摘要
Using the mesoscale model MM5, the development of initial condition uncertainties at different scales and amplitudes and their influences on the mesoscale predictability of the "0185" Shanghai heavy precipitation event are investigated. It is found that different initial conditions obtained from different globe model analyses lead to large variations in the simulated location and strength of the heavy precipitation, and the scales and amplitudes of the initial condition perturbations significantly influence the model error growth. The power spectrum evolution of the difference total energy (DTE) between a control simulation and a sensitivity experiment indicates that the error growth saturates after 12 h, which is the predictable time limit of the heavy precipitation event. The power spectrum evolution of the accumulated precipitation difference between the control and sensitivity simulations suggests a loss of the mesoscale predictability for precipitation systems of scales smaller than 300 kin, i.e., the predictable space for the heavy precipitation event is beyond 300 km. The results also show that the initial uncertainties at larger scales and amplitudes generally result in larger forecast divergence than the uncertainties at smaller scales and amplitudes. The predictable forecasting time and space can be expanded (e.g., from 12 to 15 h, and from beyond 300 kin to beyond 200 km) under properly prescribed initial perturbations at smaller scales and amplitudes.
Using the mesoscale model MM5, the development of initial condition uncertainties at different scales and amplitudes and their influences on the mesoscale predictability of the "0185" Shanghai heavy precipitation event are investigated. It is found that different initial conditions obtained from different globe model analyses lead to large variations in the simulated location and strength of the heavy precipitation, and the scales and amplitudes of the initial condition perturbations significantly influence the model error growth. The power spectrum evolution of the difference total energy (DTE) between a control simulation and a sensitivity experiment indicates that the error growth saturates after 12 h, which is the predictable time limit of the heavy precipitation event. The power spectrum evolution of the accumulated precipitation difference between the control and sensitivity simulations suggests a loss of the mesoscale predictability for precipitation systems of scales smaller than 300 kin, i.e., the predictable space for the heavy precipitation event is beyond 300 km. The results also show that the initial uncertainties at larger scales and amplitudes generally result in larger forecast divergence than the uncertainties at smaller scales and amplitudes. The predictable forecasting time and space can be expanded (e.g., from 12 to 15 h, and from beyond 300 kin to beyond 200 km) under properly prescribed initial perturbations at smaller scales and amplitudes.
基金
Supported by the China "973" Project under Grant No. 2004CB418304
the National Natural Science Foundation of China under Grant Nos. 40745026 and 40875035