INTRODUCTION

The two great outstanding difficulties are those connected with the completeness necessary in the initial observations and with the elaborateness of the subsequent process of computing. These are discussed in Ch. 11/1 and Ch. 11/2. The scheme of numerical forecasting has developed so far that it is reasonable to expect that when the smoothing of Ch. 10 has been arranged, it may give forecasts agreeing with the actual smoothed weather. When that stage has been attained, the other difficulties will tend to group themselves with questions of the desirability of weather forecasts and of their cost. We need here an estimate of the economic value of a forecast reliable for n days ahead, given as a function of n. As with improved methods n is likely to increase, so forecasts will become of more value to agriculturalists. Now the annual value of the world's food crops is at least £1000,000,000, so that a very tiny fractional saving would correspond to a large sum.

THE PROBLEM OF OBTAINING INITIAL OBSERVATIONS

Pattern

When the observations are taken at stations scattered irregularly, an interpolation has to be made to find the initial data at the centres of the chequers of our chessboard. It has been mentioned in Ch. 9/1 that this interpolation was found to be both troublesome and inaccurate, as may be evident from the map on p. 184.