A Hybrid Framework for Improving NMME Precipitation Forecasts

Tuesday, 15 December 2015: 08:30
3011 (Moscone West)
Shahrbanou Madadgar1, Linyin Cheng1, Andrew W Wood2, Amir Aghakouchak3 and Mark Svoboda4, (1)University of California Irvine, Irvine, CA, United States, (2)National Center for Atmospheric Research, Boulder, CO, United States, (3)University of California Irvine, The Henry Samueli School of Engineering, Irvine, CA, United States, (4)Univ of NE/Lincoln-Nat'l Rsrcs, Lincoln, NE, United States
Precipitation forecasting is a major challenge in most regions around the world. This study focuses on improving global dynamic model forecasts in the Southwest United States, which is often affected by multi-year droughts. The North American Multi-Model Ensemble (NMME; Kirtman, 2014), consisting of 99 ensemble members from eight dynamic models, has shown limited skill in predicting winter precipitation in the region. This study proposes a hybrid approach that combines the NMME model simulations with statistical forecast – i.e., a Bayesian-based forecast model that leverages the predictability of atmosphere-ocean teleconnections. Specifically, winter precipitation is predicted given the state of indices including the Multivariate ENSO Index (MEI), Atlantic Multi-decadal Oscillation (AMO), Pacific North American index (PNA), and North Atlantic Oscillation (NAO). An Expert Advice (EA; DeSantis et al., 1988) algorithm is utilized first to find the most skillful ensemble members from NMME, and then to combine the dynamical and statistical models, leading to improved seasonal climate predictions that can benefit drought prediction. The model showed a significant improvement of NMME precipitation predictions during the 2012-2014 droughts in CA and NV – e.g., the approach forecasted the negative precipitation anomalies in 2012-2014 over most parts of the southwest United States. Overall, the hybrid approach results are encouraging for further exploration and could potentially prove to be useful for operational climate prediction.