The impact of climatology and systematic errors upon the skill of DERF forecasts

The average skill of 61 thirty-day forecasts of 500 mb geopotential height from the Dynamic Extended Range Forecast (DERF) experiment are investigated. These forecasts were made using the National Meteorological Center's Medium Range Forecast model and starting from initial analyses during the winter of 1986/87. The impact upon extended range forecast skill of the removal or retention of systematic errors and low-frequency climate variability is studied. If the systematic error is removed a posteriori at each forecast lead time, then the skill of forecasts of time averages may be improved considerably. The magnitude of this improvement is difficult to quantify with forecasts from a single season. Nearly all of the skill at extended range in the DERF experiments arises from the successful forecast of low-frequency fluctuations in the large-scale circulation.