一种改进的相空间近邻等距模式及其在短期气候预测中的试验

An Improved Equal Distance Model of Near Neighborhood in the Phase Space Used in Short-Term Climate Prediction Experiments

现有研究表明，气候要素系统可能存在混沌现象．通过应用混沌理论中的重建相空间技术，可在相空间中揭示出传统方法无法揭示的复杂气候动力特征，为气候预测研究开创一条新途径．现有相空间理论的研究多着重于对混沌系统的识别和应用混沌相空间理论建立不同的模式进行提前预测，这些预测多为对月时间尺度的；对于如何提高预报准确率和延长预报时效，从而实现对多个月份、季、年等更长时间尺度提前预报的研究还不多见．基于混沌理论的相空间近邻等距模式，对该理论中的提前预报时间T、延迟时间τ和相空间维数d等进行讨论．提出，实际应用中根据提前预报时间T来建立新的时间序列，满足T=τ=sδt，消除了相空间时滞τ的变化对提前预报时间尺度T的影响．应用改进前、后的预报模式，以不同相空间维数d对气温和降水气候要素做提前T=1～15个月的预报试验和检验．结果发现，改进后的模式提高了预报准确率，延长了预报时效．

Many studies show that nonlinear chaotic properties may exit in climate system. By introducing phase space reconstruction techniques of chaotic theory, it can discover the conventionally unknown climate dynamics, and will break a new path to climate predictions. Previous studies mainly focus on identifying the chaotic properties of time series and carrying out forecasts month by month with different models. Few work has been clone on how to improve the predictability and extend the lead time. In this paper, an improvement on the equal distance model of near neighborhood in the phase space is proposed. In practical applications, it is found that a new time series can be established based on the lead time to meet the requirement that the lead time T and the delay time τ should be the same, that is, T=τ=sδt. This can remove the influence of τ upon T. The improved model is practiced in the prediction experiments for the time series of summertime temperature and precipitation （from June to August） in Yichang. The 5a prediction experiments （1996- 2000） show that, the improved model has a relative higher predictability than before, extending the lead time.