Satellite monitoring of visitor infrastructure pressure at heritage sites
Car parks expand, informal paths multiply, and vegetation thins under visitor pressure at heritage sites. Satellite imagery at multiple resolutions can track each of these changes before they become irreversible, feeding directly into UNESCO site-management reporting.
Sensors
- Pléiades-1A/1B: 50 cm pan-sharpened optical imagery; sufficient to resolve individual vendor stalls, parked vehicles and informal path widths above roughly 1.5 m. Tasking revisit at any given site can be as frequent as daily under favourable geometry, though cloud and scheduling contention typically yield useful acquisitions every few days to a few weeks.
- Planet Dove: 3 m multispectral, daily revisit globally. Spatial resolution is too coarse to resolve single structures but excellent for tracking the areal extent of bare-soil expansion and vegetation loss over time. The archive extends to roughly 2016 for most sites, giving a useful baseline.
- Sentinel-2 MSI: 10 m multispectral (visible and near-infrared bands), 5-day revisit at mid-latitudes. The near-infrared and red bands support NDVI time-series analysis of vegetation health. Free and open archive from 2015. Cloud cover is the principal constraint; tropical sites such as Machu Picchu can lose a large fraction of acquisitions in the wet season.
- VIIRS Day/Night Band (DNB): Approximately 750 m spatial resolution, nightly global coverage. Not useful for infrastructure mapping, but the DNB can detect anomalous radiance spikes on peak-visit nights at sites where vendors, generators or event lighting are present. Quantitative nighttime light data are available via NOAA STAR and the VIIRS Nightfire product from Colorado School of Mines.
What very-high-resolution imagery actually resolves
At 50 cm, Pléiades imagery resolves a parked car cleanly. A vendor stall of 2 m width sits at the detection threshold rather than the mapping threshold: you can see something is there, but you cannot reliably classify its function from spectral signature alone. An informal path worn to bare soil at 1 m width is detectable; a lightly trodden grass route at the same width is not. These are not caveats to hide; they define the analytic strategy.
The practical workflow for sites such as Petra, where published remote-sensing studies have mapped the expansion of tourist infrastructure in the Wadi Musa buffer zone, is to use Pléiades or equivalent very-high-resolution imagery for change detection of discrete objects: car park boundaries, paved road extensions, permanent structure footprints. Planet Dove then provides the temporal density to catch when those changes happened, even if it cannot characterise them in detail. The two sensors are complementary, not redundant.
Vegetation loss is the most honest signal
NDVI computed from Sentinel-2 red and near-infrared bands is, in practice, the most tractable metric for visitor pressure. Trampling compacts soil, reduces infiltration and kills ground-cover vegetation. The spectral consequence is a measurable NDVI decline that accumulates over seasons. A single busy season may not produce a statistically significant change against natural phenological variation; two or three consecutive seasons of decline along known visitor corridors constitute a defensible trend.
The critical methodological step is separating seasonal variation from permanent loss. At Machu Picchu, which sits in a cloud-forest environment, NDVI follows a strong wet-dry cycle. Comparing the same calendar month across years, rather than adjacent months, removes most of the phenological noise. Published work using Landsat and Sentinel-2 time-series at high-altitude Andean sites has demonstrated that this approach can detect statistically significant vegetation loss at spatial scales of tens of metres, which is the scale at which informal paths and rest areas form.
One honest limit: NDVI cannot distinguish trampling from drought stress, fungal damage or pest outbreaks without ancillary data. A site-management team that sees a declining NDVI trend needs ground-truth confirmation before attributing it to visitor pressure. Satellite analysis flags the where and when; field ecologists confirm the why.
Temporal cadence and the seasonal trap
Heritage site managers frequently present satellite evidence to UNESCO periodic-reporting processes on an annual cycle. Annual snapshots are not sufficient to separate permanent infrastructure additions from seasonal market stalls or temporary event structures. The minimum useful cadence for distinguishing permanent from ephemeral change is monthly composites across at least two full annual cycles, with additional high-resolution tasking around known peak-visit periods.
For sites in the northern hemisphere, peak visitor months typically run April through October. A Pléiades acquisition in August shows maximum infrastructure; one in February shows the minimum. The difference between those two images is not infrastructure creep; it is seasonality. The difference between two consecutive August images, if statistically significant, is a candidate for permanent change. Planet's daily archive makes it possible to build these seasonal profiles without relying on a single cloud-free acquisition at the right time of year.
Nighttime light as a crowd-density proxy
VIIRS DNB data are not a visitor-count instrument. At 750 m resolution, a heritage site and its surrounding town are often a single bright blob. What the DNB can detect is a statistically anomalous increase in radiance on specific nights relative to the baseline distribution for that pixel, month and day of week. At sites where large-scale evening events, illuminated festivals or extended vendor operations occur on predictable dates, this anomaly is real and reproducible.
The VIIRS Nightfire product, maintained by the Colorado School of Mines Earth Observation Group and distributed via eogdata.mines.edu, separates thermal combustion sources from reflected light, which matters when open fires or flares are present. For most heritage tourism contexts the standard DNB radiance product from NOAA STAR is adequate. The output is a time-series of radiance anomalies, not visitor counts. Framing it as anything more precise than that is not supported by the physics.
Feeding outputs into UNESCO reporting
UNESCO's Operational Guidelines for the Implementation of the World Heritage Convention require States Parties to submit periodic reports on the state of conservation of inscribed sites, including the condition of Outstanding Universal Value attributes and the effectiveness of management measures. Satellite-derived change maps are not yet a required element of these reports, but several States Parties have submitted them as supporting evidence, and the UNESCO World Heritage Centre has acknowledged remote sensing as a legitimate monitoring tool in its own technical guidance documents.
The practical deliverable for a site-management team is a GIS layer set: a polygon of the site buffer zone, annual or biannual raster difference layers showing bare-soil expansion and NDVI change, and a vector layer of newly detected structures or path networks with acquisition dates. These layers can be imported directly into the site's existing GIS environment and annotated for the periodic report. Satellize has built comparable change-detection pipelines for agricultural monitoring, including the Kingdom of Tonga crop-estimation programme, and the underlying methodology transfers directly to heritage buffer-zone analysis.
One structural limit worth stating plainly: satellite monitoring identifies change but does not enforce management responses. A car park that expands by 0.4 hectares over three years will appear clearly in the data. Whether the responsible authority acts on that information depends on governance, not on sensor resolution.
Typical figures
| Best spatial resolution (optical) | 50 cm (Pléiades-1, pan-sharpened) |
| Medium-resolution multispectral | 3 m (Planet Dove), 10 m (Sentinel-2 MSI) |
| Nighttime light resolution | ~750 m (VIIRS DNB) |
| Revisit cadence | Daily (Planet Dove, VIIRS); 5 days at mid-latitudes (Sentinel-2); tasked on demand (Pléiades) |
| Minimum detectable path width | ~1 m bare-soil path in Pléiades; ~5–10 m in Planet Dove |
| Minimum detectable structure footprint | ~4 m² in Pléiades; ~25 m² in Planet Dove (areal threshold, not linear) |
| NDVI change detection sensitivity | Statistically significant trend detectable over 2–3 annual cycles at 10 m pixel scale (Sentinel-2) |
| Archive depth | Sentinel-2 from 2015; Planet Dove from ~2016; Pléiades tasked archive from 2012; VIIRS DNB from 2012 |
| Cloud-cover constraint | Severe at tropical montane sites (e.g. Machu Picchu); monthly compositing and multi-year averaging required |
| Delivery formats | GeoTIFF raster layers, GeoJSON/Shapefile vector change polygons, PDF summary report for UNESCO annexes |
Analytics Satellize can run
| Bare-soil expansion map | Supervised land-cover classification and binary change detection on Planet Dove or Pléiades multitemporal stacks; training samples from known car park and path polygons | Annual GeoTIFF difference raster and polygon layer with area statistics, suitable for GIS import |
| NDVI trend surface | Pixel-wise linear regression on Sentinel-2 NDVI time-series, same-month-across-years to remove phenological noise; Mann-Kendall trend test for statistical significance | Raster of trend slope and significance per pixel, with summary statistics for defined buffer-zone polygons |
| Infrastructure footprint inventory | Object-based image analysis on Pléiades 50 cm imagery; manual verification of detected polygons against prior-year baseline | Dated vector layer of new or expanded structures with area, perimeter and first-detection date |
| Informal path network map | Spectral unmixing and linear feature extraction on Pléiades; bare-soil endmember fraction used to trace desire-path networks | Vector polyline layer of detected paths with estimated width class and date of first detection |
| Nighttime radiance anomaly time-series | Z-score analysis of VIIRS DNB monthly composites against multi-year pixel baseline; flagging of nights exceeding two standard deviations | Time-series chart of radiance anomalies with flagged peak-event dates, exported as CSV and PDF |
| UNESCO periodic-report change summary | Aggregation of all above layers into a structured narrative with before/after image pairs, area statistics and trend charts formatted to UNESCO reporting conventions | PDF annex and accompanying GIS package ready for submission with the State Party's periodic report |
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.