The performance of the probabilistic multi-model prediction (PMMP) system from the APEC Climate Center (APCC) in predicting the Asian summer monsoon (ASM) precipitation at a four-month lead (February initial condition) has been evaluated and compared with the corresponding prediction from a statistical model, based on hindcast data from 1983 to 2005 and real-time forecast data from 2006-2010. Particular attention was paid to the probabilistic prediction of precipitation in summers after the mature phase of El Niño and Southern Oscillation (ENSO) events. Taking into account the fact that the forecast skill of coupled models for the boreal spring and summer precipitation mainly comes from the models’ ability to capture the ENSO teleconnection, we developed a statistical model based on a regression method using the preceding winter ENSO condition as a predictor. Although PMMP has limited skill for the ASM precipitation, it provides more consistent and stable skill than the statistical model. First, although the PMMP has a higher skill for both above- and below-normal categories during the real-time forecast period, the cross-validated statistical forecast has a higher skill during the 23-year hindcast, which indicates that the cross-validated statistical skill may be overestimated. Second, the PMMP is a better tool than the statistical model for capturing the atypical ENSO teleconnection that has affected the ASM precipitation during the recent decade and early 1990s. Third, the statistical model is more sensitive to the ENSO phase than the PMMP, thus it better predicts the ASM precipitation after the mature phase of La Niña.

1. INTRODUCTION

2. DATA AND METHODOLOGY

2.1 The APCC PMMP system

2.2 Statistical model

2.3 Observational datasets

2.4 Forecast quality assessments

3. RESULTS

3.1 Evaluation of PMMP

3.2 Development of the statistical model using its climate contributor

3.3 Evaluation of the statistical prediction model

3.4 Dynamical versus statistical prediction

3.5 JJA 2010 forecast from the statistical model

4. DISCUSSION

5. CONCLUSION