Based on Kolomogoroff's similarity hypothesis of locally isotropic turbulence, the correlation coefficient Rt of the velocity fluctuations at time points t0 and t+t0 is written as follows: Rt=1-(t/T0)2/3 . (1) where T0 denotes the passing duration of the largest turbulent element which passes through the measuring point. Assuming that the correlation coefficient (Rt)w, which is calculated from the vertical velocity fluctuations w' measured in the air layer near the ground, is expressed by the formula (1), the dependence of T0 on the height of observation z and on the mean velocity u is obtained experimentally. From photographic records of vertical velocity fluctuations w' measured in the adiabatic atmosphere over level snow surfaces, Rt is calculated and is found to be well expressed by (1) in the range of small t's. It is concluded from the experimental data that T0 is proportional to z. _??_ Consequently, the spectrum of turbulent energy in the space of wave-number k is expressed by (k/k0)-5/3, where k0 is inversely proportional to z.
As a part of an artificial rain experiment, an observaticn of raindrop in thunderstcrm was made in August, 1952 in Kawagoe district. Observers made simultaneous observaticns being distributed in about twenty places. It was found that the falling pattern of rain and amount of rain had much d_??_fferences in these respective places, which were only 2-3km apart each other, and also that the raindrop which predcminates over the amount of rain was neither large one nor p_??_r_??_icle of small size, but moderate droplet of 1 to 4mm in diameter. There are many methods for raindrop observation which have been devised so far concerning rain research, among which no one that is available with exactness for the whole range of sizes is found. For instance, a good method for observing fog particle is not suitable for large raindrop and so on, and such a phenomena as of raindrop which size varies in large extent from fog particle to 6mm diameter droplet can not be observed satisfactorily by only one method. Then a several methods ought to be used in parallel, and an idea to use in combination the following three methods which are selected from the existing ones by comparing them each other from practical point of view may be sufficient for the observation of large ranged size drops. For drizzle: MgO method. For moderate raindrop: Special filter paper method and absorbing method. For large raindrop: Standard mesh screen method. It is desirable to make fine network of observation for highly changeful phenomena such as thunderstorm, because an observation at only one spot can hardly catch the general structure of the phenomena. Or, in the case a thunderstorm is progressing to some direction, it would be sufficient to distribute the observers on a line which is perpendicular to the path. Though the analysis made in this paper was not sufficient always as the observationaldata used for the analysis were rather poor, the followings are summar zed as the result of the analysis: 1) A general view of raindrop observations at twenty places is shown in Table I. 2) The variaticns of numbers of raindrop by time and place are shown in Fig. 9. 3) The amount of rain specified for drops of each size, being calibrated from raindrop, are shown in Figs. 10 and 11. 4) A several patterns are seen in size distribution from the observational results. 5) The raindrop which predominates over the amount of rain is droplet of 1-4mm in diameter. 6) The heavy rain which controls the amount of rain falls neither in the beginning of the thunderstorm nor in the end, but in the middle and for two or three times. 7) The rain area starts in a small one, and then expands and diminishes rapidly. 8) The maximum rainfall area was an oval of 160km2, which major and minor axes were 22 and 12km respectively.