摘要
以武汉站(5~9月)汛期降水量观测数据序列为例,将这类具有明显的不规则性(混沌特征)时间序列分解为宏观气候尺度周期的波动部分和迭加其上的微观气候尺度周期的波动部分,分别采用演化建模方法和自然基小波方法模拟逼近.特别强调由演化建模方法得到的非线性常微分方程较之传统的线性建模具有更好的分析预测能力.
The evolutionary modeling (EM), which is developed from genetic programming (GP), is a relatively new technique that is an adaptive method for solving computational problems in complex systems which are of chaotic character and nonlinear variation with time in many fields. The use of EM for the observed time series of precipitation in flood season ( May - September) at Wuhan station is studied in this paper. The time series of precipitation is split into two parts: one includes macro climatic timescale period waves that are affected by some relative steady climatic factors such as astronomical factors ( like sunspot number etc. ) as well as other factors that are known and unknown, the other includes micro climatic timescale period waves superimposed on the macro one. The evolutionary modeling (EM) is supposed to be adept at simulating the former part because it creates the nonlinear ordinary diferential equation (NODE) based upon the observed time series. The natural fractals (NF) are used to simulate the latter part. The final prediction is the sum of values from both methods, so the model can reflects multi - time scale satisfactory for climatic prediction operation. The NODE can suggest the data vary with time, which is benefit to think over climatic analysis and short - range climatic prediction. Comparison in principle between EM and a linear modeling AR(p) indicates that the EM is a much better method to simulate the complex time series being of nonlinear characteristics.
出处
《干旱气象》
2005年第4期1-6,11,共7页
Journal of Arid Meteorology
基金
国家自然科学基金项目(4207504)资助
关键词
时间序列
演化建模
短期气候预测
evolutionary modeling
NODE
time series
short- range climatic prediction