Abstract
The UK’s Earth Observation Climate Information Service (EOCIS) is a new programme of activity intended to provide policy makers and businesses in the UK with the information they need to respond to climate change. EOCIS brings together UK expertise in Earth Observation to create and make available high quality, trustworthy climate information based on measurements of Earth’s environments from space – ranging from polar ice change to carbon cycling to drought. Typical climate products, aimed at global analyses at decadal time scales, are generated from wide-swath, low-resolution sensor observations to achieve the necessary coverage. To address the needs of a wider community of both public and commercial sector climate data users at a national level EOCIS is developing a series of novel climate products, with a particular focus on integrating high-resolution sensor observations. This activity is referred to as CHUK (Climate-data at High-resolution for the UK). The objective of the work presented is to underpin the quality of the derived CHUK products by ensuring the input radiometric observations across the considered sensors are cross-calibrated to a common reference with quantified uncertainties. This will ensure a consistency between the considered high-resolution sensors and the traditional, lower resolution climate sensors. This consistency is not always achieved by these sensors’ operational calibrations, limiting the quality of derived climate products. In the visible/near-infrared sensing domain Sentinel-3A/B OLCI, Sentinel-2A/B MSI and Landsat-8/9 OLI will be considered. In the thermal infrared domain Sentinel-3A/B SLSTR and Landsat-8/9 TIRS will be considered. To achieve this, collocated sensor observations (or matchups) are identified and processed to compensate for differences in spectral and spatial sampling, as well as other observation mismatches, such as in viewing geometry. Uncertainty is propagated through these modelling steps. In addition, matchups are also found to a chosen common reference. For sensors where Level 0 data (i.e. measurement function input quantities such as detector counts) are available the recalibration itself is a large non-linear regression problem, solving for new calibration parameters in the defined measurement equation of each sensor against the reference. A novel error-in-variables (EIV) algorithm is used to perform this optimisation considering the full error-covariance structure of the matchup data to avoid biases generated by algorithms that don’t consider this. Where Level 0 data is not available a more simple recalibration process may be required. Presented here are the initial results of this activity.
Intercalibration to Facilitate the Generation of Climate Data at High-Resolution
The UK’s Earth Observation Climate Information Service (EOCIS) is a new programme of activity intended to provide policy makers and businesses in the UK with the information they need to respond to climate change. EOCIS brings together UK expertise in Earth Observation to create and make available high quality, trustworthy climate information based on measurements of Earth’s environments from space – ranging from polar ice change to carbon cycling to drought. Typical climate products, aimed at global analyses at decadal time scales, are generated from wide-swath, low-resolution sensor observations to achieve the necessary coverage. To address the needs of a wider community of both public and commercial sector climate data users at a national level EOCIS is developing a series of novel climate products, with a particular focus on integrating high-resolution sensor observations. This activity is referred to as CHUK (Climate-data at High-resolution for the UK). The objective of the work presented is to underpin the quality of the derived CHUK products by ensuring the input radiometric observations across the considered sensors are cross-calibrated to a common reference with quantified uncertainties. This will ensure a consistency between the considered high-resolution sensors and the traditional, lower resolution climate sensors. This consistency is not always achieved by these sensors’ operational calibrations, limiting the quality of derived climate products. In the visible/near-infrared sensing domain Sentinel-3A/B OLCI, Sentinel-2A/B MSI and Landsat-8/9 OLI will be considered. In the thermal infrared domain Sentinel-3A/B SLSTR and Landsat-8/9 TIRS will be considered. To achieve this, collocated sensor observations (or matchups) are identified and processed to compensate for differences in spectral and spatial sampling, as well as other observation mismatches, such as in viewing geometry. Uncertainty is propagated through these modelling steps. In addition, matchups are also found to a chosen common reference. For sensors where Level 0 data (i.e. measurement function input quantities such as detector counts) are available the recalibration itself is a large non-linear regression problem, solving for new calibration parameters in the defined measurement equation of each sensor against the reference. A novel error-in-variables (EIV) algorithm is used to perform this optimisation considering the full error-covariance structure of the matchup data to avoid biases generated by algorithms that don’t consider this. Where Level 0 data is not available a more simple recalibration process may be required. Presented here are the initial results of this activity.