Critical to the understanding of dust processes at local, regional, and teleconnected global scales is an observing system capable of resolving the spatio/temporal variability of dust globally. The most practical approach to collecting these data is the satellite platform. The current constellation of environmental satellites includes optical-spectrum radiometers capable of detecting and quantifying the properties of dust. The most capable among these sensors (i.e., offering the greatest diversity of spectral information) operate on low-earth-orbiting satellites (LEO), providing a marked tradeoff of superior characterization at the expense of poor temporal refresh rate in comparison to geostationary (GEO) satellites. However, the ‘spectral ly challenged' state of the GEO platform continues to improve as next-generation radiometers join its ranks. The first among them, the EUMETSAT Meteosat Second Generation (MSG), has introduced to GEO new spectral bands that are very useful in dust detection. The situation will continue to improve with the launch of Advanced Baseline Imagers (ABI) on the next-generation Japanese and U.S. geostationary satellites over the next several years in addition to Meteosat Third Generation (MTG).
This paper describes a new, multi-spectral satellite algorithm developed on MSG/SEVIRI that utilizes ancillary surface data in an attempt to overcome traditional challenges to the detection of dust over barren land surfaces. Specifically, the new approach incorporates the UW/CIMSS Baseline Fit surface emissivity database, derived from MODIS data and high resolution laboratory spectra for different surface types, as a means to suppressing the erroneous enhancement of land surface features while retaining the ability to detect dust above these surfaces. The algorithm is applicable to both day and nighttime conditions, over land and water, and makes use of an optimal combination of spectral information for each condition and a blend of the two across the terminator for near-seamless transition. The detection is provided quantitatively as a confidence factor [0,1], but is readily visualized as value-added imagery, presented in the context of the meteorological situation responsible for the dust lofting. In this way, the algorithm is potentially useful to both automated processes and human users alike. Examples from notable dust storms and comparisons to other detection methods will be presented.