Verification of Predictability between Global Seasonal Forecasting System Version 5 and Unified Model for 12-Day Forecast Period

Abstract:

Predictability for 12-day forecast period (fp) between high resolution seasonal forecast system, Global Seasonal Forecast System version 5 (GloSea5) and medium-range forecast system, Unified Model (UM) is verified by using analysis data of European Centre for Medium-range Weather Forecasts (ECMWF) from 1 January to 30 June, 2014.
The mean Root-Mean Square Error (RMSE) from initial time to 12-day fp of geopotential height at 500 hPa (Z500) and temperature at 850 hPa (T850) shows that GloSea5 has an increasing pattern of RMSE significantly from initial time to 3-day fp due to model stability process with spin-up time, after that forecast period the increasing rate of RMSE is decreased by 12-day fp. Whereas, the RMSE of UM shows slowly increasing pattern during the whole forecast period so that the difference of RMSE between two models is decreased gradually from 4-day fp. The spatial and meridional patterns of zonally averaged RMSE in Z500 and T850 show that it is increased rapidly and largely in middle latitude from 1-day fp in GloSea5. And the RMSE differences of two models begin to decrease from 4-day fp, appear the minimum in most regions at 12-day fp. Areas with larger RMSE of UM are also existed regionally as forecast period is increased.
In climatic aspect, the predictability of Indian monsoon index which is calculated by difference of wind speed at low and middle latitude is verified. The result of RMSE analysis appears that its pattern from initial time to 12-day fp is similar to that result of Z500 and T850. Relative variation of daily forecast value shows that UM is smaller than GloSea5 during 6 months. The correlation of Indian monsoon index to be predicted by GloSea5 with analysis values of ECMWF is less than 0.5 after 3-day fp but one by UM is more than 0.5 from 1 to 12-day fp in average. In Tropics, for the prediction of the El Nino-Southern Oscillation (ENSO), RMSE and Anomaly Correlation (AC) of forecasted SST in NINO3.4 region are investigated. Despite of only use of SST in UM background field, its SST value is more similar with analysis value of ECMWF than one of GloSea5 during short range forecast period (1~5 day). After 6-day forecast period, RMSE of GloSea5 is smaller than one of UM. That is also verified by mean distribution of AC with larger value of one in GloSea5 from 5 to 12-day fp.