Cement plant fugitive dust and kiln plume mapping from optical data

Cement production leaves optical fingerprints: kiln plumes alter aerosol columns, and fugitive dust changes surface reflectance around plant perimeters. Sentinel-2, Landsat 8/9 and MODIS MAIAC together make those fingerprints legible, within honest resolution limits.

Sensors

Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at the equator with both satellites. Detects surface reflectance changes from dust deposition and resolves individual plant structures. Aerosol-sensitive band 1 (443 nm) and band 9 (945 nm) support basic atmospheric column characterisation, though full aerosol retrieval requires ancillary processing.
Landsat 8/9 OLI: 30 m multispectral resolution; 8-day repeat per satellite, 4-day combined. Coastal/aerosol band (Band 1, 443 nm) provides a direct aerosol-sensitive channel. Archive extends to 1972 for long-term baseline construction. Slightly coarser than Sentinel-2 but radiometrically consistent across decades.
MODIS MAIAC AOD: 1 km aerosol optical depth product derived via the Multi-Angle Implementation of Atmospheric Correction algorithm. Daily revisit from Terra and Aqua combined. Resolves facility-scale AOD anomalies only for large or isolated plants; industrial clusters within 1-2 km are not separable. Useful for trend analysis and episodic plume confirmation.
Sentinel-5P TROPOMI: 3.5 x 5.5 km pixel (post-2019 upgrade) for aerosol absorbing index and UV aerosol index. Daily global coverage. Cannot attribute a plume to a single cement plant in a dense industrial zone, but confirms regional aerosol loading and flags days worth examining in higher-resolution imagery.

What the dust actually does to a pixel

Cement dust is predominantly calcium carbonate and calcium silicate, both highly reflective in the visible spectrum. When fugitive dust settles on rooftops, roads and vegetation around a plant, it raises surface reflectance across all visible bands in a spatially coherent pattern centred on the source. Sentinel-2 at 10 m resolves this deposition halo clearly enough to distinguish it from natural soil brightness, provided a pre-operational baseline image exists for differencing.

Airborne plumes behave differently. A dense kiln stack plume scatters incoming radiation, reducing apparent surface reflectance beneath it while increasing top-of-atmosphere radiance in the forward-scatter direction. This creates a characteristic asymmetric signature in multispectral imagery: bright in the plume body, shadowed below. The effect is strongest in Band 1 (443 nm) on both Sentinel-2 and Landsat OLI, where molecular and aerosol scattering is most pronounced relative to surface signal.

Surface reflectance change as a production proxy

A cement plant that is running hard produces more fugitive dust than one that is idling. The accumulation of dust on fixed surfaces, particularly the flat roofs of grinding halls and clinker stores, is measurable as a slow increase in mean reflectance over weeks. Differencing cloud-free Sentinel-2 scenes separated by 15 to 30 days produces a change map that correlates, imperfectly, with production intensity. The relationship is not linear: wind, rain and housekeeping all modulate the signal. But a plant that goes dark on reflectance change for several consecutive acquisition windows has almost certainly reduced throughput.

This proxy is most reliable for isolated facilities. In industrial clusters where multiple plants share airsheds, attributing a reflectance increase to a specific source requires wind-direction filtering and careful geometric analysis of plume orientation across multiple scenes. That is achievable but labour-intensive, and the result carries wider uncertainty bounds than a single-plant analysis.

Aerosol column retrieval: what MAIAC gives you and what it does not

The MODIS MAIAC AOD product retrieves aerosol optical depth at 1 km using multi-angle observations to separate surface and atmospheric contributions. Over cement plants, elevated AOD values above the facility footprint, sustained across multiple passes, constitute evidence of persistent particulate emission. Published studies using MAIAC over South and East Asian cement belts have identified facility-scale AOD anomalies of 0.1 to 0.3 AOD units above regional background, which is detectable but not large relative to natural dust events.

The honest limit is spatial resolution. A 1 km MAIAC pixel covers the entire footprint of a mid-sized cement plant and a substantial surrounding area. On a day with light winds, the plume may not have dispersed enough to produce a clean signal outside the source pixel. On a day with strong winds, the plume disperses rapidly and the per-pixel AOD anomaly drops below detection. MAIAC is most useful as a screening tool: it flags dates and locations worth examining in Sentinel-2 or Landsat imagery, rather than serving as a primary enforcement dataset on its own.

Resolution floors and the attribution problem in industrial clusters

Sentinel-2 at 10 m is sufficient to resolve individual kiln stacks and distinguish them from adjacent structures. It is not sufficient to quantify emission rates: that requires either in-situ monitoring or a mass-balance approach using wind-field data that is rarely available at the required resolution. What optical data can do is establish whether a plume is present, estimate its approximate horizontal extent and persistence, and flag anomalous surface changes. Quantitative emission estimates from optical data alone should be treated with considerable scepticism.

The attribution problem sharpens in regions where cement plants, lime kilns, quarries and aggregate processors sit within a few kilometres of each other. A dust plume visible in a Sentinel-2 scene may originate from any of several sources. Wind-direction analysis using ERA5 reanalysis fields narrows the candidates but rarely produces a definitive single-source attribution without corroborating evidence. Regulators using this data for enforcement should treat optical plume mapping as a trigger for ground inspection, not as a standalone proof of violation.

Cloud cover is the other hard constraint. In tropical and monsoonal regions, cement plants may be obscured for weeks at a time. A 5-day revisit means nothing if four consecutive passes are cloud-contaminated. SAR backscatter can detect structural changes at a plant during cloud cover, but it does not see aerosol plumes. The practical cloud-free revisit in many high-emission regions is closer to 20 to 40 days, which limits the temporal resolution of any production-intensity inference.

Building a defensible monitoring record

A credible cement plant monitoring programme combines three layers. First, a multi-year Landsat archive baseline establishes the plant's normal reflectance envelope and identifies seasonal patterns from raw material stockpile changes and maintenance shutdowns. Second, Sentinel-2 time series at 10 m tracks shorter-term deviations, flagging candidate plume events for analyst review. Third, MAIAC AOD provides a daily screening layer that can trigger priority re-examination of recent Sentinel-2 acquisitions even before the next cloud-free pass arrives.

Satellize applies this layered approach within its analytics platform, drawing on the same open constellations used in its Tonga crop-estimation programme and extending the methodology to industrial facility monitoring. The output is not a single image but a structured time series: per-facility reflectance anomaly scores, plume-detection flags keyed to wind direction, and AOD percentile rankings relative to a rolling 90-day regional baseline. That package gives a regulator or an environmental due-diligence team something they can act on, while being explicit about which findings warrant field verification.

What the archive cannot tell you

Optical methods detect particulate matter as a column or surface phenomenon. They do not distinguish PM2.5 from PM10, they cannot measure heavy metal content in the dust, and they say nothing about SO2 or NOx co-emissions from the kiln flame. Those questions require either in-situ sampling or complementary sensors. SO2 and NOx attribution from cement plants is covered in separate pages in this library.

Night-time kiln operation is also largely invisible to passive optical sensors, which depend on reflected sunlight. A plant running three shifts will show a daytime plume only during the morning or afternoon overpass window. Thermal infrared can detect kiln heat signatures at night, but at the spatial resolutions available from Landsat TIRS (100 m resampled to 30 m) and ASTER, individual kiln stacks are marginally resolved at best. The honest conclusion: optical data is a powerful, low-cost first screen, but it captures a fraction of the operational picture.

Typical figures

Best spatial resolution (plume mapping)	10 m (Sentinel-2 MSI visible/NIR bands)
Best spatial resolution (AOD retrieval)	1 km (MODIS MAIAC); 3.5 x 5.5 km (TROPOMI aerosol index)
Revisit (cloud-free, practical)	5-day nominal Sentinel-2; 20-40 days effective in tropical/monsoonal regions
Key aerosol-sensitive bands	Sentinel-2 Band 1 (443 nm), Band 9 (945 nm); Landsat OLI Band 1 (443 nm)
Minimum detectable AOD anomaly (MAIAC)	~0.05 AOD units above background (published uncertainty range)
Archive depth	Landsat: 1972-present; Sentinel-2: 2015-present; MODIS MAIAC: 2000-present
Latency (open data)	Sentinel-2 L2A: typically 1-3 hours after acquisition; Landsat: same-day
Surface reflectance change detection floor	~2-3% reflectance difference (instrument noise and atmospheric correction limit)
Coverage	Global; all major cement-producing regions included in open archive
Delivery formats	GeoTIFF, Cloud-Optimised GeoTIFF, GeoJSON anomaly vectors, CSV time series

Analytics Satellize can run

Plume presence and orientation flag	Multi-date top-of-atmosphere radiance differencing in Band 1 (443 nm), combined with ERA5 wind-direction filtering to assign probable source	Per-acquisition GeoJSON flag layer with plume bearing, estimated horizontal extent and wind-consistency score
Surface dust deposition anomaly score	Sentinel-2 L2A surface reflectance time series differencing against rolling 90-day baseline; z-score normalisation per facility zone	Monthly GIS layer of reflectance anomaly by facility zone, with percentile ranking against historical distribution
Production intensity proxy index	Composite of dust deposition anomaly score and plume-detection frequency across cloud-free acquisitions; calibrated against plant-area spectral change	Quarterly facility-level index report with trend chart and cloud-cover-adjusted confidence rating
MAIAC AOD screening alert	Daily MODIS MAIAC AOD percentile ranking for each facility's 5 km buffer zone; threshold exceedance triggers priority Sentinel-2 review queue	Automated daily alert feed (JSON or email digest) with facility ID, AOD value, regional percentile and recommended follow-up action
Multi-year baseline and shutdown detection	Landsat OLI archive time series (2013-present) analysed for structural reflectance shifts indicating facility construction, expansion or prolonged shutdown	Annotated change-point timeline per facility, delivered as PDF report with supporting imagery chips
Industrial cluster attribution assessment	Wind-field-conditioned plume geometry analysis across multiple Sentinel-2 scenes; Bayesian source-probability weighting for facilities within 5 km radius	Source-attribution probability map per plume event, with explicit uncertainty bounds and recommendation on whether ground inspection is warranted

Sources

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.