GOLD STANDARD · REMOTE SENSING · HOW X WORKS
Tell a slow optical change from a reversible thermal one
A scan mirror can alter radiance and polarisation as its coating ages, while temperature changes move the same calibration residual over each orbit or season. Geometry-controlled records, thermal coefficients and change-point logs help separate slow trend, reversible drift and discrete contamination events.
This guide connects the Gold Standard of Remote Sensing, Gold Standard of Education, and How X Works hub.
Chapter contents and reading routes
Choose the route closest to your present task. Every chapter remains expanded, so the article also reads naturally from start to finish.
Orient the optical trend
Build and work the separation model
Validate on orbit
Transfer and qualify
Use FAQs, evidence and next routes
CHAPTER 1 OF 12 · ORIENT THE OPTICAL TREND
1. Mirror change reaches both radiometry and polarimetry
A reflective coating has wavelength- and angle-dependent complex reflectance. Differences between orthogonal polarisation components create diattenuation and retardance. Contamination, oxidation or microstructural change can therefore alter throughput and the Mueller matrix together.
A radiometric correction alone may leave scene-dependent polarimetric bias. Conversely, an apparent polarisation trend can come from changing scan angle, scene polarisation or temperature rather than permanent coating ageing.
CHAPTER 2 OF 12 · ORIENT THE OPTICAL TREND
2. Separate slow, reversible and discrete components
Ageing is usually sought as a slow secular component. Temperature drift is often partly reversible around a thermal operating point. Decontamination heating, safe-mode recovery, manoeuvres or algorithm updates can create steps.
Write these components explicitly before fitting. A smooth polynomial in time can absorb thermal cycles and steps, producing an attractive curve with the wrong physical interpretation.
CHAPTER 3 OF 12 · BUILD THE MODEL
3. Start with a geometry-controlled residual
Build residuals only after matching wavelength, detector, mirror side, scan angle and viewing geometry. A useful first model is r = α + βt + γ(T−T0) + geometry terms + event steps + noise, where β is a candidate secular trend and γ is a thermal coefficient.
This is an identification scaffold, not a guarantee. If temperature and mission time move together, β and γ may be nearly inseparable. Planned thermal diversity or repeated calibration views are then more valuable than a more complicated regression.
CHAPTER 4 OF 12 · WORK THE CORRECTION
4. Work a temperature-adjusted leakage example
Suppose instrumental Q/I leakage was 0.30% at the reference epoch and is now 0.42%. Laboratory or on-orbit thermal characterisation gives +0.01 percentage point per kelvin, and the mirror is 6 K warmer. Temperature explains +0.06 percentage point.
| Quantity | Value | Interpretation |
|---|---|---|
| Observed change | +0.12 percentage point | 0.42% − 0.30% |
| Thermal contribution | +0.06 percentage point | 0.01 × 6 K |
| Unexplained residual | +0.06 percentage point | Candidate ageing, contamination or other drift |
The remaining 0.06 percentage point is not automatically coating ageing. It is a residual to test against contamination events, geometry errors, detector drift and calibration-target uncertainty.
CHAPTER 5 OF 12 · BUILD THE LEDGER
5. Track the Mueller elements that matter to retrievals
Do not reduce the mirror to one scalar sensitivity. Record relevant Mueller-matrix elements or instrument-level leakage terms by wavelength, field angle and mirror side. The retrieval may be most sensitive to one cross-talk element rather than total throughput.
A trend ledger should include sign convention, reference frame and rotation order. A stable physical response can appear to change sign when frames or scan geometry are handled inconsistently.
CHAPTER 6 OF 12 · VALIDATE ON ORBIT
6. Use repeated targets and independent calibration paths
Deep space, solar diffusers, lunar views, Rayleigh-scattering scenes, sun glint and cross-sensor comparisons constrain different pieces of the system. Repeated observations at matched geometry can expose time dependence; thermally varied observations estimate reversible response.
No single target is perfect. Diffusers age, lunar geometry needs modelling, ocean glint is scene-dependent and cross-sensor transfer carries the reference sensor’s uncertainty. Agreement across independent paths is stronger than a precise result from one path.
CHAPTER 7 OF 12 · VALIDATE ON ORBIT
7. Use warm–cool loops and event-aware trend tests
Within a period where permanent ageing is negligible, compare residuals on warming and cooling branches. Reversibility supports a thermal term; hysteresis suggests additional states such as contamination migration or thermo-mechanical settling.
| Pattern | Likely clue | Follow-up |
|---|---|---|
| Repeats with temperature | Reversible thermal sensitivity | Estimate γ by geometry bin |
| Persists after return to T₀ | Secular or hysteretic change | Inspect event log and later cycles |
| Jumps at one date | Maintenance or configuration step | Fit and verify a change point |
| Only one scan angle changes | Field-angle or mirror-side issue | Recheck geometry-specific calibration |
CHAPTER 8 OF 12 · TRANSFER THE METHOD
8. Changed-context transfer: satellite to airborne polarimeter
An airborne system experiences faster temperature changes, vibration and maintenance access. The same decomposition—geometry, thermal response, slow trend and event steps—still applies.
The cadence changes. Airborne calibration can use pre- and post-flight laboratory checks, while a satellite depends more heavily on celestial, atmospheric or onboard references. Do not copy an on-orbit time constant into an aircraft maintenance rule.
CHAPTER 9 OF 12 · QUALIFY THE TREND
9. Protect against collinearity and retrieval feedback
Orbital local time, season, temperature, scene type and instrument age can be correlated. Use matched subsets, designed calibration opportunities and uncertainty-aware regression. Report covariance or sensitivity rather than presenting each coefficient as independently known.
If a retrieval algorithm is updated using the same residual record, preserve an untouched diagnostic stream. Otherwise algorithm adaptation can hide hardware drift and make the corrected product look stable by construction.
CHAPTER 10 OF 12 · QUALIFY THE TREND
10. Limits and common misconceptions
A secular residual is not direct chemical proof of coating ageing. Contamination film growth, detector response, diffuser degradation and geometry error can create similar trends. A temperature coefficient measured prelaunch may also change after launch.
A small leakage percentage can still matter for weakly polarised scenes, and a large raw effect may cancel in a particular retrieval. Evaluate the calibrated measurement and the final science product, with uncertainty carried through both.
CHAPTER 11 OF 12 · CHECK AND CONTINUE
11. Questions readers often ask
Can time and temperature always be separated statistically?
No. If they are nearly collinear, the data do not contain enough independent variation. A planned thermal manoeuvre, matched seasonal revisit or external calibration reference may be necessary.
Does a stable radiometric gain prove polarimetric stability?
No. Throughput and Mueller cross-talk can evolve differently. Monitor the polarimetric elements and retrieval sensitivities that the science product actually uses.
CHAPTER 12 OF 12 · CHECK AND CONTINUE
12. Evidence, connected routes and the next check
Mirror contamination in space I: mirror modelling uses Mueller matrices and thin-film Fresnel modelling to describe time-, wavelength-, polarisation- and scan-angle-dependent degradation. Polarimetric calibration of the spaceborne Directional Polarimetric Camera installed on the GF-5 satellite develops a spaceborne calibration model. In-flight radiometric and polarimetric calibration of the Directional Polarimetric Camera onboard GaoFen-5 over the ocean demonstrates Rayleigh and sun-glint calibration routes.
Continue through polarimetric frame rotation and sign conventions, scan-mirror diattenuation and swath residuals, whitecap Mueller matrices and glint geometry, and whitecap polarisation residuals. The next check is to plot the same geometry bin against both temperature and mission time before naming a trend.