Sensors
- MODIS / Terra and Aqua: 36-band imager; CO₂-slicing retrieval uses bands near 13.3–14.2 µm at 5 km nadir resolution. Revisit roughly twice daily per satellite. Reliable for optically thick cloud (optical depth > ~1); degrades for thin cirrus below about 0.3 optical depth.
- VIIRS / NOAA-20 and Suomi-NPP: Successor to MODIS heritage; M-band thermal channels at 750 m, day-night band at 750 m. CO₂-slicing applied at 6 km aggregate resolution for cloud-top pressure. Twice-daily revisit per satellite; archive from 2012.
- MISR / Terra: Nine fixed cameras at angles from 70.5° forward to 70.5° aft, 275 m per-camera resolution in red band. Stereo-photogrammetric cloud-top height derived geometrically, independent of atmospheric temperature profiles. No thermal assumption required; works on any optically thick feature. Nine-day global repeat.
- CALIOP / CALIPSO: Two-wavelength polarisation lidar at 532 nm and 1064 nm. Vertical resolution 30 m below 8.2 km, 60 m above. Detects optically thin cirrus with optical depth as low as ~0.001, which passive thermal methods miss entirely. Narrow 70 m ground track; 16-day exact repeat. CALIPSO decommissioned 2023; archive remains the definitive thin-cloud reference dataset.
Why the cloud top matters more than the cloud itself
A cloud's visible top is not the same as its radiative top, its pressure level, or the altitude a jet aircraft must avoid. Getting these three things confused costs money in fuel diversions and, in trace-gas remote sensing, introduces systematic errors that can exceed the signal being measured. When a column-retrieval algorithm for ozone or nitrogen dioxide assumes a clear sky but encounters a cloud at 300 hPa, it misattributes the absorption feature to the wrong atmospheric layer. The bias is not random noise; it is a structured error that skews regional emission inventories.
Aviation meteorology needs cloud-top pressure for a different reason: turbulence and icing intensity peak near the overshooting tops of deep convective systems, and pilots need that altitude in pressure coordinates to match their flight-management systems. A 50 hPa error at 200 hPa translates to roughly 500 metres of altitude uncertainty near the tropopause, which is operationally significant.
CO₂ slicing: elegant in theory, honest about its limits
The CO₂-slicing method, applied operationally by both MODIS and VIIRS, exploits the fact that carbon dioxide absorbs thermal emission at known wavelengths near 14 µm. By comparing radiances in channels with different CO₂ absorption strengths, the algorithm infers the pressure level at which the cloud is emitting. The maths are well-established: the ratio of the radiance difference in an absorbing channel to that in a window channel is proportional to the cloud's effective emissivity and pressure.
The practical limit is optical depth. When a cirrus layer is optically thin (roughly below 0.3), the cloud emits too weakly in the CO₂ channels to separate its signal from the surface or lower atmosphere below it. The algorithm either misses the cloud entirely or assigns it a pressure level that is too low, meaning it appears lower in the atmosphere than it actually is. For aviation icing hazard this is the wrong direction to err. MODIS cloud-top pressure products carry an estimated uncertainty of 50–150 hPa for thin cirrus, widening to worse than 200 hPa near the tropopause where the temperature lapse rate flattens.
What MISR's nine cameras see that a single radiometer cannot
MISR's stereo approach is geometrically rather than radiometrically derived. With nine cameras imaging the same scene within about seven minutes at angles spanning 140 degrees, the apparent displacement of a cloud feature between cameras encodes its height above the surface, exactly as parallax works in binocular vision. The retrieval requires no assumption about the atmospheric temperature profile and is unaffected by the thermal emission ambiguity that limits CO₂ slicing.
The trade-off is coverage. MISR's swath is 380 km and its exact repeat is nine days, so it cannot provide the near-real-time global coverage that aviation meteorology demands. It also requires a textured cloud surface with trackable features; a featureless stratiform deck can defeat the wind-corrected stereo matcher. Where it excels is as a validation reference: MISR stereo heights have been used extensively in published studies to audit MODIS and VIIRS CO₂-slicing products, and the agreement is generally within 1 km for optically thick convective systems.
CALIOP and the cirrus problem: the only method that actually works
Thin cirrus is the pathological case for passive sensors. It is radiatively important (globally, cirrus covers roughly 30 percent of the tropics and modulates outgoing longwave radiation significantly), it is a known icing hazard for aircraft, and it is nearly invisible to CO₂-slicing methods. CALIOP, the lidar aboard CALIPSO, resolves this by sending a pulse and timing the return. Optical depth as low as 0.001 is detectable. Vertical resolution of 30 m below 8.2 km means thin multi-layer cloud structures that appear as a single feature to MODIS are resolved as distinct sheets.
The cost is spatial sampling. A 70 m ground track repeated every 16 days is a curtain through the atmosphere, not a map. CALIOP is the calibration standard for everything else, not an operational global product on its own. CALIPSO itself was decommissioned in August 2023 after more than 17 years; the archive through NASA Earthdata remains the deepest global lidar cloud record available. Its successor, the CATS instrument on the ISS, had a shorter life, and the EarthCARE mission from ESA and JAXA (launched 2024) carries ATLID, a high-spectral-resolution lidar intended to continue this capability.
Where the three methods agree, and where they do not
For deep convective cloud with optical depth well above 1, CO₂ slicing, MISR stereo, and CALIOP backscatter typically agree to within 1 km in height. The disagreements are instructive. MISR stereo can place the wind-corrected geometric top above the radiative emission top measured by MODIS because the overshooting dome is optically thick but not in thermal equilibrium with the surrounding tropopause. CALIOP frequently finds a thin cirrus layer 1 to 3 km above the MODIS-retrieved top of the same convective system.
For aviation hazard products, the conservative approach is to use the highest top estimate from whichever method is available. For trace-gas retrieval correction, the radiative top from CO₂ slicing is usually the physically appropriate quantity, since it reflects the pressure level that the spectrometer is actually sensing. Neither is universally correct; the choice depends on the application, and any system that does not state which definition it uses is worth querying.
From archived curtains to operational products
Building an operational cloud-top pressure product means fusing these methods deliberately. MODIS and VIIRS provide global coverage twice daily at coarse resolution. MISR provides periodic high-quality stereo validation. CALIOP provides the thin-cirrus ground truth. The NASA MODIS cloud product (MOD06/MYD06) and the NOAA VIIRS cloud-top properties product are both publicly archived and form the backbone of most operational aviation meteorology pipelines.
Satellize runs analytics on open constellation data including VIIRS, and can layer cloud-top pressure retrievals against flight-route geometries or trace-gas retrieval grids on client request. The analytics approach follows the same published method classes used in the Tonga crop-estimation programme: open sensor data, documented retrieval algorithms, and outputs calibrated against the best available reference. If you are assessing whether a cloud-top pressure dataset is fit for a specific trace-gas correction workflow, the right conversation starts with the optical depth threshold of your target gas retrieval, not with the sensor name.
Typical figures
| CO₂-slicing spatial resolution (MODIS) | 5 km nadir (aggregated from 1 km radiances) |
| CO₂-slicing spatial resolution (VIIRS) | 6 km aggregate from 750 m M-bands |
| MISR stereo cloud-top height resolution | 275 m per camera; stereo height precision ~500 m for trackable features |
| CALIOP vertical resolution | 30 m below 8.2 km; 60 m from 8.2–20.2 km |
| Revisit (MODIS Terra + Aqua combined) | ~4 overpasses per day at mid-latitudes; twice daily per satellite |
| MISR global repeat | 9 days; 380 km swath |
| CALIOP ground track width | 70 m; 16-day exact repeat |
| Minimum detectable optical depth (CALIOP) | ~0.001 for thin cirrus |
| CO₂-slicing pressure uncertainty (optically thick cloud) | ~50 hPa; degrades to 150–200+ hPa for thin cirrus |
| Archive depth | MODIS from 2000; VIIRS from 2012; CALIOP 2006–2023 |
Analytics Satellize can run
| Cloud-top pressure map for trace-gas correction | CO₂-slicing retrieval (MODIS MOD06 / VIIRS cloud-top properties product) | Gridded GeoTIFF or NetCDF layer at 5–6 km, matched to client retrieval grid, with per-pixel optical depth flag |
| Aviation convective hazard altitude product | VIIRS cloud-top temperature converted to pressure altitude via co-located NWP profile | Flight-route segment report with cloud-top pressure, equivalent altitude, and optical depth class; updated per overpass |
| Thin cirrus detection flag | CALIOP backscatter archive query for optical depth < 0.3 features; spatially collocated with passive sensor overpasses | Collocated flag layer indicating where CO₂-slicing retrievals are likely biased; delivered as GIS polygon layer |
| MISR stereo vs. MODIS thermal bias assessment | Geometric stereo height from MISR MINNAERT wind-corrected stereo matcher compared to MOD06 CO₂-slicing pressure | Statistical bias report by cloud regime (convective, stratiform, cirrus) for a defined region and time window |
| Multi-year cloud-top climatology | MODIS or VIIRS archive aggregation by season, regime, and pressure level | Monthly climatology NetCDF with percentile distributions; suitable for climate model evaluation |
| EarthCARE ATLID readiness assessment | Gap analysis comparing CALIOP archive statistics to planned ATLID coverage for a client's region of interest | Written technical note with coverage probability estimates and recommended fusion strategy |
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.