Suggestions in the Observational Record of Land–Atmosphere Feedback Operating at Seasonal Time Scales

Observed monthly precipitation anomalies are standardized across midlatitude land, and ergodicity is invoked to combine the spatially distributed data into probability density functions (pdfs) of precipitation conditioned on the strength of earlier anomalies. The conditional pdfs, though broad and overlapping, are indeed distinct at a high (99.9%) level of confidence. This implies a nonzero degree of predictability for midlatitude precipitation, even at 3-month leads. This behavior is reproduced by an AGCM only when land–atmosphere feedback in the model is enabled.