Investigations into the Subseasonal Predictability of Soil Moisture and Streamflow

Abstract The predictability of hydrological variables (specifically, soil moisture and streamflow) at subseasonal time scales is examined with an offline (land only) modeling system. Meteorological forcing for the offline system is derived from state-of-the-art subseasonal to seasonal (S2S) forecast products, ensuring that the offline simulations performed constitute fair forecasts in their own right—no observations from the forecast period are employed in the simulations. Validation data for the offline forecasts consist of in situ measurements of soil moisture and streamflow. The offline system’s hydrological forecasts are found to be at least as skillful as those generated by the fully coupled S2S system. The subseasonal streamflow forecasts are less skillful than the soil moisture forecasts due to the poor quality of the subseasonal precipitation predictions; we can, however, increase the streamflow forecast skill considerably by using climatological precipitation forcing during part or all of the forecast period, thereby increasing the relative contribution of soil moisture initialization to the forecasts. Skill for both soil moisture and streamflow is found to be sensitive to the depth of the modeled soil column, with deeper depths, associated with longer soil moisture memory, becoming more advantageous at longer leads. Overall, our results illustrate the potential of using the offline system for extracting additional hydrological prediction skill from standard S2S forecasts and, more generally, for examining the nature of hydrological predictability and the physical factors that control it.