Determination of fog horizontal visibility by means of NOAA-AVHRR

Determination of fog horizontal visibility by means of NOAA-AVHRRJörg BendixUniversity of Bonn, Department of GeographyMeckenheimer Allee 166, D-53115 Bonn (Germany)Tel.: ++49-228-737206, Fax: ++49-228-737506ABSTRACTTwo schemes for the determination of fog horizontal visibility using AVHRR data both, day and night, are presented. Ground truth has been performed by the comparison of calculated visibilities and simultaneous transmissometer measurements. The daytime approach is based on the calculation of fog albedo from AVHRR #1 and the relation of fog optical depth and geometrical thickness to spectral extinction and horizontal visibility. The mean deviation between observed and calculated visibilities is about 56 m. The nighttime approach uses the relation between fog and air temperature derived from AVHRR # 4 and 5 and a simple model for the approximation of the fog droplet distribution. The resulting fog extinction is directly related to horizontal visibility. Because fog optical depth is not available at night, the calculation of horizontal visibility using the nighttime approach is only valid if fog extinction is nearly homogeneous over the whole fog layer. Mean deviations between observed and calculated visibilities therefore slightly increase to about 61 m.INTRODUCTIONFog often occurs in mid latitudes during high pressure weather situations and is frequently accompanied by strongly reduced horizontal visibilities at ground. Unfortunately, the knowledge about the spatial distribution of fog and horizontal visibility from conventional meteorological observations is rather poor [1]. Fog monitoring by means of satellite images provides a complete overview of the spatial fog distribution during satellite overpass. The detection of fog using an image of temperatur difference of #3 and #4 (NOAA-AVHRR) data using a simple threshold technique is generally possible both, day and night [2] [3] [4]. However, according to the international convention fog is defined as horizontal visibility less than 1 km. The main problem of the threshold technique is therefore the discrimination of ground fog and low level stratus which is dependent on horizontal visibility at ground. The aim of this paper is to describe and evaluate two methods for the calculation of horizontal visibility from NOAA-AVHRR data. Finally, ground truth due to simultaneous transmissometer observations will be discussed. All visibility measurements used for verification purposes are taken from 19 transmissometer sensors of a traffic routing system installed along the motorway BAB 4 between Cologne and Aachen (Germany).CALCULATION OF HORIZONTAL VISIBILITY Horizontal visibility (VIS) is generally related to spectral extinction (βext) at 0.55 µm by Koschmieder's law (1) [5].VISext=⋅11βεln() (1) (ε = contrast threshold of 5%)For the calculation of horizontal visibility it is therefore necessary to calculate spectral extinction from satellite imagery. Spectral extinction can be either obtained from fog optical depth (δ) and geometrical thickness (∆z) (2) [6] orusing the fog droplet spectrum defined by the droplet radius(r) an the count of drops at the radius r, n(r), as well as the efficiency of extinction Q(λ,r) (3) [7].βδextz=∆ (2)βλπextrrQ r r n r dr=⋅⋅⋅⋅∞∫(,)()2 (3)THE DAYTIME APPROACHThe calculation scheme using the solar channel 1 (AVHRR)is based on the determination of fog optical depth from fog albedo (Fig. 1). However, before horizontal visibility can be obtained several analytical steps have to be performed [6].After fog detection [2] and calibration of # 1 (AVHRR) data several steps of radiometric correction are necessary: Anisotropic correction by means of the coefficients for low clouds [8] as well as atmospheric correction using radiosonde data of the radiosonde Essen (Germany) and the5S code [9] have been performed to obtain the fog top albedonot yet considering the albedo of the underlying surface which also contributes to the satellite signal. The albedo ofthe underlying surface has been calculated from a cloud-free winter image of the study area which has also been correctedfor anisotropic and atmospheric effects. A simple radiative transfer model taking into account fog top albedo and the albedo of the underlying surface then provides the spectral albedo of fog [10] which is directly related to fog optical depth and liquid water path [11]. Finally, fog geometrical depth has been obtained by superimposing the binary fog