Model analogues: Cheap forecasts and their implications



Ding et al (2018) and (2019) proposed an alternative forecast method (Model-Analogs) to avoid additional forecast ensemble integration. In this method, the predictions are obtained from a long preindustrial control simulation by matching their selected variables to observed fields. The forecast members at various lead times are then taken from the subsequent evolution of these states in the control simulation. They showed skillful seasonal prediction of tropical Pacific SST anomalies using Model-Analogs. The skills in some regions such as the eastern equatorial Pacific even exceed that from initialized seasonal forecasts.


Jeffree_ESMWG_23.pdf (8.0 MB)