Methodology
How we counted the nights
Every assumption is declared. Every source is cited. Every limitation is named before you find it. This analysis is descriptive — we measure, we don't model.
Core principle
This analysis makes no causal claims. We don't say hot nights cause the exodus — we say they coincide with it and every resident of Nouakchott knows the connection. We don't predict the future. We don't run regressions. We count.
The approach is the same as La Taxe Canicule: take public data, apply an international standard, and report what comes out. The methodology is transparent, the data is accessible, and anyone can reproduce the analysis from scratch.
"We didn't model anything. We counted the nights the city doesn't sleep."
01 — Temperature data
Hourly air temperature at 2 meters above ground, from the Open-Meteo Historical Weather API. The underlying model is ERA5, the reanalysis product of the European Centre for Medium-Range Weather Forecasts (ECMWF / Copernicus). It's the same model used by national weather agencies worldwide.
Source
Open-Meteo Historical Weather API. License CC BY 4.0. ERA5 model, Copernicus / ECMWF. open-meteo.com
02 — Defining a night
A "night" is the period from 6:00 PM on day N to 6:00 AM on day N+1. This 12-hour window covers sunset (around 7PM in Nouakchott) through past sunrise (around 6:30AM, varying by season).
For each night, we compute the minimum temperature over this window. If that minimum stays above the threshold, the night is classified as "hot." We require at least 8 valid data hours for a night to count.
03 — Hot night threshold
The WHO and WMO define a "tropical night" as one where the minimum temperature doesn't drop below 25°C. That threshold was designed for temperate latitudes. It doesn't reflect Nouakchott's reality, where 25°C at night feels comfortable.
We use 26°C. Still conservative for a Sahelian city, but it corresponds to the sleep disruption point documented in tropical epidemiological literature. The data pipeline initially computes at 25°C (WHO standard). The visualization script recomputes everything at 26°C. Both datasets are available.
Declared assumption
Threshold: nighttime minimum > 26°C.
Reason: WHO 25°C designed for temperate latitudes. 26°C is conservative for Nouakchott. Matches sleep disruption thresholds in tropical literature.
Alternative: 25°C (WHO, would increase counts) or 28°C (stricter, would reduce to ~3% of nights — doesn't reflect lived experience).
04 — Consecutive streaks
Consecutive hot-night streaks are computed by walking through nights chronologically. Each time the minimum temperature crosses above the threshold, a streak begins. It ends when a night drops below. Streaks are sorted by length, descending.
The longest recorded streak is 23 consecutive nights, starting September 2, 2019. For three weeks, Nouakchott's nighttime temperature never dropped below 26°C. Not a single night of relief.
05 — Declared context (not measured)
Two elements of this analysis are not measured by our data but declared as context:
Power outages. We have no access to SOMELEC outage data. No public data exists. We declare that outages are frequent and concentrate in summer based on official communiqués, press reports, and the universal lived experience of Nouakchott residents.
Seasonal exodus. No seasonal mobility data exists for Nouakchott. We declare that the exodus exists — to villages in the interior, abroad, and to legzer on the outskirts — based on local knowledge. It's a phenomenon every resident of Nouakchott knows and can confirm.
Important distinction
Temperature data is measured (Open-Meteo / ERA5). Power outages and the seasonal exodus are declared (local knowledge + press sources). Both types of information are clearly distinguished throughout the site.
06 — Limitations
We name our limitations before you find them.
This analysis does not claim a causal link between hot nights and the exodus. It does not predict future years. It covers only Nouakchott — not other Mauritanian cities. Six years of data (2019–2025) is not enough to establish a long-term climate trend. The 26°C threshold is a declared choice, not a medical absolute. ERA5 data is a reanalysis — not ground station measurements. The spatial resolution covers the whole city without distinguishing neighborhoods.
Power outages are not quantified. The seasonal exodus is not measured. Health impacts are not assessed — we have no hospital data for Nouakchott.
These limitations bound the analysis without invalidating it. We count the nights. We declare the context. We show the data. The rest belongs to those who live in the city.