On spatiotemporal series analysis and its application to predict the regional short term climate process

Based on the theory of reconstructing state space, a technique for spatiotemporal series prediction is presented. By means of this technique and NCEP/NCAR data of the monthly mean geopotential height anomaly of the 500-hPa isobaric surface in the Northern Hemisphere, a regional prediction experiment is also carried out. If using the correlation coefficientR between the observed field and the prediction field to measure the prediction accuracy, the averagedR given by 48 prediction samples reaches 21%, which corresponds to the current prediction level for the short range climate process.