Abstract
Society is becoming increasingly dependent on remotely sensed observations of the Earth to assess its health, help manage resources, monitor food security, and inform on climate change. Comprehensive global monitoring is required to support this, necessitating the use of data from the many different available sources. For datasets to be interoperable in this way, measurement biases between them must be reconciled. This is particularly critical when considering the demanding requirements of climate observation – where long time series from multiple satellites are required.
Typically, this is achieved by on-orbit calibration against common reference sites and/or other satellites, however, there often remain challenges when interpreting such results. In particular, the degree of confidence in the resultant uncertainties and their traceability to SI is not always adequate or transparent. The next generation of satellites, where high-accuracy on-board SI-traceability is embedded into the design, so-called SITSats, can therefore help to address this issue by becoming “gold standard” calibration references. This includes the ESA TRUTHS (Traceable Radiometry Underpinning Terrestrial- and Helio- Studies) mission, which will make hyperspectral observations from visible to short wave infrared with a target uncertainty of 0.3 % (k = 2).
To date, uncertainty budgets associated with intercalibration have been dominated by the uncertainty of the reference sensor. However, the unprecedented high accuracy that will be achieved by TRUTHS, and other SITSats, means that the reference sensor will no longer be the dominant source of uncertainty. The accuracy of cross-calibration will instead be ultimately limited by the inability to correct for differences between the sensor observations in comparison, e.g., spectral response, viewing geometry differences.
A matchup is defined as an event where two satellite sensors observe approximately the same location at approximately the same time, commonly referred to as a simultaneous-nadir overpass (SNO). Analysis of matchups is a key technique to compare and cross-calibrate EO sensors in flight. Several observational aspects make the sensor comparison challenging, including mismatch in spectral and spatial sampling, viewing geometry mismatch (changes in surface reflectance, polarization), and temporal mismatch (changes in illumination angle, atmosphere). These effects introduce a comparison mismatch, responsible for further uncertainty in the comparison beyond the uncertainties of the compared measurements themselves.
Sensor measurement data is processed to make them more comparable, including accounting for the spectral response function to match that of the sensor under test, and sampling the data spatially such that the two sensors observe the same field of view. Presented here is an end-to-end simulation of this cross-calibration processing chain. This is used to assess the performance of the TRUTHS-sensor cross-calibration by evaluating the uncertainty achievable in a set of representative simulated matchups for a core set of target sensors (e.g., Sentinel-2 MSI, Sentinel-3 OLCI).
End-to-end Simulation of Cross-calibration Performance for On-orbit Reference Calibration of Earth-viewing Sensors
Society is becoming increasingly dependent on remotely sensed observations of the Earth to assess its health, help manage resources, monitor food security, and inform on climate change. Comprehensive global monitoring is required to support this, necessitating the use of data from the many different available sources. For datasets to be interoperable in this way, measurement biases between them must be reconciled. This is particularly critical when considering the demanding requirements of climate observation – where long time series from multiple satellites are required.
Typically, this is achieved by on-orbit calibration against common reference sites and/or other satellites, however, there often remain challenges when interpreting such results. In particular, the degree of confidence in the resultant uncertainties and their traceability to SI is not always adequate or transparent. The next generation of satellites, where high-accuracy on-board SI-traceability is embedded into the design, so-called SITSats, can therefore help to address this issue by becoming “gold standard” calibration references. This includes the ESA TRUTHS (Traceable Radiometry Underpinning Terrestrial- and Helio- Studies) mission, which will make hyperspectral observations from visible to short wave infrared with a target uncertainty of 0.3 % (k = 2).
To date, uncertainty budgets associated with intercalibration have been dominated by the uncertainty of the reference sensor. However, the unprecedented high accuracy that will be achieved by TRUTHS, and other SITSats, means that the reference sensor will no longer be the dominant source of uncertainty. The accuracy of cross-calibration will instead be ultimately limited by the inability to correct for differences between the sensor observations in comparison, e.g., spectral response, viewing geometry differences.
A matchup is defined as an event where two satellite sensors observe approximately the same location at approximately the same time, commonly referred to as a simultaneous-nadir overpass (SNO). Analysis of matchups is a key technique to compare and cross-calibrate EO sensors in flight. Several observational aspects make the sensor comparison challenging, including mismatch in spectral and spatial sampling, viewing geometry mismatch (changes in surface reflectance, polarization), and temporal mismatch (changes in illumination angle, atmosphere). These effects introduce a comparison mismatch, responsible for further uncertainty in the comparison beyond the uncertainties of the compared measurements themselves.
Sensor measurement data is processed to make them more comparable, including accounting for the spectral response function to match that of the sensor under test, and sampling the data spatially such that the two sensors observe the same field of view. Presented here is an end-to-end simulation of this cross-calibration processing chain. This is used to assess the performance of the TRUTHS-sensor cross-calibration by evaluating the uncertainty achievable in a set of representative simulated matchups for a core set of target sensors (e.g., Sentinel-2 MSI, Sentinel-3 OLCI).