Software to harmonise fundamental climate data records

Implementing policies designed to fulfil Paris Agreement commitments will likely involve conspicuous demands for clear accountability for how decisions will be made. Such accountability will, in turn, demand a high bar for trustworthiness of data used as the basis of decisions. So, how can we be sure trusted climate data will be available in time to make a difference?

In his opening address at COP27 in Sharm el-Sheikh on 6 November, UN Climate Change Executive Secretary Simon Stiell asked governments to focus on a transformational shift to implementing the Paris Agreement and putting negotiations into concrete actions. With a call to action for principles of transparency and accountability to apply throughout the process, Mr Stiell added, “COP27 sets out a new direction for a new era of implementation: where outcomes from the formal and informal process truly begin to come together to drive greater climate progress — and accountability for that progress”.

The litmus test will be how far deliberations lead to meaningful action, but if climate progress requires accountability, a foundation for retaining commitments to net GHG emission reduction strategies will be the establishment and maintenance of trust. In turn, a vital component of trust will necessarily be global confidence in the reliability of data.

Earth observation technologies will provide the backstop for any trusted system supporting the Paris Agreement, for which prototype sensors are being developed, such as the CNES-led MICROCARB mission to be launched in 2024. However, the necessary foundational layer of trust in the reliability of climate data for such earth observation systems is not, at this point, in place.

That’s where the MetEOC-4 project is working to fit together important components of the complex jigsaw of a system to support measurements; precisely enough to provide actionable long-term climate trend data. Speaking just after its halfway point, NPL Senior Research Scientist Sam Hunt provided context for one missing part of the puzzle, the limited performance of satellite calibration and validation methods.

According to Sam, “Satellite sensors in orbit today are not in agreement. Even for the best satellite sensors, differences between measurements are in the order of one to three per cent right now. The effect of that lack of accuracy is to mask underlying trends in measurements of climate-related data.”

The question then arises, which sensors deserve to be trusted? The best answer right now is, as Sam puts it, “It’s hard to say…”.

Non-interoperability limiting data utility

Within MetEOC, Sam and colleagues are working on solutions to enable sensor-to-sensor post-launch interoperability, the absence of which limits the utility of earth observation data, especially for applications requiring high accuracy such as climate data records.

Explaining the connection, Sam said, “These differences in earth observation measurements between satellites of a certain scale will mask subtle climate effects. This limits the whole earth observation system, which could (otherwise) become an interoperable system of systems in which data from different satellites would draw the same conclusions and be used in a series to see what’s happening over time.”

A way that would bring credence to data, especially suited to the most rigorous science applications, is to apply metrological principles. For climate records that would mean the inclusion of information about uncertainties over decades, which is currently not possible.

Sam is leading the research effort within MetEOC-4 to raise upper limits of achievable performance of satellite calibration and validation infrastructures, by developing tools to iron out incompatibilities of sensor outputs, so measurements may be ‘interoperable’, and comparable in a metrological sense.

Efforts to enhance the utility of sensor data are also underway, by developing more accessible data formats.

Bringing metrology to long term climate data

Since the International System of Units (SI) was established, metrology has served science and society at large by delivering measurement stability and consistency through principles of traceability, uncertainty analysis, and comparison. The payoff is data that has associated estimates of uncertainties traceable to the SI, and the benefit of scientific defensibility of applications of data.

However, classic concepts of laboratory-based metrology don’t directly translate to the challenges of Earth Observation, so SI-traceable reference standards are not yet established in orbit.

Currently, sensors are characterised and data validated by:

• Onboard sensor calibration systems, such as reflectors characterised pre-flight, or reference blackbody sources — both subject to change from launch and in-flight
• Vicarious calibration methods, using natural phenomena such as sun glint from water, Rayleigh scattering, in-situ measurements such as RadCalNet sites or pseudo-invariant sites such as the Sahara, Antarctica, and the Moon for solar reflective sites
• Comparisons with other reference satellites, either through matchups (near-simultaneous viewing of the same scene) or using pseudo-invariant sites as transfer standards

Even when sensor calibration systems are characterised pre-flight in a fully SI-traceable manner, traceability in orbit is challenged by the stresses of launch, the harsh environment of space, and degradation over time. On-the-ground validations of geophysical quantities retrieved from EO measurements also can’t readily establish SI-traceability, while uncertainties propagated from ground measurements to the top of the atmosphere, as well as differences in ground and satellite observations, may be larger than required to differentiate climate effects. Comparisons with well-calibrated satellite sensors provided the most effective way to achieve harmonisation and if these ‘reference’ satellites have adequate stability, at some point, uncertainty to the SI could be established.

For all these comparison methods, but particularly those using reference satellites, suitable software is required to facilitate bias assessment and interoperability.

Developing Fundamental Climate Data Records

Access to good climate data matters because climate scientists require accessible, credible, and relevant climate data so they may help decision-makers plan effectively. For long-time series of earth observation data, needed to detect changes, a missing component is uncertainty information attributable to environmental changes.

Applications of earth observation data are produced using sequences of data transformations of lower-level data sets to higher-level products, to ensure that higher-level products can be relied upon. Currently, uncertainty information in higher-level datasets is absent or deficient.

Moreover, in each data transformation, the uncertainty in lower-level data propagates to the higher level. Therefore, a barrier to improved uncertainty information in higher-level products is the absence of uncertainty information in lower-level data sets.

Climate Data Records (CDRs) are designed to provide trusted information about how, where, and to what extent the land, oceans, atmosphere, and ice sheets change over time. Such records are made up of merged data – such as surface temperatures and albedo — measured from surface-, atmosphere-, and space-based systems over decades or change signals to be large enough to be detected. In some cases, efforts are being made to exploit historic data that was once collected for different purposes, for example, the US National Oceanic and Atmospheric Administration offers long-term climate records using modern data analysis to historic satellite data to clarify underlying climate trends.

Fundamental Climate Data Records (FCDRs) include sensor data (such as calibrated radiances and brightness temperatures) that have associated calibration data.

The potential value of applying metrological rigour to satellite sensor calibrations is clear: long-term climate records for which changes due to climate change can be discriminated from natural, random, or systematic errors in satellite sensors.

However, as described earlier, satellite observations often do not have robust measurement traceability to the SI at the necessary uncertainty levels. Furthermore, uncertainty analysis is challenged by large data volumes, complex data processing chains, and complex error-covariance structures of the derived spectral imagery, plus satellite observations tend not to be repeatable.

Software to deliver data interoperability for future missions

“Looking at EO data series produced by sensors in the past, we typically find sensor-to-sensor calibration differences that drift over time, impacting derived geophysical products”, said Sam.

“In MetEOC, we’re attempting to solve this problem by developing a validated method to cross-calibrate satellite sensors against a reference sensor, plus provide rigorous estimations of uncertainty. In future, such a system will make use of TRUTHS as its reference sensor and provide a means to disseminate this gold-standard calibration.”

This task dovetails with parallel tasks and will assimilate knowledge developed in related projects.

The Horizon 2020 Fidelity and Uncertainty in Climate data records from Earth Observation (FIDUCEO) project, completed in 2019, devised a way to perform metrologically rigorous uncertainty analyses for historical sensors en route to processing raw data to FCDRs. FIDUCEO made use of ‘matchups’ to recalibrate historical satellite data. Four sample FCDRs were produced, each with observational uncertainty information, from sensors spanning the visible, infrared, and microwave. The propagation of this information to derived geophysical datasets was demonstrated by creating five climate data records (CDRs) with traceable uncertainty and stability estimates, a process that combined FCDR data from different spectral channels and FCDR data values from different observational pixels.

FIDUCEO is also built on QA4EO, which is a best practice framework endorsed by The Group on Earth Observations (GEO) to establish The Global Earth Observation System of Systems — based on coordinated and harmonised processes and activities that enable interoperability.

FIDUCEO set new standards for accuracy and rigour and has been influential within the earth observation community by showing that metrological best practices can provide defensible uncertainty and stability information.

Sam helped optimise software that calibrated FIDUCEO satellite data records and in MetEOC-4 leads the development of software that will cross-calibrate future satellites. The plan envisions enhancing the FIDUCEO software to produce metrologically rigorous tools able to harmonise FCDRs and remove biases between sensors, suitable for future greenhouse gas emission sensing satellites such as the EU CO2M mission.

Currently, NPL and the University of Reading are developing software to generate sensor-to-sensor match-up datasets for comparison and inter-calibration studies.

Software features developed for MetEOC-3, QA4EO, and FIDUCEO will later be combined, along with the refined record matchup software, into one software suite for developing harmonised FCDRs.

Simulated satellite series matchups will then be used to validate the harmonisation software and to compare the results of this method with other inter-comparison methods.

After that, NPL supported by Rayference will develop methods to produce high-quality, uncertainty-quantified reference data to independently validate top-of-atmosphere (TOA) sensor data. This builds on outputs of MetEOC-3 that added understanding to how to conduct satellite-to-ground comparisons over desert sites and will extend that to vegetated sites.

NPL will also test a lunar irradiance model developed in the ESA LIME project that could offer rigorous uncertainty analysis and consider uncertainties for validating TOA sensor data. This will explore the potential of using the moon as an alternative reference source for calibrating satellite sensors.

Putting in the metrology groundwork to exploit TRUTHS data

Looking ahead to the likely impact of the work, Sam added, “We want to come up with a method to do cross-calibrations to higher accuracies than anyone currently does it, and with well-understood uncertainty”.

“The ideas and algorithms we’re developing in MetEOC are expected to be implemented in the TRUTHS mission. This sensor-to-sensor interoperability solution will make sure that, after launch, measurements remain calibrated and traceable, producing data of a quality that’s fit for purpose. This strand is one way to do that and is putting in the metrology groundwork to do TRUTHS cross-calibrations effectively.”

Thanks in part to NPL’s leadership and coordination roles over the years, MetEOC has supported the development of TRUTHS from concept to reality. The TRUTHS mission is being developed to bring metrological laboratory levels of certainty to in-flight calibration of earth observation satellites. Conceived by NPL, led by the UK Space Agency (UKSA) and delivered by the European Space Agency (ESA), the mission will be the platform for a primary standard (cryogenic radiometer) in space, mimicking methods used on the ground, allowing SI-traceable recalibrations of instruments in orbit.

In the final communiqué for the October 2022, GCOS Climate Observation Conference declared its commitment to supporting the development of a ‘comprehensive and sustainable global climate observing system’. The statement also claimed many climate observations are underexploited because of the lack of consistency, and clarity, in their processing, interoperability and usability.

Climate data collected from earth observation techniques will be required to be credible to serve rigorous climate science applications. Complementary software tools and methods being developed and implemented in time to exploit TRUTHS data. Following the launch of the mission, trustworthy information about variability and change over decades will be available for the first time, serving climate observation communities’ need for interoperable, usable, and, ultimately, trusted climate data.