Our data projectHow long does winter stay below freezing across the NWT? Python on six weather stations
Environment and Climate Change Canada publishes daily weather for stations across the territory. We downloaded the 2025 files for six airport stations, from Inuvik in the north to Fort Smith in the south, on 22 October 2026 and analysed them in Python. Every figure below is our own count from those files.
Our counts from ECCC daily files, calendar year 2025, six NWT airport stations listed north to south (downloaded 22 October 2026). A run is consecutive days whose high stayed below 0| Station | Days with no high or low | Lows of -30 or colder | Longest run of highs below 0 | What ended that run |
|---|
| Inuvik A | 0 | 43 | 77 days, 16 October to 31 December | The end of the file |
| Norman Wells A | 9 | 59 | 36 days, 2 November to 7 December | A missing reading on 8 December |
| Fort Simpson A | 4 | 40 | 52 days, 10 November to 31 December | The end of the file |
| Yellowknife A | 1 | 53 | 67 days, 27 January to 3 April | A day above 0 |
| Hay River A | 3 | 36 | 39 days, 23 November to 31 December | The end of the file |
| Fort Smith A | 8 | 42 | 39 days, 23 November to 31 December | The end of the file |
This is the script that prints the first four columns. It uses only the standard library and runs in a folder holding the six downloaded CSV files:import csv
stations = {
"Inuvik A": "en_climate_daily_NT_2202571_2025_P1D.csv",
"Norman Wells A": "en_climate_daily_NT_2202801_2025_P1D.csv",
"Fort Simpson A": "en_climate_daily_NT_2202104_2025_P1D.csv",
"Yellowknife A": "en_climate_daily_NT_2204101_2025_P1D.csv",
"Hay River A": "en_climate_daily_NT_2202403_2025_P1D.csv",
"Fort Smith A": "en_climate_daily_NT_2202203_2025_P1D.csv",
}
for name, path in stations.items():
missing = 0
below_minus_30 = 0
run = 0
longest = 0
longest_end = ""
with open(path, encoding="utf-8-sig") as weather:
for day in csv.DictReader(weather):
high = day["Max Temp (°C)"]
low = day["Min Temp (°C)"]
if high == "" or low == "":
missing += 1
run = 0
continue
if float(low) <= -30:
below_minus_30 += 1
if float(high) < 0:
run += 1
if run > longest:
longest = run
longest_end = day["Date/Time"]
else:
run = 0
print(name, "| missing", missing, "| lows of -30 or colder", below_minus_30,
"| longest run below 0", longest, "ending", longest_end)
1. A run is a counter that resets
For each station the loop keeps one number, run, that grows by one for every day below freezing and drops to zero on any warmer day. Remembering the largest value it ever reached, and the date when it did, gives the longest stretch. It is the same pattern as a winning streak in a game.
2. One warm afternoon splits a winter
Yellowknife is the station to study closely. January 2025 there holds two runs, not one: 25 days, then a single day on 26 January with a high of 2.1, then the 67-day run. Whether one mild afternoon should break the count is a question about meaning, not code, and students have to defend their choice.
3. Gaps and edges change answers
At Norman Wells the longest run stops at a missing reading on 8 December, and 23 more sub-zero days follow it, so a blank cell, not the weather, decided that number. And four of the six longest runs end on 31 December only because the file does: those winters carried on into 2026.
Yellowknife A (Climate ID 2204101) in more detail, from the same 2025 file; our counts| What we measured | Result |
|---|
| Days missing both a high and a low | 1 (23 April, flagged M) |
| Days whose high stayed below 0 | 158 |
| Coldest low in the year | -40.6 on 5 January |
| Warmest high in the year | 28.9 on 31 July |
| Run still going when the file ends | 51 days, 11 November to 31 December |
What six stations still cannot tell you
Six airports describe six points for one year. They say nothing about the land between them, nothing about whether 2025 was a typical year, and nothing about climate trends, which need many years and careful comparison. Nor does a longer run at Inuvik prove Inuvik was colder: Norman Wells had more days with a low of -30 or colder. A student who writes those limits beside the table has learned more than one who only prints it.
A good extension is to download the same six stations for 2024 and join each pair of files, so the December runs can be measured to their true end and the Norman Wells gap can be tested both ways. Data Source: Environment and Climate Change Canada, used under the ECCC Data Services End-use Licence, which allows reuse including for commercial purposes (historical climate data). Environment and Climate Change Canada has not endorsed this page or our classes.
