Dynamical Downscaling of Seasonal Climate Prediction Over East Asia for March-September 2003 at CWB Chih-Hui Shiao, Ying-Jui Chuang Central Weather Bureau, Taipei October 2003
The plan for the CWB dynamical downscaling forecast systems System I : NCEP-RSM (60-km) + IRI-ECHAM4.5 (T42L18) System II : CWB-RSM (60-km) + CWB-GCM (T42L18) 2 system X 5 members Model integration : 7 months Domain : 100。E-135。E, 10。N-35。N
The current status System I : NCEP-RSM (60-km) + IRI-ECHAM4.5 (T42L18) Two member (ASST and PSST) System II : In past studies, CWB-RSM was applied to simulate the 1998 and 2002 summer monsoon over East Asia has successfully captured not only the regional circulation but also the precipitation patterns. CWB-RSM (60-km) + CWB-GCM (T42L18) CWB-RSM (120-km) + ECHAM4.5 (T42L18) Forecast period : 4 months Domain : 100。E-135。E, 10。N-35。N
To examine the performance of NCEP RSM and CWB RSM in the dynamical downscaling precipitation forecast, based on the forecast results on the past half year. Experimental period : March-September 2003 Integration length : 4 month-long forecast for each month Domain : East Asia Base field : ECHAM4.5 (T42L18) support by International Research Institute for Climate Prediction (IRI) 20 members of 7 month-long forecast for each month a set of AMIP simulation
The evolution of precipitation on March-September 2003 by CAMS_OPI
CAMS_OPI precipitation on 2003 (average over 115。E-125。E, 15。N-30。N)
The precipitation anomaly The anomaly of analysis is the CAMS_OPI analysis in 2003 compare with the average of 1979-1995. The forecast precipitation anomaly of NCEP RSM is the forecast in 2003 compare with the simulation of NCEP RSM forced by the ECHAM4.5 AMIP simulation averaged in 1971-1986.
Precipitation Anomaly of M2+M3+M4 with ECHAM4.5 / NCEP RSM NCEP RSM ECHAM Regional downscaling forecasts is under control of GCM strictly except those mesoscale phenomena result from high-resolution topographic and coastal effects.
CAMS_OPI NCEP RSM The major difference of precipitation anomaly between forecast and analysis are the forecast is too wet along the coastline and too dry in midlatitude
One possible source of the precipitation anomaly bias may due to the difference between the AMIP simulation and the analysis. The difference of ECHAM4.5 AMIP run to GPCP in 1979-1986 mean AMJ MJJ JJA JAS
AMJ MJJ JJA JAS CAMS_OPI NCEP RSM AMIP run bias
Summary I Anomaly forecast is possible if the model climatology is ‘realistic’.
Bias Correction The precipitation difference between the GPCP analysis and ECHAM4.5/NCEP RSM downscaling AMIP simulation are used to modify the ECHAM4.5/NCEP RSM currently forecasts.
Forecast precipitation by NCEP RSM without and with bias correction (average over 115。E-125。E, 15。N-30。N) with bias correction without bias correction
(M3 and M4) without bias correction with bias correction
Downscaling forecasts by NCEP RSM with bias correction (average over M2+M3+M4 ) AMJ MJJ JJA JAS
Downscaling forecasts with CWB RSM (average over M2+M3) AS AM MJ JJ JA
R.M.S.E. of precipitation with different forecast start month (average over M1,M2,M3,M4)
Summary II Bias correction is necessary for single model forecast. How to eliminate the large-scale bias but keep local phenomena is still known.