Using artificial sky brightness to retreive Night time Aerosol optical depth

Vienna EGU 2009 meeting presentation (our very first results)

Estimating Aerosol optical depth

Figure 1: SAND-3 robotized spectrometer

The basic idea here is to find for a given site an empirical relation between the sky brightness and the Aerosol Optical Detph (AOD). This relation should only vary with geographical properties of the vincinity and on seasonal variations in ground reflectance. It may be necessary to define a relation dependance with snow cover. The relation could be constructed for each site and each reflectance pattern by comparing the last evening AERONET mearurement with the first evening SAND-3 measurement along with the first morning AERONET measurement with the last morning SAND-3 measurement. Modelling studies conducted with model ILLUMINA have shown that the relation is complex but quite linear for AOD ranging from 0 to 0.5 (@ 550 nm).

We successfully applied this idea for CARTEL (Sherbrooke, Canada) during march 2009. During this month the snow cover have changed. To get rid of the snow cover change effect, we simply divided the sky brightness measurement by the snow cover fraction. We have made the approximation that when there is some snow on the ground, most of the light comes from the reflexion on the snow and thus the sky brightness should be approximately proportionnal to the ground surface fraction covered with snow.

Figure 2: Typical data acquired by SAND-3

After applying this coarse correction, we obtained the following relation:

If we use this preliminary relationship to convert sky brightness into AOD we obtain the following figure. It is important to say that our modelling works showed that the slope of the relationship is rapidely decreasing for AOD>0.5 so that large SAND-AOD values are certainly underestimated in figure 4. On the figure 4 we can appreciate the good day-night consistency and continuity of the datasets (AERONET and SAND-3). Both datasets are level1 which means that no cloud filtering or quality controll have been applied to the data.

Figure 3: Typical relation between AOD
and Sky brightness (here for Université
de Sherbrooke (Canada) in march 2009)

Figure 4: Resulting SAND-3 AOD during march 2009