A direct representation of the range of forecasts represented by an ensemble prediction system is the spaghetti plot which displays a specific contour for each individual ensemble member on a map. The reference contours are chosen to be meteorologically significant such as the 0 degree C contour or a specific 500 mb height. The contour lines of each member can illustrate the geographical variation the ensembles; and the relative dispersion of contour lines provides a qualitative measure of the confidence. For ensembles with many members (e.g. ECMWF EPS with 51 members), plotting summary statistics – ensemble mean and standard deviation – can better highlight and more quantitatively depict uncertainty in the forecasts. These maps present contour lines of the mean forecast overlayed on shading to represent the standard deviation. These maps enable both the best forecast and its certainty to be viewed for a wide region in a single image. AER's eCast(tm) product provides both types of maps based on the ECMWF and GFS ensemble products.
Quantitative presentations of forecast confidence in an operational context lend themselves to station (point) forecasts. In addition to traditional measures of (deterministic) forecast accuracy, the probabilistic nature of ensemble forecasts requires consideration of the forecasts reliability and sharpness. eCast provides bias-corrected and calibrated station forecasts for daily maximum and minimum surface temperatures to address the needs of accuracy, reliability, and sharpness.
Operational corrected and calibrated station forecasts are presented in both graphical and tabular formats. Tabular forecasts display both the forecasts of individual members at standard positions in the ensemble (e.g. lowest/highest member, 50% and 80%), as well as the ensemble mean of all members. The full ensemble of station forecasts are available via a non-interactive interface. Ensembles of station forecasts also graphically depicted with "plume plots", showing the series of forecasts of each ensemble member for the whole forecast length. Various ensemble statistics and additional data are also highlighted in the plots (climate normal, high/low, 10%/90% members, ensemble mean).