Illegal pesticide and agrochemical dumping on agricultural land
Banned or expired agrochemicals dumped on fields leave spectral signatures that differ measurably from drought, frost or disease. Sentinel-2 red-edge and SWIR bands, combined with water-quality proxies, can surface those signatures before ground inspectors arrive.
Sensors
- Sentinel-2 MSI: 10 m visible and NIR bands, 20 m red-edge (B5, B6, B7) and SWIR (B11, B12) bands. Five-day global revisit at the equator with two satellites. Red-edge bands are the primary discriminant for chemical-stress signatures; SWIR responds to leaf-water content changes preceding visible necrosis.
- Landsat 8/9 OLI: 30 m multispectral, 16-day single-satellite revisit (8 days combined). SWIR-1 and SWIR-2 bands complement Sentinel-2 for cross-sensor validation and extend the archive to 1972 via earlier Landsat generations, giving decades of baseline for anomaly detection.
- Sentinel-3 OLCI: 300 m resolution, daily revisit. Designed for ocean and inland water colour; chlorophyll-a and turbidity retrieval algorithms (Case-2 water models) can flag algal-bloom or suspended-sediment anomalies in rivers and reservoirs downstream of suspected dump sites.
- MODIS Terra/Aqua: 250 m to 500 m, near-daily revisit. Useful for rapid triage across large agricultural regions and for establishing multi-year seasonal baselines against which Sentinel-2 anomalies are assessed. Spatial resolution too coarse to confirm small dump events but adequate for watershed-scale water-quality screening.
Why chemical stress looks different from everything else
A field that dies from drought dies slowly, and it dies uniformly within a soil-moisture gradient. Frost kill follows topography: low-lying areas first, field edges last. Fungal or bacterial disease spreads in characteristic patterns tied to spore dispersal or root-contact networks. Chemical dumping does none of these things. It produces spatially abrupt necrosis, often ignoring field boundaries, concentrated near access tracks or drainage inlets, and it can kill a crop in days rather than weeks.
The spectral consequence is equally distinctive. Healthy vegetation has a steep reflectance rise between the red and NIR bands, the so-called red edge. Chemical stress, particularly from herbicides that inhibit photosynthesis or disrupt chlorophyll synthesis, collapses chlorophyll content faster than it collapses leaf structure. The result is a red-edge shift and a NDRE (Normalised Difference Red Edge) drop that precedes, and is disproportionate to, the NDVI drop. Sentinel-2 bands B5 (705 nm), B6 (740 nm) and B7 (783 nm) resolve this shift at 20 m. Drought stress, by contrast, tends to reduce SWIR reflectance as leaf-water content falls before chlorophyll is affected. The two stress types leave different fingerprints across the same sensor.
What a time series reveals that a single image cannot
A single image showing dead vegetation is evidence of nothing specific. A time series showing a field that was healthy on day zero, partially stressed on day five, and fully necrotic on day ten, with no corresponding rainfall deficit, temperature anomaly or regional disease outbreak, is a different matter entirely. Change-vector analysis on Sentinel-2 surface-reflectance products (available through the Copernicus Data Space) can isolate pixels whose spectral trajectory diverges sharply from the seasonal norm established over three to five prior years.
The practical threshold for detection depends on the size of the affected area and the severity of the kill. Published studies using Sentinel-2 have demonstrated detection of vegetation-stress patches down to roughly 0.1 hectares when the spectral contrast against surrounding healthy crops is strong. Partial contamination, where the dose is sublethal, is harder: the red-edge signal is subtler, and distinguishing it from early-season nutrient deficiency requires careful baseline calibration. Honest answer: a skilled analyst can flag suspicious patches reliably; confirming the cause requires ground sampling.
Temporal cadence matters. Five-day revisit from Sentinel-2A and 2B combined is adequate for most agricultural contexts, but cloud cover is the binding constraint in humid regions. In persistently cloudy environments, a six-week gap in usable imagery is not unusual, which can allow a dump event to be missed entirely or for the spectral signature to be obscured by regrowth.
Following the contamination downstream
Agrochemicals dumped near drainage channels do not stay on the field. Organophosphates, atrazine, chlorpyrifos and similar compounds enter surface water within hours of heavy rain. The water-quality signal they produce is indirect: turbidity increases as disturbed soil moves, and algal blooms can follow elevated nutrient loading from certain fertiliser-class chemicals. Sentinel-3 OLCI provides daily turbidity and chlorophyll-a estimates for water bodies larger than roughly one square kilometre, using published Case-2 water algorithms. Smaller channels fall below its detection floor.
For smaller water bodies, Sentinel-2 Band 3 (green, 560 nm) and Band 4 (red, 665 nm) ratios can serve as turbidity proxies, though the 10 m pixel mixes water and bank vegetation in narrow channels and the retrieval is less reliable than purpose-built ocean-colour sensors. The practical use is triangulation: a vegetation kill patch in the field time series, coinciding temporally with a turbidity spike in the adjacent watercourse, is a much stronger enforcement signal than either observation alone.
Methodological grounding in published programmes
The EU's Farm to Fork strategy and the European Environment Agency's pesticide monitoring frameworks have driven published research into remote-sensing detection of agrochemical impacts. The FAO's work on integrated pest management and its guidance on highly hazardous pesticide reduction provide regulatory context for what constitutes an illicit dump versus licensed application. The Joint Research Centre has published on Sentinel-2 based crop-stress monitoring across EU member states, providing validated spectral indices and seasonal baselines that an analyst can adapt for enforcement purposes.
It is worth being clear about what satellite analytics cannot do here. They cannot identify the specific compound involved. They cannot distinguish an illegal dump from a licensed but misapplied treatment without ground truth. They cannot produce evidence admissible in most jurisdictions without corroborating field data. What they can do is prioritise inspector time: a region with ten thousand fields can be screened in hours, and the twenty fields showing anomalous spectral trajectories can be flagged for physical investigation. That is the genuine value proposition, and it is a significant one.
From alert to enforcement: the analytic workflow
A practical workflow begins with a seasonal baseline built from three or more years of Sentinel-2 surface-reflectance composites, cloud-masked and atmospherically corrected. Each new acquisition is compared against the baseline using change-vector analysis or a trained anomaly classifier. Flagged pixels are clustered into candidate events, filtered by spatial compactness (dump patches tend to be compact and irregular, unlike disease fronts), and ranked by the magnitude of the red-edge deviation.
Candidate events above a confidence threshold trigger a downstream check: does the adjacent watercourse show a contemporaneous turbidity anomaly in Sentinel-3 or Sentinel-2 water-quality retrievals? If yes, the event is escalated to a priority alert with a georeferenced polygon, a time-stamped spectral trajectory chart, and a comparison against the nearest weather station record to exclude drought or frost. The output is a structured report, not a verdict. The inspector who arrives on site knows where to take soil and water samples and what time window to focus on.
Satellize applies this kind of multispectral change-detection workflow on open constellations, including in agricultural contexts such as the Kingdom of Tonga crop-estimation programme. The same spectral-trajectory logic that estimates canopy condition for yield forecasting can be inverted to flag anomalous decline. The difference is in the baseline, the thresholds, and what the analyst is trained to look for.
Typical figures
| Primary spatial resolution | 10 m (Sentinel-2 visible/NIR), 20 m (Sentinel-2 red-edge/SWIR), 30 m (Landsat OLI), 300 m (Sentinel-3 OLCI) |
| Revisit cadence | 5 days (Sentinel-2A+B combined at equator); 8 days (Landsat 8+9 combined); daily (Sentinel-3, MODIS) |
| Key spectral bands | Sentinel-2 B5/B6/B7 (red-edge, 705–783 nm), B11/B12 (SWIR, 1610/2190 nm); Landsat OLI Band 5 (NIR), Bands 6–7 (SWIR) |
| Minimum detectable affected area | Approximately 0.1 ha for high-contrast necrosis in Sentinel-2 imagery; sublethal or partial stress requires larger affected areas for reliable detection |
| Latency from acquisition to alert | Sentinel-2 Level-2A surface reflectance typically available within 3–5 hours of overpass via Copernicus Data Space; analytic processing adds hours to 1 day depending on pipeline |
| Water-quality proxy coverage | Sentinel-3 OLCI reliable for inland water bodies above approximately 1 km²; Sentinel-2 water retrievals usable for smaller features with reduced accuracy |
| Archive depth | Sentinel-2: from 2015; Landsat: from 1972 (with sensor changes); MODIS: from 1999 |
| Cloud-cover limitation | Persistent cloud can create gaps of 4–8 weeks in humid agricultural regions; SAR (e.g. Sentinel-1) can partially fill temporal gaps but does not provide the spectral discrimination needed for chemical-stress typing |
| Delivery formats | GeoTIFF anomaly maps, GeoJSON event polygons, PDF enforcement-ready reports, time-series charts (PNG/SVG) |
Analytics Satellize can run
| Vegetation-stress anomaly map | Change-vector analysis on Sentinel-2 NDRE and SWIR-based indices against multi-year seasonal baseline; spatial clustering to isolate compact irregular patches | GeoTIFF layer and GeoJSON polygon file, updated each clear-sky acquisition, with per-patch anomaly score and spectral trajectory |
| Chemical-vs-other-stress classification | Spectral index combination (NDRE, NDVI, SWIR leaf-water index) with decision-tree or random-forest classifier trained on published stress-type spectral libraries | Per-patch stress-type probability score (chemical / drought / frost / disease / inconclusive) included in alert report |
| Downstream water-quality flag | Sentinel-3 OLCI chlorophyll-a and turbidity retrieval (Case-2 water model) and Sentinel-2 green/red band ratio for smaller channels, temporally aligned with field-stress events | Time-stamped turbidity and chlorophyll anomaly chart for named watercourses, appended to enforcement report |
| Priority inspection shortlist | Multi-evidence ranking combining field spectral anomaly score, water-quality co-occurrence, proximity to road access, and absence of meteorological stress explanation | Ranked table of candidate sites with coordinates, anomaly date, confidence tier, and recommended sampling window; delivered as PDF and CSV |
| Historical baseline and seasonal norm report | Multi-year Sentinel-2 and Landsat surface-reflectance composites used to establish field-level phenological curves; deviation thresholds set per crop type and region | Static baseline GeoTIFF stack and summary PDF, produced once per season or on programme initiation |
| Monitoring area coverage dashboard | Automated ingestion and cloud-mask scoring of each new Sentinel-2 and Landsat acquisition over the defined area of interest | Weekly coverage report showing percentage of area with usable imagery in the preceding 30 days, flagging zones with persistent cloud gaps requiring contingency planning |
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.