TCOscope / 2026
A five-day view of the ozone layer built from satellite chemistry, stratospheric weather and the latest observed atmospheric state—made explorable in a fast public web application.
Selected city outlook
294.8 DU · T+3
Context & objective
Short-term ozone variability matters for ultraviolet exposure, ecosystems, atmospheric circulation and scientific monitoring. Physics-based composition systems are powerful, but expensive to run and complex to turn into a focused local product.
TCOscope predicts change instead of rebuilding the atmospheric field from scratch. It fuses observations, future meteorology and regime indices, then publishes five global maps through a simple read-only delivery stack.
The protective layer that absorbs biologically damaging ultraviolet radiation.
Typical levels and variability differ sharply between the tropics, midlatitudes and polar regions.
Brewer–Dobson circulation, seasonality and temperature redistribute and alter ozone.
QBO, ENSO, solar forcing and polar chemistry influence changes across multiple time scales.
From orbit to interface
A rolling 21-day TROPOMI NRTI history provides ozone and trace-gas context. T0 must contain real observations for every required product.
GFS fields at +24 to +120 hours describe winds, temperature and geopotential height through the stratosphere.
Earthformer-Lite combines history, future meteorology and global regime indices to predict anchored residual corrections.
Five compressed grids move through PostgreSQL and FastAPI to the WebGL globe, city charts and CSV export.
Data lineage
Operational observation history
Total-column O₃ plus CO, NO₂, SO₂, HCHO and aerosol index in a rolling 21-day window.
Future meteorology
Wind, temperature and geopotential height at 200, 100, 50, 30 and 10 hPa for every lead.
Large-scale context
QBO at 30/50 hPa, F10.7 solar flux, ONI, Arctic Oscillation and North Atlantic Oscillation.
Independent ground validation
Daily total-column ozone from 38 stations that were not the model's satellite training target.
Cross-sensor validation
An external Level-3 daily ozone product used to test transfer beyond TROPOMI retrievals.
Physics-based reference
CAMS total-column ozone on matching leads, evaluated alongside the simple persistence baseline.
Model architecture
The public product
Drag, zoom and hover to inspect exact ozone values and coordinates anywhere on Earth.
Choose a city, focus the globe and read its complete T+1…T+5 ozone series.
Download one wide CSV containing coordinates and all five dated forecast columns.
Evidence, not just a demo
All five leads
Lower is better · Dobson Units (DU)
Same satellite target and observation mask used by the model pipeline
13,772 matched pairs from 38 independent Brewer/Dobson stations
Raw NASA OMPS-NPP NMTO3 observations on a common grid
TCOscope beats persistence at every lead in all three comparisons.
The advantage over CAMS remains against an external satellite and independent ground stations.
Matched leads and common observations are evaluated without test-period bias correction.
The period covers Northern Hemisphere winter and spring, not a complete annual cycle.
The benchmark is retrospective; a frozen prospective 6–12 month protocol is still required.
The claim is about total-column ozone forecasting—not superiority over CAMS as a complete composition system.
Next chapter
Register checkpoint and code hashes, then evaluate 6–12 untouched months without tuning.
Add calibrated intervals and explore ensembles with physics-based forecasts.
Increase spatial detail carefully while preserving global consistency and affordable inference.
Atmospheric Researchers
