OCR Text |
Show the inclusion of several storms of strongly differing frequency content. Figure 184 is the spectrum of a group of storms stratified by wind direc'tion, which includes several convective spring storms which were not in the sample of Figures 182 and 183. Note the spectral peak at 0. 025 cycles min. -1. This naturally occurring high spectral density at low frequencies has to be coped with in a spectral analysis scheme. The frequency of the forcing function used to delineate seeding effects has to be a compromise between two considerations. It would be desirable to use a periodic forcing function of high frequency, where the natural spectral density is low. On the other hand, it would be desirable to have a long period input so that the diffusive and transport characteristics of the atmosphere would be more likely to maintain this periodicity ·in the forcing function as it travels over the target area. The use of a practical long period requires, however, a long continuous record to obtain the desired degrees of freedom. Unfortunately, ·a typical nearly continuous precipitation period is of the order of twelve hours in duration, so that a period in the 2 to 1 / 2 cycle per hour range is dictated. One cycle per hour was selected as the best compromise. To test the sensitivity of the techniqu_~, various forcing functions were superimposed or applied on the unseeded series whose spectra are displayed in Figures 182-184, and the coherenc~ between-the time series and a square wave of forcing function frequency was computed. Figures 185-187 present the results for a series of forcing functions superimposed on the natural data at 2 cph. The natural coherence is very low at 2 cph for these long time series. In Figure 185, results are shown for superimposed 2 cph square waves with amplitu.des of + 0. 02 and + 0. 01 in/ho_ur precipitation rate, -;- 0. 02 in/ hour, .and + 20% of the natural rate~ Note that at this high frequency, where the natural .spectral density is fairly low, all of these forcing functions produce a clearly identifiable result in the calculated coherence. The harmonics occur because a strict square wave was used; it probably would not occur in real data. The effect is very definite when these changes ar~_superimposed on all three of the ti~e s·eries whose spectra were presented in Figures 182184. When a lower frequency is used, the higher natural spectral content can interfere with the sensitivity of the analysis technique. Figure 188 is the Buffalo Pass seasonal spectrum with a forcing function of + 0. 05 superimposed at 1 / 2 cph. The effect is still very clear for this rather high precipitation increase. In Figure 189 the same forcing function at 1 / 2 cycle per hour is applied to the series of Figure 184, which had a high natural component at 1 / 2 cycle per hour. Even with this rather high amplitude artificial square wave , the effect is practically nondiscernable at the fundamental frequency. The selected frequency of one cycle per hour was found to be the lowest possible frequency that was practical from an analysis standpoint, and the highest possible from a consideration of dispersion and transport characteristics. 273 |