Prediction

 CDL prediction system

 

El Nino

CDL Seasonal Prediction System


    During the last few decades, many studies have been devoted to predicting climate variability, in accordance with increased scientific and economic interest in seasonal climate prediction and predictability. To keep pace with these demands, many climate centers have developed and tried to improve seasonal forecast systems to enhance society prediction ability to cope with climate variability and therefore reduce its vulnerability by providing useful information. In order to obtain useful seasonal predictions, two SNU seasonal prediction systems are setup and operationally updated.


    1. SNU coupled GCM

    Another seasonal prediction system is based on SNU Coupled GCM (Kim et al. 2009; Ham et al. 2009).To obtain the initial conditions, the SNU coupled GCM is integrated from January 1980 by nudging the observed variables of both ocean and atmosphere. For oceans, the ocean temperature and salinity obtained from GODAS reanalysis are nudged from surface to 500m with a 5-day restoring time scale. For the atmosphere, zonal and meridional wind, temperature, and moisture fields obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) 40-Year Reanalysis (ERA40; Uppala et al., 2005) are nudged for all vertical levels with a 6-hour restoring time scale. Given the initial conditions, hindcast forecasts are carried out with a 7-month lead time. Six members, generated by one-day lag using Lagged Averaged Forecast method (LAF), are used for ensemble forecasts. Figure 1 shows the correlation skill of 20-year hindcast experiements for summer season.

     

    Figure 1. Correlation skill for JJA SST. Seasonal prediction is started from May 1st.


    2. Modified Zebiack-Cane model

    Another is based on modified Zebiak-Cane model (CZ model) for El Nino prediction. It has made using the KMA/SNU ENSO prediction system (Kang and Kug, 2000). The system is based on the intermediate ocean and statistical atmosphere model. The ocean model differs from the Cane and Zebiak (1987) model in the parameterization of subsurface temperature and the basic state. The statistical atmosphere model is developed based on the singular value decomposition (SVD) of wind stress and SST. In order to improve the western Pacific SST prediction, we introduced heat flux formula and vertical mixing parameterization to the ocean model. The initialization of the model is done by combining observed SST and wind stress. Wind stress is calculated by using 925hPa wind of NCEP/NCAR reanalysis data. Using calculated wind stress for initialization has a better forecast skill than the case of FSU wind stress in recent prediction. (Kug et al., 2001). In addition, the present prediction is attended with random noise to consider weather noise and to generate many sets of prediction. Our approach for random noise is similar to Kirtmanand Schopf (1998). Figure 2 shows seasonal SST forecast in tropical Pacific basin for summer 2009. The forecasts indicate that new El Nino evolves from Summer 2009.