UrbClim
UrbClim is a fast urban boundary layer climate model (De Ridder et al., 2015, Urban Climate 12, 41-58). It is the first of the two heat models, grouped in the handbook under VITO alongside HiREx.
How it works
UrbClim couples two things: a simplified land-surface scheme, extended to represent urban surface physics, and a three-dimensional atmospheric boundary layer module.
Synoptic-scale meteorological fields are imposed at the lateral and top boundaries. That arrangement is what lets it have things both ways. The large-scale forcing is retained, so the model stays anchored to real regional weather, while it resolves intra-urban variation in air temperature at 100 to 200 m resolution on an hourly time step.
Why its speed matters
UrbClim is roughly two orders of magnitude faster than a full mesoscale model.
That is not a convenience, it is what makes the analysis possible at all. Because it is so much cheaper to run, UrbClim can be run over long periods and across multiple climate scenarios. A model a hundred times slower could simulate a single day; this one covers a whole summer, several times over, under different futures.
Read that alongside HiREx
The trade-off runs the other way for HiREx, which is expensive enough that it runs only for a handful of carefully chosen days. The two models are paired precisely because their strengths are opposite.
Input data
The datasets below build the urban surface and geometry that the two heat models work on. The handbook lists them as a single block for both models; the Used by column shows which one each primarily serves.
Building footprints and heights
| Source | Ordnance Survey MasterMap Topography Layer (Building category), joined to the OS Building Height Attribute (BHA) table |
| Description | Polygon footprints of all buildings in the study area, each carrying the height attributes needed to construct the three-dimensional urban geometry used for shading and sky view factor calculations |
| Used by | HiREx |
Tree footprints
| Source | Bluesky National Tree Map (NTM) Canopy layer |
| Description | Polygons delineating individual tree crowns, each attributed with crown height statistics. These represent the second element of the three-dimensional urban geometry and are the principal control on daytime shading in the HiREx calculation. |
| Used by | HiREx |
Green Areas layer
| Source | OS MasterMap Greenspace; OpenStreetMap landuse layer; Sentinel-2 L2A NDVI (Copernicus Data Space Ecosystem, 10 m) |
| Description | A composite layer, not a source dataset in its own right. Low vegetation and bare areas are both derived from it. |
| Used by | UrbClim |
Low vegetation and bare areas are not obtained from separate source datasets. Both come from this single composite layer, which is constructed in three steps.
The three construction steps, in full
Step 1: Greenspace filtering and garden splitting.
OS MasterMap Greenspace polygons are retained, excluding those classified as inland water or manmade surface.
Private garden polygons (priFunc = 'private garden') frequently comprise several
distinct garden plots connected through narrow passageways such as shared gateways,
and are therefore split into topologically separate parts using a Voronoi-based
partitioning approach:
- Each polygon is eroded by 2 m.
- Where this produces a MultiPolygon, the original (uneroded) geometry is partitioned using a Voronoi diagram seeded from the representative points of each eroded part.
- The intersection of each Voronoi region with the original geometry yields the final parts, with floating-point slivers merged into the largest adjacent part.
- No buffering is applied to the outputs, so the union of all parts exactly reconstructs the original polygon geometry.
Step 2: Vegetated landuse.
A subset of the OSM landuse layer is created retaining only vegetated classes: farmland, forest, scrub, grass and meadow.
These polygons are clipped by the geometric union of all OS MasterMap Greenspace polygons, including inland water and manmade surface features, so that they do not spatially overlap the greenspace layer.
The filtered greenspace layer and the vegetated landuse layer are then merged into a single layer with new sequential integer identifiers.
Step 3: NDVI zonal statistics.
A per-polygon mean NDVI value is computed for every Green Areas polygon from the Sentinel-2 L2A summer 2023 raster (10 m resolution).
Where configured, the NDVI raster is resampled to 1 m using bilinear interpolation before zonal statistics are computed, to improve accuracy for small polygons.
Statistics are computed with the rasterstats library using pixel centres as the
sampling strategy (all_touched = False), and the result is stored as the
ndvi_mean attribute.
The 0.35 threshold
Polygons are then assigned to one of two classes using a threshold of
ndvi_mean = 0.35.
This threshold separates sparse or low-biomass vegetation, such as mown grass, scrub and low-density planting, from surfaces with negligible photosynthetic activity such as bare soil, gravel and dry vegetation.
Low vegetation
| Source | Green Areas layer, polygons with ndvi_mean > 0.35 |
| Description | Vegetated surfaces excluding tree canopy, used to represent evapotranspiration and the lower surface albedo and heat capacity of vegetated ground in the land-surface scheme |
| Used by | UrbClim |
Bare areas
| Source | Green Areas layer, polygons with ndvi_mean ≤ 0.35 |
| Description | Unsealed but effectively unvegetated surfaces: bare soil, gravel, hard-standing within greenspace parcels, and dry or senescent vegetation. Treated as permeable but non-transpiring. |
| Used by | UrbClim |
Bare areas are kept separate from low vegetation because their thermal and moisture behaviour differs materially. Both are permeable, but only one transpires, and transpiration is a significant cooling mechanism.
Sealed surfaces
| Source | OS MasterMap Greenspace; OpenStreetMap landuse layer |
| Description | Impervious surfaces including roads, pavements, car parks and paved squares |
| Used by | UrbClim |
This layer is defined by exclusion rather than by direct mapping. The full study-area bounding box is taken, and from it are removed:
- The OS MasterMap Greenspace areas, retaining polygons classified as manmade surface, which are themselves sealed
- The OSM landuse categories
grass,forest,farmland,meadow,scrub,park,recreation_ground,heath,allotments,cemeteryandorchard
Whatever remains is sealed.
Water bodies
| Source | OS MasterMap Topography Layer, Water category |
| Description | Polygons representing rivers, canals, reservoirs and other standing and flowing water |
| Used by | UrbClim |
Water is treated as a separate surface class because of its high heat capacity and evaporative behaviour.
Generated output data
UrbClim produces a single NetCDF file per scenario, for example
urbclim_manchester_output_ssp585_2050.nc, at roughly 7 GB.
| Contents | Hourly gridded urban meteorology |
| Grid | 301 × 231 cells at 200 m resolution |
| Domain | About 60 × 46 km, centred on Greater Manchester |
| Projection | EPSG:32630 (WGS 84 / UTM 30N) |
| Time axis | Summer half-year at hourly resolution: 2,927 steps, 1 June to 30 September |
| Array shape | (t, y, x) for each variable |
The four variables
| Variable | Meaning | Range | Mean |
|---|---|---|---|
T2M | 2 m air temperature (K) | 275.4 to 307.5 K (2.2 to 34.3 °C) | 290.2 K (17.1 °C) |
QV2M | 2 m specific humidity (kg/kg) | 0.0037 to 0.0168 | 0.0087 |
WS2M | 2 m wind speed (m/s) | 0.01 to 12.26 | 2.06 |
LST | Land surface temperature (K) | 274.5 to 321.5 K (1.4 to 48.4 °C) | 290.7 K (17.5 °C) |
Why LST ranges so much wider than T2M
Land surface temperature peaks at 48.4 °C while air temperature tops out at 34.3 °C. The gap reflects daytime heating of sealed surfaces: asphalt and paving get far hotter than the air above them.
That difference is exactly what HiREx goes on to resolve at street level, and it is why shade matters so much to experienced heat.