This project examines the systematic biases of the APEC Climate Center (APCC) coupled models (i.e., CCSM3 and SCoPS) in simulating the observed tropical Pacific Ocean climate and the El Niño-Southern Oscillation (ENSO). It is found that the APCC coupled models can capture the key observed features of the tropical Pacific climate and the ENSO. Particularly, some advantages are found in the SCoPS model that will replace the CCSM3 model for operationally forecasting seasonal climate. For instance, the SCoPS model can more realistically simulate the observed asymmetric features of the annual mean rainfall in the meridional direction, and the annual cycle of the sea surface temperature (SST) and zonal winds along the equator. The temporal evolution patterns of equatorial SST anomalies, the zonal wind anomalies, and the anomalous zonal currents associated with the ENSO are better simulated by the SCoPS model. However, both models have several biases with respect to the observations. These include a cold bias in the central equatorial Pacific and a warm bias in the extratropical northeastern and southeastern Pacific Ocean, a thermocline that is too diffuse, precipitation that is too strong in off-equatorial regions, an ENSO shifted to the west which extends too far westward, and the underestimation of the zonal advective feedback, thermocline feedback and thermodynamic damping among the ENSO related feedbacks. This project also attempts to improve the initial conditions in the CCSM3 model for better ENSO forecasts. With the GODAS nudging initialization method, where initial conditions are generated in a coupled simulation performed with the simulated ocean temperatures nudged to those of the reanalysis, the CCSM3 model has better ability in predicting the ENSO amplitude, global SSTs, and extratropical air temperatures. This is possibly due to the reduced initialization shock, which is a severe problem in predictions using the GODAS method, where the ocean reanalysis data is used as the initial conditions without a nudging process. To improve further the forecasting skill for strong ENSO events or ENSO amplitude, we also implement the GODAS-WIND nudging method, where not only 3-D ocean temperatures but also surface winds are nudged to their observations. However, there is no discernable improvement of the ENSO forecast with the GODAS-WIND nudging initialization method except for forecasts initialized in the late spring. It is also found that the APCC CCSM3 model has a year-to-year variation in the forecasts for the tropical Pacific and the global SST, which is due strongly to changes in the ENSO intensity. That is, a level of their forecast skill increases with the ENSO intensity. Since this result implies the importance of obtaining accurate ENSO amplitude forecasts in order to obtain the better global and tropical Pacific SST predictions, we also attempt to find possible reasons for the increasing errors of the ENSO amplitude in the APCC CCSM3 model with increases in forecasting lead times using the BJ index analysis. Based on a series of analysis including the BJ index, we conclude that the forecasting error for the ENSO amplitude may be related to the tropical Pacific cold bias, which strongly affects the intensity of the thermocline feedback.

1. INTRODUCTION

2. DATA AND METHODOLOGY

2.1 Observational Datasets

2.2 Models

2.3 Hindcast Simulations

2.3.1 Initialization Methods

2.3.2 Forecast Verification Methods

2.4 Bjerknes Stability (BJ) Index Analysis

3. LONG COUPLED SIMULATION ANALYSIS

3.1 Annual Mean States

3.1.1 SST

3.1.2 Surface Winds

3.1.3 Subsurface Ocean Temperatures

3.1.4 Precipitation

3.1.5 Seasonal Cycle

3.2 Tropical Pacific SST variability

3.2.1 Spatial Patterns

3.2.2 Temporal Patterns

3.2.3 Seasonal Phase Locking

3.2.4 ENSO Stability

3.3 ENSO Teleconnections

3.3.1 Geopotential Height at 200 hPa

3.3.2 Precipitation

4. ASSESSMENT OF PREDICTION SKILL OF THE TROPICAL PACIFIC CLIMATE

4.1 Improvement of Prediction Skill from the GODAS to the GODAS-Nudging Method

4.1.1 ENSO Prediction

4.1.2 Global Climate

4.2 Role of Surface Winds in Initial Conditions

4.3 Year-to-Year Variation of Forecast Skill

4.4 ENSO Feedbacks and Model Drift

5. SUMMARY AND CONCLUDING REMARKS

REFERENCES