| Numerical weather prediction is very sensitive to initial condition because of the chaotic nature of atmospheric system. Recent studies showed that even small scale and small amplitude initial errors could contaminate short-range deterministic mesoscale forecast. This rapid error growth was flow-dependent and highly related to moist convection. In this study, the predictability of a Meiyu heavy rainfall event which occurred during4-6July2003in eastern China is investigated by mesoscale numerical model MM5, including the practical predictability characteristics of Meiyu heavy rainfall, the down-scaling of large-scale initial errors, the evolution and spacial propagation features and mechanisms of various-scale differences.Practical predictability is related to realistic uncertainty of initial states under current condition, which can be represented by difference between initial conditions from two analysis centers. The practical initial uncertainty concentrates on large scale and gives rise to middle-and small-scale errors soon after the simulation starts. The error-growth of the experiment with large-scale practical initial errors can be divided into five stages according to its scale feature.1)0-3h, the down-scaling of large-scale difference;2)3-6h, small-scale error growth and saturation;3)6-12h, increase of middle-scale error with some equilibrium features;4)12-21h, equilibrium large-scale error growth;5) collective amplification of all scale errors.The down-scale evolution process of large-scale error is investigated detailed in this study. Under the combined impacts of large-scale perturbation pressure difference and wind advection difference, middle-scale potential vorticity (PV) error appeared near the local depression center, giving rise to precipitation deviation and inducing many isolated small-scale errors to rainwater mixing ratio. The diabatic heating from phase state transition leads to small-scale potential temperature difference, after which convection error accrues. In the process of down-scaling, diabatic heating has played very important roles.One can make much more accurate forecast by reducing the amplitude of initial error. But, as the continuous halving of error amplitude, the error grows much fast with its doubling time decreasing markedly. It is demonstrated that using the method of reducing initial error amplitude can only improve Meiyu rainfall prediction skill by less than18hours. The above conclusions would suggest an inherent, finite limit to the predictability of Meiyu front system, as successive refinements of the initial condition would bring smaller and smaller increments to skillful forecast length.The rapidly spacial propagations of initial errors are very important to local region error-growth. The spacial propagating characteristics of differences depend on scale concerned. Small-scale errors spread as isotropic circle through acoustic wave. Then, along with the evolution of moist convection, it breaks up:concentrating error energy to new developing convection-scale differences, which will spread isotropic again after amplification. Through above circulation process, small-scale differences increase and grow up-scale fast.Mesoscale errors may propagate wavily upstream and downstream through gravitational wave as well as transferring downstream together with moist convection by horizontal wind advection after forming. Moreover, propagation of error energy against the mean flow is very important to error growth, because it can even accumulates differences locally at place far upstream from initial error source and lack of moist convection. Under the effects of upstream propagation, the local error will take on a waveshape appearance.Both the small-scale acoustic wave spreading and mesoscale gravity wave propagation are independent of diabatic heating. To generalize the predictability features of Meiyu heavy rainfall from real case studies, an ideal precipitation process of Meiyu front is simulated here. The idealized initial condition is comprised of a low-level, warm-core3-dimension vortex embedded in zonal homogeneous weakly sheared upper and low level jets. The simulated front system contains many features like real Meiyu front, such as strong upper and low level jets, weak horizontal temperature contrast, high humidity gradient, cyclonic shear at low troposphere level and abundant warm moist flow transportation.Predictability of this idealized Meiyu front rainfall is studied then. The initial small-scale errors firstly grow at moist convection regions, and then quickly spread to the whole low pressure area. Under the influence of strong moist convection, the errors of all scales amplify together, among which, small-and middle-scale errors show wave like evolution in accordance with precipitation increasing and decreasing, while large-scale error becomes saturated and keeps the slow growth. From the vertical distribution point of view, dry difference total energy (DTE) is characterized by propagation from lower to upper level. The DTE is dominated by middle troposphere error which is caused by the maintenance of600hPa latent heat energy error (LHE). After24h of simulation, the low-level error of mixing ratio grows rapidly leading to a local maximum to DTE under900hPa. In order to combining DTE and LHE linearly, this study has defined a notion called moist difference total energy (MDTE) with the weight of LHE equals to0.2. The result of MDTE not only contains propagation features of DTE from700hPa to500hPa, but also shows the error maintenance features of LHE at600hPa and low-level troposphere. Moist convection plays an important role on error growth. When initial environmental relative humidity decreases, the simulated moist convection and rainfall also lessen, so the error growth is inhibited. |
PDF Full Text Request