171 are monotonic while others are not. Wind direction and time of dav are non-monotonic. In order to treat TAM 6 and the Lake Michigan shore. The joint occurrence of time and direction make it very hkely that these two variables, plots are made of SO. concentration vs, wind direction and SO. concentration ves. interesting relationships may be observed from an time of day. Examples of these charts are shown in Figures 134-137. In each of these figures, the 50th, 90th, and 95th percentile values as well as the maximum are presented. These charts and a tabulation of the data on which the charts are based were used to select time bands and wind direction bands in which the variation of SO. concentration was relatively small. The Tabulation Prediction Scheme was developed for the TAM Stations. The bands for time of day and wind direction selected for each of the TANI stations are presented in Table 63. Many interesting features are brought out by these charts which may be useful in practice to forecast SO. concentrations. The relative positions of clean areas or pollution sources with respect to the TAM stations are indicated. In Figure 187, showing the time of day vs. SO. concentration for TAM6, the usual peaks of the early morning hours as well as those of the late afternoon or early evening hours are shown in the 90th and 95th percentile curves. Note that the morning peak shows up in the 90th and 95th percentile curves, but is barely perceptible in that of the 50th percentile. This suggests that the peak is due to relatively infrequent events that can produce unusually high concentrations in the morning, possibly inversion breakup fumigations. There is more evidence here pertinent to fumigations. In Figures 136 and 137, the highest values observed (i.e., the highest values of each MAX curve) occur at 0600, 0700, or 0800 at six of the eight TAM stations. These are the hours during which fumigations may be expected. The wind direction associated with the SO. peaks are those associated with the highest peaks of the MAX curves in Figures 136 and 137. TAM7 is a good example. The highest SOz. value occurred at 0700 with a wind from 195 degrees. The joint occurrence of a very high SQ. vaiue at a morn- ing hour from the direction of known high stack sources in thedustrial corridor is consistent with the inversion breakup fumigation process. This example gives just a hint about how SOs percentile plots can be used to investigate meteorological events. Also of interest is the early afternoon peak in the TAM 6 maximum value. An examination of Figure 135, showing wind direction vs, SOQ. concentration for TAM 6, shows that the peak occurred with an east wind off the lake. (Figure 135 shows an actual case of high SO. concentration at TAM 6 during a lake breeze.) Strong SOz. sourees are located between a lake breeze eauses this high SO» value. Additional examination of the other charts. The interquartile range is a measure of the dispersion of the measurements for each combination of meteorological measurements. The amount of noise or uncertainty in the prediction is shown by the magnitude of the interquartile range. Similarly, the difference between the 95th and 7dth percentile values represents the amount of skewness present. Since the distributions approximate the log normal function, one would expect appreciable skewness. Before a final choice is made of the meteorological variables used in the Tabulation Prediction Scheme, each must be examined. This may be done in a number of ways, one of which is to plot the percentile values of SOs concentration vs. each meteorological varia- ble and examine the relationships. The 50, 90, and 95 percentile and maximum curves could be plotted for this purpose in the same format as shown in Figures 134 and 135. Where a network of stations is available, such as exists In New York City, Los Angeles, Washington, or Chicago, the receptor-oriented techniques may be applied to each of the stations to obtain isopleths of concentrations similar to that obtained in the sourceoriented model. CONSTRUCTION OF THE TABLES Meteorological variables recommended for the construction of a Tabulation Prediction Scheme are (1) wind direction, (2) ceiling height, (8) wind speed, (4) temperature, and (5) hour of the day. The rele- vance of each of these variables has been given above, The importance of atmospheric stability has been noted, but such data are usually unavailable. A sub- stitute is provided by the choice of the variables: ceiling height, hour of the day, and wind speed. Com- binations of these are closely related to stability. For example. during high winds, e.g., near 15 knots, the stability is usually neutral. This is true during the day or night. During unlimited ceilings with light winds, an inversion is highly likely at night, but un- stable conditions are to be expected during the day. Class intervals for the variables, ceiling height, wind speed and temperature are shown in Table 64. As a first step the data are stratified by season and by the presence or absence of precipitation. An interme- diate tabulation is prepared in which wind direction and hour of day are listed as Band I, Band JI, and