High spatial resolution missions as Sentinel-2 open opportunities to set-up operational Earth Observation (EO) services at local scale. However, due the frequent cloud coverage in Western Europe, combined with the lower revisit time of high spatial resolution sensors, operational services have to combine data from different missions in order to have sufficiently frequent observations. This is for instance done within the WatchITgrow platform, an EO-based service for potato monitoring and yield prediction, where Sentinel-2 and DEIMOS-1 are combined in order to ensure capturing changes in the potato fields. A seamless combination of EO products coming from different missions is not straightforward. The joint use of data from different sensors raise some clear concerns about data consistency. There might be radiometric biases at TOA (Top-of-Atmosphere) level. Furthermore, even for a perfectly cross-calibrated constellation of instruments, intrinsic differences in the Relative Spectral Response Functions (RSRFs) of comparable bands might cause discrepancies in the final products. Finally, biases in the products can also be introduced through the use of different processors and algorithms (e.g. for the atmospheric correction). One option is to normalize time series at the product level following an empirical approach. The drawback of such an approach is that the root cause of an observed bias is not identified. Furthermore, a non-linear empirical relationship could be found with associated large uncertainty. Finally, such an empirical relationship is only applicable to the sites or time periods for which it has been derived. Within the BELHARMONY project (https://belharmony.vito.be/), which we present here, a bottom-up approach is used instead in order to assess and to improve multi-mission high resolution time series consistency. First, vicarious radiometric calibration methods are used to assess the consistency at the L1 TOA level. This is done for the following sensors: Sentinel-2A/B MSI, Landsat-8 OLI, PROBA-V, and Deimos-1. The bias assessment is performed over targets with low (Bassani and Sterckx, LPS2019) as well as targets with medium radiance levels (Wolters and Sterckx , LPS 2019) . This distinction is required (but often neglected) as, particularly at very low (and also extremely high) scene brightness, the radiometric response of the optical system might not be linear. Next, to adjust the spectral response of one sensor to another, we model the difference that is related to the difference in the RSRFs , allowing to derive band- and/or index-dependent spectral adjustment functions. For this, we use simulated vegetation spectra derived from physically-based radiation transfer models that consider the leaf optical properties, the canopy structure, and the background reflectance. These simulations are completed by adding spectral libraries of non-vegetated surfaces. Finally, we generate enhanced Level 2 and Level 3 time series over the BELAIR sites using MORPHO, a multi-sensor Level 1 to Level 3 processing chain. In the frame of BELAIR, test sites have been set up in Belgium, for which targeted EO data and other measurements are collected on behalf of the Belgian and international research communities, and which may be used as calibration and validation sites for new EO missions, data, and products. Using the extensive set of EO data and in-situ reference data which have been systematically collected over the BELAIR urban, agricultural, and coastal waters sites since 2013, we will evaluate the impact of the different harmonization measures.
The BELHARMONY approach for harmonization of multi-sensor high resolution time series
Bassani C;
2019
Abstract
High spatial resolution missions as Sentinel-2 open opportunities to set-up operational Earth Observation (EO) services at local scale. However, due the frequent cloud coverage in Western Europe, combined with the lower revisit time of high spatial resolution sensors, operational services have to combine data from different missions in order to have sufficiently frequent observations. This is for instance done within the WatchITgrow platform, an EO-based service for potato monitoring and yield prediction, where Sentinel-2 and DEIMOS-1 are combined in order to ensure capturing changes in the potato fields. A seamless combination of EO products coming from different missions is not straightforward. The joint use of data from different sensors raise some clear concerns about data consistency. There might be radiometric biases at TOA (Top-of-Atmosphere) level. Furthermore, even for a perfectly cross-calibrated constellation of instruments, intrinsic differences in the Relative Spectral Response Functions (RSRFs) of comparable bands might cause discrepancies in the final products. Finally, biases in the products can also be introduced through the use of different processors and algorithms (e.g. for the atmospheric correction). One option is to normalize time series at the product level following an empirical approach. The drawback of such an approach is that the root cause of an observed bias is not identified. Furthermore, a non-linear empirical relationship could be found with associated large uncertainty. Finally, such an empirical relationship is only applicable to the sites or time periods for which it has been derived. Within the BELHARMONY project (https://belharmony.vito.be/), which we present here, a bottom-up approach is used instead in order to assess and to improve multi-mission high resolution time series consistency. First, vicarious radiometric calibration methods are used to assess the consistency at the L1 TOA level. This is done for the following sensors: Sentinel-2A/B MSI, Landsat-8 OLI, PROBA-V, and Deimos-1. The bias assessment is performed over targets with low (Bassani and Sterckx, LPS2019) as well as targets with medium radiance levels (Wolters and Sterckx , LPS 2019) . This distinction is required (but often neglected) as, particularly at very low (and also extremely high) scene brightness, the radiometric response of the optical system might not be linear. Next, to adjust the spectral response of one sensor to another, we model the difference that is related to the difference in the RSRFs , allowing to derive band- and/or index-dependent spectral adjustment functions. For this, we use simulated vegetation spectra derived from physically-based radiation transfer models that consider the leaf optical properties, the canopy structure, and the background reflectance. These simulations are completed by adding spectral libraries of non-vegetated surfaces. Finally, we generate enhanced Level 2 and Level 3 time series over the BELAIR sites using MORPHO, a multi-sensor Level 1 to Level 3 processing chain. In the frame of BELAIR, test sites have been set up in Belgium, for which targeted EO data and other measurements are collected on behalf of the Belgian and international research communities, and which may be used as calibration and validation sites for new EO missions, data, and products. Using the extensive set of EO data and in-situ reference data which have been systematically collected over the BELAIR urban, agricultural, and coastal waters sites since 2013, we will evaluate the impact of the different harmonization measures.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
