Given a plant $P$, predict its height $H$ at time $t$ since planting given some environmental variables $E_1, \dots, E_n$. How accurate are the state-of-the-art models for such thing? What data is used to make the prediction?
I've worked on a couple of these for biofuel production. The answer varies widely according to plant P and how well studied plant P is. In terms of what data is used, it includes but is not limited to: temperature, insolation, length of the day, cloud cover, rain, soil composition, soil type, soil density, other local effects like wind or pests, altitude, humidity, and pH. For corn(food) and miscanthus(experimental biofuel crop) the models are quite good if the year is pretty normal and you're in a well-studied parameter space, mostly because corn and grass grows straight up. For trees, non-crops, non-corn crops, etc the models are almost completely useless. It's a question of testing nearly every possible combination of parameters for each plant P, and how much people care about how tall it gets.
Given how much data goes in, it surprises me that even under ideal conditions(well-studied) some of the parameters the model spits out are impossible/wrong. There's a measure of leaf overlap(just as an example) that's especially poorly defined but a rough proxy might be "number of leaves that shade a specific point, on average" and it's regularly measured at <10 and regularly predicted at well north of 50.