Night-light luminosity as a property market economic proxy
VIIRS Day/Night Band radiance time-series can proxy sub-national economic activity where GDP statistics are absent or delayed, giving property analysts a monthly signal tied to real human behaviour on the ground.
Sensors
- VIIRS Day/Night Band (Suomi NPP): Panchromatic low-light sensor, 0.5–0.9 µm, 750 m native pixel at nadir. Daily global overpass near 01:30 local time. Monthly cloud-free composites published by NOAA/NGDC remove lunar and ephemeral artefacts. Operational since 2012, providing over a decade of archive.
- VIIRS Day/Night Band (NOAA-20): Identical DNB specification to Suomi NPP, launched 2017. The two satellites overpass at slightly different times, improving cloud-gap fill when compositing. Together they extend the usable archive and reduce single-night contamination.
- DMSP-OLS archive: Predecessor sensor, 1992–2013. Coarser resolution (approximately 2.7 km), 6-bit radiometric depth, and a hard saturation cap around 63 digital numbers that makes it unsuitable for dense urban cores. Useful for long-period trend baselines where VIIRS does not reach.
- VIIRS Nightfire (EOGDATA/Mines): A derived product from the VIIRS shortwave-infrared and DNB channels that separates combustion sources (gas flares, industrial furnaces) from ambient artificial light. Relevant when industrial activity in a study area would otherwise inflate apparent settlement luminosity.
What upwelling light actually measures
Artificial light escaping upward from a surface is a by-product of human activity: street lamps, commercial signage, industrial operations, residential windows, vehicle traffic. The VIIRS Day/Night Band detects radiance in the 0.5–0.9 µm panchromatic window with a noise-equivalent delta radiance of roughly 2 × 10⁻¹¹ W cm⁻² sr⁻¹, sensitive enough to resolve individual fishing vessels on open water. On land, that sensitivity translates to a measurable signal from a small market town or a rural petrol station.
The economic logic is straightforward. Wealthier households and businesses consume more electricity, keep longer commercial hours, and generate more vehicle movement at night. Published academic work, including studies using DMSP-OLS and later VIIRS, has found statistically significant correlations between night-light radiance and GDP at sub-national scales, particularly in countries where official statistics are infrequent or unreliable. For property analysts, the relevant signal is not absolute luminosity but its rate of change over time: a district brightening faster than its neighbours is accumulating economic activity, and land values tend to follow.
The saturation problem and the blooming artefact
VIIRS DNB does not saturate in the same way DMSP-OLS did, but it has its own upper-end behaviour. At very high radiance levels, typical of dense city centres in high-income countries, the sensor's automatic gain control compresses dynamic range, reducing the ability to distinguish between a moderately lit commercial district and a brightly lit one. Practical consequence: VIIRS is most discriminating in the mid-range luminosity typical of secondary cities, peri-urban growth zones, and emerging markets. Those happen to be exactly the geographies where property analysts most need a proxy for economic activity.
Blooming is a separate problem. Point sources of intense light scatter into adjacent pixels through the sensor's point-spread function, making a bright commercial node appear to illuminate surrounding dark areas. A shopping centre can appear to 'light up' a kilometre of farmland that has no actual development. Analysts must apply a spatial filter or use the monthly composite's stable-lights mask to avoid misreading blooming as genuine economic expansion at the urban fringe. This is not an exotic correction; NOAA's monthly composite processing already applies some filtering, but local validation against ground truth remains advisable.
Monthly composites versus single-night imagery
A single VIIRS DNB overpass is nearly useless for economic trend work. Cloud cover blocks upwelling light entirely, producing a false zero. Moonlight, particularly near full moon, adds a background radiance that inflates apparent luminosity. Transient events such as fires, festivals, or temporary industrial flares create spikes that have nothing to do with the underlying economic state of a district.
NOAA's monthly cloud-free composites, produced by the Earth Observation Group at Colorado School of Mines, address most of these issues by selecting the lowest-radiance cloud-free observation in each pixel over a calendar month and applying lunar illumination masking. The resulting product is stable enough for trend analysis at quarterly or annual cadence. Annual composites smooth residual noise further. The honest caveat is that monthly composites have a latency of four to six weeks after month-end, so they are not a real-time signal. They are a retrospective economic indicator, comparable in spirit to a lagging macroeconomic index.
Building a property market proxy from radiance time-series
The analytical workflow starts with a polygon layer of the property market zones of interest: administrative boundaries, catchment areas, or custom investment geographies. For each zone, mean or median DNB radiance is extracted from each monthly composite over the chosen archive period, producing a time-series going back to 2012 on VIIRS and to 1992 on DMSP-OLS (with a calibration step to bridge the two sensors). That time-series is then decomposed into trend, seasonal, and residual components. Seasonal variation in night-light is real: retail and hospitality zones brighten in the pre-Christmas period in temperate countries; agricultural processing areas spike at harvest.
The proxy gains interpretive power when compared across zones rather than read in isolation. A district whose luminosity grew 15 per cent over three years while an adjacent district was flat is a meaningful signal, even if the absolute radiance values carry uncertainty. Property analysts typically combine this signal with daytime optical indicators and any available transaction data, using the night-light trend as a leading or confirming indicator rather than a standalone valuation input. In genuinely data-sparse regions, such as secondary cities in sub-Saharan Africa or Pacific island economies, it may be the only consistent time-series available at district level.
Honest limits before you commit to a methodology
Night-light radiance cannot distinguish the source of luminosity. A new data centre, a prison, and a thriving retail strip may produce similar radiance signatures. Industrial facilities that operate at night, including gas flares and port lighting, can dominate a pixel and mask the residential or commercial signal an analyst actually wants. VIIRS Nightfire can help isolate combustion sources, but separating industrial from commercial lighting in mixed-use zones requires additional layers.
At 750 m native resolution, VIIRS cannot resolve individual parcels or even city blocks in most urban environments. It is a district-level or corridor-level tool. It will not tell you which side of a road is brightening faster. Cloud-persistent regions, including tropical coasts and monsoon-affected markets, will have months with no usable composite pixels even after cloud-free compositing, creating gaps in the time-series that require interpolation. And in any market where outdoor lighting is heavily subsidised or subject to curfew, the link between luminosity and economic activity weakens considerably.
Satellize runs VIIRS DNB time-series extraction and trend decomposition as a standard analytic layer, applicable to any geography a client defines, drawing on the same open archive used in its Tonga crop-estimation programme for sub-national activity baselining. The output is a zone-level radiance index delivered as a GIS layer and a structured data feed, ready to enter a valuation model.
Typical figures
| Native spatial resolution | 750 m at nadir (VIIRS DNB); ~2.7 km (DMSP-OLS archive) |
| Revisit cadence | Daily overpass per satellite; two satellites (Suomi NPP + NOAA-20) provide near-daily redundancy |
| Monthly composite latency | 4–6 weeks after month-end (NOAA/EOG product) |
| Spectral band | 0.5–0.9 µm panchromatic (DNB); shortwave-infrared channels additionally used in Nightfire product |
| Radiometric sensitivity | Noise-equivalent delta radiance ~2 × 10⁻¹¹ W cm⁻² sr⁻¹ (VIIRS DNB) |
| Saturation behaviour | Dynamic range compressed at very high urban radiance; most discriminating in mid-luminosity (secondary city) range |
| Archive depth | VIIRS: 2012–present; DMSP-OLS: 1992–2013 (requires inter-calibration to bridge sensors) |
| Global coverage | Full global daily coverage; polar regions subject to solar contamination in summer months |
| Minimum detectable feature | Small settlements and isolated industrial sites detectable; individual parcels below resolution floor |
| Delivery formats | GeoTIFF monthly composites (open source); zone-level radiance index as CSV, GeoJSON, or GIS layer |
Analytics Satellize can run
| Zone-level radiance index time-series | Zonal statistics extraction from NOAA monthly cloud-free composites; seasonal decomposition (STL or classical additive model) | Structured CSV or GeoJSON time-series per defined property market zone, updated monthly |
| Economic activity trend score | Linear or Mann-Kendall trend test on annual radiance means; percentile ranking across peer zones | District-level trend score layer (GIS polygon) with confidence intervals, delivered as a quarterly report |
| Blooming-corrected lit-area extent | Spatial filtering and stable-lights thresholding to remove point-source scatter; comparison of lit-area perimeter across time steps | Annual lit-area boundary shapefile per zone, showing expansion or contraction of economically active footprint |
| Industrial light source separation | VIIRS Nightfire combustion-source mask applied before zonal extraction to isolate residential and commercial luminosity | Adjusted radiance index excluding flagged industrial pixels, with a pixel-count quality flag per zone |
| Cross-sensor long-period baseline | DMSP-OLS to VIIRS inter-calibration using published regression coefficients; spliced time-series from 1992 | 30-year radiance trend dataset for markets where historical context is required for valuation modelling |
| Peer-market luminosity benchmarking | Comparative zonal statistics across a defined set of analogous markets; normalised by zone area | Benchmarking table and ranked chart showing target market position relative to comparators, in a PDF briefing |
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.