Mine tailings dust emission mapping for silicosis exposure burden
Dry mine tailings and haul roads emit respirable crystalline silica that causes irreversible silicosis. Sentinel-2 spectral indices, MODIS aerosol optical depth and ERA5 wind fields can map emission potential and downwind exposure corridors around active and legacy sites.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands, 20 m in shortwave infrared; 5-day revisit at the equator with two satellites. Used to compute bare-soil indices (BSI), iron-oxide ratios (band 4 / band 2 or band 11 / band 8A) and tailings-extent mapping. Cannot penetrate cloud; dry-season compositing is essential for arid tailings sites.
- Landsat 8/9 OLI: 30 m multispectral resolution, 16-day revisit per satellite (8-day combined). Provides a continuous archive back to 1972 via earlier Landsat missions, enabling multi-decade tailings-extent change detection. Band 6 (SWIR-1) and band 7 (SWIR-2) are particularly sensitive to moisture content and mineral composition of tailings surfaces.
- MODIS MAIAC AOD: MODIS Multi-Angle Implementation of Atmospheric Correction (MAIAC) delivers aerosol optical depth at 1 km resolution with daily global coverage. Anomaly detection against a rolling baseline can flag elevated dust loading downwind of tailings. Uncertainty is higher over bright desert surfaces; retrievals over very light-coloured tailings can be unreliable without ancillary surface-reflectance constraints.
- ERA5 reanalysis wind fields: ECMWF ERA5 provides hourly 10 m and 100 m wind speed and direction at approximately 31 km horizontal resolution, with an archive from 1940. Used to construct downwind exposure corridors and to weight emission-potential scores by prevailing transport direction. Spatial resolution is coarse relative to local topographic channelling around pit walls and waste dumps.
- Sentinel-1 SAR C-band: 10 m resolution, 6-day revisit (two satellites), cloud-independent. Backscatter change between passes can indicate surface disturbance on tailings facilities, flagging newly exposed dry material even when optical sensors are obscured. Not a direct dust measurement, but a useful proxy for surface activity.
What the tailings surface actually tells a sensor
A mine tailings storage facility is not a uniform grey slab. Its spectral signature varies with moisture content, particle size, mineralogy and crust development. Sentinel-2 shortwave-infrared bands (band 11 at 1610 nm, band 12 at 2190 nm) are sensitive to surface moisture: wet tailings absorb more strongly, dry tailings reflect more. That contrast is operationally useful because dust emission potential rises sharply as surface moisture falls below roughly 2 to 4 percent by mass, a threshold established in wind-tunnel studies of mine waste.
Iron-oxide content also matters. Tailings from gold, copper and iron-ore operations often carry elevated iron-oxide concentrations that produce a characteristic red-to-ochre tone in band 4 (red) relative to band 2 (blue). The band 4 / band 2 ratio, or the more formalised ferric-iron index using band 4 and band 8A, helps discriminate tailings from surrounding soils and road surfaces. Silica itself is spectrally bland in the visible and near-infrared, so the satellite cannot directly measure crystalline silica content. What it can do is map the spatial extent and dryness of the source material, which is the first input to any emission model.
From spectral index to emission-potential score
The standard analytical chain starts with a cloud-free Sentinel-2 composite, typically a dry-season median over 30 to 90 days. Bare-soil index (BSI) and modified soil-adjusted vegetation index (MSAVI) separate vegetated from unvegetated surfaces. Tailings polygons are then extracted by combining BSI thresholds with the iron-oxide ratio and, where available, cross-referencing against mine-permit boundaries or Landsat change-detection layers showing historical land-cover conversion.
Each tailings polygon receives a dust-emission potential score based on surface dryness (SWIR reflectance proxy for moisture), estimated fetch length (the unobstructed upwind distance within the polygon), and the absence of surface crust indicators. ERA5 wind roses, computed over a multi-year period at the nearest grid point, are then used to rotate the emission score into a directional exposure corridor. The result is a probability-weighted downwind footprint: not a precise concentration field, but a ranked map of which residential or agricultural areas are most likely to receive elevated dust loading on days when wind speed exceeds the erosion threshold for fine particles, generally around 5 to 8 m/s for uncrusted tailings.
MODIS MAIAC AOD anomalies add a temporal validation layer. On days when AOD at 550 nm is elevated by more than one standard deviation above the site-specific baseline and ERA5 wind vectors point from the tailings toward the anomaly, the event is flagged as a probable dust episode. This is circumstantial rather than definitive: MODIS cannot distinguish silica dust from other aerosol types, and the 1 km footprint blurs point sources in complex terrain.
Legacy sites are the harder problem
Active mines have operators, permits and, in many jurisdictions, dust-suppression obligations. Legacy tailings facilities, often abandoned decades ago, have none of those controls and frequently lack even accurate spatial records. The Landsat archive, which extends back to 1972 with Landsat 1 MSS data, allows analysts to date when a tailings facility was last actively managed, how much its surface area has changed, and whether vegetation has colonised and stabilised portions of it.
A facility that was wet-processed and then abandoned in a semi-arid climate will typically dry to a dust-emitting surface within one to three dry seasons. Landsat time series can identify that transition. The practical output for a public-health agency is a ranked inventory of legacy sites by emission potential and proximity to populated areas, something that is very difficult to compile from ground surveys alone when sites number in the hundreds across a mining district.
What the method cannot do, and why that matters
Crystalline silica content in airborne dust cannot be determined from space. The satellite contribution stops at mapping source extent, dryness and probable transport direction. Actual personal exposure requires ground-level gravimetric sampling or optical particle counters, and the proportion of respirable crystalline silica in tailings dust varies widely by ore type and processing method. Some tailings carry less than 1 percent free silica; others, particularly from hard-rock gold operations, can exceed 20 percent.
Cloud cover is a persistent constraint. In tropical mining regions with long wet seasons, optical compositing windows may be only 60 to 90 days per year, and the driest, most emission-prone period often coincides with the clearest skies. That alignment helps, but a single severe dust event during an otherwise cloudy month will be missed by MODIS AOD if the event occurs under cloud.
Spatial resolution also limits haul-road analysis. Haul roads are a major dust source in active mines, but at 10 m Sentinel-2 resolution a 20 m wide road is only two pixels across, making spectral characterisation noisy. Sentinel-1 SAR backscatter can detect surface disturbance more reliably on narrow linear features, but still cannot quantify emission rates.
Connecting the map to a health assessment
The exposure-corridor layer becomes actionable when overlaid with census population grids (WorldPop or similar), school and clinic locations, and land-use data showing where people spend extended outdoor time. A simple proximity-weighted exposure index, combining emission potential score, modelled transport probability and receptor population density, produces a ranked list of communities for priority ground-level monitoring. That ranking does not replace epidemiological study, but it tells a health ministry where to deploy its limited air-quality monitors first.
Occupational exposure inside the mine boundary is a separate question governed by industrial hygiene standards rather than community health frameworks, and satellite data has less to add there. The clearest satellite contribution is to the community burden: the villages, schools and smallholder farms that sit downwind of tailings facilities and have no regulatory standing to demand dust suppression. Satellize has applied comparable spectral-index and wind-field workflows to agricultural land-cover analysis, including the Kingdom of Tonga crop-estimation programme, and the same compositing and anomaly-detection logic transfers directly to dust-source characterisation.
A completed assessment typically delivers three outputs: a GIS layer of ranked emission-potential polygons, an annual time series of MODIS AOD anomaly events attributed to each facility, and a downwind population exposure report. Those three products together give a health authority the spatial and temporal evidence base to prioritise remediation, demand monitoring and, where legal frameworks allow, pursue liability.
Typical figures
| Tailings mapping spatial resolution | 10 m (Sentinel-2 visible/NIR), 20 m (Sentinel-2 SWIR), 30 m (Landsat 8/9 OLI) |
| Optical revisit (cloud-free composite window) | 5 days (Sentinel-2 two-satellite), 8 days (Landsat 8+9 combined); dry-season composites typically 30–90 days |
| AOD product resolution | 1 km (MODIS MAIAC, daily global) |
| Wind-field resolution | ~31 km horizontal, hourly (ERA5 reanalysis) |
| Spectral bands used | Sentinel-2 B2 (490 nm), B4 (665 nm), B8A (865 nm), B11 (1610 nm), B12 (2190 nm); Landsat OLI B6, B7 (SWIR) |
| Minimum mappable tailings polygon | ~0.1 ha at 10 m resolution; smaller features below reliable detection |
| AOD anomaly detection threshold | >1 standard deviation above site baseline at 550 nm (MAIAC); coarser aerosol types not discriminated |
| Archive depth | Sentinel-2 from 2015; Landsat from 1972; ERA5 from 1940; MODIS MAIAC from 2000 |
| Crystalline silica detection | Not possible from orbit; satellite maps source extent and dryness only |
| Delivery formats | GeoTIFF, GeoPackage, Cloud-Optimised GeoTIFF (COG), PDF report, CSV anomaly event log |
Analytics Satellize can run
| Tailings extent and dryness map | Sentinel-2 dry-season median composite; BSI, MSAVI and iron-oxide ratio thresholding; Landsat change detection for legacy site dating | GeoTIFF polygon layer with per-facility dryness score and estimated surface area, updated seasonally |
| Dust emission potential ranking | Multi-factor score combining SWIR-derived surface moisture proxy, fetch length within polygon, and absence of vegetation or crust; published wind-erosion threshold literature used for calibration | Ranked GIS layer of tailings and haul-road segments by emission potential, with metadata table |
| Downwind exposure corridor model | ERA5 multi-year wind rose per facility; directional weighting of emission score to produce probability-weighted transport footprint | Vector exposure-corridor layer per facility, clipped to configurable downwind distance (e.g. 5 km, 15 km) |
| MODIS AOD anomaly event log | MAIAC AOD daily retrieval; rolling 3-year baseline per 1 km cell; events flagged when AOD exceeds baseline by >1 SD and ERA5 wind vectors are consistent with source attribution | CSV and GIS point layer of attributed dust episodes with date, magnitude and source-facility ID |
| Community exposure index | Overlay of downwind corridor probability with WorldPop population grid and OpenStreetMap school/clinic locations; weighted sum to produce receptor-population exposure score | PDF report with ranked community list and map; GeoPackage for client GIS integration |
| Legacy site inventory | Landsat time-series analysis (1972 to present) to date facility abandonment and track vegetation colonisation; combined with current Sentinel-2 dryness assessment | Tabular inventory with facility coordinates, estimated abandonment date, current emission potential score and nearest settlement distance |
| Sentinel-1 surface disturbance alerts | Coherence change detection between consecutive Sentinel-1 SAR passes over tailings and haul-road surfaces; cloud-independent flagging of newly exposed dry material | Weekly alert GeoTIFF with disturbance probability layer; email notification when threshold exceeded at nominated facilities |
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.