The visual range, or visibility, is an understandable and for many purposes, an appropriate measure of the optical environment. It is inversely proportional to aerosol concentration. Another measure of haziness is the extinction coefficient, Bext, defined as Bext=K/visual range. K is the Koschmieder constant. The value of K is determined by both the threshold sensitivity of the human eye as well as by the contrast of the visible objects against the horizon sky. In this report, we have taken K=1.9 in accordance with the data of Griffing (1980). The extinction coefficient is in units of km-1 and is proportional to the concentration of light scattering and absorbing aerosols and gases.
Data Sources and Processing
The following discussion will be limited to those items that are directly relevant to the climatic maps and
trend graphs. The sources and the gross features of visibility data have been described in the past by
almost all investigators dealing with the subject.
The visibility database consists of 280 U.S. stations for which computerized data exist. Their locations are depicted in Figure 2. The temporal coverage for most stations is from 1960. The main database of hourly weather observations resides on 2 CD-ROMS (1960-1990) and an Exabyte 8500 tape (1991-1992). The data were obtained from the NOAA's National Climatic Data Center.
Trends of Haze at Selected Sites
This section illustrates the haze pattern at selected sites. The primary purpose of this detailed examination
is to illustrate the overall pattern of the daily haze signal and also to illustrate some of the problems
associated with the visual range data.
The daily extinction coefficient (excluding fog and precipitation) for Houston, TX and New Orleans, LA is shown in Figures 3a and 3b. The 32 year daily time series exhibits a significant amount of noise but there is also evidence of long-term trends. For example, visual examination reveals that in the 1970s the Bext was generally higher than in the 1980s for both sites. The long-term time series also reveals systematic problems due to the visibility threshold effect. For example, during 1970-1973 in Houston, TX the lowest extinction coefficient was 0.16 km-1. This threshold is caused by a change in the maximum visual range reported at that site. A similar threshold problem is evident for New Orleans, LA during 1980-1988.
Correction of these threshold anomalies is possible if over 25% of the data is above the threshold. For the above two sites, this appeared to be the case. Consequently, the 75th percentile of the Bext was used for observing the overall pattern.
Data Quality Control
Data quality control was implemented by: (1) visually inspecting daily time series for each station, (2) observing
the percentile trends, and (3) examining the spatial pattern.
Whenever at least 50% of the observations were above the Bext threshold, the median can be established. when less than 50% of the observations are above the threshold an acceptable median can be established by extrapolation. A major advantage of use of quintiles in specifying trend of extinction is that such nonparametric statistics require no assumptions about the form of the actual distribution function. Trijonis and Yuan (1978) have successfully used this approach. The excellent spatial and temporal coverage of the visibility data base can be utilized only after careful site by site scrutiny for anomalous behavior such as this. The following results were compiled after extensive examination of each site's percentile trends. The advantageous feature of systematic threshold offset is that the properly presented data can identify its limitations and provide remedies for them.
There are other sites particularly in the western U.S. where the extinction coefficient is so low that virtually the entire signal is below that threshold. A good illustration for such a station is Bismarck, ND. The data indicate (Figure 4) that the extinction threshold has changed over the years. Furthermore, it is evident that most of the signal is at or below that threshold. The procedure to detect the threshold problem was the examination of the difference between the high and low percentiles. If for instance, the 90th and 25th percentile Bext have the same value, that is an indication that the lower percentiles are at the sensitivity threshold. Such is the case for Bismarck, ND (Figure 4).
The utility of the computed percentile values is demonstrated in Figure 5. For Sioux City, IA, the trend graph shows that from 1948 to 1968, the threshold extinction coefficient (FX) was at 0.08, and then dropped to 0.04 km-1. Since the 50th percentile was near the high initial threshold, this median would indicate a drop in the later periods; the mean would show a strong decrease in the late 1970s. The 90th percentile, however, appears to be a reliable, robust measure which is above the threshold influence; it indicates a clear increase in extinction over the entire period. Thus, depending on whether one follows the 50th or 90th percentiles, one would arrive at opposite conclusions about the trend.
Data Processing and Presentation
The specific parameter that is plotted for the haze maps is the 75th percentile. While this is unconventional, it constitutes
the safest approach in that it does not require any extrapolation or other adjustments to the data. More conventional
statistical measures, e.g. the mean, can be estimated as follows: from previous research, e.g. Husar et al. (1979),
the extinction coefficient is roughly lognormally distributed with a typical logarithmic standard deviation of 2.5. For such a
distribution, the 50th percentile is 0.5 times the 75th percentile, and the mean is 0.76 times the 75th percentile. Thus, if one wishes
to convert the maps, the scales of the intervals must be multiplied by the appropriate fractions. We recognize that even if the
haze is lognormally distributed everywhere, its log standard deviation will tend to vary geographically and seasonally. The available
data suggest, however, that it ranges between 1.6 and 3.4.
Figure 1 shows 16 haze climatic maps representing four time periods and four seasons. The time periods are selected to center around 1960, 1970, 1980, and 1990, while the quarters are January-March, April-June, July-September, and October-December.
The spatial patterns are presented on contour maps. The contours were derived from the station-point observations using a spatial extrapolation scheme. In the first step, the data from the random locations were projected to a uniform grid (120x180) that covers the conterminous U.S. The gridding used inverse distance squared (1/r2) as the station weighing factor. The extrapolations outside the U.S. boundaries were clipped to eliminate spurious extrapolations. The resulting contours are shown in Figure 1. The shades are 0.03 km-1 apart. The darkest shade (black) has an extinction coefficient (75th) percentile of >0.2 km-1 which corresponds to 1.9/0.2=9.5 km visual range. The lowest contour is set at 0.05 km-1which corresponds to 1.9/0.05=38 km visual range. Since these values represent the 75th percentile of Bext the median and the mean can be estimated by assuming a logarithmic standard deviation, e.g. 2.5 for such a distribution. The median of the contour range is between 18 and 76 km, while the mean is between 12.5 and 50 km.
Data Filtering and Aggregation
For purposes of spatial-temporal trend analysis, the raw visibility observations were summarized as quarterly averages of
noontime light extinction coefficient. For each month and station, three different extinction coefficients were calculated.
The first set included all visibility data regardless of weather and pollutant conditions (BX). The second group (FX) is
composed of extinction coefficients excluding precipitation and fog. For the third group (RH) a relative humidity correction
was performed to compensate for water vapor effects. This latter parameter is closely related to the dry fine particle aerosol
mass concentration. The functional form of this correction factor is given in Table 1.
Role of Meteorological Parameters
Meteorological parameters, e.g. temperature, humidity, wind direction, wind speed, solar radiation, etc. influence visibility
through dispersion of aerosols, or by changing their properties, formation, and removal rates. Temperature and relative
humidity may also influence the emission rate of aerosols or their precursor gases, e.g. for electric utility plants at high summer
temperature and relative humidity.
The relative humidity dependence of light scattering has been studied extensively (Waggoner and Weiss, 1986). The relative humidity correction factor used in this work was derived from field and laboratory data.
The extinction coefficient calculated from the visual range is influenced by both haze and natural obstructions to vision such as rain, fog, and snow. The role of these natural obstructions can be eliminated by discarding the values that occur when these meteorological phenomena occur. Figure 6 illustrates the extinction signal with and without the natural visibility obstructions. Figures 6a and 6b show an extinction coefficient for Raleigh/Durham airport for 1961 and 1992, respectively. These figures clearly indicate high excursions of extinction coefficient during both summer and winter seasons. Figures 6c and 6d illustrate the corresponding Bext values with the natural visibility obstructions removed. The data clearly indicate that the high Bext values are eliminated and that there has been a substantial increase in the extinction coefficient between 1961 and 1992, particularly during the summer season.