Longer memory climate processes help to produce reliable climate forecasts at decision making time scales. For example, sea surface temperature anomalies can be reliably predicted several months to years in advance. Land surface integrates random weather and climate variability by providing large storage capacity for water and its slow release to the atmosphere (evapotranspiration). Therefore, the soil moisture variability has a red spectrum similar to the SST variability. Hence, we hypothesize that soil moisture can also be predicted several months to year in advance despite limited predictability of precipitation. In this project, I will investigate (1) potential soil moisture predictability in the CESM-DPLE, and (2) explore if soil moisture prediction can be improved further by providing realistic soil moisture initializations.