Fog Detection with Terra-modis and Msg-seviri

A method is proposed to discriminate bewteen ground and uplifted fog by means of TERRA-MODIS data. First results of the MODIS fog/low stratus detection scheme are presented. The method is based on radiative transfer calculations which provide minimum and maximum fog albedo for 7 spectral bands. Threshold functions which are dependent on the solar zenith angle are used for 7 initial tests. The mask is completed by means of a snow-/ice cloud and aerosol removal test and a check against contaminations due to mid level water clouds. The method also recognizes pixels partly covered by fog. Fog geometrical thickness is determined by means of a 2D-iteration method based on RTC-derived lookup tables and least square variance analysis. Finally, the potential of MSG-SEVIRI for the nowcasting of fog/low-level stratus both day and night is investigated. 1. THE FOG PROBLEM IN REMOTE SENSING Fog is a common obstacle for land, air and sea traffic. However, no reliable method is available for nowcasting of fog on a spatio-temporal basis. The observation network of horizontal visibility is sparse and numerical models are still too slow and not yet sufficiently accurate to provide for the spatio-temporal nowcasting of fog extent, fog formation and fog clearance (COST722 2003). Common techniques for space-borne fog/low stratus detection are using the temperature difference between IR11μm and IR3.7μm aboard the NOAA satellites, both day (only possible until NOAA-14) and night (EYRE et al. 1984, TURNER et al. 1986, BENDIX & BACHMANN 1991, DYBBROE 1993, GREENWALD & CHRISTOPHER 2000, BENDIX 2002). Because the old Meteosat system does not provide IR3.7 μm data, fog detection could only be performed by means of the broadband solar channel, with reduced accuracy especially in snowy conditions (GÜLS & BENDIX 1996). The IR3.7–based algorithm was recently successfully implemented on the basis of the new GOES-series (GOES-8+) which provides all requried spectral bands to GEO systems for the first time (ELLROD 1995, LEE et al. 1997). Unfortunately, one major problem in fog detection by satellite data is not solved until today. The mentioned threshold-techniques are not able to discriminate between ground fog (visibility < 1km at ground level) and low level stratus (=uplifted ground fog, hill fog, visibility at ground > 1 km) (Fig. 1). On the other hand, new LEO/GEO-instruments give new opportunities to discriminate between ground fog and low-level stratus. In the current study, the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the TERRA and AQUA platforms is used. This instrument provides 29 spectral bands in 1 km, 5 bands in 500 m and two bands in 250 m resolution (KING et al. 1992). Figure 1: The fog problem in remote sensing 2. OVERALL METHODOLOGY A comprehensive scheme to detect ground fog from MODIS data should address four main problems (Fig. 2). (Part I) The scheme must start with a gross check on fog/low stratus-covered pixels. (Part II) To discriminate between fog and low level stratus, fog top and base heights must be determined. If LWP and LWC and/or extinction and optical depth are known, fog geometrical thickness could be easily derived by (BENDIX 1995b, HUTCHISON 2002).