Branches with more data per tree entry will fill more baskets than branches with less data per tree entry.

Once you have learned to see these things it is impossible to unsee them. We will never look at a tree the same way again. It was a lightbulb moment! I thought I knew my local woods—I walk there almost every day. But it’s a thrill to see it through fresh eyes, to develop a much deeper understanding.”—Peter Gibbs, Chair BBC Gardeners’ Question Time

I have read several of Tristan Gooley’s fascinating ‘how to read nature’ books. In fact one is open on my desk right now—The Lost Art of Reading Nature’s Signs. All of Gooley’s books are chatty, readable explorations of nature for the armchair enthusiast. How to Read a Tree (2023) examines trees in the same relentless depth, digging into facets that sound obvious (like trunks and leaves) as well as those most of us don’t see but should (like the tree’s shape). Did you know that trees grow bigger on their southern side, or that a strong pale line down the middle of the leaves means there’s water nearby? Me either! Tristan not only noticed, but studied it with an enviable passion for understanding why, carried away by what goes on around him.

Multiple updates of these headers can often be found in files ( treename;1, treename;2 etc, called cycles, see → Opening and inspecting a ROOT file). The entire book was a joy to read and both information dense and effortless/fun. There are moments of profundity throughout. Compare a silver birch, with its narrow shape and light and airy crown, to the broadly spreading crown of an oak.The root node also gives us two more pieces of information ‘ petal width (cm) ≤ 0.8’ and ‘ gini = 0.664’. We will discuss what they mean now… 😃 If the split level is set to 0, there is no splitting: all data members are stored in the same branch. On the plot above we can see the training data (represented by o) on which the Decision Tree has been trained (and has overfitted) and the the testing data (represented by x).