Le Cozannet, G.; Bulteau, T., Garcin, M., Garnier, C., Müller, H., Hoareau, A. and Mallet, C., 2016. Detecting errors in coastal databases using Bayesian Networks. In: Vila-Concejo, A.; Bruce, E.; Kennedy, D.M., and McCarroll, R.J. (eds.), Proceedings of the 14th International Coastal Symposium (Sydney, Australia). Journal of Coastal Research, Special Issue, No. 75, pp. 1162 - 1166. Coconut Creek (Florida), ISSN 0749-0208.

With growing concerns regarding coastal planning, management and adaptation, coastal observatories are collecting an increasing amount of geographical datasets such as historical shorelines, descriptions of hydrodynamic processes, coastal geomorphology or hydrogeology. Evaluating the integrity of such databases is difficult in practice due to the volume and heterogeneity of data, as well as the number of persons involved in data collection and management. Here, we adress the question of error detection in coastal databases. We test the utility of Bayesian Networks to guide users toward subsets of the database, where errors are more likely. The approach is applied to two coastal databases: (1) the French Basque Coast geographical database, managed by the Aquitaine Coastal Observatory; (2) the Eurosion database, describing European coasts and resulting from the synthesis and aggregation of multiple datasets. We show that the approach is more useful in the second case, but that the method can also help to derive a probabilistic information from a regional deterministic database. Finally, we highlight the utility of improving current coastal data models, especially for improving the description of coastal cliffs evolutions in large aggregated databases.