The science and engineering behind RefCal.

Making spectral data from different sensors comparable is a measurement problem first. Every reading carries the imprint of the atmospheric state, the illumination and viewing geometry, the sensor's spectral response and its degradation over time. RefCal is built to separate those effects, factor them out where the evidence supports it, and state how much confidence remains in the result.

Our work is organised in eleven areas, from the physics of a single measurement to whether the result helps someone decide. Each sets out what makes the problem hard, why it matters to the people who use satellite data, and how we are tackling it.

§ 01

Areas of work

Eleven areas in four parts, from how a single measurement is understood to whether the result helps someone decide. Each links to its own page.

Current focus as of 24 Sept 2026

No. Area and objective Stage

The measurement engine Learning what changes a measurement, and how sure to be about it.

P01 What changes a measurement Describe how the atmospheric state, the illumination and viewing geometry and the sensor's own spectral response shape what a satellite measures, and how much uncertainty each of them contributes. Observed in real data 1 result P02 Telling effects apart Current focus Combine independent kinds of evidence to separate effects that any single source leaves entangled, such as sensor degradation and a change in the atmosphere. Observed in real data 1 result P03 Effects that act together Learn how effects combine, such as illumination geometry and atmospheric water vapour acting together, wherever the evidence supports it. Scoped P04 Knowledge that carries over Current focus Carry what is learned at one place, time or set of conditions over to measurements elsewhere, wherever the same effects apply. Scoped P05 Factoring out an effect Current focus Where the evidence identifies an effect, factor it out so that the measurement moves closer to independent reference values, within its stated uncertainty. Mechanism shown 1 result P06 Confidence that means what it says Current focus State confidence that follows the actual evidence and conditions, so that over many comparisons the errors behave as the stated uncertainty says they should. Observed in real data 1 result

A single sensor What one sensor can and cannot settle on its own.

P07 What one sensor can settle For a single sensor, state how strongly its measurements support different ground values at a chosen wavelength, including the ambiguity that remains because different surfaces can produce the same reading. Scoped

Several sensors together What becomes more certain when sensors are compared at the same wavelength.

P08 Combining sensors Current focus Combine sensors so that a useful conclusion about the ground becomes more confident, for example that a low value is unlikely at a given wavelength, even when several higher values remain possible. Observed in real data 1 result

Delivery and use Keeping meaning intact from measurement to map to decision.

P09 Results on a common grid Current focus Deliver results on a common map grid on which every measurement keeps its meaning, its uncertainty and the record of what shaped it. Observed in real data 1 result P10 Traceable updates Update every affected result when new knowledge arrives, and keep a reproducible record of what changed and why. Mechanism shown 1 result P11 Better-supported decisions Apply confidence information to real questions about real areas, so that the person asking can back a signal, weaken an unsupported explanation, expose an alternative, or see which evidence would help most next. Scoped

§ 02

Approach

No single source of evidence separates every effect in a measurement. Each physical contrast below constrains a different part of the problem; combined, they narrow what any one of them leaves open.

Lunar observations The Moon is a stable reference that a sensor views with no atmosphere in between. Repeated lunar observations isolate the instrument's own response and its drift over time.

Instrumented reference sites Sites such as those of RadCalNet measure surface reflectance and atmospheric state at the time of overpass, with stated uncertainties, giving an independent reference traceable to SI units.

Independent atmospheric measurements Ground-based measurements of aerosol and water vapour constrain the atmospheric state separately from the satellite signal.

Illumination and viewing geometry The same surface seen under different sun and view angles separates effects that depend on geometry from those that do not.

Atmospheric moisture Drier and wetter conditions over the same target expose the spectral imprint of water vapour absorption, and how it changes with geometry.

Cross-calibration between sensors Sensors measuring at the same wavelength over the same ground, close in time, expose differences between instruments, provided shared errors are not counted as independent evidence.

§ 03

Stages

Each area advances through defined stages, from an established method to a result confirmed outside SpectraWorks. A stage is reached when the supporting results are published, with their data, conditions and limits, on a dated page.

Stage Meaning

Scoped The question and method are defined; results are not yet published.

Mechanism shown Demonstrated on designed test data where the right answer is known.

Observed in real data Observed in real satellite measurements, with the conditions recorded.

Validated under stated conditions Checked against independent reference measurements, under the conditions stated.

Independently reproduced Confirmed by a team outside SpectraWorks.

Didn't hold / Open problem A result that went against the objective, or a question still open, stated as plainly as a success. Results that went against the objective are published in the same way, labelled as such, together with what they taught us.

§ Contact

Working on the same problems?

If you research calibration, uncertainty or cross-sensor comparison, or hold reference data that bears on this work, contact us.