## The openair project » Examples of usage » Model Evaluation

Frequently, the evaluation of models is limited to a few numeric statistics e.g. mean bias, correlation coefficient. However, almost all the functions in openair that are used for analysing air pollution measurement data can also be applied directly to model output data. Applying these functions and comparing modelled-measured outputs can greatly improve the evaluation of models. Many of the functions can help reveal why a model performs as it does, rather than only providing a measure of agreement.

For example, bivariate polar plots can help capture the wind speed dependence of modelled values, can indicate missing sources, and when combined with the flexible 'type' option, reveal far more detail on model output characteristics.

There are also a growing number of specific model evaluation functions such as the conditionalQuantile function type 'help(conditionalQuantile)' into R to find out more.

load("~/openair/Data/CMAQozone.RData")

conditionalQuantile(CMAQ.KCL, obs = "o3", mod = "mod")

### Conditional Quantile