中文摘要：

针对已有的时空模型研究深度不够的问题,该文主要研究两类时空理论变异函数模型,即分离和非分离模型,并分析其对时空预测精度的影响。首先,阐述了两类模型中较为流行的6种时空理论子模型及其参数物理意义;最后以2014年山东省PM2.5日均浓度为例,给出各模型理论表达式,并在Matlab中对各模型运用遗传算法进行参数拟合和时空预测精度验证。实验结果显示,6种模型的整体预测精度顺序与模型拟合精度顺序相当,说明时空理论模型同空间理论模型一样,其模型拟合精度与后续预测精度成正相关。

英文摘要：

Aiming at the lack of study on the spatio-temporal model, two spatio-temporal variation functions（separation model and non-separation model）and its influence on the spatio-temporal prediction accuracy were analyzed in this paper. Six sub models and its physical meaning of the parameters of two spario-temporal models were elaborated. Taking the PM2.5 daily concentration of Shandong province in 2014 as an example, the theoretical expression of each model was proposed, and each model with genetic algorithm theoretical was fitted and the prediction accuracy were calculated by Matlab. The results showed that for the model fitting accuracy, the order of RMSE and MAE was same as relative accuracy, which showed that the prediction accuracy had positive correlation with relative accuracy.