The APEC Climate Center (APCC) has provided the real-time Boreal summer Intraseasonal oscillation (BsIso) forecast service because of the prominent role of BsIso in subseasonal to seasonal predictions (s2s) predictions. This study focuses on in terms of the improvement of the real- time BsIso forecast system through additional forecast contents and the construction of the BSISO real-time verification system at APCC. The added forecast content, a 5-day mean of outgoing Longwave Radiation anomalies for 20-day forecasts, demystifies the BSISO activity and makes it more intuitive. The real-time verification system at the APCC produces verification information about BSISO indices using five skill metrics such as correlation, root-mean-square-error, phase error, and amplitude error and releases this information on the APCC website.

In addition, the ability of five operational models to predict BSISO indices is evaluated using information from the previous two years, 2013– 2014. BsIso forecasting performance indicates that BsIso indices are generally predictable from 1 week to over 20-day forecast lead time, and the predictability of BsIso 2 is better than that of BsIso 1 in the many phases. ECMWF yields the best performance in most skill metrics, but it is worse than the other models in representing BsIso phase amplitude. Predictability increases somewhat for forecasts with well-developed BsIso at the initial time. This suggests that a more skillful forecast for the BSISO may be associated with a stronger initial amplitude. According to the performance at each phase, phases 7 and 8 are more predictable for most of the models. The potential BSISO predictability in each model far exceeds the actual skill while the upper bound for predictability varies across the models. That means the BSISO forecast skill can be improved by using better initial conditions and model physics. In conclusion, the models’ performance in predicting the BsIso varies with the initial phase and amplitude of BsIso. A multi-model ensemble system is therefore advantageous in BsIso predictions.

1. INTRODUCTION

2. Data and Methodology

2.1 Data

2.1.1 Forecast Data

2.1.2 Reanalysis Data

2.2 Methodology

2.2.1 Real-time BsIso indices

2.2.2 Verification Metrics for BSISO Forecasts

3. RESEARCH RESULTS

3.1 supplement of forecast contents

3.2 Construction of real-time verification system

3.2.1 Procedure for the skill assessment of BSISO prediction

3.2.2 Skill Assessment

3.2.2.1 Individual model

3.2.2.2 Intercomparison

4. CONCLUDING REMARKS

REFERENCES

FIGURE CAPTIONS

TABLES

FIGURES

OPERATIONAL MANUAL