Research and Application of Operational MJO Monitoring and Prediction Products in Beijing Climate Center

Both Madden-Julian oscillation (MJO) and the boreal summer intra-seasonal oscillation (BSISO) have great impacts on the global weather and climate events, which are the most important predictability source of sub-seasonal to seasonal (S2S) prediction. However, the monitoring of MJO/BSISO in China National Climate Center/Beijing Climate Center (NCC/BCC) entirely depends on external data, and the prediction skill of the introduced statistical forecast model is also much lower than dynamic mode, and the practical real-time operation ability has not been established. Therefore, based on CMA (China Meteorological Administration) global analysis data of T639 model, OLR (outgoing long-wave radiation) data of FY-3B satellite and the real-time forecast data of BCC atmospheric general circulation model system (BCC_AGCM2.2), applying the real-time multivariate MJO (RMM) index and BSISO index, BCC develops the MJO real-time monitoring and forecast technology, and establishes the trial ISV (intra-seasonal variability)/MJO prediction system (IMPRESS1.0).In comparison, monitoring results based on T639 wind analysis and FY-3B OLR data is generally consistent with the operational products from other centers, suggesting the capability of characterizing the oscillation and evolution of MJO/BSISO index accurately. Case study for the typical strong MJO event in March 2015 indicates that the amplitude peak of RMM index based on T639 and FY-3B OLR data is weaker than monitoring results based on NCEP and NOAA OLR data. Further analysis for three variables U850 (zonal wind at 850 hPa), U200 (zonal wind at 200 hPa) and OLR show that the convection monitored by FY-3B satellite is more consistent with NOAA's result, while the projection amplitude of the U850 based on T639 analysis against MJO mode is slightly weaker than NCEP/NCAR reanalysis data, which leads to weaker RMM index amplitude. The forecast skill verification shows the IMPRESS1.0 is able to provide correct evolution and intensity information of MJO at least 16 days in advance, and the skill of operational forecast in 2015 reach 18 days. The rolling prediction skill could be improved continually as the evolution of MJO event, and the predicted RMM index phase space track is closer to reality. Meanwhile, the verification of hindcasts by using correlation skill (COR), root mean square error (RMSE) and mean square skill score (MSSS) shows that the IMPRESS1.0 has useful prediction skill for about 12 days for MJO index and 8 days for BSISO1 and BSISO2 index, respectively. The case study for BSISO event in July 2015 also shows prediction skill, the reconstructed anomaly circulation and convection against BSISO index clearly demonstrate the dominant mode and northward propagation of BSISO. Therefore, the unified monitoring and forecast productions based on IMPRESS1.0 can provide important references for extended-range prediction, and offer certain help for operation and research.