Extracting signals of commonness and rarity from noisy data: North Sea fish as a case study

The concept of species commonness is usually related to abundance and conservation status. Analysing and predicting commonness trends can result in discovering indicators for ecosystem status and in preventing sudden changes in biodiversity. However, there is large disagreement among biologists when stating if a species is common in a certain area. Several and mostly unknown variables should be taken into account. The result is that a formal separation between common and uncommon species is very hard to trace. A definition of species commonness by means of numeric parameters is missing and is likely to be very hard to find. Thisc paper addresses such problem and presents two approaches to automatically characterize species commonness. The former searches for an analytical function of biodiversity parameters, producing a degree of commonness for a species in a certain area. The latter uses a Clustering technique to automatically classify species commonness. Both the approaches take into account abundance as well as geographical and temporal aspects of species distributions. They rely on the OBIS database, and try to overcome the sampling biases from which large biodiversity collections are affected. Classification agreement was calculated on some North Sea species, with respect to two experts opinions. This benchmark revealed the hardness of the problem and the complementary benefits brought by our approaches.