Abstract: The inconsistency of the forecast reflects evolution characteristics with time of the prediction error of continuous multiple prediction at a fixed time in the future. In order to explore the inconsistency of forecasting products in operational numerical forecasting applications, 12 h precipitation and 2 m ground temperature, which are predicted by three numerical models GQEC, T639 and GQJP from November 2015 to October 2016 are analyzed. Quantitative calculation method of Jumpiness index is adopted by considering the sensitivity of the index to the target region.The inconsistency of the numerical model in three different regions is studied by means of statistical analysis and typical case study. Results show that for statistical average, the inconsistency of the numerical models increases with the extension of the forecast time. Long time prediction inconsistency is greater. Jumpiness index of precipitation and temperature is related to the magnitude of the change, and Jumpiness index of precipitation is larger than that of temperature. It also shows that results of two consecutive temperature forecast are more consistent. The temperature prediction ability of the model is better than that of precipitation forecast. The comparison of different numerical models shows that GQEC has obvious advantages in many aspects. Although Jumpiness index of GQJP is less than T639, its jumping frequency is greater, indicating its prediction consistency is inferior to T639. There are seasonal differences in jumping frequency of model products. Both the jump frequency of precipitation and temperature is the highest in summer and the lowest in winter. The inconsistency test of two typical cases of rainstorm and cold air cooling process further corroborates statistical analysis results. Results also show that the forecast inconsistency of the numerical model is not only related to the geographical position, but also to the selected area size. The larger the region is, the smaller Jumpiness index becomes, and vice versa. In addition, the spatial distribution of Jumpiness index in the region is related to geographical location and topography. In general, where elements change bigger, Jumpiness index becomes greater there. The regional distribution of Jumpiness index of different meteorological elements is different. The index value of Jumpiness index of 12 h precipitation forecast increases gradually from north to south in region Ⅰ. While 2 m ground temperature prediction, Jumpiness index from north to south in region Ⅰ gradually decreases.