The selection of model-generated CO2 data: a case study with seasonal biospheric sources

We assess whether data selection is required when comparing modelled concentrations with observed data that have been selected for ‘background’ conditions. To do this, the Melbourne University Tracer Model was forced with a set of seasonally varying biospheric CO2sources. The variation in predicted seasonal cycles of CO2 concentrations up to 500 km from CO2 monitoring sites was used as an indication of how regionally representative the model data is for each location. Where the variation is due to land influences this was reduced by removing air with recent land contact from the model output. Modelled radon-222 concentration from a simulation using the same wind fields as the CO2 case was found to provide an effective selection method. Selection generally reduces regional variability. The impact on the seasonal cycle, however, is relatively small and there are only a small number of sites where selection clearly improves the comparison with the observed seasonal cycle.