grid power solution for it, so it can run with no grid connection at all.

The rest is cheap. With 30 years or more of daily rainfall from the Meteorological Department and a few simple rain gauges at the planting areas, we could set a sensible planting trigger, for example 50 mm of rain since September or soil wet to 20 cm.

I don’t know yet whether water from the air will become part of how we plant trees in Aruba, but I’m curious enough to keep testing it, and I hope others are too. If you have experience watering native trees here, run a water truck, or have a site that could host a trial plot, I’d like to hear from you.

Thank you to the National Lab of the Rockies, the Aruba Hotel and Tourism Association (AHATA), Iberostar and Penn State for supporting this research. Thanks also to Dr. Brian Fronk and Dr. Margaret Busse at Penn State for their help, and to Brian and Bradley for their work this summer.