1078 
from observations that imcluded many cases of flow 
over a smooth surface. With \ = 27.5, curve (3) would 
nearly coincide with curve (2). 
The wide spread of the curves (2) to (5) clearly 
demonstrates the importance of the assumptions as to 
the diffusivity near the boundary surface. Assumptions 
(1), (4), and (5) appear too extreme, but none can be 
rejected offhand since we have no basis for describing 
what actually takes place, but are trying to introduce 
a model which gives results in consistent agreement 
with observed conditions. In order to advance a step 
further we have to compare the theoretical conclusions 
with the empirical results. Our empirical data are of 
two kinds. On the one hand we have available a few 
series of measurements of humidity at different heights 
above the sea surface from which I, can be found, and 
on the other hand we have series of accurate ships’ 
observations of meteorological conditions and sea sur- 
face temperatures from which we can compute evapora- 
tion, using different I, values, and we can compare 
the results with those derived from energy considera- 
tions. In order to facilitate computations and com- 
parisons we shall use vapor pressure instead of specific 
humidity. Approximately, e = qp/0.621 or with p 
= 1000 mb, e = 161g. We then write equation (15) 
in the form 
H= Ka (es = (36) 
and shall call K, the evaporation factor. We shall use 
units such that with pressure in millibars and wind 
velocity in meters per second we obtain the evapora- 
tion in millimeters per 24 hours. The curves for K, 
(a = 800 cm) are shown in Fig. 3. The heavy curves 
Ca) Wa, 
¢ FROM "METEOR" OBSYV, 
—— USED BY JACOBS 
O 2 «4 36°68 On Te 6s & 2 
WIND VELOCITY (M SEC.7!) 
Fie. 3.—Theoretical values of the evaporation factor K as 
function of the wind velocity at 800 em, computed on the 
basis of different assumptions, and empirical values shown by 
dots and a dashed line. 
should be used for computing evaporation from single 
observations. If climatological averages are available, 
Wwe must use some transition curves in an interval of 
velocities around 6-7 m sec, the width of which 
depends upon the spread of values upon which the 
averages are based. In Fig. 2, arbitrary transition 
curves are indicated for the interval 2-11 m sec—. 
Discussion. The theoretical considerations lead to 
such different evaporation factors that the theoretical 
approach is useless until there can be advanced ade- 
MARINE METEOROLOGY 
quate arguments for accepting one of the many models 
which at present can be used to describe the character 
of the eddy diffusion of water vapor close to the sea 
surface. The theory apparently leads to consistent re- 
sults at wind velocities up to 6 m sec for individual 
cases, or up to 4 m sec for average conditions, but it 
should be emphasized that this consistency arises be- 
cause it has been accepted that the sea surface is 
smooth at Iow wind velocities. The empirical evidence 
for this feature is, however, so meager that its reality 
is badly in need of confirmation. Should it not be con- 
firmed, we must admit that the theoretical computation 
of the evaporation factor fails at all wind velocities. 
The only remaining conclusion is that such a factor can 
probably be established, either on a theoretical basis 
Gf a new foundation for the development of a theory 
is found) or on an empirical basis. 
In trying to do the latter we may follow one of three 
courses: (a) the evaporation coefficient T may be de- 
termined from measurements of humidity gradients 
over the sea; (b) where accurate meteorological ob- 
servations are available, and where evaporation can be 
computed from energy considerations, the evaporation 
factor may be determined; or (c) where climatological 
charts are available, based on observations that may 
contain small systematic errors, such as temperature 
observations on shipboard, the procedure under (6) 
can be used. We shall examine these procedures more 
closely. 
In Fig. 2 the values of T have been plotted which 
have been derived from observations of humidity gradi- 
ents. Most of the observations were made at wind 
velocities below 6 m see and agree fairly well with the 
theoretical values for a smooth surface. The few ob- 
servations at high wind velocities agree quite well with 
curve (4) for a rough surface, and this feature led the 
writer [24] to accept that approach although he pre- 
viously [23] had adhered to different concepts. It will 
be shown here that the agreement between the em- 
pirical T values and the theoretical curve (4) must be 
accidental because the evaporation factor K, based on 
curve (4), gives far too high evaporation values. 
The second method (6), can be used only if very 
accurate meteorological observations are available for 
a large area and a sufficiently long time. The Meteor 
observations in the Atlantic [12] satisfy this require- 
ment, and from these the data in Table II have been 
obtained. The factor K, which has been plotted in 
Fig. 3, lies between 0.07 and 0.09, and agrees fairly 
well with values derived from curve (2). The result is 
in sharp contrast to that derived from examination of 
humidity gradients and indicates that of the different 
theoretical assumptions number (2) is the most nearly 
correct. It should be borne in mind that the computa- 
tion of the evaporation factor on the above basis is 
somewhat uncertain, because the meteorological ob- 
servations are extended over only from 114 to 3 
months in summer whereas the “energy values” apply 
to the whole year. The errors that may be introduced 
because of this circumstance should, however, not be 
serious, because in the tropics the annual variation in 
