Fig. 5 Growth curves of rms vector wind error in m/s for resolution T170 (dotted) vs T126, as a function of forecast length, averaged over all February cases (N=21).