Prediction of summer surface air temperature over Northern Hemisphere continents by a physically based empirical model

Summer surface air temperature (SAT) variability over Northern Hemisphere (NH) continents can profoundly impact human society, yet its seasonal prediction remains challenging, partly due to the limited prediction skill of dynamical models, especially over extratropical and high-latitude areas. Previous research has defined five indices associated with different atmospheric circulation patterns, which have important contributions to variations of summer SAT. This study further establishes a physically based empirical model (P–E model) using the Bayesian dynamic linear model method for the prediction of the indices, and uses the predicted indices to reconstruct the summer SAT anomaly field. Results show that the P–E model can reasonably well predict the five indices during 1950–2021. Combining this with the linear trend, the total summer SAT anomaly is also reconstructed. The high cross-validated hindcast skill for the period of 1950–2021 and independent forecast skill of 2022 indicate that the summer SAT over NH continents can be reasonably predicted by the P–E model.