For the uncertainty of heavy rain belt forecast of the ECMWF model in the Yangtze-Huaihe River Basin in China during the extremely long Meiyu period in 2020, 10 typical long and narrow heavy rain belts are selected to analyze the location forecast bias, forecast stability and continuity of forecast bias using the method for object-based diagnostic evaluation (MODE). Ensemble sensitivity analysis (ESA) is also used to analyze the key synoptic systems which caused the typical forecast bias of the heavy rain belts. The results show that forecast uncertainty of the eastern part of rain belts is generally higher than that of the western part, especially during the medium range in which forecast uncertainty further strengthens. During the short range, both the western and eastern parts of rain belts have obvious northward systematic forecast bias. The forecast of the western part of rain belts has relatively lower missing rate, better stabil-ity and continuity than the eastern part. During the medium range, the forecast error of the eastern part of rain belts grows obviously and the average latitude variation of the eastern part is also higher than that of the western part. The results of ESA indicate that the northward forecast bias of the eastern part of rain belts is caused by the joint effect of the 500 hPa high-level trough, the subtropical high, the 850 hPa low-level jet, the shear line and the unrealistic positive feedback between them and precipitation latent heat releasing, of which the development and strengthen of high-level trough and low-level jet is one of the background characteristics of the northward forecast bias of the eastern part of rain belts.