Vancouver data lessonA street tree 66 metres wide? Finding outliers in Vancouver’s public trees file
The City of Vancouver publishes its public tree inventory as open data. Statistics 12 students summarise the trunk diameters, set a fence for outliers and find out what one impossible number does to each statistic. Every figure here is ours.
The file is called Public trees. The City describes it as a listing of public trees on boulevards and public trees in parks, in the City of Vancouver, and warns that private trees are not included in the inventory and that it may be several years between updates for some attributes. Our export on 23 October 2026 held 184,268 rows, each with a genus, a species, a height in metres, a trunk diameter in centimetres and one of the city’s 22 local areas.
Five-number summary of trunk diameter, our calculation on the 183,791 trees with a diameter above zero. City of Vancouver Open Data, Public trees, downloaded 23 October 2026; quartiles by Python’s inclusive method| Statistic | Diameter (cm) |
|---|
| Minimum | 0.6 |
| First quartile, Q1 | 10.2 |
| Median | 25.4 |
| Third quartile, Q3 | 45.7 |
| Maximum | 6,673.0 |
| Interquartile range | 35.5 |
| Upper fence, Q3 + 1.5 × IQR | 98.95 |
| Mean | 32.48 |
The maximum is the first thing a careful student questions. A trunk 6,673 cm across is a tree more than 66 metres wide; the record, asset 68485, is a pin oak in Mount Pleasant, and the next largest tree in the file measures 779.0 cm. One extra digit typed at entry would explain it, but the file cannot tell us, so we flag it rather than fix it. What the value does is instructive: drop it, and the mean moves only from 32.48 to 32.45 cm, because one value among 183,791 has little pull on an average. The maximum and the range, though, are ruined by it.
The Python behind the table reads the export, keeps positive diameters, finds the quartiles and the fence, and counts what lies beyond it. Only the standard library is needed:import csv, statistics
with open("public-trees.csv", encoding="utf-8-sig") as f:
rows = list(csv.DictReader(f, delimiter=";"))
sizes = sorted(float(r["diameter_cm"]) for r in rows
if float(r["diameter_cm"]) > 0)
q1, median, q3 = statistics.quantiles(sizes, n=4, method="inclusive")
fence = q3 + 1.5 * (q3 - q1)
print("five-number summary:", sizes[0], q1, median, q3, sizes[-1])
print("upper fence:", round(fence, 2))
print("trees beyond it:", sum(1 for s in sizes if s > fence))
print("mean:", round(statistics.mean(sizes), 2))
Outside the fence is not wrong
Our script finds 3,925 trees above the 98.95 cm fence. Almost all are simply big old trees, which Vancouver has plenty of. The 1.5 IQR rule marks values worth a second look, not values to delete, and students must say that in words.
Zeros are missing data
There are 477 rows with a diameter of 0 and 3 with a height of 0. A tree with no trunk is not a tiny tree; it is a measurement nobody recorded. Leaving zeros in would drag the first quartile down for no good reason.
Heights come in steps
The file holds only 84 distinct heights, and the eight commonest cover 93.7 percent of rows: 4.6, 7.6, 10.7, 13.7, 16.8 and 19.8 m lead. The steps are about 3 m apart, which we read as ten-foot bands converted to metres; that reading is ours, not the City’s.
Correlation, before and after cleaning
Across 183,788 trees with both measurements above zero, our correlation between diameter and height is r = 0.651. Keep only trees inside the fence and it rises to 0.755. For maples alone it is 0.807; for the Prunus genus, cherries and plums, 0.636. Thicker trees tend to be taller, but the height steps blur the pattern, and nothing here says diameter causes height: both grow with age, the lurking variable Statistics 12 warns about.
A follow-up session compares genera. Our medians are 22.9 cm for maples (Acer, the commonest of 171 genus labels, with 41,976 trees), 33.0 cm for Prunus and 36.8 cm for lindens (Tilia). Side-by-side box plots make that comparison in one picture. Contains information licensed under the Open Government Licence - Vancouver. We use the file under those terms, which permit commercial reuse (licence, dataset page). The City of Vancouver does not endorse this page or our classes.
