- Sensitivity analysis of compositing strategies : modelling and experimental investigations

Context and objectives

Earth observation with optical sensors based on satellite platforms is known to be limited by the interference of clouds and atmospheric constituents like ozone, water vapour and aerosols. High temporal resolution satellites, such as meteorological satellites or the new instrument VEGETATION, provide multiple images of the same site over short periods of time. When several images from the same sensor are available for an area, individual images can be merged to create a temporal composite image in order to reduce the atmospheric perturbations and produce image products devoted to land applications.

The existing compositing strategies produce radiometric artefacts in the spectral bands. This study aims at investigating systematically, the main issues related to the temporal synthesis production using VEGETATION simulated data first and subsequently actual VEGETATION data after launch. The final objective is to proposed new compositing algorithms for the VEGETATION time series production which keep the signal consistency over space and time.

Project outcome

Expected scientific results

The analysis of the 1-year time series shows three nested scales of signal variation : a 5-day cycle related to the viewing angle due to the wide swath sensor, a 26-day cycle corresponding to the satellite orbit and the seasonal sun cycle variable according to the latitude. The perturbing factors have been ranked according to their impact on the signal. The sensitivity analysis highlighted the larger effect of the viewing angle than the atmosphere variability with regards to the day-to-day variation. However the various perturbing factors always showed a coupled effect on the signal. discussed while the comprehensive assessment has to be completed for the global simulated data set, i. e. for most existing, coupled surface-atmosphere environments.

The conclusions drawn from both modelling and experimental approaches have resulted in the proposition of two new image compositing strategies. The MC-FUME proposed by the VITO is based on statistical estimation techniques. The NARASE one has been developed by UCL based on a normalisation of the directional effects. The respective performances of these different compositing algorithms is currently assessed using actual VEGETATION data.

Project leader(s): UCL - Environmental Sciences
Location: Region:
  • World