Geostatistical space-time modeling for temperature estimation

Air temperature is one of important environmental phenomena with both spatial and temporal characteristics. In order to realize spatial-temporal interpolation at any point in space-time field, a kind of practical product-sum covariance for spatial-temporal modeling is chosen for monthly average air temperature in the three provinces (Heilongjiang, Jilin and Liaoning Province) of Northeast China from January 1972 to December 2008. As every station´s temperature can be regarded as a time series, decomposition is processed and seasonal part is removed, and the residuals are generated for further analysis in space-time. Spatial-temporal variogram is built based on the ones of pure space and pure time. Extending 2d-kriging to 3d, the monthly average temperatures of all stations in January 2008 are estimated, and the effect is compared with spatial kriging. The result of compare shows that spatial-temporal interpolation is practical, and because of considering the correlation of both space and time, its accuracy is better than that of spatial kriging.