Data were gathered from two main sources: literature and Natural History Collections (NHCs). To collect literature data, a comprehensive search was performed on the public databases Scopus and Web of Science. In addition to this, we also searched data from journals specialised on Mediterranean marine fauna, namely Iberus and all the volumes of both journals of the Italian Society of Malacology (Società Italiana di Malacologia, SIM): Bollettino Malacologico and Alleryana. Since until the publication of the Checklist of the Italian Fauna (Bedulli et al. 1995a, Bedulli et al. 1995b, Bedulli et al. 1995c, Bello 1995, Bodon et al. 1995, Manganelli et al. 1995), no unified standard existed for Italian molluscan taxonomy and nomenclature - and verifying the accuracy of identifications reported in literature would have been difficult without direct check of the actual specimens - the literature search was restricted to publications issued after the first edition of the Checklist of the Italian Fauna. Species distribution information published in various formats (e.g. data tables in supplementary materials or within the paper, species lists, statements reporting the species occurrence) were considered as potential raw data. Papers with data already published in public databases were excluded. In order to avoid collecting the same record several times, only papers with new data were considered (i.e. new data derived from a dedicated sampling or non-new data published for the first time). In Table 1, search information are summarised. The full list of papers from which data were gathered is provided in Suppl. material 1. Data from NHCs were collected by direct request to private collectors (Giannini and Oliverio 2022) and institutions. The number of raw records initially received from each NHC owner and the final number of standardised records are listed in Suppl. material 2, together with the code used to identify the institution - derived from the Global Registry of Scientific Collections (Liu et al. 2024) - or private collector in the dataset. From both sources, records were included in the dataset if at least the occurrence locality and a taxonomic identification at the genus level (or more specific) were stated.

To remove human-readable leading, trailing, double spaces and non-printable characters, the entire dataset was run through the Excel TRIM, CLEAN and SUBSTITUTE functions. Carriage returns were checked and removed using Notepad++ software (Ho 2011). To define the consistency and quality of the collected occurrences, a series of cleaning filters and manual checks were applied, which are described in detail in the next section (i.e. Step Description). To establish the quality of the taxonomic data, the resulting list of species in the dataset was extracted and compared with the Checklist of Italian Fauna (Renda et al. 2022). The species present in the dataset, but not included in the latter, were checked by expert taxonomists (under the supervision of Marco Oliverio and Luigi Romani), who solved problems and helped remove dubious records. The records underwent two geographic quality checks: the first one pre-georeferencing (Step description section, second step) - performed using the point-radius method (Wieczorek et al. 2010) - in which all records with too vague textual location were removed; and the second one post-georeferencing, in which all records with an uncertainty radius > 5000 m were removed (Step description section, seventh step). The double spatial filtering was necessary as high levels of uncertainty in georeferencing can create problems for spatial analyses (Reside et al. 2011). On the other hand, removing data with high spatial uncertainty can excessively reduce the sample size and, thus, the accuracy of the analyses (Smith et al. 2023) for which the data can be used. The uncertainty limit of 5000 m was chosen as a compromise between the completeness of the dataset and the quality and usability of the data. Some records from private NHCs refer to shells collected by fishermen and it was decided to keep them in the dataset, only if the fishing area/location was recorded and the geographical reliability of such records was acceptable, specifying that the collector is a fisherman in the dwc:recordedBy column to ensure transparency. In these cases, raw records are frequently accompanied by two locations: the port of arrival of the fishing vessel or the place where the collector acquired the specimen and the recorded collection area. If only the location of acquisition (and not the actual collection area) was explicitly indicated, the record was excluded from the dataset. This approach was consistent with the treatment applied to specimens obtained through purchase or as gifts, if only the acquisition location was provided. Like all other records, locations that were too vague were excluded during the double geographical quality check.

1. Firstly, data were merged and formatted in a Darwin Core scheme (Darwin Core Task Group 2009), using the Biodiversity Data Cleaning toolkit package (Ribeiro et al. 2022) in R (R Core Team 2022, ver. 4.2.2).

2. With the same package, a first filter was performed to clean the dataset from duplicates and records lacking essential information (i.e. identification or locality/coordinates). Then, data were manually filtered to retrieve records that were: out of scope (i.e. occurrences outside the Italian Marine Exclusive Economic Zone, fossils, non-marine species), too vague (i.e. broad locality, specimens with a higher level of identification than the genus) or dubious (dubious locality, ambiguous and/or unclear identification). All these cleaning steps have been consolidated into the "Invalid Records Filter" block in Fig. 2.

Figure 2.  

Flow chart of the standardisation pipeline, with the number of input records and the number of discarded ones at each step. With the term "invalid", we identify all records out of scope or with problems (i.e. duplicates, records lacking essential information, fossils, non-marine species, specimens with a higher level of identification than the genus, ambiguous and/or unclear identification, records with dubious locality, occurrences outside study area or with broad locality).

3. Taxonomy was aligned to the one proposed by the World Register of Marine Species (WoRMS Editorial Board 2024) using the taxon-match Life Watch webservice (Fuentes and Fiore 2014), also extracting the WoRMS Life Science Identifiers (Vandepitte et al. 2015) for each valid scientific name to trace as far as possible rehashes of taxonomy, which in marine molluscs are quite common, especially through molecular evidence (e.g. Chiappa et al. (2022), Furfaro et al. (2022), Nocella et al. (2024)).

4. The remaining dubious taxonomy that was not automatically validated was checked manually and then submitted to experts, which resulted in the removal of other records with dubious identification.

5. Open Nomenclature (ON) qualifiers (Richter 1948, Matthews 1973, Bengtson 1988) were used to set uncertainty and provisional statuses for taxonomic identifications, applied following Sigovini et al. (2016) guidelines for ON terms harmonisation.

6. Subsequently, records were classified in seven different groups based on the type of the geographic information they had, in order to georeference them by the most appropriate method. Georeferencing was performed following the point-radius method (Wieczorek et al. 2010) and the Zermoglio et al. (2020) protocol, using GEOLocate web-based collaborative client (Rios and Bart 2010) and QGIS (QGIS.org 2024, ver. 3.22.13). Each final processed record has associated coordinates expressed in WGS84 decimal degrees and an uncertainty measure in metres. Table 2 details the method and, where applicable, protocol of georeferencing applied to each of the seven original geographic information classes. The GEBCO_2022 global terrain model was used to georeference depth data correctly (GEBCO Bathymetric Compilation Group 2022).

7. During the georeferencing process it was possible to remove other data occurring outside study boundaries. We then excluded records with >5000 m of uncertainty radius.

8. As raw temporal data from NHCs arrived in various formats, this information was handled with the R package lubridate (Grolemund and Wickham 2011) and converted to ISO 8601 format. The temporal information collected spans various degrees of resolution, from the exact date to time ranges between years. We decided to mantain also records without temporal information. No hourly data were collected. Standardision and cleaning pipeline main steps are reported in Fig. 2.

Collected data occurred within the Italian Exclusive Economic Zone (EEZ), that consists of a marine area of 538,216 km² (Khalfallah et al. 2020) situated in the Mediterranean Sea. This area is divided into nine biogeographical sectors based on the distribution of coastal taxa and the presence of physical, hydrological and physiological barriers (Bianchi 2004, Giacobbe and Oliverio 2024): Ligurian Sea, Northern Tyrrhenian, Southern Tyrrhenian and Messina Strait on the west coast of Italy, Northern Adriatic, Southern Adriatic and Mid-Adriatic on the east coast and Southern Mediterranean and Ionian Sea in the southern part. This subdivision has already been used to study the distribution of both marine plant and animal species in the marine flora and fauna checklist of the Società Italiana di Biologia Marina (Relini 2008, Relini 2010) and for the latest version of the Italian checklist of the marine malacofauna (Renda et al. 2022) and has been proposed as a standard in the Marine Regions Gazetteer (Flanders Marine Institute 2024). Sectors with the highest number of recorded occurrences are the Tyrrhenian Sea (with 10480 occurrences for the Southern area and 10076 for the Northern one), followed by the Ionian Sea (7284), Ligurian Sea (5384), Northern Adriatic (2652), Mid-Adriatic (2647), Messina Strait (2212), Southern Mediterranean (1776) and Southern Adriatic (1585). Since sector areas are very diverse, from the smallest (the Messina Strait) to the largest (the Northern Tyrrhenian Sea) data coverage for each sector may vary depending on its size. Despite this general trend, the Strait of Messina, which covers only 0.2% of the Italian sea surface, contains about 5.0% of the dataset's occurrences, proving to be a particularly well-studied area (Fig. 3). The distribution of collected, cleaned and georeferenced occurrences is shown in Fig. 4.

Figure 3.  

Data abundance (%) per biogeographical sector compared with its extent (%) in the Italian Marine Exclusive Economic Zone.

Figure 4.  

Distribution of collected occurrences.

35.06440614922465 and 45.80891370810167 Latitude; 5.889722222129848 and 18.99523827942329 Longitude.

The dataset includes 44096 occurrences of 1513 Italian marine mollusc species, covering 85% of the Italian malacofauna and six out of the eight classes reported in Italy (Renda et al. 2022): Gastropoda (1110 species, 32832 occurrences), Bivalvia (326 species, 9410 occurrences), Polyplacophora (28 species, 820 occurrences), Cephalopoda (35 species, 682 occurrences), Scaphopoda (12 species, 331 occurrences) and Monoplacophora (2 species, 21 occurrences) (Table 3). Of the specimens reported in the dataset, 97% are identified at least at species level, while the remaining part is identified at genus level. The complete list of species can be viewed in Suppl. material 3, while Suppl. material 4 contains annotations on the inclusion of some ambiguous taxa (i.e. taxonomic entities that are difficult to interpret/not completely resolved, possible fossil specimens and occasional alien species for the Italian fauna) in the dataset.

Creative Commons Attribution Non-Commercial (CC-BY-NC 4.0)