Inside the Tree Lab: Designing Trees with AI
Before we spend years and real water on a tree, we want to ask it some questions. How much water will it need in its fifth Eloy summer? Does a deeper pot really help? What happens if we push for faster growth at the cost of drought tolerance? The Tree Lab is our tool for asking those questions on a computer first.
One promise up front, because it matters: everything the Tree Lab produces is a model estimate, not a lab result. No real tree was grown, edited, or measured to make these numbers. The Tree Lab is a thinking tool, and our tall-pot trials are what will test it.
What the Tree Lab does
The Tree Lab is an online simulator. It runs in a web browser and has four parts.
1. Designer. You choose a base tree: giant sequoia, Aleppo pine, river red gum, or mesquite. Then you move sliders for traits such as root depth, drought tolerance, and growth rate. Traits trade off against each other the way they tend to in real plants. Pushing hard on fast growth, for example, can cost you survival under stress.
2. Test. You pick a container (a standard pot or a deep tall pot) and a weekly water budget in gallons, then run ten years. The simulation uses Eloy-area climate averages from NASA POWER, a free NASA data service, for the point at about 32.76 N, 111.55 W, using 2001 to 2025 averages. A 3D tree grows on screen, and you get an estimated height, estimated survival for a batch of 20 trees, and total water used.
3. Genome Explorer. Click a trait and see which gene families are known in the scientific literature to influence it, with links to sources. This is at the level of a textbook or review paper. It explains biology. It doesn't give lab instructions.
4. Collection. Save your favorite designs, compare them, and export them as a file so nothing is lost.
What a first run showed us
In an early test with no trait changes, the model's giant sequoia reached roughly 20 feet in 10 years on 15 gallons a week in a tall pot. On 5 gallons a week, none of the 20 virtual trees survived. The model's river red gum reached roughly 48 feet in the same span.
Those numbers line up with the broader picture from our research on desert giants: sequoia is thirsty, and red gum is a proven desert grower. But they are rough. The model's settings are starting guesses, and we expect to change them once real measurements come in. Please don't quote them as predictions.
Where the AI comes in
"AI" can mean a lot of things, so here is what it means for us right now.
Reading the literature. We use search tools such as Semantic Scholar and PubMed to find and summarize papers on drought, heat, and root growth in conifers. Every claim is checked against the original paper before it goes in our files.
Comparing genes across species. Scientists know many stress-related genes in crops and model plants. Free comparison tools like PLAZA and NCBI BLAST help find the matching candidates in the public giant sequoia genome (Scott et al. 2020). A match is a candidate, not proof. Research suggests drought tolerance in sequoia is spread across many genes with small effects (De La Torre et al. 2022).
Looking at protein shapes. The AlphaFold database offers predicted 3D protein structures. We use these to understand and illustrate biology, not to design changes.
Building and running models. AI helps us write and check the code behind the simulator.
What's next for the model
The current Tree Lab uses simple rules. Our AI-first research plan lays out a stronger path:
A real forestry growth model. Tools like 3-PG, a widely used stand-growth model with a free R version (r3PG), estimate growth from weather, soil water, and sunlight. We plan to run it on Eloy weather and feed the results into the Tree Lab.
Better climate and soil inputs. Daily weather from Daymet, long-term normals from PRISM, and soil data from SoilGrids.
Calibration with real trees. Monthly measurements from our tall-pot trials will tell us where the model is too optimistic or too gloomy.
Seed-source ranking. Comparing Eloy's climate with the hottest, driest wild sequoia groves to choose the best seed to start from.
Why build this at all?
Three reasons. First, it saves water, money, and trees. We can rule out bad ideas on a screen. Second, it makes our thinking visible. Anyone can see the assumptions and challenge them. Third, it invites people in. A student, a grower, or a scientist can try a design in a few minutes and tell us what we got wrong.
The honest limits
A simulator is only as good as its assumptions. Ours doesn't know about your soil's caliche layer, a neighbor's shade, a bad batch of seed, or a broken drip line in July. It can't tell us whether a gene change would work in a real tree. That work, if it ever happens, belongs in a licensed partner lab under federal permits.
What the Tree Lab can do is help us ask better questions, and that's where good research starts. We plan to add it to this site as its own page. Every number it gives is a model estimate, not a lab result.
Comments