HiREx
HiREx, short for High Resolution heat Exposure, is a downscaling module. It takes the hourly UrbClim meteorology as input and resolves heat exposure at metre scale (Souverijns et al., 2023).
Where UrbClim answers how hot is this neighbourhood, HiREx answers how hot is this spot on this pavement, which is the question that actually matters to a person waiting for a bus.
How it works
For each time step, HiREx performs three calculations:
- Solar position. It calculates the solar zenith and azimuth angle, establishing exactly where the sun is in the sky.
- Shadows. From that solar position and the three-dimensional urban geometry, it derives the shadow cast by individual buildings and tree crowns.
- Sky view factor. It computes the sky view factor for every grid cell: how much of the open sky that point can actually see.
From these, together with the UrbClim meteorology, it derives high-resolution heat-stress information at metre scale.
Why sky view factor matters
A narrow street between tall buildings sees little sky. That cuts direct sun during the day, but it also traps outgoing radiation at night, which is a large part of why dense urban areas stay warm after dark.
Sky view factor captures both effects in a single number per grid cell.
The geometry these calculations depend on comes from two datasets documented on the UrbClim page: building footprints and heights, and tree footprints. Tree crowns are the principal control on daytime shading.
Generated output data
HiREx produces one NetCDF file at 1 m resolution for each combination of intervention scenario, time horizon and climate scenario, covering one hot day and one typical day in the summer.
Each file holds a single variable:
| Variable | Meaning |
|---|---|
Tg | Black globe temperature (°C) |
Tg is the radiant-load component of WBGT. It resolves the effect of building and
tree shade on what a person standing at that point actually experiences, rather than
what a weather station in an open field would record.
Choosing representative days
Because HiREx runs at 1 m resolution are computationally expensive, each is executed for a single diurnal cycle rather than a continuous season.
Two days are selected per scenario and time horizon:
| Day | Represents |
|---|---|
| Typical day | Median summer conditions |
| Hot day | The upper tail of the heat-stress distribution |
Selecting them well matters, because these two days stand in for an entire season.
The three metrics
Selection uses the UrbClim T2M, QV2M and WS2M fields over the summer season.
For each day, three metrics are derived:
| Metric | How it is derived |
|---|---|
| Daily maximum air temperature | Directly from T2M |
| Daily mean relative humidity | From specific humidity and temperature |
| Apparent temperature | Combines temperature, humidity and wind into a single measure of perceived heat |
Reducing each metric to one number per day
Each metric is reduced to a single daily value by taking the 90th percentile across urban grid cells, rather than the domain mean.
Why the 90th percentile, not the mean
The UrbClim domain is about 60 × 46 km, so it contains a great deal of countryside around the city. A domain mean would let cool rural surroundings dilute the signal, and the chosen "hot day" would not be the day that was hottest where people actually are.
The 90th percentile makes the selection reflect the warmer, more built-up parts of the domain.
Picking the hot day
The hot day is chosen in two stages:
- Screen. The hottest 15% of days are retained on the basis of daily maximum temperature.
- Rank. From this subset, the day with the highest combined score is selected, weighted primarily on apparent temperature, then maximum temperature, then relative humidity.
Screening on temperature before ranking on apparent temperature is deliberate. It ensures the selected day is genuinely hot rather than merely humid, since a warm, very humid day can score highly on apparent temperature without being an extreme heat event.
Picking the typical day
The typical day is the day whose temperature, humidity and apparent temperature lie closest to the seasonal median across all three metrics.