The sensitivity of ensemble simulations of deep convective events in the southeastern United States to initial-condition (IC) errors is examined by imposing idealized moisture perturbations at small and large scales. Four severe weather events are considered, ranging from a springtime frontal system to convection driven almost exclusively by daytime heating. Events with strong synoptic-scale forcing were insensitive to the scale of IC errors, but weakly forced events exhibited greater sensitivity to small-scale than large-scale IC errors. Additional ensemble simulations of idealized convective systems suggest that the greater sensitivity to small-scale IC errors of the weakly forced cases arises from their higher sensitivity to the strength and location of the first convective elements. Ensemble spread and predictability are characterized by two measures: the ratio of the perturbation kinetic energy (KE) about the ensemble mean to the background KE and the neighborhood-based fractions skill score (FSS) of hourly precipitation with respect to that in an unperturbed reference simulation. For simulations of both observed events and idealized convective systems, the FSS appears to be a more discriminating indicator of differences in predictability between different convective events.
KEYWORDSconvective systems, error growth, initial conditions, numerical weather prediction, predictability, severe weather 1 et al., 2003;Selz and Craig, 2014;Sun and Zhang, 2016). Recent work has demonstrated, however, that relatively small initial-condition (IC) errors on much larger scales of (100) km can propagate rapidly downscale and subsequently reduce predictability in a similar manner to that produced by relatively large errors in the conditions on small scales, thereby making the identification of the actual initial scale of errors responsible for forecast degradation ambiguous (Durran et al., 2013;Durran and Gingrich, 2014;Durran and Weyn, 2016;Weyn and Durran, 2017). In particular, Weyn and Durran (2017, hereafter WD17) showed that intrinsic predictability in simulations of idealized mesoscale convective systems (MCSs) was independent of the horizontal scale of equal-absolute-amplitude IC errors across a range of different MCS types. This study seeks to answer the question of whether this scale independence is also present in Q J R Meteorol Soc. 2019;145 (Suppl. 1):57-74. wileyonlinelibrary.com/journal/qj How to cite this article: Weyn JA, Durran DR. The scale dependence of initial-condition sensitivities in simulations of convective systems over the southeastern United States. Q J R Meteorol Soc. 2019;145 (Suppl. 1):57-74. https://doi.