On periods and symmetry points in pressure as aids to forecasting

(a) The movements of the atmosphere are on such a scale as to produce much coherence between the pressures on successive days. Persistence is especially marked in high values of pressure, which usually last for a week or thereabout, and will appear on a graph as surges. (b) Two surges m days apart will make a considerable contribution to the computed amplitude of a wave whose period is m days; and a third surge or hollow at an appropriate time may further increase it. Thus mere coherence produces an appreciable amount of periodicity. (c) The existence of coherence in pressure series makes the ordinary Schuster criterion for “reality” inapplicable. A test suitable for harmonic analysis follows from a previous paper by the author, but one adapted for correlograms is much needed. (d) The amount of periodicity to be expected in a coherent random series may be examined theoretically or experimentally. It does not differ appreciably from that found in actual pressure series. Hence if a wave system appears there is no special ground for expecting the continuation which is necessary as the basis of a weather forecast. During a stretch of 144 days there was no evidence that a periodicity of, say, p days observed over a stretch of 3 p days had any appreciable tendency to persist over another 3 p days. (e) Surges and hollows also tend to make points of positive or negative symmetry, and experiments with coherent random series produce as much symmetry as we find in actual pressure series. Thus the symmetry that is observed, like the periodicity, is due to ordinary atmospheric conditions; and even when the symmetry is strongly marked continuity cannot be relied upon to justify prediction.