Enhancing sub-seasonal soil moisture forecasts through land initialization

Abstract

We assess the relative contributions of land, atmosphere, and oceanic initializations to the forecast skill of root zone soil moisture (SM) utilizing the Community Earth System Model version 2 Sub to Seasonal climate forecast experiments (CESM2-S2S). Using eight sensitivity experiments, we disentangle the individual impacts of these three components and their interactions on the forecast skill for the contiguous United States. The CESM2-S2S experiment, in which land states are initialized while atmosphere and ocean remain in their climatological states, contributes 91 ± 3% of the total sub-seasonal forecast skill across varying soil moisture conditions during summer and winter. Most SM predictability stems from the soil moisture memory effect. Additionally, land-atmosphere coupling contributes 50% of the land-driven soil moisture predictability. A comparative analysis of the CESM2-S2S SM forecast skills against two other climate models highlights the potential for enhancing soil moisture forecast accuracy by improving the representation of soil moisture-precipitation feedback.