North Vancouver data projectHow tall is the tallest street tree? 615 metres, says the first draft
The District of North Vancouver publishes an inventory of its registered street trees, with species, trunk diameter and height. We asked an AI-style script for the tallest one, then ran checks we had written down first. Everything in this section, code and numbers alike, is our own work on the District's file as downloaded on 23 October 2026.
The prompt: “Using this CSV, which standing street tree in North Vancouver is the tallest?” The obvious script filters out removed trees, takes the maximum of HEIGHT_M and prints it. It prints European Hornbeam, 615 m, STREE01036, a tree in Braemar with a 30 cm trunk. That is a data-entry slip in the file, almost certainly a misplaced digit, and a careless agent would report it as a fact about North Vancouver.
The draft answer against our checks (our run on the District file, 23 October 2026; review limits are ours, not the District's)| Step | What it found |
|---|
| Rows in the file | 4,289 trees, of which 3,671 are marked not removed |
| Standing trees with a height value | 3,634 |
| Draft answer (plain maximum) | European Hornbeam, 615 m, trunk 30 cm |
| Check 1: any height over our 40 m review limit | Exactly one row, the 615 m hornbeam |
| Check 2: trunk far too wide for the height (our ratio rule) | 194 standing trees, to be looked at by a person |
| Checked answer after setting the one impossible row aside | 30 m, shared by a Red Alder and a Big Leaf Maple |
The whole exercise, with the limits stated in the code before the maximum is ever printed:# Tallest street tree: a quick draft, then our plausibility checks (teaching example for this page)
import csv
def num(text):
try:
return float(text)
except ValueError:
return None
trees = [r for r in csv.DictReader(open("EnvStreetTree_nonpersonal_subset.csv", encoding="utf-8"))
if r["REMOVED_RE"] == "No" and num(r["HEIGHT_M"]) is not None]
# The draft: what a quick AI-written answer does. Take the maximum and report it.
top = max(trees, key=lambda r: num(r["HEIGHT_M"]))
print("draft answer:", top["COMMONNAME"], top["HEIGHT_M"], "m", top["ASSET_ID"])
# The checks: our own review limits, chosen by us and written down before looking at the answer.
REVIEW_HEIGHT_M = 40 # any street tree taller than this goes to a person
REVIEW_RATIO = 0.6 # trunk diameter in cm per metre of height above which we look again
flagged = [r for r in trees if num(r["HEIGHT_M"]) > REVIEW_HEIGHT_M]
odd_shape = [r for r in trees if num(r["DBH_CM"]) and num(r["HEIGHT_M"]) and
num(r["HEIGHT_M"]) <= REVIEW_HEIGHT_M and num(r["HEIGHT_M"]) * REVIEW_RATIO * 10 < num(r["DBH_CM"])]
print("flagged over", REVIEW_HEIGHT_M, "m:", [(r["ASSET_ID"], r["COMMONNAME"], r["HEIGHT_M"]) for r in flagged])
print("trunk far too wide for the height:", len(odd_shape))
kept = [r for r in trees if r not in flagged]
tallest = max(kept, key=lambda r: num(r["HEIGHT_M"]))
print("checked answer:", tallest["HEIGHT_M"], "m")
Limits before looking
The 40 m limit and the trunk ratio were written into the script first. Choosing limits after seeing the answer is how people talk themselves into believing it. Our limits are teaching choices, and a forester might pick different ones.
Flag, do not delete
The hornbeam row is set aside and reported, not removed from the file. Its real height is unknown to us; the District is the only party who can correct it. An honest answer says that, rather than replacing 615 with a guess.
Ratio checks find more
Range checks catch the absurd. Ratio checks catch the quietly odd: a Vine Maple listed with a 200 cm trunk and only 8 m of height, which also has no ASSET_ID. Either may be a typing slip; neither should feed a chart unread.
Keys matter as much as values
The TAGNUMBER column repeats across many rows, while ASSET_ID is unique where filled. An agent told to look up a tree by its tag number would pick the first match and sound sure of itself. In class, learners find the repeats themselves before they write any lookup tool.
This file is the District of North Vancouver's, so it covers District streets, not the separate City of North Vancouver. Before using it we removed the columns holding staff usernames, house numbers and free-text comments; the licence does not cover personal information, and none is needed for this exercise. Reused under the Open Government Licence - North Vancouver, whose grant extends to commercial purposes (District data page and licence). Contains information licensed under the Open Government Licence - North Vancouver. The District does not endorse this page.
