An Exploration of the NRL Global Surface Observation Database Rudolf B. Husar CAPITA Washington University, ST. Louis, MO 63130 July, 2000 Background Surface observations provide valuable data for the testing and verification of atmospheric chemical transport models. In particular, the reported estimate of the surface visual range provides an estimate of the horizontal extinction coefficient along the line of sight, which in turn is related to the concentration of aerosols. Additional parameters in the surface meteorological record include a coded description of the current weather. There are specific codes for dust, smoke, and haze, as well as for a large variety of weather-related obstructions to vision. The combination of the visual range and the weather code allows the estimation of the surface aerosol concentration as well as its causes. Given the successful operation of the NRL-NAPS global aerosol model it became desirable to verify and/or augment the NAPS model using the available surface observations over land. Such verification is particularly important since the current operational satellite-based aerosol retrievals operate well over the oceans but do not provide information over land. A major problem with the visibility observations is that the obstructions to vision include rain, fog, clouds, and other forms of hydrometeors. In order to detect the aerosol contribution the weather related phenomena need to be filtered so that the aerosol signal is not obscured by hydrometeors. In 1999, NRL has contracted with R. Husar to implement the weather filters for the calculation of the aerosol extinction coefficient, and to implement the ideal code in the NAPS system that would perform the appropriate filtering and plotting algorithms. In summer of 1999, these algorithms were developed and tested using a global data set for one six hour period. Since September 1999, the Bext filtering algorithm was incorporated in the production of six hourly surface weather maps. The aerosol extinction coefficient was drawn as a color-coded circle, with radius proportional to the extinction coefficient and color-coded according to dust, smoke, and haze. The experience of the past 9-month suggested that both the weather filter and the display rendering should be re-examined to better represent the surface aerosol pattern. This report summarizes this second effort which consisted of: 1. Assessment of the NRL surface observation over a longer period. 2. Evaluation of surface extinction coefficient data over Africa , March-June 2000. 3. Recommendation for additional activities that would improve the utility of the surface visibility data. Assessment of the NRL surface observation over a longer period The NRL global surface observations were collected over the period March 28-June8, 2000. The NRL data were plotted on the global map (Figure 1). On any particular day (e.g. May 1, 2000), the station data are coded according to four different categories. Red circles, represent stations for which valid Bext data were available. Blue circles represent stations that are eliminated because of the weather flag, i.e. when the relative humidity was (>90%). The yellow circles indicate stations that were filtered due to other weather flags (ww code>13). The small black circles show the stations for which visibility data were not available at that particular hour. Since the weather filters are highly diurnal, the pattern of these stations is shown in Figure 1 through AVI animation representing the 24-hours May 1, 2000. This evaluation indicates that most of the global network provides data every 3 hours. Eastern Europe, Japan and Australia have data for every hour. The majority of the U.S. has data are in 6-hour increments at 00:00, 06:00, 12:00, 18:00. The spatial pattern of the available data indicates that globally 30-40% of the stations provides valid aerosol Bext values. Stations in Europe and Southeast Asia have low fraction of valid Bext stations. However, the reason there is the high relative humidity, particularly in the early morning hours. The fraction of the valid stations is very low in Africa, between Sahara and South Africa, where <20%?? of the available stations report valid values. The cause of the missing Bext data in the mid-section of Africa is almost exclusively missing data. The spatial pattern of the stations in North Africa for the 24-hour period is illustrated in Figure 2. Evaluation of surface extinction coefficient data The resulting bext values for Africa, north of the Equator, Europe, and Middle East are shown in Figure 3. The yellow circles are proportional to the surface extinction coefficient, 1/km. The animation represents daily pattern at noon (12:00 GMT) The daily data over Africa indicate a rather consistent pattern in that high extinction coefficients tend to occur over cluster of stations. This is consistent with the existence of regional dust clouds that occur over 500-1000 km areas. The main problem in Africa is the large number of missing values. Recommendation for additional activities Based on the above discussion it is concluded that the weather filter incorporated in the current operational algorithm is not too stringent. The station data are eliminated due to high humidity (RH>90%) and precipitation. A major problem, particularly in Africa is lack of data. In order to improve the station filters and derive meaningful Bext values for model comparison it would also be necessary to evaluate each station whether it is suitable based on additional statistical criteria. Many stations in Africa and elsewhere report constant visibilities throughout the record. Since, it is impossible to have a constant extinction coefficient in the atmosphere the invariant visibility data suggest that the observers did not bother to report the actual visual range. This is troublesome for model comparison since the existence of dust clouds do not show up in all of the data and the model comparison would be meaningless. Given sufficient length of a data record, say 3-month, it would be possible to identify and reject those stations that report constant visibility. In fact, this latter approach was applied in deriving a global continental haze climatology, published recently by Husar et al., 2000.