Online Python classes in Remscheid

import sqlite3

db = sqlite3.connect("remscheid_trees.gpkg")      # a GeoPackage is an SQLite file
rows = db.execute("SELECT height_max FROM Baumstandorte_Mittelpunkte_Remscheid "
                  "WHERE inside = 1")
heights = sorted(h for (h,) in rows)
n = len(heights)
print("Trees inside the city:", n)

def percentile(p):                     # the value p percent of the way along the sorted list
    return round(heights[round(p / 100 * (n - 1))], 1)

for p in (0, 25, 50, 75, 90, 99, 100):
    print(f"{p:>3}th percentile:", percentile(p), "m")
print("Mean:", round(sum(heights) / n, 1), "m")

Sorting once does all the work. After that, the 50th percentile, the median, is simply the value halfway along the list: 21.3 metres. A quarter of the trees are below 14.7 metres and a quarter above 25.9, so the middle half of all the trees the method found lies between those two heights. The mean, 20.2 metres, is a little lower than the median, because the list has a long tail of short trees and young growth below it.

low, high = percentile(25), percentile(75)
print("Middle half of all trees:", low, "to", high, "m")
for limit in (40, 50, 60):
    taller = n - sum(1 for h in heights if h < limit)
    print(f"Trees of {limit} m or more:", taller)

Data: Stadt Wuppertal, Einzelbaum Stammpositionen aus Luftbildern (2025) Remscheid und Solingen, Remscheid GeoPackage, CC BY 4.0, column height_max, rows with inside = 1; downloaded 3 November 2026. The percentiles, the counts and the reading of them are ours.

Now the part that makes learners sit up. Ninety-nine trees in a hundred are 34.2 metres or shorter, yet the tallest row says 108.6 metres, and 19 rows claim 60 metres or more. No tree in the Bergisches Land is anywhere near that tall. Those rows are almost certainly something else the height model saw from the air, a mast or a chimney or a bad patch of data, that slipped through the vegetation mask. The percentiles barely notice them; the maximum is made of them. That is the second half of the lesson: the far end of a sorted list is where data errors gather, so look there before you quote a record.

The crown templates tell a quieter story. Gaussian and spherical shapes, which the publisher assigns to broadleaf trees, account for 210,521 of the trees found; parabolic, hyperbolic and conical shapes, used for conifers, account for 189,009; 5,955 rows have no shape at all. Those are our sums from a second query, and the template is the method's own guess, not a botanist's.

1. Ages 9 to 12. Measure the heights of ten family members, write them on cards, sort the cards and find the middle one. Then add a pretend giant of 3 metres. Which moves more, the middle card or the average?
2. Teenagers. Run both versions, then work out the heights for the 10th and 95th percentiles and draw a box plot by hand from your numbers.
3. Adults. Choose your own cut-off for "probably not a tree", count what it removes, and say in one sentence why you chose it.
4. Going further. Compare the median height of broadleaf and conifer templates. Do the two groups overlap more than you expected?