Accuracy Quantitative Assessment Methods

The data files of weather forecast applications are usually in binary formats such as NetCDF and GRIB. These data files contain multiple meteorological elements (such as temperature, wind, and precipitation). The consistency of these meteorological elements can be assessed by using the two methods:

• Quantitative assessment method: Based on meteorological elements in a specific area, this method calculates four metrics, including root mean square error (RMSE), mean error (ME), mean absolute error (MAE), and Pearson and space correlation coefficients (Pearson/SCC).
  • RMSE: indicates the deviation between two sets of data. A smaller value indicates a closer proximity between two sets of data, that is, a better forecast result. The value 0 indicates that a result is completely correct.

    

  • ME: indicates the average error of two sets of data. Generally, a value closer to 0 indicates a smaller error.

    

  • MAE: A smaller value indicates a smaller error. 0 indicates that two sets of data are consistent.

    

  • Pearson/SCC: When the value is -1, it indicates complete negative correlation between two sets of data. When the value is +1, it indicates complete positive correlation between two sets of data. When the value is 0, there is no correlation.

    

• Graphical assessment method: This method uses graphical software such as NCAR Command Language (NCL) to draw the value distribution of meteorological elements in an area and allows observation of the result differences between different platforms.