EXPERIMENTAL ANALOGIES TO ATMOSPHERIC MOTIONS 
checked. These all are in qualitative agreement in 
giving a relatively constant velocity c for a fluid layer 
of density p1, advancing under its own excess weight 
through another of density p2. The velocity is given by 
c=K 
where h is approximately the height of the nose of the 
front, and where K has a numerical value from 0.6 
to 0.9 for the total depths (two or more times /) used 
by these experimenters. The fact that the experimental 
case, during the steady part of the motion, represents a 
balance between a supply of gravitational energy and 
frictional dissipation alone, without any influence of 
tendencies toward Margules-type geostrophic equilibria, 
suggests that little is to be expected from the result on 
a large scale. It may, however, be significant in smaller 
occurrences such as the pseudo-fronts beneath thunder- 
storms or in very unsteady motions where Coriolis 
effects have little chance to develop. 
The second group of investigations consists of work 
in air by Sipinen [59] im 1924 and Harwood [81] in 
1945. Sipinen’s work is particularly interesting because 
of the use he makes of the very thin, warm, smoky 
layers which can be put on a smooth surface by puffing 
cigarette smoke gently over it. Slight currents then 
induce vortices at the boundaries between clear (cool) 
and smoky (warm) air layers. Or they can easily be 
developed and seen via the smoke by slight warming 
(e.g., by the hand) of a spot on the surface. 
Harwood’s work, however, is the more significant 
because it is capable of immediate quantitative ex- 
tensions. He utilized a wind channel whose base was 
divided longitudinally into two equal parts. One half 
was cooled while the other was heated. Titanium tetra- 
chloride smoke served as a tracer in either the cool or 
the warm layers. In certain ranges of air velocity, 
channel depths, and rates of heating and cooling, he was 
able to obtain either stable wave corrugations on the 
interface or families of vortices with members spaced 
at regular mtervals along the deformed discontinuity 
surface. Some extensions which would be simple in 
principle, although perhaps difficult in detail, should 
make it possible to study wave propagation and in- 
stability rates for measured thermal and velocity con- 
ditions. Such studies, if sufficiently successful in yielding 
the required type of measurements, would have very 
obvious utility in checking and suggesting alterations 
in the many existing theoretical calculations for such 
wave properties. 
The practical problems are such that the nonrotating 
case would obviously have to be tackled first. But there 
is no reason, provided the scale of the apparatus is 
sufficiently small (Harwood’s working section was 1 yd. 
long by 16 in. wide), why the whole equipment could 
not be put on a rotating table of the moderate size 
being constructed at present at the University of Chi- 
cago, or certainly in a rotating room on the scale of 
that at Gottingen [49]. An arrangement, in fact, could 
be achieved here in which it certainly ought to be 
possible to investigate systems of light and heavy fluid 
1239 
arranged side by side, similar to those of Margules 
[44] and Starr [61] which arose from a discussion of the 
energy transformations in the atmosphere. The prin- 
cipal obstacles preventing a close approximation to 
Starr’s system are the infinite extent, up- and down- 
stream, of his east-west currents, and the assumption 
of passage through equilibrium states at every stage. 
The first could probably be obviated to a large extent 
by arranging the initially adjacent regions of heavy and 
light fluid along a narrow rectangular container suffi- 
ciently long to make the end effects rather small, and 
the second by using very small density differences 
(0.001 in specific gravity is quite feasible). 
Thermal Convection Studies 
In the field of thermal convection two groups of 
investigations of meteorological interest from our pres- 
ent viewpoint may be recognized. The first includes a 
number in which rotational effects were not considered 
in the experimental plan, and the second includes some 
in which they were incorporated. The latter group will 
be considered in connection with the general circulation. 
The first class includes work by Aitken [2] between 
1870 and 1880, by Terada and Hattori [65] in 1926, the 
investigation by Kobayasi and Sasaki [84] in 1932, and 
a long series of experiments described in papers by 
Sandstrém [53-55] between 1908 and 1919, V. Bjerknes 
[10] in 1916, Jeffreys [35] in 1925, and Godske [28] in 
1936. The experiments considered in this latter series 
and in the convectional part of Terada’s work consisted 
of various arrangements of heat and cold sources in 
rectangular containers of water. The important me- 
teorological idea discussed in the Bjerknes-Sandstrém- 
Jeffreys controversy was that, on an atmospheric scale, 
a distribution of heat sources at levels higher than the 
corresponding cold sources would lead to a much less 
vigorous atmospheric circulation than the reverse case 
of, for example, cold sources on high plateaus versus 
heat sources at the ocean surface. This result was ap- 
parently verified by Sandstrém’s experiments which 
showed much more vigorous and different types of 
circulations in the latter case than in the former. The 
theoretical arguments on this point depend on whether 
one adopts the assumptions of Bjerknes or those of 
Jeffreys or takes some intermediate position. Experi- 
mentally, Sandstrém obtained velocities which even- 
tually became hardly perceptible when the heat source 
was above the cold source. However, unpublished work 
done by J. G. Phillips at the University of Chicago 
from 1942 to 1944 has shown definite back-and-forth 
currents in the zone between the hot and cold sources 
at quite low total temperature differences when the 
hot source is higher. The experimental conclusion de- 
pends on what is taken as the proper similarity criterion 
in this case (e.g., on the relative importance of various 
types of horizontal and vertical heat transfer) and on 
whether Phillips’ velocities correspond to small or great 
velocities in the atmosphere. Offhand one would expect 
that the relative vertical positions of the predominantly 
side-by-side thermodynamic sources affecting the at- 
mosphere would not produce great effects, because of 
