摘要
本文把1958—1984年天山乌鲁木齐河源1号冰川的融水径流总量的时间序列X(t),分解为趋势项L(t)、周期项P(t)、平稳项S(t)、随机项ε(t),使该时间序列表示为X(t)=L(t)+P(t)+S(t)+ε(t)。采用非线性回归提取L(t),用谱分析和Fourier级数提取P(t)余差用自回归方程建模,用上述项的叠加作出预报,按相关指数公式计算R=0.90,效果良好。
The main purpose of this paper is to decompose the time series of snowmelt runoff X(t) into four terms: (a) trend or longterm movement L(t); (b)sessonality about the trend P(t); (c) stationary term 5(0; (d) white noise term ε(t1), so that the series can be expressed
as
X(t) = L(t)+ P(t)+ S(t)+ ε(t).
L(t) is obtained by nonlinear regression and P(t) by spectral analysis and Fourier Series. After L(t) and P(t) have been eliminated, we can establish an autoregressive model for the remainders. By superimposing these terms, we can make the simulation as well as predication. The results are satisfactory.
出处
《冰川冻土》
CSCD
北大核心
1990年第2期161-165,共5页
Journal of Glaciology and Geocryology
关键词
冰川
年径流总量
乌鲁木齐河源
时间序列模拟
The decompossition of timeseries, spectral analysis, autoregressive model, checking and forecasting
