Abstract:Global operational numerical models have the problem of occasional occurrences of extreme large medium-range forecast errors, and tracing these error sources could provide important references for improving the model itself and data assimilation system. In this study, errors of the operational forecasts in medium range (6 days) from the China Meteorological Administration high-resolution global assimilation and forecast system (CMA-GFS) and global ensemble prediction system (CMA-GEPS) with a lower resolution are analyzed during the period spanning from January to February 2020. The error origin for a particular case (initialized on 12 UTC 8 February 2020) with extreme large medium-range forecast errors over East Asia is investigated by employing the method of ensemble sensitivity analysis. From the spatial-temporal evolution characteristics of forecast errors in CMA-GFS and results from ensemble sensitivity analysis based on CMA-GEPS, a preliminary deduction was acquired about the key forecast error source region, i.e., an up-stream region of East Asia located in the Atlantic Ocean and western Europe (20°-90°N, 90°W-60°E). When the initial conditions of the control forecast of CMA-GEPS are replaced with that of the best ensemble member but confined to the above-mentioned key error source region, the medium-range forecast error of 500 hPa potential height was reduced greatly over East Asia, less than 50% of the original control forecast error. This further confirms the effectiveness of the identified key error source region.