Air quality forecasting using a numerical weather prediction model: Potential sources of error
Potential sources of error when using an operational numerical weather prediction model to forecast air quality are also investigated. In the UM a semi-Lagrangian advection scheme is used with polynomial interpolation. This can result in unrealistic negative values of tracer which, when combined with a positive definite scheme in which negative values of tracer are set to zero, results in a failure of the model to conserve tracer mass. This problem occurs where sharp gradients of tracer concentration exist, such as are found close to source locations and leads to an overprediction of tracer concentrations.