RECONSTRUCTING FINE-SCALE AIR POLLUTION STRUCTURES FROM COARSELY RESOLVED SATELLITE OBSERVATIONS

The distribution of short-lived air pollutants such as NO2 is closely tied to geographically fixed emission sources. Variations due to changing weather conditions or emission strengths may be considered as noisy deviations from a mean picture. In this study we investigate to what extent fine scale details of (NO2) air pollution structures can be reconstructed from coarsely resolved satellite observations. For this we set up an idealized test environment where both the original distribution to be reconstructed and the level of noise of the observations are known, and apply a number of iterative image reconstruction algorithms originally developed for application in computer tomography. In the case of noise-free observations, the original distribution can be completely recovered except for spatial frequencies where the spectrum of the aperture function (the pixel rectangle) exhibits nulls. For noisy observations the situation is more complicated and the success of reconstruction critically depends on the level of noise and the spatial density of the sample.