연구보고서
- 저자
- 유진호 박사
- 작성일
- 2016.02.29
- 조회
- 276
- 요약
- 목차
This study assessed the systematic error and predictability of KMA’s global seasonal forecast model (GloSea5). The spatial pattern of systematic error for 500hPa geopotential height forecast has a strong zonally symmetric structure indicating increased meridional temperature gradient, and thus increased zonal wind. The Eddy component of systematic error resembles the circumglobal teleconnection (CGT) pattern with wavenumber-5 structure, which seems to be associated with enhanced convection near the Maritime continent as well as increased extratropical zonal wind. The initial structure of systematic error remains and strengthens with increasing lead time, suggesting that the model’s mean biases were determined within a few days of integration.
In the subseasonal time scale, forecast skill is evaluated in terms of TCC (temporal correlation coefficient) and ACC (Anomaly pattern Correlation Coefficient) of the ensemble mean forecast. In general, forecast skill linearly decreases with lead time and it reaches marginal level after three weeks. However, there are large skill differences between the cases and it turns out that there are particular spatial patterns that bear higher forecast skill. These spatial patterns resemble the inverse of systematic error in Pacific North Atlantic sector, and they are unchanged with season during the initial two weeks of forecast. However, after three weeks, consistency drops between the forecasts and observations for relatively well predicted patterns. There are cases when the forecast skill is consistently high up to the fourth week, and these cases tend to occur sequentially. This indicates that the slow varying remote forcing has an impact on the East Asian weekly mean circulation forecast. Nevertherless, there is marginal skill in the weekly mean forecast for the east Asia region after three weeks. Therefore, it is necessary to consider this skill limitation during operation. In order to increase the skill and utility of KMA’s 1 month forecast, it is suggested to modify current operational system to include the first two weeks information and to revisit the contents of the forecast information for non-specified user regarding the current uncertainty of the forecast.