However, the calibration technique does not work well in all circumstances. For some seasons and variables there is no skill improvement for the week-2 forecast. Figure 6 shows there is almost no improvement for the 1000-hPa height field RPSS score after day 7. Previous studies have indicated that there remains room for improvement in the week-2 forecasts, as can be seen when comparing the calibrated and the optimal ensembles displayed in Fig. 2. How can we improve the current calibration technique? Do we need a hindcast for the calibration of the week-2 forecast? These questions will be discussed in the next section.