集合预报物理基础的探讨

将集合预报中的每次积分算程视为非平衡统计物理理论中的准粒子轨迹，由此对Lorenz模型 进行了数值试验，计算了初值位于不同性质平衡态附近时准粒子数处于基态和第一激发态随 时间的演化.结果证明：(1)若动力系统在整个相空间内存在稳定的平衡态，在稳定的平衡态 附近，系统随时间长期演化行为是可预测的.(2)若动力系统在整个相空间内不存在任何稳定 的平衡态，初值位于远离非稳定的平衡态，则在1—2周内准粒子多数分布在低能量态，即预 报是最可几率的.(3)若初始状态位于非稳定平衡态附近，系统随时间的演化几乎是不可预测 的.这从理论上说明了作大量积分算程的集合预报其效果会比单一初值的单程积分要好.这就 从物理上对集合预报能提高准确率提供了一种解释. 关键词： 集合预报 Lorenz模型 正则分布 概率密度分布

On physical basis of ensemble prediction

By viewing each integral process in the ensemble prediction as a locus of quasi-particle in the nonequilibrium statistical physics theory,numerical experiments of Lorenz model are performed,and under the circumstance that the initial value of the model is near the equilibrium state of different properties,the temporal evolution of the number of quasi-particles in the ground state and the first excited state is calculated in this paper.The results confirm that if the dynamic system has stable equilibrium states in the whole phase space,the long-range temporal evolutional behavior of the system in the vicinity of the stable equilibrium states is predictable.If the dynamic system has not any stable equilibrium state in the whole phase space and the initial value of the system is far away from the unstable equilibrium states,then most of the quasi-particles lie in the low-energy state within 1-2 weeks,i.e.it is most probable.If the initial value lies in the vicinity of the unstable equilibrium states,the temporal evolution of the system is almost unpredictable.This proves theoretically that the effect of the ensemble prediction obtained after performing a large number of integral processes in better than that from the single integral process of a single init ial value.This offers a physical explanation why the ensemble prediction has a h igher accuracy.