By: Jonathan Mittaz
Metrology is the science of measurement which both defines the System International (SI, The International System of Units, 2019) as well as mathematical frameworks for measurement uncertainties (for example see the GUM: Guide to the expression of Uncertainty in Measurement, 2008). In an ideal world all measurements would be linked back to the SI and hence to a fixed, unchanging reference. Such traceability would, of course, also be the gold standard for satellite climate data records but we are yet to attain this.
Over the past 10-15 years there has been a move to incorporate metrological principles into Earth Observation data. At first there were the QA4EO principles (Quality Assurance Framework for Earth Observation: Principles 2010) leading to an initial principle that:
“Data and derived products shall have associated with them a fully traceable indicator of their quality”
Fully traceable quality was not fully defined but it basically means that all data needs information to help users determine how good the data is in a way that traces back to a reference, ideally to the SI. Since then, a number of initiatives have started to fill in the details of how to provide quality information for climate data. For example, the ESA Climate Change Initiative (ESA CCI) now has a requirement on providing uncertainties (something that was generally lacking in previous data), and projects such as the Copernicus Climate Change Service (C3S) Data Store provides quality information for some datasets including independent quality assessments. From a metrological perspective, however, what is needed is proper traceability where all uncertainties are traced from their physical origins through to the final signal in a bottom-up manner. In the case of earth observation (EO) data this means tracing errors from both the instrumentation (which measures the incoming signals) as well as fully understanding the errors associated with any retrieval process (where instrument measurements (such as radiances) are converted into the particular geophysical parameter of interest (such as a temperature)).
Work is now in progress to provide metrological traceability to earth observation data. For example, the Horizon 2020 FIDUCEO project has defined metrological methods for satellite data (see the FIDUCEO web site and Mittaz, Merchant & Woolliams 2019) and the Horizon 2020 GAIA-CLIM project (see the GAIA-CLIM website) looked at applying metrological principles to non-satellite measurements such as in-situ reference data. What these projects have shown is that uncertainties are, in fact, complex with both space and time variable uncertainties which also have a range of error correlations (which arise when parts of the errors used to determine the uncertainties are correlated over time or space and which must to be taken into account when propagating uncertainties see Mittaz, Merchant & Woolliams 2019). Some examples of the complexity of error in a sensor are shown in Figure 1 and Figure 2. Figure 1 provides an example of a FIDUCEO uncertainty tree for the SLSTR instrument (Sea and Land Surface Temperature Radiometer, Smith et al. 2021) which demonstrates that the uncertainties are invariably complex because the fundamental sources of error (the outer parts of the uncertainty tree in black text) are many. Figure 2 shows different uncertainties (independent, structured and common) from the Advanced Very High Resolution Radiometer (AVHRR) sensor and shows that different components of uncertainty can also have very different temporal characteristics which are also very different between different versions of the same instrument.
Figure 1: An uncertainty tree diagram for the SLSTR sensor (taken from Smith et al. 2021
Figure 2: Plots showing three different types of uncertainty (independent (red), structured (blue) and common (green)) for an infrared channel (10.8µm) of selected AVHRRs as a function of time.
The metrological community through the National Metrological Institutes (NMIs, of which the National Physical Laboratory is the UKs) have also been leading initiatives. These include the MetEOC projects (Metrology for Earth Observation and Climate), which aims to contribute to the establishment of a metrology infrastructure tailored to climate needs in readiness for its use in climate observing systems. At a European level there is also the European Metrology Network (EMN) for Climate and Ocean Observation, which is a network of European NMIs and affiliated partners to support the application of metrology to climate and ocean observations. Both these projects aim to continue applying metrological techniques to new instrumentation, new models, improved in-situ reference data and an improved understanding of satellite uncertainties.
Finally, there is the question of linking EO and climate data to the SI. Some in-situ references are already in the process of being traced to the SI (such the GCOS Reference Upper-Air Network (GRUAN) and the ESA Fiducial Reference Measurements for validation of Surface Temperatures data (FRM4STS)). But for satellite data there is currently no way do this as there are no in-orbit SI traceable references. But this is about to change. The upcoming TRUTHS mission (to be launched around 2028) and the similarly specified CLARREO Pathfinder mission (due for deployment in 2024) both aim to provide in-orbit SI traceable references in the reflectance (visible) domain. Therefore, within the next 10 years we should have the capability of providing satellite climate data which are traced to the SI, have metrologically based uncertainties and provide, for the first time, the best climate data records possible.