Evaluation on the CFSv2 forecasts of three cold waves during 2015–2016 in China.

In this paper, by using forecasts of the highest and lowest daily surface air temperatures from the Coupled Forecast System version 2 (CFSv2) over a period of 9 months, NCEP/DOE reanalysis data from the National Centers for Environmental Prediction (NCEP), and data from the U.S. NOAA Climate Prediction Center (CPC), three cold waves are identified during 2015–2016 in China, and the CFSv2 forecasts of these cold waves are evaluated. The results show that the high-resolution CPC data and the low-resolution NCEP/DOE data both very clearly show the temperature drops and temperature anomalies of these cold waves. The earlier the forecast lead time is, the greater the errors in the beginning and ending dates of the cold waves between the CFSv2 forecasts and the observations; there are errors of 1–2 days in the durations of the cold waves between the CFSv2 forecasts and the observations. CFSv2 exhibits certain abilities to forecast the overall temperature drops and temperature anomalies of cold waves at forecast lead times of 0, 5, 10, and 15 days, but the CFSv2 prediction ability is poor at forecast lead times exceeding 20 days. CFSv2 can predict the spatial distribution of temperature drops better than it can predict that of temperature anomalies, but the absolute errors in the observed temperature drops are greater than those in the observed temperature anomalies. The temperature drops from the CFSv2 forecasts are obviously underestimated in North China, South China, and their nearby areas but overestimated in Mongolia and Northeast China. The temperature anomalies from the CFSv2 forecasts are obviously underestimated in the inland areas of China but overestimated in the surrounding oceans and adjacent coastal areas. The difference fields at forecast lead times of 0 and 30 days are basically the same, but the absolute values of the differences increase with increases in the forecast lead time (with a few exceptions). [ABSTRACT FROM AUTHOR]