Urban heat island mapping as a property amenity indicator
Land surface temperature retrieved from Landsat and ECOSTRESS thermal bands reveals intra-urban heat variation at block level, giving property analysts a quantitative habitability input that listing data never captures.
Sensors
- Landsat 8/9 TIRS: Two thermal infrared bands (Band 10 at 10.6–11.19 µm, Band 11 at 11.5–12.51 µm) with a native ground sampling distance of 100 m, resampled and delivered at 30 m. Revisit 16 days per satellite; combined Landsat 8 and 9 gives roughly 8-day repeat. The dual-band arrangement enables the split-window algorithm for atmospheric correction. Free archive back to 2013 (Landsat 8) and 2021 (Landsat 9).
- ECOSTRESS (ISS-mounted): Five thermal infrared bands centred near 8.3, 8.8, 9.2, 10.5 and 12.0 µm, at approximately 70 m spatial resolution. Non-sun-synchronous orbit means acquisitions occur at varying local times, which is useful for capturing peak afternoon heat. Revisit is irregular, roughly every few days over a given city, but the variable overpass time adds diurnal context that Landsat's fixed mid-morning pass cannot provide.
- Sentinel-3 SLSTR: Dual-view thermal imaging at 1 km resolution in two TIR channels (10.85 µm and 12.0 µm). Too coarse for block-level property analysis but useful for city-wide baseline climatology, trend monitoring and cross-calibration with higher-resolution sensors. Daily revisit over most latitudes.
- ASTER TIR: Five thermal bands between 8.125 and 11.65 µm at 90 m resolution, with on-demand tasking. The multi-band arrangement supports temperature-emissivity separation more precisely than two-band approaches. Archive extends to 1999. Acquisition is not systematic; scenes must be requested or found in the existing catalogue.
What a thermal image actually measures, and what it does not
Thermal infrared sensors record emitted radiance from surfaces: asphalt, concrete, roofing membrane, soil, leaf canopy. After atmospheric correction, this yields land surface temperature (LST). LST is not the same as the air temperature a resident feels walking down the street. The two correlate, but the relationship depends on wind speed, humidity, surface geometry and the ratio of radiating surface to shaded volume. A canyon of tall buildings can have a very hot roof plane and a relatively tolerable street-level microclimate. Conversely, a low-density suburb with dark roofing and no tree cover can produce high LST and high near-surface air temperature simultaneously.
This distinction matters for property analysis. LST from satellite is a reliable, repeatable, comparable metric across neighbourhoods. It is a proxy for habitability, not a direct measure of it. Used carefully, it identifies which blocks consistently run hotter than the city median, which is a defensible amenity input. Used carelessly, it overstates the lived experience of any specific parcel.
How the split-window algorithm turns two bands into a temperature map
Landsat 8 and 9 TIRS carry two thermal bands specifically to enable split-window retrieval. The method exploits the fact that the atmosphere attenuates radiance differently at 10.9 µm and 12.0 µm. By differencing the brightness temperatures from the two channels and applying coefficients derived from radiative transfer modelling, analysts can remove most of the atmospheric water-vapour signal and recover LST with an accuracy typically cited in the literature at around 1 to 2 K under clear-sky conditions. Emissivity, the surface's efficiency as a thermal emitter, must also be estimated. A common approach derives emissivity from the Normalised Difference Vegetation Index: vegetated surfaces have higher emissivity (closer to 1.0) than bare soil or concrete (around 0.92 to 0.96).
The practical limit is cloud. Thermal retrieval requires a clear line of sight; even thin cirrus degrades results. In humid tropical cities or persistently overcast maritime climates, cloud-free acquisitions may be rare. Building a seasonally representative LST composite requires screening multiple Landsat passes over months, which means the map reflects a temporal average rather than any single day's conditions. That is not necessarily a disadvantage for property analysis, where long-run habitability matters more than yesterday's temperature.
The 100 m floor: what block-level means in practice
Landsat TIRS has a native instantaneous field of view of 100 m, resampled to the 30 m grid shared with the optical bands. The resampling is a geometric convenience, not additional information. In practice, the effective spatial resolution for thermal features is closer to 100 m. That means a single pixel can contain a mix of roof, road, garden and tree canopy in any dense urban setting.
For property intelligence, this resolves to block-level differentiation at best. You can reliably distinguish a park-adjacent residential block from a heat-exposed arterial corridor two blocks away. You cannot distinguish one terraced house from its neighbour, or reliably attribute temperature to a specific rooftop. Analysts who claim parcel-level LST from Landsat are misrepresenting the physics. ECOSTRESS at 70 m is marginally better but faces the same mixed-pixel problem in dense urban fabric. For true parcel-level thermal work, airborne thermal surveys or very-high-resolution commercial thermal sensors are the appropriate tool, and they sit outside the scope of open-constellation analytics.
Turning temperature differentials into an amenity score
The raw output of LST retrieval is a raster of surface temperatures in Kelvin or Celsius. For property analysis, the useful transformation is relative: how many degrees does a given block deviate from the city-wide or district-wide median LST under comparable conditions? A block that consistently runs 4 to 6 K above the urban median in summer composites is a quantifiable negative amenity. One that sits 2 to 3 K below, typically because of tree canopy, water proximity or light-coloured surfaces, carries a measurable thermal advantage.
This differential can be expressed as a z-score or percentile rank and joined to a property dataset by block or postcode. The result is a heat-exposure layer that sits alongside flood risk, noise contour and green-space proximity in a multi-factor habitability index. Correlating LST percentile ranks with transaction price data is a legitimate research exercise; several peer-reviewed studies in journals such as Remote Sensing (MDPI) have found statistically significant associations between urban heat exposure and residential price discounts, though the magnitude varies considerably by city, climate zone and the controls used.
Satellize builds LST composites of this type as GIS layers delivered to property analytics teams, drawing on the same open-constellation pipeline used in the Tonga crop-estimation programme, adapted for urban thermal retrieval.
Archive depth, change over time, and the greening signal
Landsat's continuous archive from 1984 (Landsat 5 TM) through to the present is one of the most underused assets in urban analysis. For thermal work, Landsat 8 from 2013 onward is the practical starting point because TIRS offers the two-band split-window capability. That still gives more than a decade of consistent thermal observations across most cities worldwide.
A ten-year LST trend for a neighbourhood tells a different story from a single-date snapshot. Greening initiatives, albedo-raising interventions such as cool roofs, and urban densification all leave thermal signatures. A district that has cooled by 1.5 K relative to the city median over eight years, coinciding with a documented tree-planting programme, is a more defensible investment narrative than one resting on a single summer image. Equally, a neighbourhood warming faster than the city average, perhaps because of infill development replacing green space, is a risk signal worth flagging in a long-hold property portfolio.
Typical figures
| Native thermal resolution (Landsat 8/9 TIRS) | 100 m (delivered on 30 m grid; effective thermal detail ~100 m) |
| Native thermal resolution (ECOSTRESS) | ~70 m |
| Native thermal resolution (Sentinel-3 SLSTR) | 1 km (city-scale climatology only) |
| Revisit (Landsat 8 + 9 combined) | ~8 days per path; cloud-free composite may require 1–3 months of passes |
| Spectral bands used | TIR Band 10 (10.6–11.19 µm) and Band 11 (11.5–12.51 µm) for split-window LST |
| LST retrieval accuracy (clear sky) | Approximately 1–2 K under clear-sky conditions; degrades with cloud, haze or uncertain emissivity |
| Minimum detectable thermal contrast | Approximately 1 K between spectrally homogeneous blocks; sub-kelvin differences are below reliable detection at 100 m |
| Archive depth (Landsat 8/9 TIRS) | Landsat 8 from February 2013; Landsat 9 from October 2021 |
| Cloud limitation | Thermal retrieval requires clear sky; persistent cloud cover may limit usable scenes to 4–8 per year in wet-tropical cities |
| Delivery formats | GeoTIFF LST raster, block/postcode-aggregated CSV with percentile ranks, QGIS/ArcGIS-compatible layer package |
Analytics Satellize can run
| Seasonal LST composite map | Split-window algorithm applied to Landsat 8/9 TIRS Band 10 and Band 11; cloud-masked median compositing over 3-month window | GeoTIFF raster of mean summer and winter LST at 30 m grid, clipped to study city boundary |
| Urban heat anomaly layer | Z-score normalisation of LST relative to city-wide median for matched acquisition conditions | Block-level GIS polygon layer with heat anomaly score (standard deviations from median) joinable to property datasets |
| Decadal LST trend analysis | Pixel-wise linear regression over annual summer composites from Landsat 8 archive (2013 to present) | Trend-rate raster (K per year) and summary table of warming/cooling by neighbourhood polygon |
| Vegetation-heat correlation report | Pearson correlation and scatter plots of NDVI (Landsat OLI Band 4/5) against LST at block level | PDF report with mapped correlation coefficients and ranked block table; supports green-premium and heat-discount narratives |
| Multi-sensor diurnal heat profile | Fusion of ECOSTRESS variable-overpass acquisitions with Landsat mid-morning baseline to approximate afternoon peak LST | Time-of-day LST comparison table per district; flags neighbourhoods where afternoon heat substantially exceeds morning readings |
| Heat-exposure amenity index | Weighted composite of LST percentile rank, distance to water, tree canopy fraction and impervious surface fraction | Scored polygon layer (0–100 index) at block or postcode level, formatted for integration into property valuation models |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.