The statistical modelling of joint species abundance is currently receiving increasing attention in the ecological literature because it complements earlier descriptive accounts of the effect of disturbance on community change with quantitative measures of spatial covariation among species at different spatial scales. Using a hierarchical Bayesian approach, we fitted both temporal and spatial plant cover data to a joint distribution model of plant abundance of the dominant plants in wet heathlands (Erica tetralix, Calluna vulgaris and Molinia caerulea) undergoing nitrogen deposition over several decades. The spatial analysis was based on pin-point cover data from 114 Danish sites with a total of 1509 randomly placed plots along a nitrogen deposition gradient. The temporal analysis was based on pin-point cover data from 22 sites with a total of 986 observations from plots that had been resampled at least three times since 2007. The spatial variation was partially explained by nitrogen deposition, as the cover of Erica decreased significantly with nitrogen deposition, whereas the cover of both Calluna and Molinia increased (albeit non-significantly) with nitrogen deposition. There was a strong and significant positive spatial covariation between Erica and Calluna. Oppositely, there was a strong and significant negative spatial covariation between Calluna and Molinia. The spatial covariation between Erica and Molinia was not significantly different from zero. There were no significant changes in the cover of any of the three species since 2007. However, by including data from 2004, the previously reported result of a significant decreasing cover of Erica was confirmed, and the cover of Calluna increased significantly. We conclude that the statistical modelling of joint species abundance is a potentially powerful tool to understand the effect of the alteration of nitrogen dynamics on community composition. This is essential to design management practices aligned with the predicted effects of varying levels of nitrogen deposition on community structure. The study also demonstrated the insight gained on the community dynamics by the use of hierarchical Bayesian models.