Statistical characteristics of the observational errors in total ozone measurements by Dobson spectrophotometers and M-83 filter ozonometers are examined by the use of the structure functions. In order to explain the seasonal variation of the functions, it is necessary to consider not only random errors but also systematic errors which are correlated with both the total ozone amount and the errors at other stations. The optimum interpolation scheme is modified in case of the existence of systematic errors. By this method, grid point values of total ozone fluctuations are obtained together with the estimates of errors. Spatial mean values are computed by the average of grid point values. When the interpolation networks overlap, estimates of errors in spatial averages are calculated as variances of statistical samples which are not independent each other. By these procedures, an objective treatment is possible for observational data by Dobson spectro-photometers and M-83 filter ozonometers with the estimates of errors both in grid point values and in spatial averages. The interpolation errors thus obtained are found to be small enough to discuss the fluctuation of total ozone on a global basis.
The fluctuations of monthly mean total ozone from 1962 to 1976 are analyzed on a global basis, using the longitude-latitude grid point values obtained by the method described in part I (Hasebe, 1980), From these values, global mean, hemispheric mean, zonal mean, and meridional mean values for northern midlatitudes are calculated with the estimates of errors. With the application of numerical filters, non-stationary annual oscillation (NSA), quasi-biennial oscillation (QBO), and long-term variations (LTV) are examined. The results are summarized as follows: (i) NSA; The large amplitude of the oscillation in northern midlatitudes is situated near 40°E and 160°F where the time averaged distribution of total ozone shows large gradient; east of the maxima. In the northern hemisphere around 1970 to 1971, an apparent phase reversal is found, corresponding to the reduction of about 10% in the amplitude of the annual oscillation. (ii) QBO; Significant oscillation has been detected with larger amplitude in the northern hemisphere. The oscillation cores in northern midlatitudes are seen near 140°E and 20°W; on and to the west of the time averaged maxima of total ozone. The poleward propagation of the QBO phase is not always seen, because of the difference in the QBO period between low and high latitudes. When the QBO phase exhibits poleward propagation, total ozone maxima in midlatitudes are seen about π/2 later than the quasi-biennial west wind maxima at 50mb in the tropics. However, this relation does not exist when the ozone phase propagates equatorward. (iii) LTV; The existence of the four-year oscillation is found at high latitudes, particularly in the northern hemisphere. The positions of the oscillation core in northern midlatitudes resemble to those of the NSA, but enhanced near 40°E longitude. Maxima and minima of this oscillation are found in late winter or early spring of even years in both hemispheres, while the phase in the southern hemisphere is preceded by about 3π/4 to that in the northern hemisphere. The earlier investigations of long-term trends (e.g., Angell and Korshover, 1976) are qualitatively confirmed in this analysis. It remains still uncertain, however, whether such long-term trends are essentially explained by an autoregressive process because of the short period of coverage of the observational data.
Inertio-gravity waves of intrinsic frequency ω propagate through a shear flow in a stratified, rotating fluid with Coriolis parameter f. The question of when a surface ω=f is an absorber (as found by Jones 1967 and Miyahara 1976) and when it is a reflector (as for equatorially-trapped waves) is addressed by means of calculations based on ray tracing for a vertically-sheared mean flow at large Richardson number. These calculations show that the rays reflect-or more properly refract-away from the surface ω=f, in all cases except the f-plane case studied by Jones. However, on a β-plane with limitingly small the time for a wave packet to get away from the surface ω=f becomes arbitrarily long. This means that for sufficiently small β there is absorption in practice if not in principle, since some dissipation will always be present. For parameters typical of mountain lee waves in mid-latitudes, for example, molecular dissipation alone would be more than enough to destroy a wave packet near ω=f, the appropriate dimensionless values of β being extremely small in such cases. We thus suggest that the absorption found by Miyahara (1976) in a numerical model with finite dissipation and small β should not be regarded as surprising, nor as conflicting with the conclusions of earlier work by Matsuno (1966), Lindzen (1970) and others on equatorial trapping of waves for values of β which are effectively much larger.
The decay rate of the non-molecular radar cross section βM of the post-Fuego stratospheric aerosols observed by lidar was much smaller than that deduced by using the wellknown Gudiksen's 2-dimensional atmospheric model. The discrepancy can be explained by a kinetic aerosol model including condensation growth of the aerosols. A closed incompressible air parcel model is proposed to discuss chemical and physical processes of the volcanic substances separately from the global transport. The kinetic aerosol model is used to examine temporal variations of aerosol size distributions in a volcanic air parcel. Sulfur cycle in the model is described by three equations. The first one expresses conversion of SO2 to H2SO4 vapor of which concentration is described by the second one. The third one describes the time evolution of the size distribution in terms of condensation and coagulation processes. Assuming two initial size distributions, Haze H and log normal, these equations are solved numerically for the aerosols with size range 0.01μm<r<0.4μm. In the case of the Haze H initial size distribution the discrepancy can be explained if we take appropriate values of conversion rate and initial concentration of SO2. On the contrary, in the case of log normal, the observed decay rate can not be explained by the kinetic aerosol model in this paper.
The dissolution process of ice crystals in replica solution was studied. The dissolution form of ice crystals depends on the concentration of the solution. For example, in lower concentration a hexagonal plate dissolves through the dodecagonal plate with more pointed corners in the ‹1120› directions, while in higher concentration it dissolves through the dodecagonal plate with more pointed corners in the ‹1010› directions and thereafter through nearly regular dodecagonal plate. The dissolution form of ice crystals depends on the dimension of dissolving ice crystals, too.