Apart from the disconcerting news about his own demise, the old king's ghost in Hamlet is chiefly remembered for his salutation: "Hamlet, Hamlet, I am thy father's spirit!"; or, as I am told it goes - phonetically - in Afrikaans: Omelette, Omelette, ekk iss di Poppi's spook! But Hamlet pere had a sorry tale to tell. He was, he said,
Cut off even in the blossoms of my sin
Unhousel'd, disappointed, unanel'd,
No reckoning made, but sent to my account
With all my imperfections on my head.
Now "cut-off" has a different meaning for a meteorologist, but it is a matter of almost just as much concern. It refers to the time at which the computer is told to get on with the business of working out a forecast, after which all further observations are ignored.
The foundation on which all computerised predictions of the weather are constructed is a "snapshot" of the atmosphere at a certain instant, based on weather observations carried out at perhaps a thousand places all around the world. Armed with the appropriate mathematical equations, and given this "starting picture" of the distribution of pressure, temperature and humidity, the computer calculates expected values of these elements at some future time. But some weather observations arrive at their destination more quickly than do others; 50 per cent, perhaps, may arrive within the hour, another 20 per cent may come in the succeeding hour, and the stragglers then arrive in dribs and drabs throughout the remainder of the day.
This poses a dilemma for the meteorologist. The more weather observations the computer has, the more detailed will be its initial picture of the atmosphere - and therefore the more accurate the ultimate prediction. But if one waits too long before telling the machine to proceed and do its calculations, the forecast will become available too late to be of any use. The "cut-off" therefore - the time when the meteorologist must say "No more! We must get on with it" - is a necessary compromise which may have a direct bearing on the quality of forecasts.
Normally for, say, a five-day forecast, which requires observations from all around the world, but where, on the other hand, early availability is not so critical, the cut-off may be six hours or more from the nominal time of the original observations. For short-range predictions, on the other hand, where timely access by the forecaster is of the essence, the cut-off must be reduced to, say, 120 minutes; the computer model, like old king Hamlet, is
sent to my account
With all my imperfections
on my head.