Sparse Sentinel
Built by Sparse Supernova
Your monitoring network already produces the signals needed for earlier warning. Sentinel-for-Air-Quality learns normal behaviour across pollutants, sites, and time of day, detects deviation early, predicts likely episodes, and shows operators and planners where to look before health-relevant thresholds are breached or complaints escalate.
It sits above existing reference stations, micro-sensors, and SCADA/building systems. Sentinel does not replace regulatory monitors. It adds earlier warning, clearer evidence, state-aware anomaly scoring, and impact metrics linked to health guidelines and local standards.
Sentinel also checks whether sensor channels remain trustworthy, helping teams distinguish real pollution episodes from sensor drift, calibration problems, and communication faults across PM, gases, and meteorology.
Over time, each deployment builds structured air-quality intelligence that can improve network design, maintenance, planning, and cross-site benchmarking of exposure and exceedance days.
This prototype is a cross-sector Tech/AI innovation, designed to complement WHO air quality guidelines and national indices, not to replace statutory reporting systems.
Earlier warning for ambient and indoor air quality across reference stations, sensor networks, and meteorology.
Start air quality onboarding Upload site details and monitoring data for Sentinel calibration.
Download air results (JSON) — sample public results for integrations or offline review
This card explains what Sentinel-for-Air-Quality actually does: it learns normal air patterns at each site, spots unusual changes early, and highlights where problems are likely to appear.
Learn normal. Detect deviation early. Predict likely episodes. Show clear warnings on the UI.
This card shows how early Sentinel can see a pollution episode developing compared with traditional thresholds, giving you extra time to respond before guideline levels are breached.
Lead times shown are demo scenario outputs and will vary by site, signal quality, pollutant mix, weather conditions, and episode type.
This card summarises Sentinel’s best guess about why a pattern looks abnormal, combining several signals to describe likely causes such as traffic, weather, or airport activity.
Sentinel combines pollutant behaviour, weather, site state, and nearby context to produce a structured explanation of what may be driving a developing episode.
This card shows how today’s air quality looks in health terms, and how Sentinel’s earlier warnings could reduce the time people spend breathing air above guideline levels.
Impact values are indicative demo outputs for showing how earlier warning can support exposure reduction and planning. They are not statutory compliance determinations.
This card is a snapshot of key pollutants and weather at this site right now, so you can see which substances are driving the air-quality picture at a glance.
This table compares when a standard limit or index would trigger versus when Sentinel raised an early warning, and how much extra warning time that created.
Threshold means the hour when a conventional 1-hour or 24-hour guideline, index band, or alert condition would be flagged. Sentinel shows the earlier hour when deviation was already visible.
| Episode | Threshold | Sentinel | Lead | Confidence |
|---|
This card shows which pollution sources and dispersion patterns—like traffic, industry, or background air—look normal, and which are behaving unusually.
Sentinel groups likely source and dispersion behaviours into a small set of operational patterns so teams can quickly see where the most unusual influence may be coming from.
This card breaks the day into states such as night, morning peak, or weekend, and shows which states are most often linked to air-quality issues at this site.
Rather than treating every hour as the same, Sentinel learns different normal patterns for states such as night, morning peak, daytime, evening peak, weekend, and stagnant conditions.
This card looks at alert patterns over time and suggests which alerts are genuinely useful and which are just noise, helping to reduce alarm fatigue.
Sentinel can help teams separate meaningful multi-signal episodes from repeated threshold chatter or short-lived spikes that add noise without improving decisions.
| Severity | Pattern | Code | Freq | Outcome |
|---|
This card shows how healthy each sensor channel is, so you can quickly see whether a strange reading is likely to be real air pollution or a sensor problem.
Air-quality interpretation depends on trustworthy channels. Sentinel checks for drift, humidity artefacts, intermittent communication problems, flat-lining, noise, and missing data.
This card summarises nearby traffic conditions—how many vehicles, how congested, how many heavy vehicles—to help explain traffic-related spikes in pollution.
Traffic is often a strong driver of short-term PM and NO2 behaviour. Sentinel uses nearby traffic context to test whether changing road conditions line up with pollutant episodes.
This card shows recent flight activity and whether the site is currently downwind of the runway, helping to explain pollution patterns near airports.
Where sites are near airports, Sentinel can compare air-quality change with recent runway use, aircraft movements, taxiing, and wind direction to identify likely aviation-related influence.
This card compares different sites or zones, highlighting where air quality is consistently better or worse and helping to prioritise attention and action.
Sentinel compares sites using a consistent structure so operators and planners can spot persistent differences in AQI, PM2.5, NO2, and exceedance days across the network.
| Site | Region | AQI | PM2.5 | NO2 | Exceedance days | Status |
|---|
