The 1908–1910 "Thor" expedition focused on the Mediterranean Sea and the adjacent waters of the Atlantic Ocean, the Sea of Marmara and the Black Sea; additional stations from the Atlantic are included in the digitized datasets since they were reported in the individual biological publications.

During the two core expeditions of 1908–09 and 1910, 1,102 samples from 261 stations were collected. However, as the publications aimed at providing a more comprehensive picture of hydrographic and biological patterns across different geographic areas, they also included selected data from previous "Thor" expeditions in the North Sea and the North-Eastern Atlantic (1905–1906) and from other commercial and research vessels in the Mediterranean (1911–1912). These publications are also included in Schmidt (1912) and thus the introductory volume includes 1,566 samples from 443 stations.

The individual biological publications did not strictly analyse only the data from these stations but combine samples from various oceanographic expeditions from 1904 to 1930, including those of the research vessel “Dana” from its global campaign in the years 1921 and 1930 (Fig. 2). Thus, the combined datasets digitized and re-published herein finally comprise information from a total of 567 stations and 2,043 samples, which partially overlap (Fig. 3).

Overview of the four digitized datasets

Introductory table (sampling and environmental information)

The introduction to the Danish Oceanographical Expedition by Schmidt (1912) includes a series of tables containing information on all samples collected during the core expedition of the "Thor" to the Mediterranean in 1908-1909 and 1910 (pages 25–41), additional tables listing samples on the related expedition of the "Thor" in the Atlantic Ocean in 1905 and 1906 (pages 41–47) and supplementary samples taken in 1911 and 1912 by other vessels in several locations in the Mediterranean Sea (pages 47–49). Additionally, the introduction contains information on the characteristics of the research vessel, details on the sampling gears and a description of sampling methods. The main table describes all samples and lists relevant metadata such as station number and position, depths, sampling date and time, bottom characteristics, gear used (including the length of wire for nets and the duration of the haul) and environmental parameters (see also below, section "Sampling methods").

Pelagic polychaetes (Wesenberg-Lund 1939)

This publication describes the pelagic polychaetes (excluding Tomopteridae), collected from various cruises of the "Thor" and additional vessels (1905–1911), as well as samples from the "Dana" expeditions (1921 and 1930), including a few samples from the tropical Eastern Atlantic. The paper focuses on taxonomic treatments, including descriptions and illustrations for each species, but provides full details on the samples each species was found in.

Fish (Clupeiformes) (Fage 1920)

This publication includes data from various "Thor" expeditions from 1905–1910, but also includes information from other vessels, encompassing in total a period from 1904–1912 and a geographic area including the Atlantic, the Mediterranean and the Suez Canal. The paper focuses on the life cycle and development of commercially important species of the familes Engraulidae and Clupeidae (e.g. sardines, sprats, anchovies) and analyses only larval stages. Abundances per larval size class are reported for each station, aiming at assessing seasonal fluctuations and productivity of different regions in the study area. Samples, their metadata, and abundances per size class are reported in tabular format for each species.

In addition to the data collected by the research vessels, Fage (1920) also presents a time-series of twelve years of fish captures (“sardines”) by the fishermen of Colliure (southern France), showing defined seasonality and allowing for a comparison, within the same report, between basic data related to larval growth and data related to commercial fishing. The digitization of these "fishermen data" was considered also of historical value and is included here as Suppl. material 5.

Calcareous algae (Lemoine 1915)

This dataset includes taxonomic descriptions, drawings, global distribution maps and other remarks for the calcareous algae collected in the framework of the expedition. The paper focuses on taxonomic treatments, including descriptions and illustrations of species and listing the date and location it was found, but omitting additional information on coordinates, sampling dates, depth or gear. These metadata could in some cases be inferred from the introductory table, but in most cases this proved impossible due to lack of details given.

Table 2 provides an overview of the characteristics and coverage of the four individual datasets; Figs 4, 5, 6, 7 show the sampling stations of each of the datasets. The datasets can also be viewed through the MedOBIS data viewer (account required) and downloaded in various formats.

This work was supported by the LifeWatchGreece Research Infrastructure (MIS 384676), funded by the Greek Government under the General Secretariat of Research and Technology (GSRT), ESFRI Projects, National Strategic Reference Framework (NSRF). The original "Thor" expedition was funded by the Carlsberg Foundation (Copenhagen).

The research steamer “Thor” was 35 m long and 6.5 m wide and was equipped with laboratory facilities for oceanographic surveys. The main types of sampling gear used during the “Thor” expedition and specifically for the four examined datasets are described below and listed in Table 3. In addition, film material recorded during the Danish Oceanographic Expedition Around The World (1928-1930) on the research vessel “Dana” documents the equipment of the ship and the use of the gear, much of which was used in a similar fashion during the "Thor" expeditions. Schmidt (1912) provides an extensive description of different types of sampling gear and their use on board.

For horizontal bottom and pelagic trawling, two models of young fish trawls ("Petersen's young fish trawl") were used, one 7 m long net with an opening of 200 cm diameter (Y 200), the other 8.5 m long with an opening of 330 cm (Y 330), both with a mesh size of 1 mm. Less frequently, a Monaco trawl (M) with a length of 6 m and a mesh size of 3 cm was used.

To collect fish eggs, larvae and other plankton, either open conical stramin nets (three different opening diameters; 100 cm, 150 cm and 200 cm – S 100, S 150 and S 200, respectively) with a mesh size of 1 mm, or fine-meshed open silk nets (P 100 and P 30) with mesh sizes of 333 µm and 76 µm were used. The silk nets were either used for surface sampling or they were attached to a trawl-wire used in greater depths.

Vertical plankton hauls were performed with Nansen closing nets (opening diameters of 30 cm (N 30) and 50 cm (N 50)) and a mesh size of 76 μm.

Four different types of ring trawls were used, with diameters from 130 cm to 1000 cm (C 200, C 130, E 300, E 1000 – the latter two only during the Dana expedition) and varying mesh sizes. The ring trawls were used for horizontal trawling, but could be closed at a specific depth before hauling them to the surface, thus allowing sampling at selective depths.

Bottom dredges comprised one with a rectangular opening (D 1), one with a triangular opening (D 2), and a small hand-dredge (H) with dimensions of 18 x 14 cm.

Additional types of sampling gear used occasionally during the "Thor" expedition included a small type of otter-trawl, a Danish eel hand-seine dragged by hand (mesh size of 1–2.5 cm) and shrimp nets used during wading from the beach for capturing various small fish, among others (Table 3).

The most commonly used gears during the expeditions were the young fish trawls, stramin nets, silk nets and Nansen nets (Table 4).

Several samples are not included in Table 4 for the following reasons:

• Wesenberg-Lund (1939) lists information on the sampling gear for each sample of pelagic polychaetes; however, in some cases it seems that samples from different gears were merged, as two different gears are listed per sample. These cannot be allocated unambigously.
• Fage (1920) lists sampling gears used to catch fish (Clupeiformes) in great detail when related to the "Thor" expeditions 1908–1910; however, for several samples proceeding from complementary expeditions the only information given is "Petersen's young fish trawl", without further specification. By matching samples to the information in the introductory volume, information on the opening diameter could be assigned to most samples. The remaining samples are listed in Table 4 as "Y".
• No information on the sampling gear used for the collection of calcareous algae is given by Lemoine (1915); however, by matching the samples to the information given in the introductory table (Schmidt 1912), information could be inferred for several of the samples. However, samples could not always be unambiguously matched, thus explaining the low numbers in Table 4.

Environmental parameters

During the "Thor" expedition, environmental conditions occurring at the time of the sampling were measured in great detail. Measurements are reported in parts of Volume 1 and Volume 3 of the series (Table 1). These were not digitized here, and the only environmental parameters contained in the dataset are those reported in the introductory table by Schmidt (1912). Additional information on the methodology was obtained from Nielsen (1912), same volume (Table 5).

Individual datasets were digitized manually from scanned copies of the original documents. In general, the quality control procedures applied through the data management process, were based on the Principles of Data Quality according to Chapman (2005)​ and the OBIS node manual. Moreover, as the digitization of historical datasets scattered over different manuscripts in a heterogeneous format presents a special challenge, more details on the digitization steps, the structure of the data and the quality control measures are given below (section "Step description").

Digitization steps

1. Manual digitization of the information on samples and their associated metadata from the “Table of Stations” (page 25 and onwards) in the introductory volume (Schmidt 1912). These data were enriched with information from the main text (e.g. sampling methods and gear) when necessary.
2. Independent manual digitization of the three biological publications, including all available data on samples, sampling metadata, species occurences, abundances, etc.
3. Independent quality control of each dataset, e.g. check if location of coordinates falls on land or outside the study area; standardize taxon names against WoRMS (WoRMS Editorial Board 2016); check for inconsistencies in dates, depths, locations, sample numbers, abundances, use of gear, etc.
4. Sampling information in the biological publications (e.g. location, date, time, gear) were matched against the introductory volume which was used as a reference dataset, and unique sample IDs were ascribed to these samples. In cases where the biological publications included samples from additional expeditions which were not covered in the introductory volume, samples could not be matched (Fig. 3). There were also samples mentioned in the introductory volume which were not associated with any biological information, since they are described in other publications of the series that were not digitized yet.
5. A second round of quality control was performed on the integrated datasets. In cases of inconsistencies of information between the biological datasets and the introductory volume it was assumed that the information in the introductory table is the correct one, and a note was kept in the remarks field of the dataset.
6. Missing data in a dataset (on e.g. coordinates, sampling gear, sampling time, etc.) were either derived from other datasets or from other scientific and historical publications whenever possible.
7. Creation of metadata for each dataset.

This strategy resulted in a uniform dataset which allowed the seamless integration of data, as they all resulted from the same core expedition. However, to pay respect to the individual data publications and to allow users to access the data under a specific taxonomic focus, the datasets were published separately, with separate metadata and individual URLs. All datasets are available as Darwin Core Archives under a CC-Zero waiver via the Integrated Publishing Toolkit (IPT) of the Mediterranean Ocean Biogeographic Information System, supported by the LifeWatchGreece Infrastructure (see section "Data resources" below for the URLs of each dataset).

Data have to be formatted according to the DarwinCore schema (DwC) for dissemination through global biogeographic databases (e.g. Ocean Biogeographic Information System (OBIS), Global Biodiversity Information Facility (GBIF)), thus some modifications of the or enrichments were unavoidable. As this format is not necessarily user-friendly for human users (e.g. information might be merged into one field), a version of the datasets in a non-standardized but more easily usable format is provided as Suppl. materials 1, 2, 3, 4

Difficulties and problems encountered during the procedure

Same station codes for different locations

The samples of the core expedition of "Thor" in 1908–1910 had been given incrementing numbers in the introductory volume. However, the same identification codes had been used for stations of the 1905–1906 expedition, thus Schmidt created unique station codes by appending the last two digits of the year to these stations (e.g. "59.06" for station 59 in the year 1906). This practice was followed during the digitization for all years and stations from all publications where only the station code was listed. In cases where this was not sufficient to create a unique station ID, an additional letter was added to the station code (e.g. 283a.11 and 283b.11). This was done in two cases: (a) where the same code was used for different stations on different expeditions in the same year, and (b) where in the list of stations two different spatial points (coordinates) were listed under the same station code.

Samples not identified unambiguously in the manuscript

No unique sample codes were given in any of the publications. During the digitization process, unique codes were created for all samples (a sample was assumed to represent a unique combination of position, date, time, gear, length of wire and haul duration) by appending a letter to the station number (e.g. 10.08-a, 10.08-b, etc). In one case, where the same sample (all data identical) was listed twice in the introductory table, this was treated as two different samples to be in accordance with the manuscript; however, the biological data could not be mapped unambiguously to either of the two. During the assignment of sample IDs, records from the individual dataset were matched to the introductory table which served as a reference. This allowed to identify samples unambiguously and to supplement sampling metadata where it was missing in some cases. However, not all samples in the datasets “Pelagic Polychaetes” and “Calcareous algae” could be matched unambiguously to a sample code in the introductory table, even if they were taken at the same day and at the same station. In the polychaete dataset, many samples (gears/times) seem to have been merged (i.e. two sampling gears/times reported), or the sampling gear might have been matched to one sample and the sampling time to another. The dataset on calcareous algae did not report information on sampling time and gear. As the database requires a unique sample code, however, new “artificial” sample codes were created for these ambiguous samples. This probably increases the number of sampling events artificially, but as the total number of these “artificial” samples is low (20), this was considered as the best practice.

Coordinates format and precision, georeferencing

Latitude and Longitude are given in Degrees-Minutes in the original publications. This corresponds —roughly, depending on latitude— to a radius of uncertainty of around 2,000 m around the given point. In the datasets, coordinates were converted to decimal degrees with two decimals places to account for the originally given precision (no seconds given). The radius of uncertainty was recorded in the field "coordinateUncertaintyInMeters" in the DwC files. However, many of the samples were trawl catches and the ship was moving during the time of sampling over a distance of 1–2 nautical miles per hour [“… towed after the ship for a shorter or longer time, usually ½ to 1 hour, the ship sailing about 2 knots in the hour…”, in Schmidt (1912), p. 13]. These trawls were not represented in the data (only one point of coordinates was given, presumably the starting point). This additional uncertainty of the position resulting from the trawling has not been incorporated in the data, and the position of each sampling station must be considerd as an approximation.

In a few cases, no coordinates but a named location was given. These locations were georeferenced, providing coordinates for latitude and longitude, as well as a radius of uncertainty (“coordinateUncertaintyInMeters”), using the Edit Point location tool.

Several reported coordinates fell on land. Some of those represented errors in the manuscript that could be resolved from the context (e.g. East and West were interchanged) and these were corrected (original coordinates reported in the remarks field), but some stations still fell on land and their correct position could not be determined. All the subsequent publications have used the same, obviously false, coordinates, and the included maps often do not have a sufficient precision to georeference the stations correctly.

In addition, some stations were listed with the same coordinates in the introductory table but were actually placed at different positions according to the map in the introductory volume. As it could not be determined whether the map or the table were correct, and the map does not allow georeferencing with a great accuracy, the coordinates were kept as they were listed in the original table with a remark.

Time format

The publications were not consistent in their use of “am” and “pm” when noon/midnight were concerned. Times could be given as 0 pm or 12 pm for noon and 12 am or 0 am for midnight, but not always in the same manner. The correct conversion into the 24 h format could however be determined from the sequence of samples.

Depths

Although depths were given for each station, it was not always clear whether these were bottom depths or sampling depths. In most cases, they appeared to be bottom depths, while in other cases they were listed for each sample and differed per station, or they were clearly sampling depths (e.g. 0–1 m at a deep water station). Data exchange schemes such as Darwin Core only provide fields for sampling depths. As a compromise, and to avoid false conclusions (i.e. interpreting bottom depth as sampling depth), sampling depths were filled in only when unambiguous, using complementary information from the depth column, the gear (e.g. benthic, pelagic) and the length of the wire used. All other depths were included in the remarks field of each sample, leaving their interpretation to the users of the data.

Often, depth ranges, or either minimum or maximum depths only (e.g. >3,000 m) were recorded. This may have resulted from difficulties associated with precise depth measurements during that era: although depth could be accurately determined by the sounding of a lead ball, the interference of other conditions such as weather, currents, the available wire length and the nature of the bottom often might have allowed only for estimates, ranges or minimum values. In addition, multiple depth values per station may have resulted from the working practice on board (see also Schmidt 1912 for a description of a typical working day on board). Coordinates were usually recorded once, before sampling started. During subsequent measurements, the boat moved away from its initial position, either actively during trawling, or passively (drifting). Any subsequent depth measurement in the area was still considered to belong to the same station, but was effectively performed at a different spatial location. Geographically, any location on earth has only one depth or altitude, thus when translating the information from the manuscript into a data schema, information loss occurred, as only one bottom depth can be associated with one position.

Environmental information

Data for the "nature of bottom", air temperature and chlorine concentration were missing for stations after ID 297.11 in the middle of an expedition (page 48), and since no explanation was provided in the manuscript this is suspected to be a typesetting error. In some cases, environmental parameters were reported but their measurement protocols were unknown (Table 5), or the data were derived from subjective estimates (e.g. weather and sea state reports); nevertheless, all this information was made available even with limitations and/or bias, considering this would offer an added value to the completeness of the dataset.

Ditto marks and abbreviations

Many table cells of the original manuscript contained ditto marks (") (Fig. 8), a typesetting symbol meaning "same value as above". In the "Thor" publications this made sense in many cases (e.g. for the station number, date or coordinates), but in other cases the ditto marks seemed to represent "no data" (e.g. when the first rows of the table contain ditto marks for environmental measurements without any previous value being present). Sometimes (e.g. for gear, duration of haul, and depth), it was difficult to understand whether the symbol represented "same as above" or "no data". During the data digitization process the ditto marks were always treated as "same as above" for the station number, date, coordinates, nature of the bottom, time, depth, sampling gear, duration of haul and wire length, and they were replaced by the values of the preceding record in the introductory table (when present). However, for weather, wind force and direction, sea force and direction and environmental parameters, the ditto marks were interpreted as "no data". In all cases where there seemed to be errors in the original manuscript regarding the use of ditto marks (e.g. some depths were impossible or did not make sense in the context) a remarks field was included in the dataset, leaving the interpretation of the data to the user. Ditto marks were also present in the paper on Engraulidae and Clupeidae, but there a distinction was made in the use of the symbols: "no data" was signified by two dots (..), and "see above" by ditto marks (").

Size classes

The paper of Fage (1920) on Clupeiformes contained larval size classes. These were reported as distinct records in the dataset. The size class was added in the field “lifeStage". Size classes were consistent within some species (e.g. Sardina pilchardus [as Clupea pilchardus in publication]), but not within others (e.g. Sprattus sprattus [as Clupea sprattus in publication]), neither between species. Size classes used for sprat larvae were different from size classes for other species.

Inconsistencies in the introductory table

In the samples collected from the Strait of Messina in 1911 (page 47), depths were reported in brackets, and often as ranges (e.g. “10–60”). Although the meaning of this notation remains unclear, presumably these were an estimation of sampling depths, as bathymetric maps show these sampling areas to be much deeper than 60 m. In many of these samples, station codes do not correspond to coordinates: one station codes may have several different coordinates, or the same coordinates were sometimes given to different station codes. During digitization, new station codes were assigned following the strategy described above (paragraph "Same station codes for different locations").

The number of rows per column in the introductory table was not always consistent; containing, for instance, three different depths and sampling times and four different sampling gears. In these cases, it was often unclear which information belonged to which sampling event. Usually, the additional information row was considered to belong to the previous one. Although this strategy may have introduced wrong information into the dataset there was no other possible solution for such issue. Similar issues were also encountered in some of the biological publications.

Data inconsistencies within and across datasets

The structure of the publications is prone to create inconsistencies in the data, both across publications and within publications. Data on the same samples were published across several volumes over many years. Each publication repeated all or parts of the sampling information mentioned in the introductory table, but often with small differences (e.g. slightly different sampling times, depths, gears) or even with typographic errors. As it is unknown from where individual authors received this metadata on sampling —from the introduction of Schmidt (1912) which all subsequent authors consulted as a reference, or from handwritten notes which were re-transcribed and published over and over again— the correct information for a sample cannot be determined. Similarly, inconsistencies occurred also within some datasets: as occurrences are typically grouped by taxon, metadata on a sample (e.g. date, time, depth, sampling gear) were repeated over and over again for many taxa, thus introducing errors and inconsistencies. This was particularly pronounced in the dataset “Pelagic polychaetes”. During the digitization of these datasets, the introductory table has been treated as a “master table” and was assumed to be correct, and all other samples were matched to this reference.

Coordinates given here are for the four datasets combined. For information on the coverage of individual datasets see Table 2

0.52 and 59.32 Latitude; 33.05 and -29.88 Longitude.

The datasets examined in this paper include two families of Actinopteri, five families of Polychaeta and three families of Rhodophyta (Fig. 9). Differences were observed between the taxonomic names of the species under which they were reported in the original biological publications and in their currently accepted status according to WoRMS.

In the Polychaeta, several taxa are nowadays reported under different names, but the number of taxa per family reported remains stable: 10 species belonged to the family Alciopidae, 5 to the Polynoidae, 5 to the Lopadorrhynchidae, 1 to the Amphinomidae and 1 to the Typhloscolecidae.

In the Actinopteri, 4 species belong to the Clupeidae and 1 to the Engraulidae. Records for Clupea sprattus and C. sulinae were merged under the name Sprattus sprattus, according to their currently accepted taxonomy. Taxa not identified to species level were reported under the lowest possible rank (e.g. Clupeiformes).

The most significant reclassifications were observed in the Rhodophyta where only 6 species (out of 21) had still the same scientific name and were members of the same family. The 11 species that were originally part of the Corallinaceae family in Rhodophyta have increased to 13, while the 10 species of Hapalidiaceae have decreased to 7 species. Three species that were members of the Hapalidiaceae family belong now in the Corallinaceae family, while 6 species that used to be part of Corallinaceae are now in Hapalidiaceae.

Report on the Danish Oceanographical Expeditions 1908-1910 to the Mediterranean and adjacent seas

http://ipt.medobis.eu/resource?r=thorexpeditionintroduction

4