The pseudo-invariant calibration sites (PICS) as deserts have been exploited for decades for multi-temporal monitoring and cross-calibration in the solar spectral range. The PICS algorithm aims to simulate the TOA reflectance in the visible to near-infrared (NIR) spectral range of the sensor under test (e.g. OLCI) over a predefined desert site (e.g. CEOS PICS sites). This method makes it possible to perform multi-temporal analysis, as well as a comparison of multiple sensors on the same site over the same or different observation periods.
The PICS method supported by DIMITRI has been successfully performed. The figure below displays time-series of the elementary ratio (estimated gain coefficients) -defined as observed reflectance to the simulated one- over all the PICS calval sites. First, a clear consistency over all the 6 PICS could be seen. The averaged ratios over the time-series –depicted in the top of each plot- show bias values within <4% and uncertainty about ±2.5% for almost all the spectral range of OLCI from Oa01-Oa19 even some bands display about 2% bais (e.g. Oa02, Oa03, Oa05, Oa08 and and Oa17; Figure below). Note that the bias is strongly impacted by the high ratios observed during the first 6 months of the mission.
Time-series of the elementary ratios (observed/simulated) signal from S3A/OLCI for (top) band Oa8 and (bottom) band Oa17 over six PICS Cal/Val sites. Dashed-green and orange lines indicate the 2% and 5% bias respectively. Error bars indicate the elementary uncertainty of PICS-methodology.
Verification of the geo-location accuracy is done with a specific tool (GEOCAL) running at the MPC/CC. June results confirm very good performance and that OLCI is compliant with mission requirements: the centroid of the geolocation error is around 0.2 pixel in both along-track and across-track directions.
Histograms of geolocation errors for the along-track (left) and across-track (right) directions.
Geo-referencing error in along-track (left) and across-track (right) directions for all the GCPs.
Completion of the time series (started using the partial reprocessing dedicated to validation: 4 days every month between 26/04/16 and 12/03/2017) and continued afterwards is displayed in the following figure.