Autoregressive Behavior of Aviation Weather Forecasts

We analyze U.S. Terminal Aerodrome Forecasts (TAFs) for autocorrelation of forecast changes for a given aerodrome and hour between successive forecasts. We find a pessimistic and overreacting bias. When forecasters adjust a TAF, they tend to back out some of their changes in a later TAF. This tendency is much more pronounced when forecasters worsen a forecast than when they improve it.