Detection of dust over deserts using satellite data in the solar wavelengths

Dust is a dominant feature in satellite images and is suspected to extract large radiative forcing of climate. While remote sensing of dust over the d ark oceans is feasible, adequate techniques for remote sensing over the lan d still have to be developed. Here, similar to remote sensing of aerosol ov er vegetated regions, we use a combination of visible and mid-IR solar chan nels to detect dust over the desert. Analysis of Landsat TM images over Sen egal taken in 1987 show that the surface reflectance at 0.64 mu m is betwee n 0.54 +/- 0.05 of the reflectance at 2.1 mu m, and reflectance at 0.47 mu m is 0.26 +/- 0.03 of that at 2.1 mu m, surprisingly similar to relationshi ps in non-desert sites. We also found that dust have only a small effect on the surface + atmosphere reflectance at 2.1 mu m over the desert. Therefor e, in the presence of dust, we use the Landsat TM data at 2.1 mu m channel to predict the surface reflectance at 0.64 and 0.47 mu m. The difference be tween the satellite-measured reflectances of surface + atmosphere and the p redicted surface reflectances is used to derive the dust-optical thickness r at 0.64 and 0.47 mu m. Results show that r can be derived within Delta ta u = +/-0.5 for the range of 0 < tau < 2.5, thus enabling detection of dust sources and the estimation of three to five levels of dust opacity over the desert. The method is very sensitive to the correct knowledge of the dust absorption and is equally sensitive to dust in the entire atmospheric colum n. It is best applied in the red part of the spectrum (around 0.64 mu m), w here dust was found to be weak-absorbing or nonabsorbing, We plan to use th is method as part of the dust monitoring from the Earth-observing system MO DIS instrument.