Error Calculation Methods

Different HPC programs use different methods to measure the accuracy of results. The following describes general error calculation methods, which are also used by WRF.

In the following formulas, x and y indicate two datasets to be compared.

Mean error (ME): A smaller value indicates a smaller error.
$\text{[math]}$
Mean absolute error (MAE): A smaller value indicates a smaller error. 0 indicates that the two datasets are consistent.
$\text{[math]}$
Root mean square error (RMSE): indicates the relative deviation between two datasets. A smaller value indicates a closer similarity. 0 indicates that the two datasets are consistent.
$\text{[math]}$
Space correlation coefficient (SCC, also called Pearson correlation coefficient): It is used to measure the linear correlation between two datasets, that is, the linear correlation between equal-interval variables. If the correlation coefficient is 0, there is no linear correlation between x and y. The absolute value of the correlation coefficient is positively correlated with the correlation. If the correlation coefficient is close to 1 or -1, the correlation is strong; if the correlation coefficient is close to 0, the correlation is weak.
$\text{[math]}$

$\text{[math]}$

WRF Application Assessment Tool

diffwrf: It is a tool provided by WRF to compare the differences between two WRF datasets.
Use diffwrf to compare two WRF running result files.

Figure 1 diffwrf running result
Next Time: weather forecast result at a specified time.
Field: meteorological elements of WRF. For details about the meaning of each variable, see official WRF documents.

Based on the quantitative evaluation script of the NetCDF library, diffwrf cannot display the SCC indicator. You can implement an evaluation script based on the mathematical definition of SCC. Use the quantitative evaluation script to compare two WRF running result files.

Figure 2 Result of running the quantitative evaluation script