| As for weather forecast, the predictability of precipitation has always been a general concerned focus. Especially in East Asia, precipitation forecast is a very challenging thing. Recent years, there are some studies to research on the climate model predictability of synoptic processes in order to advance the model physical parameterization and data assimilation in Atmospheric General Circulation Model (AGCM). In addition, numerical weather prediction model or climate model is tended to be developing a unified approach to weather and climate prediction. In the paper, a hindcast for the heavy precipitation event occurring in the region of the Yangtze and Huaihe rivers in China during 24 June to 3 July 1998 was performed using the Beijing Climate Center Atmospheric General Circulation Model version 2.0.1 (BCC_AGCM2.0.1) of China Meteorological Administration. This paper discussed the coordinated initial value in climate model while without data assimilation, analysed the predictability of the global 500hPa geopotential heights from BCC_AGCM, examined the predictability of heavy rainfall in China, and discussed the effect of ensemble forecast for different initial time on weather predictability. The main conclusions are as follows:(1) The coordination of initial value in numerical model always plays an important role in the improvement of weather predictability. We used the hourly temperature, vorticity, and divergence data from NCEP reanalyses to spin-up integrate BCC_AGCM for a period time before the beginning of forecast. The results of hindcast experiment shown that the forecasted 500hPa geopotential heights over the whole globe after spin-up time more than 1 days are obviously better than that without any spin-up and when the spin-up time is up to 10 days the forecast performance at the first three days of forecast are the best. In addition, we found that the forecast of convective precipitation and large-scale precipitation in China could reach a stable level after spinning up time above 5 days. So, the spin-up time period needs more than 5 days to gain the good coordinated model initial data.(2) The results of daily hindcast experiment after 10 days spin-up from June to August in 1998 shown that the predictability timescale of the 500hPa geopotential heights over the whole globe is about 4~7 days, BCC_AGCM could capture the variations of trough and ridge at 500hPa in middle to higher latitudes over the Northern Hemisphere (NH) during the first 6 days forecast. If new weather systems generate abviously in the middle latitudes during the first 6 days, the predictability timescale will be shortened. BCC_AGCM could also simulate the development of trough and ridge well in tropical and subtropical regions over NH in the first 4 days. Prediction error of the 500hPa geopotential heights over the Sorthern Hemisphere (SH) is larger than that over the NH. Prediction errors in regional scales develops basically from the small error at the beginning forecast, and always happens in high and low pressure system centers in the middle to higher latitudes especially larger gradient area of isobar over the SH.(3) The predictability of heavy rainfall in the region of the Yangtze and Huaihe rivers during 24 June and 3 July was examined. The BCC_AGCM model was run from 0000 UTC 24 June after spinning up for 10 days. The results of hindcast shown that BCC_AGCM could forecast the position and development of rain belt in East of China (110°E~120°E) from 24 June to 3 July, but the intensity of precipitation is weaker than the observation. The spatial distribution of forecasted rainfall over China is nearly coincident to the observation during the first 3 days after initial forecast time. However, the rainfall intensity has some regional differences between the hindcast and the observation. After 5 days, the forecast skill falls obviously. The predictability of daily precipitation at different thresholds was tested using BIA, ETS, and HK scores. In the 2 days of forecast, the BCC_AGCM model has high predictability for the daily precipitations above 5mm and 10mm, and the ETS and HK scores are above 0.25 and above 0.4, respectively. The locations of forecasted precipitation are near coincident to the observations and the BIA score is close to 1.0. But for the daily precipitation larger than 30 mm, the model has less predictability and the BIA score is far from 1.0 and the ETS and HK scores are less. Overall, BCC_AGCM model has high predictability of daily precipitation in the first three days of forecast.(4) We found the results of ensemble forecast could improve the weather predictability. The ensemble is made for four hindcast experiments running from 0000 UTC 21 June, 22 June, 23 June, and 24 June, 1998 respectively. The results shown that the predictability of the ensemble of the global 500hPa geopotential height forecast during first five days, is not higher than that from the case only running from 0000 UTC 24 June. But after 5 days forecast later, the ensemble forecast is better than that without ensemble forecast. In addition, the ensemble forecast could improve the forecast skill of the location and its variations of rain belt in the eastern part of China (110°E~120°E) from 28 June to 3 July obviously. During this period, the BIA score for the daily precipitation above 5mm is close to 1.0 than that without ensemble forecast, and ETS and HK score above 5mm and 10mm increase obviously. |
