Modélisation de l'évolution spatiale et temporelle de l'épaisseur optique des aérosols à l'échelle régionale
Monitoring of aerosol optical depth (AOD) is of particular importance due to the significant role of aerosols in the atmospheric radiative budget. AOD is a key parameter in studies related to global climatology, atmospheric pollutants, forest fires, and for performing atmospheric corrections on remotely sensed imagery of surface scenes. We attempt to fill gaps in spatio-temporal AOD measurements using a new methodology that links AOD measurements and particulate matter Transport Model (TM) using a data assimilation approach. This new modelling package (AODSEM for A[barbelow]erosol O[barbelow]ptical D[barbelow]epth S[barbelow]patiotemporal E[barbelow]volution M[barbelow]odel) uses a size and aerosol type segregated semi-Lagrangian-Eulerian trajectory algorithm driven by analysed meteorological data. Its novelty resides in the fact that the model evolution is tied to available AOD measurements and all physical processes have been optimized to track this important but crude parameter. We present in this paper a sensitiviity study to AODSEM's important parameters or processes (spatial and temporal resolution, hygroscopic effects, spin-up time, precipitations, size distribution). We also present the first validation results for this new model applied to North America during June 1997 and august 1998. The results show the potential of this approach especially when used with remotely sensed AOD. Residuals between AODSEM analysis and measurements are smaller than typical errors associated to remotely sensed AOD. AODSEM also give better results than classical interpolation schemes. This result is especially evident when the available number of AOD measurements is small. Our results shows that AODSEM can be efficient to correct for error induced by simplifications in the physical parametrization and by error in emission inventories. Our results also shows that in the context of low polluted zone and period, AODSEM can be used to provide AOD spatio-temporal forecasts. In that case, correlation between AODSEM and independent AERONET data can reach r = 0,86 with a mean residual of [left angle bracket][delta]T[subscript a] [right angle bracket] = 0,02 and [sigma] = 0,04. For a specific pollution event, AOD forecast can also be achieved when an episodic emission inventory is available"--Résumé abrégé par UMI.