Abstract:Based on the National Gridded Real-time Observation Datasets covering Jiangsu Province released by the National Meteorological Information Centre of CMA and observations of automatic stations, the consistency and accuracy of hourly 2 m temperature, 2 m relative humidity, 10 m wind and precipitation from July 2017 to June 2018 were evaluated in details by the error statistical analysis, skill score and other methods. The MODE method was applied to reveal the spatial deviation of precipitation between gridded products and observed rainfall records. The results indicate that the mean absolute error of 2 m temperature is between 0.5 and 0.8℃, the root mean square error is around 0.8℃ and the 2 m maximum temperature exhibits a better accuracy than the minimum temperature. The root mean square error range of 2 m relative humidity is 6%-7%, which means that the gridded 2 m temperature and 2 m relative humidity data are well consistent with observation. The accuracy of gridded 10 m wind direction is about 70% while that of wind speed is only 56%, showing a big difference from observation. The verification results of precipitation show that the gridded data with accuracy 90%-98% performs well in forecasting rain or no-rain. Nonetheless, it may still have a great impact on the precipitation frequency evaluation. TS score of light rain is higher than those in other classes, but it declines sharply when rainfall magnitude increases. Moderate rain or above has a relatively higher probability of detection, which means that the precipitation event is less detected than observed. Therefore, for the quantitative precipitation verification, it is not suitable to replace the observation data by the gridded data. Further study on 24 h accumulated rainfall bias between gridded data and observation indicates that the spatial structure of precipitation can be well described by gridded data. The spatial scores of precipitation designed in this article is above 0.9, which reflects the spatial distribution of actual precipitation. Generally, the gridded data in Jiangsu Plain Region can basically replace the automatic stations as the real-time meteorological field for forecast and model verification. However, there are still some problems as follows: (1) the 2 m temperature and 2 m relative humidity have large errors in the hilly areas of Jiangsu Province, and precipitation product from island stations has a lower accuracy; (2) the intensity of precipitation above heavy rain is weakened by the gridded data; (3) the wind speed value is lower than the observation, leaving a gap with the requirement of forecasting operation.