IERS Bulletin A gives an expression for the uncertainty of its UT1-UTC data predictions:
S t = 0.00025 (MJD-today)**0.75 where "today" is the MJD of the bulletin's publication. The Bulletin only predicts a year ahead. Applying that formula gives an uncertainty a year ahead of 21 ms. It certainly ought to be possible, based on such a prediction, to decide with certainty whether a leap second will be required within that year. With the six-month scheduling cadence and a one-year prediction, we'd expect the kind of scheduling that we actually see: shortly after each leap opportunity they can look ahead to the next opportunity but one, and decide whether there needs to be a leap at the next opportunity. It seems to me that a switch to a monthly scheduling cadence, as Rob Seaman advocates, would have the immediate benefit of allowing a ten or eleven month scheduling lead instead of the current five or six months, without any advance in predictive ability. Immediately after each leap opportunity they could look *twelve* leap opportunities ahead, and thus decide whether the eleventh would be a good time for a leap. This is in addition to the ability to keep UT1-UTC within tighter bounds, which Rob's proposal describes. But I digress. I'm wondering how good UT1 predictions further ahead are. If the formula remains valid, it suggests that UT1 could be predicted to within 100 ms as far as eight years ahead. 100 ms is certainly a good enough prediction to schedule leap seconds on. My assumption there is highly suspect, though. Anyone know better? Does IERS publish any EOP predictions more than a year ahead? -zefram