Drought in California is very variable from year-to-year and is highly influenced by precipitation in winter months, causing up to billions of dollars of damage in a single drought year. Improved understanding of the variability of drought on decadal and longer timescales is then essential to regional water resources planning and management in this U.S. state. This is a predictability study of the Palmer Drought Severity Index (PDSI) based on a time-series of annual data, started in 1801 to 2014, and projected for the time-horizon 2015-2054. An ensemble smoothing forecast was developed with an exponential smoothing model with seasonal component. The model was implemented for forecasting 40 years of the PDSI index. Results were compared with a linear transfer function model approach in where the Pacific Decadal Oscillation (PDO) index and Nino3.4 index are both used as input time series in the transfer function model. Both approaches result in a Mean Absolute Standardized Error (MASE) lower than 1 and similar Root Mean Square error (RMSE).