roswell

20.
However, when weather conditions indicated a markedly dis-
continuo,,s variation of temperature (i.e. a "front"), appropriate
subjective modification of the objective linear interpolation
technique was applied.
The vertical distribution of wind was determined mainly from
direct observations (pilot-balloon and radio wind-sounding
measurements) of free-air wind& at weather stations in the area
of each experiment. However, actual measnrements of winds in
the upper half of the troposphere often are scarce or completely
lacking, and it was frequently necessary to make use of an
indirect method of estimating the wind at elevations greater than
5 km. Charts of the distribution of atmospheric pressure(as
given by radiosonde observations) at selected levels between
5 km and 15 lan were constl"1lcted, and the wind direction and
speeds at these levels were computed from the well-known geostrophic
wind equation, which relates the wind to the horizontal distribu-
tion or pressure •
For-the experiments carried out off the east coast of the u.s.A.
between 1 August 1946 and 1 August 1947, it seemed feasible to
show the distributions of both temperature and wind in vertical
cross-section. This was due to the fact that these experiments
were made, and the res1.1lts of same recorded, within a fairly
narrow band centered close to a line between Lakehurst, N. J.,
and Nantucket, Mass., at which points radiosonde and upper-wind
observations are taken regularly. Hcmever, vertical cross-sections
of temperature and wind were abandoned as a method of representa-
tion of the distrj_bution pertaining to all subsequent experiments.
There were several reasons for this decision. In the first place,
the sites and character of later experiments did not fit into the
existing weather-observing network in a manner favorable to cross-
sectional representation. In the second place, experi~nce brought
about the conclusion that the horizontal gradient of temperature
is usually so small that, 'Within the area encompassed by an experiment,
the difference in temperature at a given level between points at
the ends of a cross-section is no greater than the a~erage error
of the radiosonde measurements. Thirdly, it was soon realized that
the variability of the wind in space and time is such that an individual
pilot-balloon or rawinsonde ascent is not representative of the
average vertical distribution of velocity during the interval
occupied by a single experiment. Furthel"lllore, as mentioned above,
the wind at high levels in the troposphere often had to be in-
-ferred by indirect means. Since the true wind usually deviates
somewhat fran the theoretical. geostrophic wind (the latter being
derived under certain simplifying assumptions) and since the
geometry of the pressure field is subject to some uncertainty
owing to inaccuracies in the radiosonde observations, it became
apparent that the assignment of a single velocity value at any