1244 
velocity distribution, on the other hand, is not strongly 
altered. The westerly current at the rim is weaker and 
narrower and the easterly current, at say 0.77, is 
stronger and better defined with wu/Cz as much as 
-++0.04 or more. In this case, narrow, strong, easterly 
currents are set up iward toward the center as ob- 
served earlier by F. Hall. One of the important differ- 
ences between this and Vettin’s experiment is the high 
conductivity of the pan (his base plate was of glass). 
This leads to static instability over the entire bottom 
that is of course relatively stronger near the burner 
position. Thus, ink observations suggest that systematic 
vertical motions, if present at all, are very vaguely 
defined and that the rising motions are rather irregu- 
larly distributed over the pan from a Bénard-like hot 
layer at the very bottom. At any rate, considerable 
additional doubt (see also below) is thrown on any 
simple reasoning of the usual textbook variety concern- 
ing the effects of simple heat-source distributions and 
the conservation of angular momentum in the general 
circulation. Such reasoning certainly ought to apply to 
a simple disc if anywhere, but it is very hard to conceive 
of an axially symmetrical circulation which would ac- 
count for both the westerlies and the easterlies at the 
top surface in this experiment without imtroducing 
various epicycles that have no obvious relation to the 
heat sources. 
Figure 12 illustrates another striking effect which was 
accidentally observed in the rotating pan. This effect is 
a result of the rotational property stated by Taylor 
[68, 64] (see also a more recent discussion by Gortler 
[29]) that slow relative motions in a fluid initially 
rotating as a solid should tend to be two-dimensional 
in planes perpendicular to the axis of rotation. In Fig. 
12, ink is seen a minute or so after being poured into 
the rotating pan in a quite arbitrary manner. The mass 
of ink, instead of spreading out in the usual eddying 
fashion, rapidly forms vertical walls and the cylindrical 
surfaces gradually wind around one another, retaining 
their parallelism to the axis.’ They diffuse much more 
slowly than in a nonrotating pan. Such an effect of 
rotation would obviously be of the greatest importance 
in diffusion or convection problems in establishing pre- 
ferred directions with radically different coefficients of 
transfer. In the apparatus the effect is still strongly 
present when the rotational instability parameter g/aQ? 
is as high as 140 (6 rpm) and is even evident to a lesser 
extent when heating is going on with a metal bottom 
(e.g., whole vertical streamers of ink transfer laterally 
without tilting much). The rotational stability is ob- 
viously much higher and the hydrostatic stability prob- 
ably much less in this case than for the earth. However, 
certain other preliminary experiments with the same 
dishpan apparatus show that this effect has some very 
important implications from a meteorological stand- 
point. The bottom of the rotating dishpan was insu- 
lated with a layer of plaster and heating carried out as 
7. One is reminded of the Margules-Helmholtz slope condi- 
tion for a discontinuity surface of wind which reduces to 
parallelism to the earth’s axis when the density difference is 
zero. 
LABORATORY INVESTIGATIONS 
before at the rim [25]. The top surface then shows a 
wide meandering band of strong westerlies (u/Cz of 
order 0.1) with rather larger vortices than before. To- 
Fie. 12.—Ink initially of slightly greater density, a couple 
of minutes after being poured into a pan of water in solid rota- 
tion at 15 rpm. Note the vertical walls. As in Taylor [63] the 
walls remain vertical for many minutes and gradually draw 
out into surfaces which are sharply divided by clear areas when 
viewed end-on. The same effect is very marked to at least as 
low as 5 rpm and occurs under the heating conditions of Fig. 
11 when the pan bottom is insulated by a layer of plaster. 
ward the small central area the average motion is 
uncertain but is at least very weak westerly or perhaps 
easterly. If the Taylor ink columns are produced before 
heating is started and then allowed to remain, they 
retain their cylindrical character while moving with the 
thermally induced currents. In other words, the motions 
appear to remain quasi-barotropic and are certainly 
in addition quasi-geostrophic. On the other hand, the 
temperature field shows strong static stability and is 
strongly baroclinic with a dome of cold water occupying 
the center bottom. The rim-center temperature differ- 
ences run to the order of 5C. Tentatively then, one can 
conclude that it is quite possible to have a strongly 
baroclinic and reasonable pole-equator temperature dis- 
tribution in this experiment and yet also have quasi- 
barotropic currents and disturbances.® 
8. There are a number of observational difficulties concern- 
ing the representativeness of the top surface motions and 
other matters that make it necessary to regard any detailed 
