平均环流参数化预报的初步试验

本文提出一个对包括下垫面温度在内的未知总强迫项进行参数化作平均环流的预报方法.该参数化方案是假设未知作用力距平与流函数距平成正比,其比例系数是空间位置的函数,得到一个平均环流距平的预报方程.该方程表明,预报结果的好坏不仅取决于初始状态和动力学的考虑,还取决于环流的历史演变情况. 根据历史资料反求其比例系数,所作的平均环流距平的历史拟合结果令人满意.独立样本的预报结果比不上历史拟合的结果.试验结果还表明,如果加长历史资料的时间序列,把310天的历史资料增加到620天,则独立样本的预报结果有较大的改善,不仅有很多旬平均和月平均的个例非常成功,而且预报的半均准确率已超过一般惯性预报的水平.

Preliminary experiments of parameterized forecasts of time-mean circulations

A scheme of parameterizing all unknown forcings for time-mean forecasts is proposed. It is assumed that the relationship between all unknown forcings and the anomaly streamfunction isspatially dependent, a parameterized model for time-mean circulation forecasts which depends not only upon the initial state and the dynamics but also upon the historic evolution of circulations is developed.The parameterized forecasts show significant improvement in a dependent data set. But the results of independent data set are worse than those of dependent one. The experiments here also show that the more the data, the better the forecasts in the independent data set. When the time series of the data set are long enough, the skills of forecasts are very successful in some cases and an superior to the persistence on an average sense.