For a purpose such as numerical prediction by finite differences, meteorological observations are useless if they are not very complete. The most complete sets of observations are those which have been made on “international balloon days,” and if one looks at the charts of them issued by V. Bjerknes and R. Wenger from Leipzig one notices the conspicuous gaps where clouds have prevented the measurement of the upper wind by means of pilot balloons. More wind observations above clouds would, no doubt, be useful in ways other than that mentioned above. Some have been obtained recently by sound-ranging or by sending instruments up on kite-balloons. The chief advantages of this new method of observing are: (i) that it can be employed when there is fog or low cloud which would hide a pilot balloon; (ii) that it is much cheaper than sound-ranging or than captive ballooning at least for the height of 650 metres to which experiments have so far extended. Minor advantages are (iii) that the steel sphere is much less affected than a pilot balloon by vertical air currents; (iv) that these projectiles give the wind almost directly overhead, while pilot balloons drift far horizontally from the nominal site of the observation; (v) that the operations are easily performed at night; (vi) that it is well suited to making repeated observations of wind at short intervals of time, either in order to study the variations, or in order to remove them by averaging.