Improved methodology for measuring pore patterns in the benthic foraminiferal genus Ammonia

(1)

Research paper

Improved methodology for measuring pore patterns in the benthic foraminiferal genus Ammonia

Jassin Petersen

^a

, Bettina Riedel

^a,b

, Christine Barras

^a

, Olivier Pays

^c

, Anaïs Guihéneuf

^a

, Guillaume Mabilleau

^d

, Magali Schweizer

^a

, Filip J.R. Meysman

^e

, Frans J. Jorissen

^a,

⁎

aLPG-BIAF UMR CNRS 6112, University of Nantes, University of Angers, UFR Sciences, 2 Boulevard Lavoisier, 49045 Angers Cedex 01, France

bDepartment of Limnology and Oceanography, University of Vienna, Althanstrasse 14, 1090 Vienna, Austria

cUMR 6554 CNRS—LETG-Angers, University of Angers, UFR Sciences, 2 Boulevard Lavoisier, 49045 Angers Cedex 01, France

dSFR 4208 SCIAM, IRIS-IBS, CHU Angers, University of Angers, 4 rue Larrey, 49933 Angers Cedex 09, France

eDepartment of Analytical, Environmental, and Geochemistry, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussel, Belgium

a b s t r a c t a r t i c l e i n f o

Article history:

Received 26 April 2016

Received in revised form 8 August 2016 Accepted 16 August 2016

Available online 18 August 2016

Benthic foraminiferal pores are considered to play an important role in facilitating the gas exchange between the organism and the environment, with pore size and density supposed to be related to gas exchange intensity. Re- cent studies have therefore attempted to establish relationships between pore patterns and redox conditions, such as bottom water oxygen and nitrate concentrations. However, a prerequisite for such an attempt is the development of a practical and reliable methodology for measuring pore patterns. The aim of this study is to present a semi-automated pore measurement method forAmmonia, a dominant taxon of temperate coastal environments that are increasingly affected by seasonal hypoxia (bottom water oxygen concentrationb63μM). The approach is based on image analyses of a measurement frame positioned on SEM images with 1000 × magniﬁcation. Statistical analyses show that the surface area of the pores in the frame has a normal distribution.

Therefore, a mean pore surface area can be used to describe the pores in the measurement frame. We observed small but signiﬁcant ontogenetic changes in pore density (number of pores per frame) and pore surface area. Ac- cordingly, it seems preferable to limit pore measurements to size windows on chambers representing the same ontogenetic stage.

In order to demonstrate the efﬁciency of the method, we applied it in two case studies. Firstly, a study of living Ammoniain Lake Grevelingen (Netherlands) revealed a clear difference in pore patterns between three studied stations characterised by different seasonal bottom water oxygenation patterns. Secondly, a sediment core from the same site clearly showed the presence of two morphotypes ofAmmonia; one with numerous, small pores and the other with fewer but much larger pores, resulting in a higher porosity (larger part of the test covered by pores). Since the man-made closure of Lake Grevelingen in 1971, the latter morphotype has progressively replaced the former one. Finally, a summary of the measurements on 870 specimens with both pore patterns shows a strong relationship between pore density and pore surface area, suggesting a physical control of the in- teraction between these two parameters.

Keywords:

Benthic foraminifera Shell porosity Morphology Hypoxia

Coastal environment

1. Introduction

Studies focusing on pores in foraminifera are relatively scarce.

Hofker (1950)was probably theﬁrst to study foraminiferal pore patterns in a semi-quantitative way. He described morphological

differences between species of the same genus from different geologic times (Hofker, 1950, 1951).Bé (1968)studied pore characteristics of various recent planktonic foraminiferal species, and found that species of speciﬁc climate zones have similar pore patterns. Pores are formed at an early stage during chamber formation, giving rise to discussion on their role in the calciﬁcation process (Banner and Williams, 1973;

Berthold, 1976; Hemleben et al., 1977; Spero, 1988). For example, it has been suggested that they can serve as a site of osmotic exchange between the exterior and interior of a newly formed chamber (Banner and Williams, 1973). More importantly, pores of benthic foraminifera (BF) are supposed to be related to gas exchange, and speciﬁcally oxygen up- take, partly because of the concentration of mitochondrial clusters and

⁎ Corresponding author.

E-mail addresses:[email protected](J. Petersen),

[email protected](B. Riedel),[email protected](C. Barras), [email protected](O. Pays),[email protected](A. Guihéneuf), [email protected](G. Mabilleau),[email protected] (M. Schweizer),ﬁ[email protected](F.J.R. Meysman),[email protected] (F.J. Jorissen).

Contents lists available atScienceDirect

Marine Micropaleontology

j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / m a r m i c r o

(2)

ectobionts in the proximity of the pore openings (Leutenegger and Hansen, 1979; Bernhard et al., 2010). This led to the hypothesis of increased porosity in taxa and/or populations living in low oxygen environments (e.g.,Gary et al., 1989; Sen Gupta and Machain-Castillo, 1993).Perez-Cruz and Machain-Castillo (1990), for example, observed thatHanzawaia nitidulahad larger and more numerous pores in an East Pacific Oxygen Minimum Zone (OMZ) than in more oxygenated waters. Recently, a significant correlation has been shown between pore density and bottom water oxygen (BWO) and/or nitrate concentration in several deep-sea species (Glock et al., 2011; Kuhnt et al., 2013). However, important interspecific differences in pore parameters also exist (Gooday and Alve, 2001). In fact, molecular data have shown that certain pseudocryptic species show clear differences in porosity, e.g., inAmmoniaspp. (Holzmann and Pawlowski, 1997; Hayward et al., 2004),Cibicides/Cibicidoides(Schweizer et al., 2009), as well as in the planktonic taxaOrbulina universa(de Vargas et al., 1999; Morard et al., 2009) andGlobigerinella siphonifera(Huber et al., 1997). There- fore, although some phenotypic plasticity appears to exist, porosity is not only dependent on the environment, but is also genetically encoded.

In coastal environments of temperate climate zones, representatives of theAmmonia tepidacomplex are often the dominating benthic foraminiferal communities. They are capable of living in extremely variable environments and tolerating diverse biological stress factors (Murray, 2006). Therefore, numerous studies addressed their morphology (e.g., Bermudez, 1952; Banner and Williams, 1973; Poag, 1978; Jorissen, 1988), life strategies (e.g.,Bradshaw, 1957; Geslin et al., 1998; Stouff et al., 1999; Moodley et al., 2000; Thibault De Chanvalon et al., 2015;

Cesbron et al., 2016) and genetic variability (e.g., Holzmann and Pawlowski, 1997; Hayward et al., 2004; Schweizer et al., 2011; Saad and Wade, 2016). The morphologicalA. tepidacomplex includes at least three different pseudocryptic species in Europe as shown by molecular studies (Hayward et al., 2004). A morphological reexamination of specimens from known phylotypes allowedﬁnding slight differences (making them pseudocryptic instead of cryptic species), but it is very difﬁcult to discriminate these species solely based on the morphology of the test (e.g.,Hayward et al., 2004). Therefore, the nameA. tepidais used here, knowing that it designates a species complex including several pseudocryptic species. Concerning a potential relationship with bottom water oxygenation,Moodley and Hess (1992)found thatA.

tepidafrom the southern North Sea (determined asA. beccarii) survives anoxic (no measurable BWO concentration) periods and exhibits higher porosity under low oxygen conditions than under normoxic conditions in laboratory experiments.Kitazato and Tsuchiya (1999)conducted similar culture studies and found that the pore diameter increased with lower dissolved oxygen concentrations.Ammonia tepida(determined asA. parkinsoniana) assemblages dominate BF faunas on shelf environments in the Gulf of Mexico with seasonally hypoxic bottom waters (BWO concentrationb63μM;Rabalais et al., 1996, 2002; Sen Gupta et al., 1996). Recently, it has been conﬁrmed by culture experiments thatA. tepidais able to survive and calcify under hypoxic as well as anoxic conditions (Geslin et al., 2014; Nardelli et al., 2014).

In many coastal areas worldwide low BWO concentration is present- ly intensified by increasing eutrophication, as well as by global warming, which leads to reduced dissolved oxygen solubility and en- hanced water column stratification. Ecological consequences are increased duration, extension and intensity of hypoxia, inducing severe stress on benthic faunas (Diaz and Rosenberg, 2008; Riedel et al., 2016). In view of thesefindings, it becomes urgent to further analyse the relationship between foraminiferal pore patterns and seafloor redox conditions, and to explore the potential of pores as a palaeoceanographic proxy.

The aim of our study was to develop a practical, reliable and easily reproducible method for quantitatively describing pore patterns in Ammonia. In this paper, we willﬁrst discuss previously used methodologies for pore measurements (e.g.,Wiles, 1965; Moodley and Hess, 1992; Glock et al., 2011; Kuhnt et al., 2013, 2014; Weiner et al., 2015).

In most cases these methods are adapted to a certain species or context, and are not necessarily suited for measuring pores inAmmoniaand gen- era with a similar morphology and pore pattern. Most of these methods are not efficient, when measuring large quantities of specimens. The semi-automated method presented here has been specifically designed forA. tepidaand species with a similar morphology. We will illustrate the efficiency and reliability of this new method in two case studies.

2. Methodological developments

2.1. Overview of previously published methodologies for measuring pore patterns in foraminifera

Pore analysis has been applied to foraminiferal tests since the 1950s (Hofker, 1950, 1951).Wiles (1965)presented a detailed description of his methodology. He used the inner surface of crushed, fossil shells from a single planktonic species (Globigerina eggeri), oriented parallel between two glass slides, on which he measured the pores with a petro- graphic microscope. At a magnification of 270×, a square frame with a length of 45μm was used to delimit the area for pore measurements and to calculate the pore density, defined as the number of pores per surface area (seeTable 1for definition of pore parameters). In his pore analyses, no correction was made for partial pores, touching the border of the frame. For each specimen, the pores of at least three fragments were counted at least three times and an average of twenty specimens was considered representative for the pore characteristics of the analysed species (Wiles, 1965).Bé (1968)followed this procedure for analysing the pores of various living planktonic species but used 600×

magniﬁcation and a square frame with a length of 25μm. Besides pore number and density, the pore diameter was reported as well, allowing the calculation of the porosity (the percentage of frame area occupied by pores,Bé, 1968; Bé et al., 1973).

The studies of Frerichs and co-workers (Frerichs et al., 1972; Frerichs and Ely, 1978) used a similar approach for recent planktonic species (frame area = 1240μm², 450× magnification). Moreover,Frerichs et al. (1972)specified that only the last chamber of the largest specimen of each species was used. On the basis of a comparison of mean, maximum and minimum diameters of single pores, they concluded that in allfive analysed species the minimum pore diameter, and therefore the porosity, decreases with increasing distance to the equator (Frerichs et al., 1972). Earlier,Lutze (1962)had indicated that the preci- sion of pore measurements made under a binocular microscope is lim- ited by difficulties in adjusting the focus on the pore openings. More recently, for benthic foraminifera, it has become usual to measure the pores on the outside of the test, using SEM images (e.g.,Moodley and Hess, 1992).Holzmann and Pawlowski (1997)andHayward et al.

(2004)used pore analyses of SEM images together with molecular analyses to discriminate species of the genusAmmonia. InHayward et al.

(2004), measurements were systematically performed on 10 pores in the penultimate chamber, closest to the junction of spiral and chamber sutures (with the last chamber). Contrasting to the approach using a frame,Glock et al. (2011)counted all pores of one side of tests of Bolivina spissaon SEM images. This number was considered as half of

Table 1

Glossary of terms related to pore measurements used in this study.

Number of pores [ ]

Total number of pores (Np) automatically counted in the measurement frame, after manual correction for partial pores, double pores and pores of exceptionally small size Pore density

[Np/μm²]

Number of pores per surface area of measurement frame (the method presented here uses a frame with a surface area of 562μm²)

Pore area [μm²]

The mean surface area of all pores, calculated as the total surface area occupied by pores, divided by the corrected number of pores, expressed inμm²

Porosity [%]

Percentage of the surface in the measurement frame covered by pores.

(3)

the total number of pores (for the whole specimen). For Bolivina seminudaa large measurement frame (10,000μm²) was used (Glock et al., 2011). In a comparison between measurements of pore diameter on the outer and inner surfaces of fossil planktonic foraminifera, Constandache et al. (2013)found important differences, but no systematic trend of larger or smaller diameters for one of the two respective surfaces. In order to avoid distortion of the pore area, they used SEM images (1200×–6000× magniﬁcation) taken with a viewing angle perpendicular to the centre of the image and they analysed only the pores in the im- mediate proximity of the centre of the image (Constandache et al., 2013).

Fisher et al. (2003)wereﬁrst to introduce a method to measure the porosity with a software package automatically calculating the pore areas, based on a grey scale threshold applied to a frame with an area of 1216μm². The SEM images had a magniﬁcation of 5000 × and in case of low contrast between pores and the surrounding calcitic areas, images were optimised manually. In order to increase the contrast between pores and the shell,Morard et al. (2009)manually blackened pore surfaces and used the digitised drawings for image treatment.

This approach was further optimised by introducing a correction factor for the bias due to non-linear overestimation of the pore areas by the software (Morard et al., 2009).

Kuhnt et al. (2013)manually counted pore numbers forBolivina pa- cifica,Chilostomella oolinaandFursenkoina mexicanaon SEM images with 1500×–5000× magnification, depending on the species. In a subsequent publication,Kuhnt et al. (2014)compared manual pore counts with an automated method using software that counts the number of pores and calculates the pore area, again on the basis of a grey scale threshold. SEM images of entireGlobobulimina turgidaspecimens were taken at a constant magnification of 2000×, but only the last chamber was analysed. The size of the frame used for the automated method (300μm²) was a compromise imposed by the need to avoid distorted pore surfaces due to the curvature of the test and to obtain a statistically significant number of counts. The entire procedure for one specimen included successive analyses, with the frame positioned manually from the aperture downwards excluding pores withfillings and partial pores at the border (Kuhnt et al., 2014). Finally, in an approach to measure pore density for planktonic foraminiferal species with highly inflated chambers,Weiner et al. (2015)used image analysis software capable of extracting the pore diameter (on SEM images with a magnification of 4000×). In this case, the porosity was calculated for a square with four ideally rounded pores, on the basis of average pore diameter and the average distance between neighboured pore centres (Weiner et al., 2015).

2.1.1. Advantages and pitfalls of earlier used methods

Manual measurements of broken tests (Wiles, 1965; Bé, 1968;

Frerichs et al., 1972; Constandache et al., 2013) present several advantages:firstly, the pores can be measured on the inside of the test, thereby avoiding problems such as infills with detritical material, partially covered pores of earlier chambers by newly formed calcite layers, or (partial) dissolution at the outside of the test. Furthermore, the fragments are positioned parallel to the glass plates, thereby avoiding measuring problems due to the curvature of the shell. In fact,Wiles (1965)reported that the inner surfaces of the planktonic foraminifera he measured are gently concave and that they orient themselves parallel to the surface of the glass slides. The pore surface is not distorted and the measurement of the diameter should correspond to the real diameter. Finally, pores can be inspected individually, and in case of funnel-, vase- or hourglass- shaped pores, the measurement can be adjusted (Bé, 1968). As each shell fragment is measured repeatedly and several fragments per specimen are considered, the mean pore diameter is representative of the entire test (Wiles, 1965). However, this procedure is extremely time consuming, and it is also impossible to evaluate a putative ontogenetic effect on pore parameters, as the broken fragments cannot be traced back to a specific chamber. Furthermore,Lutze (1962)andBé (1968)indicated that the different shapes of the pores and difficulties in focusing lead to an elevated uncertainty in their pore diameter measurements.

The use of SEM images improved the resolution of measurements. At a magniﬁcation of 1000× or greater, pores of most foraminiferal species are distinct from the surrounding calcite and are precisely measurable.

In addition, it is possible to keep records of the emplacement of the pore analysis frame and its associated data. When the aim is to count the pores of an entire side of the test (Glock et al., 2011), it is much easier to have aﬁxed image than to move the sample under the microscope. However, for most non-ﬂattened species (e.g.,Ammoniaspp.) or very large specimens, it is impossible to obtain an image of the com- plete side of a test and to use it for pore data analysis. In such cases pore measurements have to be performed on a smaller part of the test (e.g., one chamber only or part of one chamber;Moodley and Hess, 1992;

Hayward et al., 2004; Constandache et al., 2013). Consequently, the representativeness of the selected area may pose a problem.Frerichs et al.

(1972)werefirst to limit the pore analyses to a specific part of the test, in their case the last formed chamber. When specimens of a similar size are used, this restriction has the advantage of limiting bias of the pore measurements due to ontogenetic effects. The use of the latest chamber also increases the probability that the pores are representative for the environmental circumstances measured while sampling. Another important aspect is the distortion of the pores, which may become important on SEM images taken with a viewing angle not perpendicular to (parts of) the image (Constandache et al., 2013). This is particularly important in case of strongly inflated tests, for which the curvature necessarily increases towards the outer parts of the image. These problems can cause a serious bias in pore size measurements.

Concerning automated pore measurements, the most important advantage is the short processing time, facilitating rapid production of a large amount of representative data (Kuhnt et al., 2014). The manual adjustment of the grey scale value, used to determine the boundary between pores and the surrounding calcite, is an efficient way of standardising the pore measurements for all pores in the frame. In this way, individual errors in pore area measurements are avoided. Howev- er, in cases of inclined, non-perpendicular pores, or pores with a (partial) infill with a similar colour as the surrounding test, it is possible that the pore area is under- or overestimated. Therefore, it may be necessary to test several grey scale values before producing thefinal results, or to modify the contrast of the SEM image in order to obtain a better contrast between pores and calcite (Fisher et al., 2003; Morard et al., 2009). Moreover, automated methods also introduce some systematic errors, among which the most important ones are,firstly, counting partial pores on the edge of the frame and secondly, the underestimation of the pore area of oblique pores. Partial pores are considered in more de- tail inSection 2.2. As to the problem of erroneous pore measurements due to the convexity of the test, which may be important in case of very inflated species, the mathematical approach described byWeiner et al. (2015)presents a possible solution. However, this method approx- imates the porosity with a circular shape of the pores and a regular pore distance, and is therefore less suitable for realistically representing the porosity of species with a heterogeneous distribution of pore distance or a variable pore shape.

In conclusion, among the methods introduced so far, the automated approach favours a fast processing of a large amount of data. However, species with a highly convex shape have to be treated differently, so that the method has to be adapted to the test form under consideration.

Finally, manual post-treatment of the data may be necessary to avoid bias introduced by the automated approach.

2.2. Development of a standardised pore measurement method for Ammonia

Pores are present on the majority of the curved surface on the dorsal side of each chamber ofAmmonia.In spite of the convex test shape, it is possible to measure pores on a relativelyﬂat surface because of the more or less horizontal character of the inner part of the chambers (close to the spiral suture,Fig. 1) on the dorsal side. In fact, our SEM

(4)

images taken at 1000× magnification show only distorted pores on the outer part of the chambers (away from the spiral suture,Fig. 1). The pore openings are generally well defined and clearly distinguishable from the smooth surrounding calcite. Since no clear differences are visible in the pore diameter on the inside and outside of the test (Banner and Williams, 1973; Geslin et al., 1998), and since we want to quantify the pore patterns of a single, specific chamber, we decided to use the outer surface. Also the previous studies on pore patterns ofAmmoniawere based on measurements of SEM images of the outer surface (Moodley and Hess, 1992; Holzmann and Pawlowski, 1997; Hayward et al., 2004).

The method to measure pore parameters ofAmmonia tepidapresent- ed here combines automated pore measurements on 1000× magnification SEM images in a manually positioned measurement frame on the dorsal side of the test with a post-treatment of the data. First, it is essen- tial to select a surface as horizontal as possible. Next, it is critical to as- sure that the selected area is representative at least for the whole chamber, and preferably for the whole test. Since inA. tepidatheflattest part of the test is generally found on the inner part of the chambers, we decided to use a rectangular measurement frame, and to position it close and parallel to the spiral suture (Fig. 1B). Because the aim is to compare the pore parameters with the environmental parameters during calcification, it is logical to measure one of the last formed chambers.

In case element concentrations need to be measured by laser ablation ICP-MS on the same tests, and on the same chambers as the pore measurements, it is often impossible to use the last chamber, because it is very thin, and it is often broken or even absent. For this reason, the penultimate chamber was chosen for pore measurements.

Aflowchart including the main methodological aspects is presented in Appendix A (Fig. A.1) and a detailed manual of the method is given in Appendix B, with a step by step example of the measurement procedure. We developed a macro (see Appendix B) for the online freely available, image processing software ImageJ (Schneider et al., 2012), allowing us to place a measurement frame of afixed size on the SEM image and to choose an appropriate grey scale threshold value, optimising the distinction between pores and test surface. The ImageJ software then automatically provides the number of pores and the porosity for the selected frame, as well as the pore area, perimeter and circularity indexes for each individual pore. However, in view of the very dense pore coverage inA. tepida, it is unavoidable that some pores are cut by the frame (Fig. 1B). Because their number and size can bias the results, we identified and counted these partial pores manually, and divided their number by two for the count of the total number of pores and for the calculation of the mean pore area. Another bias from the

automated procedure comes from pores that are so close that their contour lines are connected, and that are consequently counted as single pores. We identified all such“double pores”(sometimes“triple pores”) and manually corrected the number of pores by adding one unit (two units in case of triple pores) in each case. Finally, we decided to withdraw pore surfaces of exceptionally small size from the counting because such outliers largely influenced the mean pore area (seeSection 2.2.1). This concerns very small parts of pores cut by the measurement frame, small dark micro-particles accidentally attached to the test, and discrete pores of exceptionally small size (compared to the majority of pores in the frame). The limit for such exceptionally small pores wasfixed at a perimeter of 2μm (corresponding to a pore area of 0.32μm²). However, this correction was only applied when such pores of exceptionally small size representedb10% of all pores. Thefinal calculation of the mean pore area is performed after these manual corrections.

With this new, semi-automated analytical procedure, large amounts of reliable data can be produced in a fairly short amount of time. An ad- ditional advantage is that several size parameters (pore area, perimeter and circularity indexes) are obtained for each individual pore, allowing sophisticated statistical analyses.

2.2.1. Representativeness of the measurement frame

Previous studies hardly presented any statistical tests, neither for the distribution of the pores on the shell surface, nor for the variability of the pore parameters within the measurement frame. OnlyMorard et al.

(2009)performed statistical analyses on pore area data to differentiate between cryptic species ofOrbulina universa. However, statistical tests are necessary to verify the representativeness of the data obtained in the measurement frame (pore density, pore area or porosity). Firstly, it is primordial to test for a normal distribution of pore area, in order to know whether mean values and standard deviations can be used to compare results. Another important aspect is the optimal size of the measurement frame. The ideal frame should be small enough to include as little distorted pores due to curvature as possible, but it should still be large enough to be representative for the whole chamber. In order to ﬁnd this optimum, we testedﬁve different frame sizes, varying from 250 to ~ 2500μm². Thep-value of the non-parametric Kolmogorov– Smirnov (KS) test determines the probability of obtaining the observed distribution in case of a normal distribution of the population. In fact, the KS test produces a series of simulated random normal distributions, all with the same mean and standard deviation as the sample, and gives an estimate of the maximum distance (D) between the observed distribution of the pore area and the simulated normal distribution. Thus, a

Fig. 1.SEM images ofAmmonia tepida. A: View of an entire specimen from Lake Grevelingen. B: 1000× magniﬁcation of the penultimate chamber of the same specimen. Typical image used for pore measurements. The rectangle represents the pore measurement frame.

(5)

value for D close to zero indicates that the pore area is normally distributed. To evaluate the variability of D in the simulations the test was repeated ten times, resulting in a boxplot of the D-values.

Fig. 2andTable 2present the data of one tested specimen. Frames 1 to 5 decreased gradually in size (Fig. 2A) and the data for the pore area were compared between the frames after the post-treatment,

Fig. 2.Example in which frames with varying sizes have been applied on the same SEM image of a specimen ofAmmonia tepidawith large pores. A: Size of different frames, positioned on penultimate chamber. 1: largest frame, size decreasing until frame 5. B: Results presented as boxplots (with median,ﬁrst and third quartile) for pore area (after post treatment of datasets).

C: Results of one-sample Kolmogorov–Smirnov (KS) test for normal distribution of pore area dataset (10 KS test runs for each frame size).

(6)

concerning correction for partial, double and exceptionally small pores (Fig. 2B). Thep-value of the KS test proved a normal distribution of the pore area for all frames (Table 2). The boxplots inFig. 2C represent the values for D from the repeated test runs. The results show that although the pore area was normally distributed for each frame size, D seemed to increase when frame size decreased. In other words, the largest frame 1 (~ 2500μm²) presented a pore distribution closest to an ideal normal distribution. For frames 2 to 5 (decreasing in size from

~1500μm²to 250μm²), the distance to the simulated ideal normal distribution tended to increase progressively (Table 2,Fig. 2C). According to Student'st-test with Bonferroni correction (run as post-hoc test of an ANOVA model), allfive frames have significantly different D-values, except for frame 3 vs. frame 4 (Table 2). However, as mentioned above, even in the smallest frame (number 5), for which D is visually largely offset compared to the other frames (Fig. 2C), the pore areas of individual pores still represent a population with a normal distribution (Table 2). These results confirm that mean pore areas with standard deviations can be used to describe the pore size, whatever the size of the frame. Furthermore, normally distributed data permit to test for statistical differences in mean pore area between the frames (including correction for partial, double and exceptionally small pores) with an ANOVA model and Student'st-test as post-hoc test. In the example presented inFig. 2, differences between the frames are not significant at a significance level ofα= 0.05 (p-value = 0.147 for ANOVA model, no significantp-values for post-hoc test).

This test procedure with different frame sizes was repeated for two other specimens, one with much smaller pores than the specimen in Fig. 2and one with similarly large pores (Appendix A, Fig. A.2A, A.2B and Fig. A.3A, A.3B). In both cases all tested frame sizes yielded a normal distribution of pore area. For both specimens, D-values were higher for the smallest frame than for the larger frames (Fig. A.2C and Fig. A.3C).

Together, the results of the three tested specimens indicate that the smallest frame (number 5) has a higher probability of generating less normally distributed data than the other frames. A possible explanation is that, in a smaller dataset (smaller frame size), single pores of very large or small size have more inﬂuence on the distribution than in larger datasets. For the individual with numerous small pores, the pore area was smaller in the three largest frames compared to the smaller frames 4 and 5 (Fig. A.2B). For the second specimen with large pores, the pore area was strongly decreasing from frame 3 to the largest frame 1 (Fig.

A.3B). In both cases, the largest frames, which included more convex parts of the chamber, presented the lowest mean pore areas, suggesting that the difference could be due to an underestimation of the surface area of some pores on the curved surface at the outer part of the frame (towards the periphery of the chamber).

Finally, concerning the pore density for the specimen presented in Fig. 2, values for frames 1, 2 and 3 varied from 0.049 to 0.510 Np/μm², whereas the pore density was 0.046 and 0.044 Np/μm²for frames 4 and 5, respectively (Table 2). In fact, frame 5 included only 11 pores,

excluding partial pores. For the other tested specimen with large pores (Fig. A.3), there was a similar trend of decreasing pore density for smaller frames. For the specimen with small pores only (Fig. A.2), frame 5 had a smaller pore density than the other frames. This trend of increasing number of pores per surface area in the larger frames cor- roborates our hypothesis of a measurement bias caused by the fact that larger frames include a larger part of the more inclined outer part of the chamber. Apparently, in the outer part of the test, the horizontal projec- tion of an inclined surface not only leads to an underestimation of the pore area, but also to an overestimation of the number of pores. The pore area and pore density together determine the porosity (proportion of the surface covered by pores). In fact, porosity shows only little vari- ation between tested frame sizes (Table 2). This is probably explained by the overestimation of pore density in the larger frames, which com- pensates for underestimation of pore surface.

In conclusion, in order to combine maximum representativeness and maximal probability of a normal distribution, a larger frame is preferable. The smallest frame size tested (250μm²) is least suitable because distributions are farthest away from a perfectly normal distribution.

However, with the objective of avoiding bias due to the measurement of pores on the inclined outer parts of the chamber, also the larger frame sizes (~1000–~2500μm²) are less suitable. In view of these results, it appears that frame 4, with a surface area of 562μm², is the best compromise.

2.2.2. Ontogenetic trends in pore patterns

Having decided on a suitable size for the measurement frame, another important question was whether pore measurements applied on a single chamber ofAmmonia tepidacan be considered representative for the whole test. This implied investigating whether there is an ontogenetic variability in pore parameters.Fig. 3presents the three main pore parameters (number of pores, pore area and porosity) measured on the penultimate chamber, as a function of the maximal diameter of the entire test of the specimen analysed. The maximal diameter of the specimen (MDS), which can be considered as an approximation of ontogenetic stage, was measured on SEM images for 433 specimens of living A. tepida(N125μm size fraction) sampled in Lake Grevelingen (Nether- lands) in 2011 and 2012.

The MDS varied between ~150 and ~600μm. The number of pores showed a significant decrease with increasing MDS, although the data yield a fairly low determination coefficientR²(Fig. 3A,R²= 0.132, pb0.0001). The mean pore area increased significantly with increasing MDS, but once again theR²of the linear correlation was low (Fig. 3B, R²= 0.0734,pb0.0001). Conversely, porosity did not show a significant change related to MDS (Fig. 3C,R²= 0.00575,p= 0.115).

Our data suggest that the penultimate chambers of larger specimens tend to have fewer pores of a larger mean size, without a statistically signiﬁcant impact on porosity. In fact, for every 100μm size increase, on average, 2.9 pores are added per 562μm²frame, and the average Table 2

Results of statistical tests performed on the specimen discussed inSection 2.2.1(specimen shown inFig. 2). Size and surface area of frames 1 to 5. Number of pores, pore density, pore area as mean value with standard deviation and porosity for each frame (dataset after post treatment).P-values and D for Kolmogorov–Smirnov (KS) normality tests of dataset for pore area.

The KS test was run 10 times for each dataset; here the mean value of the 10 runs is presented. Comparison of D from different frames;p-values of Student'st-test with Bonferroni correction (post-hoc test of ANOVA model). Onlyp-values for comparisons of frame 4 with other frames are shown, because this frame wasfinally chosen (significant differences in bold, at significance levelα= 0.05). However, thep-values between the other frames systematically revealed significant differences.

Frame size (μm∗μm)

1 2 3 4 5

44.69∗55.87 33.52∗44.69 25.70∗39.11 16.76∗33.52 11.17∗22.35

Surface area of frame (μm²) 2497 1498 1005 562 250

Number of pores 126 74 51 26 11

Pore density [Np/μm²] 0.050 0.049 0.051 0.046 0.044

Pore area (mean + SD) [μm²] 3.07 ± 1.02 2.98 ± 1.07 2.70 ± 0.91 3.05 ± 1.11 2.52 ± 0.88

Porosity (%) 16.5 16.1 15.4 16.3 16.7

KSp-value (mean) 0.890 0.841 0.617 0.810 0.477

KS D (mean) 0.053 0.072 0.109 0.126 0.256

Studentst-testp-value b0.0001 b0.0001 0.092 b0.0001

(7)

pore size increases by 0.79μm². However, the large scatter of data makes it impossible to apply a constant correction factor. Consequently, it appears that specimens of the same size (i.e., ontogenetic stage) should preferably be used for the comparison of differentAmmoniapop- ulations. When this is not possible, for example due to a lack of sufﬁcient specimens, porosity appears to be the most reliable parameter.

3. Case studies

In the next two sections, we briefly present two case studies in order to show the efficiency and reliability of the proposed method. In thefirst example, we study specimens from nearby sites with differences in some of the physico-chemical parameters, demonstrating possible con- nections between pore characteristics and environmental conditions. In the second example, we compare living and fossil specimens from the same locality to show the existence of temporal changes in pore patterns. Finally we compare the results of all specimens analysed until now and discuss relations between the studied pore parameters. The aim of these case studies is not to discuss the observed differences in de- tail, but rather to show that with this pore measurement method, it is possible to rapidly obtain reliable quantitative results.

3.1. Application of the pore measurement method to living specimens of Ammonia tepida from Lake Grevelingen

Lake Grevelingen is a former branch of the Rhine–Meuse–Scheldt delta in the Netherlands. The artiﬁcial lake was formed after the con- struction of dams in 1964 (connection to the rivers cut off) and 1971 (connection with the North Sea cut off) and has only a minor seawater inﬂow through a small sluice (Bannink et al., 1984). The salty bottom

waters (salinity = 31–32,Hagens et al., 2015) of the lake are subject to seasonal hypoxia (BWO concentrationb63μM) and anoxia (no measurable BWO concentration), strongly affecting the benthic ecosystem (Sulu-Gambari et al., 2016). Live benthic foraminiferal faunas (recognised by CellTracker Green (CTG),Bernhard et al., 2006) were sampled seasonally in 2011 and 2012 at three stations (location map in Appendix A, Fig. A.4, water depth: 34.0 m, 22.3 m and 17.2 m for stations Grev-1, Grev-2 and Grev-3, respectively) with contrasting seasonal BWO variability. The benthic environment of Lake Grevelingen contains dense but low diversity foraminiferal assemblages, strongly dominated byAmmonia tepidaandElphidium excavatum. To investigate whether there could be a relationship between oxygen concentrations and pore patterns, the pore parameters of livingA. tepidasampled in May 2012 were analysed (in Fig. 4, we present the data for May 2012). The deepest station is much more affected by seasonal hypoxia/anoxia than the shallower stations, with a longer duration of hypoxia/anoxia and much lower oxygen concentrations in summer and early autumn (Table 3). In May 2012, at the very beginning of the hypoxic season, all three stations were inhabited by dense populations ofA.

tepidaand had oxygenated bottom waters (Table 3).

Pore patterns were measured on 21 to 30 living individuals ofA.

tepida.When comparing the three stations, the pore parameters show some significant differences between station Grev-1 and the two other stations (Fig. 4,Table 4). The number of pores (in the 562μm²measurement frame) was significantly higher at station Grev-1 (22.7 ± 5.4) than at stations Grev-2 and Grev-3 (15.5 ± 3.8 and 16.9 ± 3.1,Table 4). Al- though the pore area was somewhat smaller at station Grev-1, differences between the stations were not significant at the 95% confidence level (5.87 ± 1.65μm²for Grev-1 versus 6.57 ± 1.63μm²and 6.24 ± 1.50μm²for Grev-2 and Grev-3,Table 4). The porosity, which has Fig. 3.Pore parameters in relation to maximal diameter of the specimen. The dataset includes 433 living specimens ofAmmonia tepidafrom Lake Grevelingen sampled at 3 stations in 2011 and 2012. A: Number of pores (for measurement frame of 562μm²). B: Pore area. C: Porosity. In red are the equations for the linear regression, determination coefficients andp-values.

(8)

been measured automatically, but which basically is the product of the number of pores and pore area, shows a signiﬁcant difference between station Grev-1 and the two other stations (22.6 ± 3.7% versus 17.4 ± 3.6% and 18.3 ± 3.3% for Grev-2 and Grev-3,Table 4). Finally, the Grev-1 specimens are signiﬁcantly smaller than those of the other stations (304 ± 57μm versus 378 ± 58μm and 386 ± 39μm,Table 4).

Several earlier studies have tried to relate benthic foraminiferal pore patterns to bottom water oxygen concentration. Perez-Cruz and Machain-Castillo (1990)mentioned thatHanzawaia nitidulahas larger and more numerous pores in populations from an East Pacific OMZ compared to populations from more oxygenated waters, without giving detailed information. Only recently, the putative relation between pore patterns and oxygen concentration has been subject to more extensive quantitative studies.Glock et al. (2011), for example, studied the pore density ofBolivina spissafrom the Peruvian continental margin. They found a significant negative correlation between pore density and BWO but also observed that the negative correlation with nitrate concentrations yielded a higher coefficient of determination (R²= 0.79 for BWO versusR² = 0.95 for nitrate concentration;Glock et al.,

2011).Kuhnt et al. (2013, 2014)reported similar negative correlations between pore density and both BWO and nitrate concentrations. How- ever, these relationships were more robust for some species than for others. In conclusion, these authors suggested that pore patterns in Bolivina paciﬁca, Fursenkoina mexicana and Globobulimina turgida could potentially serve as a BWO proxy in OMZs (Kuhnt et al., 2013, 2014). ConcerningAmmonia tepida,Geslin et al. (2014)showed that under laboratory conditions, it is able to survive and calcify under hypoxic conditions, andNardelli et al. (2014)reported that specimens ofA.

tepidaare even capable of calcifying under anoxic conditions. Very few data exist today concerning the relation between the pore patterns of Ammoniaand oxygen concentration. OnlyMoodley and Hess (1992)described that individuals ofA. tepida(determined asA. beccarii), cultured in strongly hypoxic conditions, have a higher porosity than specimens grown in normoxic waters. Their observations are based on measurements of nine and eight specimens that calciﬁed new chambers respectively under high (225μM) and low (b12.5μM) oxygen conditions.

They concluded that the increased porosity under low oxygen conditions (12% versus 6% for low and high oxygen concentration,

Table 3

Details for sampling sites in Lake Grevelingen. Oxygen (O2) concentrations were all measured at the same location, in the proximity of the three stations (51°44.7756 N, 3°53.3688 E). The O2concentrations reported here correspond to the value measured in the water column at the depth of each of the three stations. Duration of hypoxia and values for minimal O2were measured during the hypoxic season in summer 2011. The foraminiferal samples were collected in May 2012 from the sediment surface (0–1 cm), size fraction 150–315μm.

Station Depth [m] Latitude Longitude Duration

Hypoxia [d]

(summer 2011)

Min.

O2[μM]

(Summer 2011)

O2[μM]

(May 2012)

n (number specimen analysed, May 2012)

Grev-1 34.0 51°44.834 N 03°53.401 E 141 0.00 163 26

Grev-2 23.1 51°44.956 N 03°53.826 E 58 0.61 177 21

Grev-3 17.2 51°44.856 N 03°53.942 E 28 62.10 189 30

Fig. 4.Comparison of pore parameters and maximum diameter of specimen for three stations in Lake Grevelingen sampled in May 2012. SeeTable 4for statistical details. All specimens are assumed to have been alive when sampled (CTG). Data based onn= 25 for station Grev-1,n= 21 for station Grev-2 andn= 30 for station Grev-3 (specimens per station, one image per specimen). A: Number of pores. B: Pore area. C: Porosity. D: Maximum diameter of specimens. Boxplots with median,ﬁrst and third quartile.

(9)

respectively) is a result of a larger pore size, and not of an increase in pore density (Moodley and Hess, 1992).

As mentioned above, the deepest station Grev-1 is subject to more severe and longer hypoxic/anoxic events than the other two stations.

In May 2012, shortly before sampling of the present material, oxygen concentration measured in the water column (measurement close to station Grev-1) was 163μM at 34.0 m, 177μM at 23.1 m and 189μM at 17.2 m water depth. In 2012, at the deepest site, substantial numbers of living foraminifera were only observed in spring (March–July, Jorissen et al., in prep.), whereas at the other two stations rich populations were found throughout the year. Consequently, it is probable that the smaller size of the specimens found at station 1 (Fig. 4D) is due to the fact that they are in an earlier ontogenetic stage. Since there was a slight increase of the number of pores, and a slight decrease of the pore area with increasing test size (Fig. 3A, B), part of the signiﬁ- cant differences between the specimens of station Grev-1 versus Grev-2 and Grev-3 could be the result of a size difference. Concerning the number of pores, the size decrease of about 80μm between specimens from Grev-1 compared to Grev-2 and Grev-3 (Table 4) should correspond to an increase of about 2.3 pores per measurement frame (applying the linear regression equation ofFig. 3A), whereas in reality, there is an increase by 7.3 pores (difference between mean values of Grev-1 and Grev-2,Table 4) and 5.3 pores (Grev-1 versus Grev-3), respectively.

For pore area there was no signiﬁcant difference between the stations and for porosity no signiﬁcant ontogenetic trend was found.

Summarising, there is a clear difference in the pore patterns of the populations of the three sites, which cannot be explained by differences in ontogenetic stage alone. Our data suggest that the higher number of pores at the deepest station, leading to a higher porosity, could be an adaptation to the lower oxygen concentration at this site. The fact that the oxygen concentration at the time of sampling was similar at all three stations does not contradict this hypothesis because the higher porosity at the deepest station could be an adaptation to a long-term environmental stress.Murray (2006)points out that foraminifera react to changes in environmental conditions on longer timescales and only when a critical threshold in such a parameter is approached. Moreover, it is very probable that the pore parameters are inﬂuenced by other environmental parameters as well (e.g., changes in salinity, temperature, nitrate concentration, etc.). Further discussion is clearly beyond the scope of this paper, which illustrates that with our pore measurement method, it is possible to quantify the presence of different pore patterns under slightly different environmental conditions.

3.2. Application of the pore measurement method to specimens of Ammonia tepida from four depth intervals in a sediment core taken at station Grev-1

In order to study the evolution of the benthic foraminiferal faunas after the closure of Lake Grevelingen in 1971 (Bannink et al., 1984), a 90 cm long sediment core was taken at station Grev-1 (corer UWITEC on board R/VLuctor, sampled in May 2012). The deepest 10 cm, deposited most likely before the closure in 1971, consisted of sandy deposits, whereas the upper 80 cm contained homogeneous clayey sediments.

The core was sliced every 0.5 cm and fossil foraminifera were picked from everyﬁfth sample. Foraminiferal densities were often very low,

and therefore successive samples had to be grouped together to reach at least 20 measurable tests. For this preliminary study we measured 83 individuals, found in four composite depth intervals (Fig. 5).

The histograms for the pore parameters presented inFig. 5show sometimes a normal distribution (confirmed by Kolmogorov–Smirnov tests), sometimes a skewed distribution, and in some rare cases a clear bimodal pattern. Concerning the number of pores, in the 83.5–90 cm interval, a large majority of specimens had 75 to 125 pores per 562μm². In the 62.5–77.5 cm interval, some specimens with much fewer pores (25– 50) appeared, leading to a visually bimodal distribution. The group with smaller pores had clearly become dominant in the 32.5–45 cm interval, and the number of specimens with a high number of pores further decreased in the 0–22.5 cm level. Concerning pore area, in the 83.5– 90 cm interval all specimens except one had a pore size below 1μm². Towards younger core intervals, specimens with larger pores (i.e., pore areaN2μm²) appeared, and became progressively more frequent, finally accounting for about half of the assemblage in the 0–22.5 cm interval. These trends resulted in a substantial increase of porosity (with the median increasing from 11.3 and 10.8 for the two lower intervals to 19.0 and 17.8 for the two upper intervals), although these histograms do not show the bimodal distribution for the number of pores, described before (Fig. 5). Because normal distributions are not found for all datasets, we used Kruskal–Wallis tests with the Mann–Whitney– Wilcoxon test as post-hoc analyses to verify which intervals are statistically different from other ones (Table 5). The tests confirm that the two lower levels are significantly different from the two upper levels.

According to historical data, seasonal hypoxia/anoxia appeared im- mediately after the closure of Lake Grevelingen in 1971 (Bannink et al., 1984). In the sediment core, this event is characterised by a major lithological shift at 80 cm depth, from sandy to silty/clayey sediments and is conﬁrmed by the presence of a¹³⁷Cs peak between 60 and 65 cm depth, attributed to the Chernobyl accident of 1986, and a sedi- mentation rate of about 2 cm/year, as indicated by ²¹⁰Pb data (Donders et al., 2012; Seitaj et al., in press). One out of the many impacts on the ecosystem was the major decrease in foraminiferal densities.

Concerning the pore patterns, in the 77.5 to 62.5 cm interval, deposited shortly after the closure of the basin, someA. tepidaspecimens with fewer but larger pores appeared. In the two subsequent intervals such specimens became more frequent, whereas specimens with numerous, small pores became increasingly rare, and almost completely disap- peared in the recent living faunas (Fig. 4).

In conclusion, the pore measurements clearly showed the presence of two morphotypes ofA. tepida, with different pore patterns. Initially, only one morphotype, with numerous small pores was present. After the closure of Lake Grevelingen in 1971, specimens with fewer but larger pores appeared (from the 62.5 to 77.5 cm level on), and progressively replaced theﬁrst morphotype. The living faunas sampled in 2012 consisted almost entirely of specimens with large pores (Fig. 4).

Our newly developed pore measurement method yields results describing this trend very precisely. The question whether this ecological replacement is related to the onset of seasonal hypoxia/anoxia or another environmental parameter (e.g., salinity and hydrodynamism), cannot be clariﬁed sufﬁciently at this point and is the subject of ongoing studies.

Table 4

Comparison of pore parameters for the specimens from three stations in Lake Grevelingen, sampled in May 2012. Mean + SD values for all pore parameters and maximum diameter of specimens. Statistical comparison with ANOVA model and Student'st-test with Bonferroni correction (post-hoc test). Signiﬁcant results in bold, for signiﬁcance levelα= 0.05.

Grev-1 Grev-2 Grev-3 All stations Grev-1 vs Grev-2 Grev-1 vs Grev-3 Grev-2 vs Grev-3

Mean + SD Mean + SD Mean + SD ANOVAp-value Student'st-test

p-value

Student'st-test p-value

Number of pores 22.7 ± 5.4 15.5 ± 3.8 16.9 ± 3.1 b0.0001 b0.0001 b0.0001 0.710

Pore area [μm²] 5.87 ± 1.65 6.57 ± 1.63 6.24 ± 1.50 0.338 0.430 1.000 1.000

Porosity [%] 22.6 ± 3.7 17.4 ± 3.6 18.3 ± 3.3 b0.0001 b0.0001 b0.0001 1.000

Specimen diameter [μm] 304 ± 57 378 ± 58 386 ± 39 b0.0001 b0.0001 b0.0001 1.000

(10)

3.3. Interdependence of measured pore parameters

Since the development of this new pore measurement method, we have analysed a total of 870 specimens ofA. tepidafrom different European coastal settings. The dataset contains both living and fossil specimens. Fossil specimens originate from the sediment core of Lake Grevelingen (n= 149), shallow marine stations from the Adriatic Sea (n= 37;Jorissen, 1987) and from a sediment core taken at the mouth of the Loire estuary (n= 6). Living specimens (veriﬁed by CTG or pseu- dopodial movement) were sampled in the Netherlands: Lake Grevelingen (n= 434), Texel (n= 6), Westerschelde (n= 29) and Oosterschelde estuaries (n= 103), as well as at the French Atlantic coast: Loire estuary (n= 38), Baie de Bourgneuf (n= 35) and Baie de l'Aiguillon (n= 33).

Fig. 6shows the number of pores (Fig. 6A) and the porosity (Fig. 6B) as a function of the pore area for the entire dataset (for easier

differentiation, fossil and living specimens are represented by different symbols). For both graphs there is a strong significant correlation for fossil and living specimens taken together, with a power function yield- ing high determination coefficients and low estimations of error comparing thefitted curve to the data (R²= 0.894, forFig. 6A andR²= 0.683, forFig. 6B, mean square error = 0.046 for both). The data clearly fall apart into two rather well separated assemblages. Afirst assemblage (group 1 inFig. 6A), which regroups most of the living individuals, includes specimens with ~ 10 to ~ 40 pores per 562μm²measurement frame and a pore area between 1 and 17μm². The second assemblage (group 2 inFig. 6A) includes mainly dead specimens with ~50 to ~150 pores per 562μm², and an average pore surface of 0.3 to 1.75μm². In Fig. 6B the total porosity (a dependant variable, the product of the number of pores and pore area) is plotted in the function of the pore area.

The same two assemblages can be recognised, but with an overlap in porosity, i.e., group 1 with a porosity of ~10 to ~30% and group 2 with a porosity of ~2.5 to ~15%.

Groups 1 and 2 correspond to the two different morphotypes observed in the core of Lake Grevelingen. It is surprising that inFig. 6A, the individual measurements of both groups are aligned along the same curve indicated by a power function. This strongly suggests an un- derlying physical constraint controlling the relationship between number of pores and pore area.Bé (1968)found similar relationships between the pore density and pore diameter for recent planktonic foraminiferal species. The fact that the same relationship is observed in planktonic and benthic foraminifera underlines that a comprehensive biomechanical model controls the porosity of shells.

The higher porosity in group 1 is the consequence of the much larger pore size, which is only partly counterbalanced by the decreased pore density. In other words, the increased porosity of Table 5

Statistical tests for the pore parameters of all specimens from the sediment core presented inFig. 5. Line 1: results for Kruskal–Wallis test comparing all intervals (p-value). Lines 2–4:

post-hoc analyses with Wilcoxon–Mann–Whitney test (p-value) for comparison of parameters from adjacent intervals. Significant results in bold, for significance levelα= 0.05.P-values of comparison between non-adjacent intervals yielded systematically significant differences.

Number of pores Pore area [μm²] Porosity [%]

Kruskal–Wallis, all intervals 0.001 0.0001 0.001

0–22.5 cm vs 32.5–45 cm 0.876 0.735 0.561

32.5–45 cm vs 62.5–77.5 cm 0.023 0.004 0.008

62.5–77.5 cm vs 83.5–90 cm 0.273 0.457 0.852

Fig. 5.Distribution of pore parameters (number of pores, pore area and porosity) for 4 composite depth intervals in a sediment core from Lake Grevelingen, station Grev-1.

(11)

group 1 (Fig. 6) is a consequence of much larger but fewer pores. The data suggest an upper threshold of porosity at about 30% (Fig. 6B). Fossil assemblages contain both morphotypes (as shown in the core of Lake Grevelingen,Fig. 5), whereas living individuals cluster almost all in group 1. Except for a single specimen in Lake Grevelingen, living specimens with small pores were only found at the site Zandkreek in the Oosterschelde.

4. Conclusion

In this paper we present a new method to accurately measure the pore parameters ofAmmonia, a dominant benthic foraminiferal genus of temperate climate zones, capable of living in a wide range of different coastal environments. Our method is based on the semi-automated measurement of pore parameters (number of pores, pore area, and Fig. 6.A: Number of pores (per measurement frame of 562μm²) as a function of pore area. Indication of two distinct groups 1 and 2. B: Porosity as a function of pore area. Different symbols indicate fossil specimens (from sediment cores) and living specimens (sampled between 2011 and 2013) for stations in European coastal areas. MSE represents the mean square error.