Multiscale Nonlinearity of Error Growth in High-Resolution Forecasts for Mesoscale Convective Systems

Abstract In this study, the nonlinearity in a weather forecast, defined as the extent to which the initial opposite-sign state vector does not keep the same magnitude and opposite direction in a forecast time, was examined in an environment containing a multiscale convective system. A pair of 18-h forecast experiments with initial perturbations of different signs was conducted for a heavy rainfall event in western Japan on 13 August 2021. Despite the initially different signs, the perturbations at a given forecast time had the same sign, which was emphasized by the low-pass filter. This result suggested that the nonlinearity was found not only on the convective scale but also on the meso-α scale. An additional forecast experiment with the initial perturbation only for water vapor applied in a limited area indicated that this meso-α scale nonlinearity originated from the asymmetry of water vapor perturbations to avoid supersaturation in the wet environment. The nonlinearity was propagated by large-scale gravity waves emitted from the perturbed convective areas. This explanation was supported by a pair of forecast experiments with dry and moisture perturbation areas on a scale larger than that of convection, in which the state vectors kept the opposite sign at a given forecast time.