Contribution de Rihab MECHRI, Catherine OTTLÉ, Olivier PANNEKOUCKE, Abdelaziz KALLEL & Ahmed BEN HAMIDA:


Sub-pixel temperatures estimation based on the assimilation of coarse resolution Thermal Infrared data using particle filtering



Thermal infrared (TIR) data is efficiently used for surface fluxes estimation giving the possibility to assess energy budgets through surface temperature. However, an accurate knowledge of such data at high spatial/temporal resolution is not possible considering the present instruments on board satellites. In fact, available instruments allow either the high spatial resolution versus a low temporal one (e.g. ASTER: repeat cycle of 16 days/spatial resolution of 15m to 90 m) or the high temporal resolution with a coarse spatial one (e.g. SEVIRI: repeat cycle of 15 min/spatial resolution of 3km). Then, it is necessary to develop methodologies to combine these multi-scale and multi-temporal data to better monitor fluxes at appropriate scale. Our approach consists in the development of a new downscaling method based on particle filtering to extract sub-pixel variables from large scale data measurements. This methodology consists in constraining surface temperatures trajectories simulated by a dynamic model and aggregated at the scale of the observations. It is based on the use of the SETHYS land surface model (Coudert et al., 2006) and was developed on a synthetic database based on the French "La Crau" region landscape and climate. First step is to generate ensemble of end-members temperatures (first guess temperature of each land cover class present in the TIR pixel). Second step is to aggregate the temperature for each ensemble member given a high resolution land cover mapping which results in a new ensemble containing random coarse spatial resolution temperatures. Last step consists in the selection of the optimal large scale temperature estimations that fit the observed temperatures using the particle filter. In this way, we can get for each end-member present in the TIR pixel, the best estimates of sub-pixel temperatures. Firstly, a sensitivity analysis has been performed to extract for each land cover class the most sensible model parameters on the surface temperature. Secondly, the new downscaling approach will be presented and its performances will be analyzed in terms of errors on the model and on the observations. Finally the performances and robustness of our approach will be tested on actual data and compared to classical Bayesian inversions.

up arrow