Downscaling 15-day Ensemble Weather Forecasts and Extension to Short-term Climate Outlooks
A diagnostic downscaling system was applied to daily output from the NCEP GFS Ensemble Prediction System, to provide an ensemble of estimates of daily climate variables at 114 New Zealand climate stations for a 15-day forecast interval. Downscaled forecasts were validated on a day-by-day basis and for time averages through the 15-day forecast interval. Fifteen-day average forecast anomalies of temperature and precipitation were used to estimate probabilities of monthly tercile outcomes for six climate regions of New Zealand, using a conditional climatology approach. Daily forecasts exhibited useful (and statistically significant) levels of skill through the first week of the forecast interval, while five-day averaged forecasts generally exhibited useful skill out to the day 6-10 average. Forecast probabilities of wet and dry spells were found to have a small amount of skill through week two of the forecast interval, in the sense of predicting whether the week will be generally wet or dry. Based on forecast average 15-day anomalies, one-month forecasts of mean temperature exhibited positive skill, while precipitation forecasts exhibited no skill. The downscaling system appears to be a useful tool for downscaling medium-range forecast output from a global NWP model.