Urban heat island intensification from construction albedo change
Large-scale construction strips vegetated surfaces and replaces them with low-albedo impervious cover, measurably raising daytime land surface temperature. Landsat thermal infrared and Sentinel-2 multispectral time-series make that thermal trajectory visible phase by phase.
Sensors
- Landsat 8 and 9 (TIRS): Thermal Infrared Sensor bands 10 and 11 retrieve land surface temperature at 100 m native resolution (resampled to 30 m in standard products). Sixteen-day repeat per satellite; combined Landsat 8/9 constellation gives an eight-day revisit at mid-latitudes. Archive extends to 1984 for Landsat 5, giving multi-decade baseline capability. Cloud cover is the principal operational constraint.
- Sentinel-2 MSI: Ten-metre resolution in visible and near-infrared bands; 20 m in red-edge and shortwave infrared. Five-day revisit (combined Sentinel-2A/2B). No thermal band, but shortwave infrared bands 11 and 12 (1.6 µm and 2.2 µm) support albedo estimation and bare-soil fraction mapping. Used to disaggregate Landsat thermal data spatially via downscaling methods.
- ECOSTRESS (ISS-mounted): Five thermal infrared bands between 8 and 12.5 µm at approximately 70 m resolution. Non-sun-synchronous orbit on the International Space Station produces acquisitions at variable local times, which is useful for capturing diurnal LST variation that polar-orbiting sensors miss. Revisit is irregular, typically three to five days at mid-latitudes, and coverage gaps occur.
- ASTER (Terra): Five thermal infrared bands at 90 m resolution with emissivity separation capability. Particularly useful for surface emissivity mapping, which is required to convert thermal radiance to true surface temperature. Terra has operated since 1999, providing a long archive, though ASTER acquires on demand rather than continuously and the sensor is now in a reduced-operations mode.
What a bare construction site actually does to surface temperature
The urban heat island effect is well-documented, but its intensification during active construction is less often quantified. When demolition or earthworks strip a vegetated or built surface, the replacement bare soil typically has a broadband albedo of 0.10 to 0.20, depending on moisture content and mineralogy. Dry, compacted construction subgrade can fall below 0.10. Grass and light-coloured roofing membranes commonly sit between 0.20 and 0.35. The physics is straightforward: lower albedo means more absorbed shortwave radiation, which drives higher surface temperatures when evapotranspiration is absent.
Published studies using Landsat TIRS data over rapidly urbanising areas in China, India and the Middle East have documented daytime land surface temperature (LST) anomalies of 3 to 8 degrees Celsius above surrounding built fabric during peak construction phases. The anomaly is largest in summer and in arid climates where soil moisture is minimal. Once a site receives its final surface treatment, whether asphalt, concrete, or green roof, the LST signature shifts again, sometimes downward if high-albedo or vegetated finishes are applied, sometimes upward if dark impervious cover replaces bare soil.
The three thermal phases a development zone passes through
Satellite time-series reveal a consistent thermal sequence for large construction projects. Phase one is demolition and clearance: existing structures are removed, vegetation is stripped, and LST typically rises as the cooling effect of trees and grass disappears. Phase two is active earthworks and foundation construction: bare soil dominates, albedo is at its minimum, and daytime LST reaches its peak anomaly relative to the surrounding city. Phase three is surface completion: asphalt roads, concrete slabs, and roofing are applied. LST may remain elevated if dark surfaces dominate, or partially recover if the design includes reflective materials or planted areas.
Tracking this sequence matters because the peak thermal impact on adjacent neighbourhoods occurs during phase two, often lasting months to years on large sites. Urban planners who rely only on pre-construction and post-completion assessments miss the worst of the thermal loading. Sixteen-day Landsat composites, cloud-permitting, are sufficient to characterise the phase transitions on sites larger than roughly two to three hectares.
Albedo retrieval and LST: what the numbers mean and where they break down
Broadband surface albedo is estimated from Sentinel-2 by converting top-of-atmosphere reflectance to surface reflectance using atmospheric correction (Sen2Cor or equivalent), then applying narrowband-to-broadband conversion coefficients published by Liang (2001) and refined in subsequent work. The result is accurate to roughly plus or minus 0.02 to 0.03 over homogeneous surfaces. Mixed pixels at construction boundaries, where bare soil, machinery, stockpiles and standing water coexist within a single 10 or 20 m cell, introduce retrieval uncertainty that should be acknowledged in any delivered product.
LST from Landsat TIRS requires emissivity correction. Emissivity varies across construction materials: concrete is approximately 0.95, dry bare soil 0.90 to 0.95, and metallic roofing as low as 0.20. Using a fixed emissivity assumption across a heterogeneous construction site will bias LST retrievals by one to three degrees Celsius. The NDVI-based emissivity estimation method (Sobrino et al.) is widely applied and is adequate for most urban monitoring, but sites with significant metal or glass coverage need more careful treatment. ASTER's multi-band thermal capability allows explicit emissivity separation, which is the more defensible approach for detailed site-level analysis.
Sentinel-2 and Landsat working together: spatial disaggregation
Landsat TIRS provides the thermal signal but at 100 m native resolution, which is coarse relative to a city block. A large construction site may contain sub-zones with very different thermal behaviour: a freshly poured concrete slab, a waterlogged excavation pit, and a dry stockpile can sit within a single Landsat pixel. Sentinel-2 provides 10 m multispectral data that correlates strongly with surface temperature through indices such as NDVI, NDBI (Normalised Difference Built-up Index) and bare-soil fraction.
Thermal sharpening algorithms, including the widely used TsHARP method and its successors, use this correlation to disaggregate Landsat LST to 10 or 20 m resolution. The result is not a direct thermal measurement at fine scale; it is a statistically derived estimate. Accuracy degrades when the relationship between spectral indices and temperature breaks down locally, as it can over water bodies or highly reflective surfaces. Delivered products should carry explicit uncertainty bounds rather than presenting sharpened LST as equivalent to a direct fine-resolution thermal observation.
What urban planners can do with a thermal time-series
A quarterly LST anomaly map over a development zone, derived from cloud-screened Landsat composites, gives planners a quantitative record of how construction is affecting the thermal environment of adjacent streets and buildings. That record has several practical applications. It can inform requirements for construction-phase mitigation measures, such as temporary ground cover or dust suppression that incidentally raises albedo. It can support post-occupancy evaluation of whether high-albedo roofing or green-roof requirements achieved their intended LST reduction. And it can feed into urban heat vulnerability assessments by identifying which residential areas adjacent to construction zones experienced the largest temperature increases.
The honest caveat is that satellite LST is a surface skin temperature measured at the moment of overpass, typically mid-morning for Landsat. It is not air temperature, and it is not a 24-hour average. ECOSTRESS's variable overpass times add diurnal sampling but do not provide continuous coverage. For planning decisions that require air temperature or indoor thermal comfort estimates, LST data must be combined with urban energy balance models, which carry their own assumptions. Satellize runs this class of analysis on open Landsat and Sentinel-2 archives, with one published analytics engagement in the Pacific Islands context (the Kingdom of Tonga crop-estimation programme) demonstrating the organisation's approach to multi-temporal optical analysis over varied land-cover types.
Archive depth and what it makes possible
Landsat's archive back to 1984 is a genuine analytical asset for cities that have undergone sustained construction booms. A city that expanded its impervious surface fraction from 30 to 60 per cent over three decades will show a corresponding LST trend in the archive, separable from broader climate warming signals with careful analysis. Sentinel-2 data is available from 2015 (Sentinel-2A) and 2017 (Sentinel-2B), which covers the most recent phase of rapid urbanisation in many emerging-market cities.
Cloud cover remains the dominant operational constraint in tropical and monsoon climates. Monthly or seasonal compositing mitigates this but reduces temporal resolution. In persistently cloudy regions, a single usable clear-sky Landsat acquisition per quarter may be the realistic expectation, which is sufficient for phase-level construction tracking but not for week-by-week monitoring. Buyers should calibrate expectations against the cloud climatology of their specific geography before commissioning a monitoring programme.
Typical figures
| Thermal spatial resolution (Landsat TIRS) | 100 m native, 30 m resampled product; sharpened to ~10–20 m using Sentinel-2 spectral disaggregation |
| Multispectral resolution (Sentinel-2) | 10 m (visible/NIR), 20 m (red-edge/SWIR) |
| Revisit cadence | 8 days (combined Landsat 8+9 at mid-latitudes); 5 days (combined Sentinel-2A+2B); ECOSTRESS irregular, ~3–5 days |
| LST retrieval accuracy | ±1–2 °C over homogeneous surfaces with correct emissivity; ±2–4 °C over mixed construction pixels |
| Albedo retrieval accuracy | ±0.02–0.03 broadband over homogeneous surfaces; higher uncertainty over mixed or metallic surfaces |
| Minimum detectable site size | ~2–3 ha for reliable Landsat thermal anomaly; ~0.5 ha for Sentinel-2 albedo change |
| Archive depth | Landsat: 1984–present; Sentinel-2: 2015–present; ASTER: 1999–present (reduced operations) |
| Key spectral bands | Landsat TIRS bands 10–11 (10.6–12.5 µm); Sentinel-2 bands 11–12 (SWIR, 1.6 and 2.2 µm); ASTER TIR bands 10–14 (8–12.5 µm) |
| Cloud constraint | Optical and thermal sensors are cloud-opaque; tropical/monsoon sites may yield fewer than 4 clear acquisitions per quarter |
| Delivery formats | GeoTIFF LST anomaly rasters, GeoPackage polygon layers, PDF quarterly reports, time-series CSV exports |
Analytics Satellize can run
| Construction-phase LST anomaly map | Single-channel or split-window LST retrieval from Landsat TIRS with NDVI-based emissivity correction; cloud-screened seasonal compositing | Quarterly GeoTIFF raster showing LST deviation from a pre-construction baseline, with per-pixel uncertainty estimate |
| Broadband surface albedo time-series | Sentinel-2 surface reflectance retrieval with narrowband-to-broadband conversion (Liang coefficients); change detection against pre-construction reference | Monthly albedo change layer (GeoTIFF) and site-level summary statistics in CSV |
| Impervious surface fraction progression | Sentinel-2 NDBI and bare-soil index classification; supervised or spectral-unmixing approach to fractional cover per pixel | Phase-by-phase GIS polygon layer with impervious fraction values and area statistics |
| Sharpened LST product at 10–20 m | TsHARP or regression-based thermal disaggregation using Sentinel-2 spectral predictors; uncertainty bounds derived from residual analysis | GeoTIFF with sharpened LST and accompanying uncertainty raster; flagged pixels where disaggregation confidence is low |
| Adjacent neighbourhood thermal impact assessment | Spatial buffer analysis around construction perimeter; LST trend extraction for residential zones within 500 m and 1 km buffers | PDF report with time-series plots and ranked list of affected street blocks by peak LST anomaly |
| Post-completion albedo and LST verification | Before/after comparison of Landsat LST and Sentinel-2 albedo against design specifications for high-albedo or green-roof requirements | Compliance summary table and annotated map layer for planning authority sign-off |
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.