Revision and quality analyses of the Helgoland Reede long-term phytoplankton data archive

Abstract

The Biological Station on Helgoland has one of the longest data series for phytoplankton species composition in the world. Since 1962, phytoplankton has been counted to a species level on a daily basis (weekdays). One of the main reasons why this data set has rarely been used is that it has never been subjected to proper quality control or inventoried properly. This paper describes the quality control which we have now carried out on this data set and also represents the first inventory of the data and its meta-information. A comparison of the electronic database where the data is archived and the original recorded lists from the different persons dealing with the samples was an integral part of the evaluation. Apart from recording inconsistencies in species identification, we also recorded and corrected the many differences found between the electronic and paper data. The evaluation was carried out in a direct comparative manner to discern taxonomic discrepancies, and in order to find random errors in the transfer of data from paper to computer random checks of the data were carried out. This paper serves to illustrate the level of quality control required when dealing with archived phytoplankton data and illustrates the typical problems encountered.

Introduction

Long-term monitoring of biological, chemical and physical parameters has been carried out continuously at Helgoland Reede on a daily basis (workdays) since 1962 (e.g. Hagmeier 1978; Radach and Berg 1986; Gillbricht 1988; Radach and Bohle-Carbonell 1988; Hickel et al. 1992, 1993, 1994). Thus, the Helgoland long-term series is one of the longest aquatic data sets in history. What makes it unique are the phytoplankton species numbers which have been counted 5 days a week in this time period. The methods used did not vary over time. Surface water samples were taken (usually before 9 a.m.) on workdays at the “Kabeltonne” site (54°11.3′N, 7°54.0′E) between the two islands at Helgoland using a bucket. This sample was mixed well and subsampled into a glass bottle for future analyses of nutrients and phytoplankton. The phytoplankton samples taken from this were preserved in brown glass bottles using Lugols solution. Using the Utermöhl method (25 ml settled out), the samples were counted daily to species level under an inverted microscope. The time-series is often cited and taken for the discussion of changes evinced in the North Sea pelagic system over recent decades. It is also regularly a fixed item in lectures on marine ecology and is very important for the parameterisation and validation of mathematical ecosystem models. This data set, in its electronic form, is becoming increasingly more valuable within the framework of biodiversity and global change considerations (Wiltshire and Manley 2004). However, as it had never been validated or subjected to rigorous quality control, its reliability was questionable. It was our aim to evaluate the data set for its usability and to document its contents and correct it where errors were found. This paper describes the outcome of this procedure and documents the meta-information required to understand the data set.

Methods and results

Data inspection and stock-taking

The first step which had to be taken in order to understand which data were available was the collection of all the relevant data in one easily manageable archive. Thus, all electronically archived Helgoland Reede data since 1962 available in different formats (www.pangaea.de) were combined in an MS Access database. This provided the uniform structure needed for comparative purposes. The data included the meteorological, physical, chemical and biological data. The latter contained abundance and biomass data for phytoplankton down to a species level. The other data was used here as meta-information. Once the data were accessible, we set about modernising the taxonomical nomenclature, whereby the conversions from out-dated names and synonyms were also carefully documented in the names list. Then the evaluation of the documented occurrence of 40 years of phytoplankton species could proceed.

Two different species “checklists” for planktonic organisms had been used over the last 40 years for the time-series species documentation (see Appendix). The older one (last updated in 1970) was divided into two groups of phytoplankton (diatoms and flagellates) and several groups of zooplankton (which are not shown here). For each species, genus or group, a size class and corresponding carbon content per cell was given. Up to ten size classes for a single species or genus were sometimes distinguished in this list. Some genera, like Chaetoceros, Coscinodiscus and Thalassiosira, were only present as different size classes and no species names were documented. When comparing the 1970 checklist with the data on paper we found that it was used in an irregular manner. Species which appear in the electronic database for the early years are also found on the original data sheets, but not necessarily on the checklist. Examples are: Chaetoceros decipiens Cleve; Coscinodiscus concinnus W. Smith; Dactyliosolen fragilissimus (Bergon) Hasle; Thalassiosira nordenskioeldii Cleve; Dinophysis acuta Ehrenberg; and some more.

The second checklist was used from around 1976 to 2001 and was last updated in 1991. In 2001 a new checklist was started and is presented in Hoppenrath (2004). In the second list, planktonic organisms are divided into five groups (centric diatoms, pennate diatoms, flagellates, protozoa and metazoa). It contains more organisms on a species level, fewer different size classes for single species and fewer indeterminable species (at the genus level) than the earlier list. A size, in µm, with one or two dimensions, and fixed carbon content per cell was assigned to each species, size class and group.

The two lists were compared to determine their congruencies. Together with the data on paper and the different electronic versions, a species inventory was built up. This resulted in two different groups of species data:

1:

Species data, seemingly accurate, based on our knowledge, literature information and questioning of the original sampling and counting persons (who luckily are still contactable). Such species appear in both of the original lists and they seem to have been recognized and counted correctly; and

2:

species where data had to be further checked before accepting them into the database.

Time charts of the abundances since 1962 were made for all species, size classes and groups found in the electronic database. All of these charts were carefully examined for inconsistencies, conspicuousness and possible mistakes in connection with the existing checklists, with the experience of the authors, and other results from phytoplankton investigations from the area around Helgoland since 1960. Figure 1 shows two of these charts as an example, and Table 1 provides some comments for a few species examples.

Fig. 1

Time charts of Coscinodiscus wailesii and Dinophysis acuta from the original Helgoland Reede data with comments from after the inspection

Table 1 Some examples of comments resulting from the inspection of the electronic Helgoland Reede phytoplankton data

Four time periods with obviously different counting results were differentiated, mainly related to the different counters. These time periods were 1962–1975, 1976–1993, 1994–1997 and 1998–2001, whereby the last period resembles the first two periods in many respects, for example in using the same size classes for species or genera. From the available data it can be concluded that a significant change in organism recognition in connection with the checklists probably took place in 1976. There is an evident change in data composition in 1994, although the person dealing with the samples only changed in 1995.

During quality control we found that a number of species which are regularly found in the German Bight (Hoppenrath 2004) never appeared in the Helgoland Reede data archives, e.g. Chaetoceros eibenii Grunow, Chaetoceros diadema (Ehrenberg) Gran, Chaetoceros danicus Cleve, Protoperidinium curtipes (Jørgensen) Balech.

A serious anomaly found in the data was related to the fact that in the electronic data archive more species were found than had originally been recorded on paper and in the checklists. It seemed that undocumented retrospect changes had been made and previously unnamed species renamed by assigning specific size classes with speculative names. A conspicuous example of this problem was the alga Coscinodiscus wailesii Gran and Angst. The electronic database had the occurrence of Coscinodiscus wailesii archived from the early 1970s onwards. However, this was really only recorded in the German Bight in the late 1970s (Der Helgoländer 1977). Indeed, this species probably reached French waters via the introduction of Indopacific oyster cultures in the 1970s (Rincé and Plaumier 1986). From then on it spread slowly into the North Sea and appeared in very high numbers in the German Bight at the beginning of the 1980s (Der Helgoländer 1984; Dürselen and Rick 1999). This data versus reality anomaly raised the question of where the data came from within the Helgoland Reede data set for Coscinodiscus wailesii in the time period before 1977. When we went back to the paper data it actually hadn’t been recorded until the 1980s. Thus, the entries in the electronic data were considered erroneous. Because of this evidence that algae counted as indeterminate species in size classes at the genus level were later reclassified to a species, we then had to verify the occurrence of all the species entered in the electronic data with the original data sheets. Other examples of retrospect naming included most of the species from the genera Chaetoceros, Coscinodiscus, Thalassiosira, Dinophysis, Gymnodinium, Gyrodinium and Protoperidinium. Luckily, this retrospect renaming seemed to be the exception rather than the rule (see Table 2).

Table 2 The most significant later changes which were found in the Helgoland Reede phytoplankton database compared with the original records

Because of the potential of this data set for examining plankton biodiversity against the backdrop of global change, one of the assessment criteria was the long-term credibility of the microalgal identification. Thus, those species were identified where the data are acceptable as a long-term series over the whole 40 years without any restriction or with some small restrictions. These are summarised in Table 3. Using this new information, species charts were made allowing easy comparisons of their long-term occurrence over the past 40 years (Fig. 2).

Table 3 Species and genera from the Helgoland Reede sampling whose data have been classified as usable after inspection

Fig. 2

Guinardia delicatula (Cleve) Hasle as an excellent example of meaningful abundance data from a long-term series

In addition to the above evaluations, randomised examination of the original lists revealed that some species from the original lists were included in the electronic database only as “sp.” at the genus level (Table 4). It seems that it was difficult to electronically archive and change the list to accommodate new species. The structure of the original database program PLANKTI used (based on DOS) was relatively inflexible to change. Over and above these inconsistencies, others such as those documented in Table 5 were found during the evaluation of the original lists. We now hope we have recorded most of these anomalies, but with a database of this size (around 8,800 entries) we are sure to have missed some errors.

Table 4 Species found during the spot check control of the original records which had been changed to the genus level in the electronic database

Table 5 More inconsistencies between the electronic database and the original lists

Changes in the electronic database

After having checked the data, it was decided to make some changes in the electronic database, returning the data to a copy of the original recordings on the data sheets. However, this in itself proved to be a difficult procedure. First, some species were deleted from the electronic database, because no data had been entered for them. Unfortunately this also included the algae Thalassiosira anguste-lineata and Thalassiosira eccentrica which, although they intermittently appear in the original data sheets (as already mentioned in Table 4), were never entered into the database. Thus, in the electronic database the data from these species only appear as size classes from the genera. To change all this back to the original recordings would have meant the evaluation of the nearly 8,800 original lists: this is currently not a practical solution.

All records of Coscinodiscus concinnus and Coscinodiscus wailesii were checked and changed back correspondingly, if the species was only written down as Coscinodiscus sp. with a special size class in the original lists. During this procedure it was noticed that most of the entries for Coscinodiscus sp. in the different size classes from the original data sheets are not available in the electronic database. For the dates where Coscinodiscus concinnus and Coscinodiscus wailesii were checked and additional missing Coscinodiscus sp. entries were noticed, these data were added to the electronic database. As with the algae mentioned above, the original data sheets would have to be re-entered to ensure a perfect copy of the data.

Gonyaulax spinifera (Claparède and Lachmann) Diesing from the electronic database did not appear in the original lists. It was always written down as Gonyaulax sp.: therefore, it was changed back to this term in the electronic version. The same applied to Pyrophacus horologicum Stein.

Great data confusion was detected for Gymnodinium and Gyrodinium, which very often seemed to have been used as synonyms by the counters. One example from 1980 illustrates this dilemma: data for Gymnodinium sp. size class 75 µm, from the original lists was found in the electronic version as Gyrodinium sp. size class 75 µm. On the other hand, Gymnodinium sp. size class 40 µm, was found as Gymnodinium sp. size class 50 µm. These data would also have to be re-entered to ensure absolute accuracy.

Carbon biomass

The evaluation of the phytoplankton archive would not be complete without dealing with the problem of carbon biomass calculated from cell size, as these data have been continuously cited and used in the literature (e.g. Eppley et al. 1970; Strickland and Parsons 1972; Verity et al. 1992).

Estimations of cell sizes had been made since 1962 by nearly all of the different counting individuals and were written down in the original data lists. However, these data were unfortunately not taken into the electronic database. Instead, fixed carbon contents (see checklists in the Appendix) were assigned to each species for the 40 years of sampling. This is unfortunate, as algal size varies considerably and, consequently, carbon relationships vary as well. In particular, diatoms are highly variable in their size and proportions, causing a factor of two or even three in the biovolume of the same species at different time periods. Thus, to achieve a proper estimation of biomass and carbon for each sample, the dimensions of the actual organisms should have been used, rather than the fixed values. With the original data, a more effective estimation of the carbon biomass would be possible and this should, if possible, be rectified in the future.

Discussion

The quality control of the data was imperative as there is an immense pressure on the release of the species data for evaluation of the time-series regarding long-term changes in the North Sea. In particular, the flagellate data needed to be evaluated as it has often been used to show an increase of algal biomass in the German Bight. This can be summed up as follows:

The evaluation of the Helgoland Reede data has shown a significant increase in the phytoplankton biomass since 1962. Hagmeier (1978) already presented an annual mean increase of biomass from 1962 until 1974 of from 25 μg C l⁻¹ to more than 50 μg C l⁻¹. Gillbricht (1983) calculated a mean increase of up to 200 μg C l⁻¹ during the vegetation period (April to September) between 1962 and 1981. Radach and Berg (1986) analysed the data up until 1984 using trend analysis, which also showed a significant increase in phytoplankton biomass (from 8.3 μg C l⁻¹ to 34.8 μg C l⁻¹, including all the summer and winter data) with a mean exponential growth rate of 0.027 a⁻¹ over 23 years. Greater phytoplankton biomass was especially found in water with lower salinity (increase of 10.5 μg C l⁻¹ to 56.2 μg C l⁻¹). An increase was found for the summer data (from 28.2 μg C l⁻¹ to 75.9 μg C l⁻¹) as well as for the winter data (from 1.8 μg C l⁻¹ to 21.9 μg C l⁻¹ in water with lower salinity and from 2.3 μg C l⁻¹ to 15.5 μg C l⁻¹ in water with higher salinity). Comparing the development of diatoms with that of flagellates, it is obvious that the increase is mainly due to a growth of flagellates, whose biomass was ten times higher in 1984 than in 1962 (25 μg C l⁻¹ versus 2.5 μg C l⁻¹). The biomass of diatoms, in contrast, had not increased by 1986 (Radach and Berg 1986), and nor has it increased since (Hickel et al. 1992, 1994; Wiltshire and Manly 2004). Nevertheless, an extremely high biomass of flagellates has been observed in some years (Hickel et al. 1992).

The observed flagellate biomass increase at the end of the 1970s could possibly be due to the method of determination: the fixed samples taken during the first half of the time-series were stored for longer before being evaluated (Hickel et al. 1993). Meanwhile, it is known that many flagellates are destroyed by the act of fixation (e.g. Reid et al. 1990) and that successfully fixed organisms can fall apart after being stored for a longer period of time. Since the end of the 1970s, samples taken at Helgoland Reede have been evaluated within a few days. It is unlikely that this affected the diatom counts, however, as these have remained stable over the years (Wiltshire and Manly 2004). These storage effects on the flagellate numbers have been further compounded by the introduction of a new microscope in 1998. The result was a quantum leap in numbers of flagellates recognised. Consequently, these effects must always be kept in mind when considering these results.

Another aspect which has to be regarded when discussing the Helgoland Reede phytoplankton data is the frequent change in the counting persons (Fig. 3). Where possible, people always counted in parallel for a few months before the first person retired. However, the personal element involved in the recognition of microalgal species cannot be eliminated completely, and it is clear from the data that leaps in numbers often coincided with personnel changes. It is unfortunate that some of the phenomena particularly related to the flagellate numbers discussed in the literature, for example a concomitant increase in flagellates with an increase in nutrients in the German Bight, could also potentially be ascribed to a change in the counting person (Fig. 4). This naturally does not mean that all of the ideas are wrong. The variability and heterogeneity of the pelagic system in the German Bight is well known. Sometimes, unexpected phenomena appear. Without the control of the original water samples, many of the questions concerning the counting by different people cannot be reconstructed.

Fig. 3

Time periods in which ten different investigators counted the phytoplankton sample from Helgoland Reede

Fig. 4

Time chart of total flagellates. The arrows mark changes in investigator

Especially for the dinoflagellates and other flagellates, it became evident during this check that there was a large difference in the taxonomic knowledge of the different investigators. Although we have now cleaned up the database, there are still likely to be random errors which currently are impossible to eliminate completely as it would require the revision of more than 8,300 samplings. All the notes in the lists drawn up are only based on the experience and estimates of the authors, the literature on the basis of the available data, and detailed questioning of the original investigators.

The quality control we carried out on the Helgoland phytoplankton time-series data has been particularly difficult because of the lack of documented meta-information, the lack of permanent preparations and the unavailability of good photographic evidence. Had it not been for the fact that most of the original sampling and counting personnel were extremely helpful and still contactable, our task would have been next to impossible. We have, as a result of these problems, rethought our strategies for long-term assessments of phytoplankton species at Helgoland. We propose the following improvements:

• 1. Samples should be counted immediately.

• 2. The person counting the samples should take part in a national counting and identification quality control programme.

• 3. For taxonomic evaluation (e.g. to update the checklists), twice weekly net haul samples should be taken for detailed observations on live material.

• 4. All meta-information and count data should be regularly archived in www.pangaea.de.

• 5. Accurate photographs (under an excellent microscope) of the species should be taken, if necessary.

• 6. Samples should be kept for 10 years, which although not fail-safe due to the disintegrative processes allows some backtracking.

• 7. Samples should be regularly preserved as permanent slides so that the species spectra are archived for posterity, allowing retrospect taxonomic evaluations.

• 8. The taxonomic characteristics and classification criteria are currently being made available to all via the Helgoland Plankton*net web site and a handbook should be published shortly.

The very arduous quality control of this historic data set is not yet complete. We still have to estimate if it would be worthwhile adding the species missing from the electronic data set but recorded on paper to the set. Isolated species may still be inaccurate and should be removed. Perhaps the most important task ahead is to completely recalculate the carbon concentrations based on the size classes of the algae. However, we are now confident that we can free up large sections of this historic data set for use in biodiversity questions and questions on change in the North Sea.

Appendix

Tables 6 and 7 show two Helgoland Reede checklists, updated in 1970 and 1991, respectively.

Table 6 The earliest Helgoland Reede checklist (last update 1970)

Table 7 The second Helgoland Reede checklist (last update 1991). H heterotrophic