The number of marine seaweeds outside their natural boundaries has increased in the last decades generating impacts on biodiversity and economy. This makes the development of management tools necessary, where species distribution models (SDMs) play a crucial role. SDMs can help in the early detection of invasions and predict the extent of the potential spread. However, modelling non-native marine species distributions is still challenging in terms of model building, evaluation and selection. This product aims to predict the European distribution of four widespread introduced seaweed species selecting the best model building process.

The large databases of EMODNET Biology only store confirmed presences of taxon. However, when mapping taxon distribution, it is also important where the taxon did not occur: there is at least as much information in absences as in presences. Inferring absences from presence-only databases is difficult and always involves some guesswork. In this product we have used as much meta-information as possible to guide us in inferring absences. There is important meta-information at two different levels: the level of the data set, and the level of the taxon. Datasets can contain implicit information on absences when they have uniformly searched for the same taxon over a number of sample locations. Normally, if the taxon would have been present there, it would have been recorded. Other datasets, however, are not informative at all about absences. Typical examples are museum collections. The fact that a specimen is found at a particular place confirms that it lived there, but does not give information on any other taxon being present or absent in the same spot. A difficulty is that some datasets have searched for a restricted part of the total community, e.g. only sampled shellfish but no worms. In this case, absence of a shellfish taxon is relevant, but absence of a worm is not. The dataset can only be used to infer absence for the taxa it has targeted. Here we implicitly assume that a dataset inventorying the endomacrobenthos, is targeting all taxa belonging to this functional group. Usually, the distinction can be made on the basis of the metadata. It is also helpful to plot the total number of taxa versus the total number of samples. Incomplete datasets have far less taxa than expected for their size, compared to 'complete' datasets. At the taxon level, taxonomic registers such as WoRMS (WoRMS Editorial Board, 2021) give information on the functional group the taxon belongs to. This information is present for many taxa, but it is most likely incomplete. The size of the register excludes any easy test of completeness of the traits. However, even if incomplete, the register trait data can be used to select the most useful datasets. If one were to use an incomplete register directly to restrict the taxa to be used in mapping, that would cause loss of interesting information. Therefore the present workflow contains additional steps using the identified promising datasets rather than the taxon list based on the register’s traits.

This product builds on the EMODnet Biology data product Presence/absence data of macrozoobenthos in the European Seas to derive estimates of temporal turnover in benthic communities on a spatial grid across European seas. This product only uses species-level records, and only uses sampling events where the full macrobenthic community was surveyed (i.e. where there are no 'NA' values in the presence/absence dataset for any species). Six time periods are considered, based on data availability: before 1990, 1990-1999, 2000-2004, 2005-2009, 2010-2014, and 2015 and after. A 1 degree grid is used to obtain reasonable numbers of repeat samples per grid cell. The code below could be adapted to set different time periods and/or a different grid resolution. This readme describes the product structure, including the workflow to generate the required derived datasets and the process for turning them into gridded maps of community turnover.