AbstractAccurate modelling and prediction of the local to continental-scale hydroclimate response to global warming is essential given the strong impact of hydroclimate on ecosystem functioning, crop yields, water resources, and economic security1, 2, 3, 4. However, uncertainty in hydroclimate projections remains large5, 6, 7, in part due to the short length of instrumental measurements available with which to assess climate models. Here we present a spatial reconstruction of hydroclimate variability over the past twelve centuries across the Northern Hemisphere derived from a network of 196 at least millennium-long proxy records. We use this reconstruction to place recent hydrological changes8, 9 and future precipitation scenarios7, 10, 11 in a long-term context of spatially resolved and temporally persistent hydroclimate patterns. We find a larger percentage of land area with relatively wetter conditions in the ninth to eleventh and the twentieth centuries, whereas drier conditions are more widespread between the twelfth and nineteenth centuries. Our reconstruction reveals that prominent seesaw patterns of alternating moisture regimes observed in instrumental data12, 13, 14 across the Mediterranean, western USA, and China have operated consistently over the past twelve centuries. Using an updated compilation of 128 temperature proxy records15, we assess the relationship between the reconstructed centennial-scale Northern Hemisphere hydroclimate and temperature variability. Even though dry and wet conditions occurred over extensive areas under both warm and cold climate regimes, a statistically significant co-variability of hydroclimate and temperature is evident for particular regions. We compare the reconstructed hydroclimate anomalies with coupled atmosphere–ocean general circulation model simulations and find reasonable agreement during pre-industrial times. However, the intensification of the twentieth-century-mean hydroclimate anomalies in the simulations, as compared to previous centuries, is not supported by our new multi-proxy reconstruction. This finding suggests that much work remains before we can model hydroclimate variability accurately, and highlights the importance of using palaeoclimate data to place recent and predicted hydroclimate changes in a millennium-long context.