Modelling reactive nitrogen concentrations and deposition on a local scale: comparison of eight models and their ensembles with three measurement campaigns

Every year, RIVM maps the amount of nitrogen in the air (concentration) and in the soil (deposition) in the Netherlands. It uses models and measurements for this. The government uses the same scientific models to estimate how much a particular source contributes to nitrogen deposition in nature reserves. Such model calculations are necessary before the source can be issued with a permit. This report discusses these calculations.

In some cases, models calculate a concentration or deposition for a specific area that is too high or too low. RIVM therefore investigated whether simultaneous use of multiple models (ensemble modelling) would result in more accurate outcomes and provide greater insight into uncertainties. To this end, RIVM compared the outcomes of models and ensembles of models with measurements. It did not study the possible drawbacks of using ensemble modelling to decide on the issue of permits.

For this study, eight scientific models were used. These models are used at the national or international level to calculate the concentration and deposition close to the source. Furthermore, measurements from three measurement campaigns were used: the first campaign monitored the concentration and deposition in the vicinity of two poultry houses, the second the concentration near a motorway and the third the concentration in the vicinity of industry. Most monitoring locations were near the source; the greatest distance was 570 metres.

The comparison shows that calculated concentrations in the air more closely matched the measured results than calculated depositions did. There are two reasons for this. First, calculating concentrations is simpler than calculating depositions. Second, the concentration measurements used were more accurate than the deposition measurements used.

Combinations of models (ensembles) matched measurements as closely as the best individual models did. Additionally, ensemble modelling provided more insight into the uncertainties of outcomes.