Computer analysis of benthic foraminiferal associations in a tidal New Zealand inlet

Census data on benthic foraminiferal tests in 45 surface sediment samples from Pauatahanui Inlet, Wellington, New Zealand, are analysed by Correspondence Analysis and Non-Hierarchical classification techniques. The faunas are grouped into 7 associations: (A) Trochamminita irregularis/Miliammina fusca - at high tide level in a small tidal creek at the limits of salt water influence; (B) Trochammina inflata/Jadammina macrescens - in an extreme high tidal pool, close to the mouth of a small stream; (C) Miliammina fusca/Haplophragmoides wilberti/Trochammina inflata - intertidal and shallow subtidal (to 0.6 m depth), muddy sand over a large area in the upper reaches of the inlet, where most freshwater runoff enters; (D) Elphidium excavatum/Miliammina fusca - intertidal muddy sand associated with shelly beaches on the fringe of association C; (E) Ammonia beccarii/Haynesina depressula - in a wide variety of intertidal and shallow subtidal (to 3 m depth) sediments that form a belt between the more brackish associations (A–D) and the more normal salinity associations (F–G); (F) Bolivina cf. translucens/Textularia earlandi/Bolivina subexcavata - in mud to muddy, very fine sand in a shallow basin (1–2.5 m deep) in the middle of the inlet and in a small, sheltered backwater; (G) Elphidium charlottensis/Patellinella inconspicua/Quinqueloculina seminula - in sandy mud and muddy fine sand, intertidal to 10 m depth, in the mouth, entrance channel and adjacent outer and middle parts of the inlet, where a flush of normal salinity water enters during each tidal cycle. Using Canonical Correspondence Analysis, the factors most influential in determining the faunal distribution are, in decreasing importance: freshwater influence (salinity), exposure to the air during tidal cycles, proximity to the open sea, tidal current strength and percentage of mud in the substrate.


INTRODUCTION
New Zealand's fossil and modern, foraminiferal faunas from normal marine environments, are well documented, but brackish foraminifera have received little attention.
This paper describes the benthic foraminiferal associations in 45 surface sediment samples from Pauatahanui Inlet (Fig.  1). The samples come from a broad range of shallow environments (high tide to 10 m depth) extending from the extreme limits of brackish influence t o normal salinity. Pauatahanui Inlet is the eastern arm of Porirua Harbour (latitude 41" 05'S, longitude 175" 50'E), on the northwest coast of Wellington region, New Zealand (Fig. 1). Most of the 500 ha inlet is shallower than 2.5 m depth ( Fig. 1) with a wide (100-800 m), intertidal fringe, especially around its upper reaches where a large area of salt marsh is present. Sandy, intertidal shoals cover a large portion of the northwest, lower reaches of the inlet. The greatest depths (3-10 m) are in the narrow, current-swept, entrance channel, which snakes its way for 2 k m in towards the middle of the inlet (Fig. 1). The mouth of Pauatahanui Inlet is filled by a Holocene sand spit, leaving only the 120 m-wide, entrance channel to provide communication with the open sea. Brackish conditions are concentrated around the upper reaches of the inlet, where 95% or more of freshwater runoff enters via five streams, with Horokiwi and Pauatahanui Streams, contributing 75% of this. Sediments (Fig. 2a) Pauatahanui Inlet's central, shallow basin (1-2.5 m) has a muddy floor with sandy mud around its fringes. Muddy to slightly muddy, very fine sand covers much of the northern and southern sides of the middle and lower reaches, becoming fine sand around the upper reaches ( 0 -l m depth). The intertidal and shallow subtidal shoals in the northwest are composed of similar, slightly muddy, fine sand. Pebbly and shelly sand forms the beaches around many parts of the inlet and shelly fine sand lines the entrance channel and seafloor outside its mouth.

METHODS
Samples 1-34 were collected by New Zealand Oceanographic Institute scuba divers, who scooped the upper 5-10cm of bottom sediment into a jar using a small hand shovel (McDougall, 1976). Intertidal samples 35-45 were collected in the same way by the author during low tide.
Foraminifera1 tests (dead plus live) were concentrated by flotation with carbon tetrachloride and divided using a microsplitter to give approximately 100 benthic forms for total assemblage analysis. Any planktic foraminifera encountered were also picked.
Previous studies indicate that 100 specimens provide a sufficiently accurate assessment of faunal composition for use in identifying and mapping associations as the computer  Irwin, 1978) and sample locations in Pauatahanui Inlet (d), an arm of Porirua Harbour (c) on the northwest coast of Wellington region (b), New Zealand (a).
programmes employed are primarily influenced by the dominants in each fauna. The extra work in picking 200 or 300 benthic tests cannot be justified in a study of this sort.
All faunas are held in the Micropaleontology Section of the New Zealand Geological Survey (samples F201992-202037). Figured specimens (Plates 1, 2) have catalogue numbers prefixed by FP. A full list of foraminifera identified in Pauatahanui Inlet is given in Appendix 1.

STATISTICAL METHODS
The data consist of counts of 99 species in 45 samples. Because the number of individuals picked was standardized at around 100 benthic forms, the data are more analogous to species proportions than species densities. The data matrix was standardized by converting counts to proportions of sample totals. The technique of detrended correspondence analysis, using the programme CANOCO (Ter Braak, 1985), was used to summarize the data. A representation of the 45 samples in 4-dimensional space was produced. The co-ordinates of a sample in each of the 4 dimensions are weighted averages of all the species proportions. This representation was used as an input to a technique using an exchange algorithm non-hierarchical clustering, with the distance between the points defined to be Euclidian distance (Banfield & Bassill, 1977), using the program GENSTAT (Payne et al., 1987).
Association score (Table 1) To determine which species characterize each of the 7 faunal associations, the 99 taxa were ranked for each association using a value (association score) calculated to reflect their importance, based on a combination of 5 criteria (modified after McCloskey, 1970 andGrange, 1979): 1. Dominance (Dom). The 10 most abundant taxa of each station in an association were scored with most abundant species given a score of 10, the second most abundant a score of 9, and so on. The dominance of a taxon within an association is given by the mean score across all stations. 2. Fidelity (Fid), or degree to which a taxon is restricted to an association expressed as the proportion of stations within the association in which the taxon occurs less the proportion of stations outside the association in which it occurs. 3. Abundance (Abund), given as the mean abundance of the taxon within the association. 4. Relative abundance (Rel), expressed as the mean a. Sediments LEGEND fine sand (muddy to slightly muddy) very fine sand (muddy to slightly muddy) s a n d y mud mud shelly b. Freshwater influence I(ahoe Stm ,o% Duck Stm 10% :. Surface tidal current velocitv  , 1978). (b) Contoured freshwater influence map of Pauatahanui Inlet. Produced by calculating a value for each station, being the sum of the reciprocals of the distances between the station and the mouths of the seven main streams that enter the inlet, each having been first multiplied by their percentage contribution of freshwater runoff to the inlet (data from Healy, 1980). (c) Contoured surface tidal current velocity map of Pauatahanui Inlet. Produced using the model data of Healy (1980: 163). abundance of the taxon within the association less its mean abundance throughout all the stations. 5. Persistence (Pers), given as the proportion of the stations within the association in which the taxon occurs. The various criteria were weighted and combined to give an empirical Association score for each species in each association, with a maximum value of 100. Association scores were calculated using the formula: 4(0.3 Dom + 2 Fid + 0.11 Abund + 0.08 Re1 + Pers).
Weightings have been assigned to each criterion to make their values more nearly equal but to give greater weight to some criteria in the following decreasing order: Abundance, Relative Abundance, Dominance, Fideiity, Persistence.

Species diversity
Three measures of species diversity have been calculated for each fauna, and also the mean values for each of the associations: 1. Number of species, S, in the fauna.

Shannon-Wiener Information Function, H(s, = C P, log, P,
where is the proportion of the ith species (MacArthur & MacArthur, 1961;Gibson & Buzas, 1973). Unlike S, the Information Function places little weight on rarer and very abundant species. The value of H depends on a combination of the evenness of species counts together to a lesser extent with the number of species present. 3. Evenness, E = e"/S, is a measure solely of evenness of species counts within a fauna, irrespective of the number of species present (Buzas & Gibson, 1969;Hill, 1973).

Environmental factors
To relate the compositions of the samples to the environmental factors observed, the technique of canonical correspondence analysis (Ter Braak, 1987), using the program CANOCO, was used.
The following environmental 'factors' were scored for each station and utilized in the CANOCO analysis to determine which were most likely to be producing the foraminifera1 faunal pattern: 1. Gravelpercentage of sediment composed of gravel fraction. 2. Sandpercentage of sediment composed of sand fraction. 3. Mudpercentage of sediment composed of mud fraction. 4. Depthwater depth below mean spring high tide. 5. Exposureproportion of time exposed above water level during tidal cycles. 6. Open sea-distance of the station from the inlet mouth expressed as a proportion of the distance between the head of the inlet and the mouth. 7. Currentsurface tidal current velocity obtained by measuring the distance travelled by floating confetti during time lapse photography of the filling of a model of the inlet (Healy, 1980: 163), expressed as a proportion of the greatest current strength in the inlet mouth (Fig. 2c). 8. Freshwater influence-calculated as the sum of x / d for each of the seven main streams feeding into Pauatahanui Inlet, where x = t h e percentage of the total stream inflow contributed by stream i (Fig. 2b, from Healy, 1980) and d = distance of the station from the mouth of stream i. Index values have been standardized between 0 and 100 ( Fig. 2b). This method of assessing the influence of salinity is believed to be more appropriate than a set of one-off salinity measurements, the values of which will depend on the state of the tide and the level of freshwater input from the streams.

Wave
where NW = the distance between the station and the inlet coastline in a north-northwest direction (the dominant direction of winds over 10 knots, Healy, 1980: 150) as a proportion of the greatest possible distance; S = the distance between the station and the coastline in a southerly direction (the subdominant direction of winds over 10 knots, Healy, 1980: 150) as a proportion of the greatest possible distance; E = exposure, above. No values for nutrients or oxygen were available, but both are generally strongly correlated positively and negatively, respectively, with mud content. Table   A plot of the sample scores on the first 2 dimensions of Correspondence Analysis is given in Fig. 3. After examination of the output from non-hierarchical classification, the samples were split into two groupsone brackish, the other closer to normal salinity. The data set was split, and separate non-hierarchical classification clustering was performed on the 9 brackish samples and the 36 near-normal salinity samples. The following 4 brackish associations (A-D) and 3 near normal salinity associations (E-G) were recognized.
This association is recorded from only one station, located at the head of a small tidal creek, and undoubtedly is the lowest salinity sample.
The agglutinated foraminifera Trocharnminita irregularis and Miliarnmina fcisca are the dominant and characterising species. Trochamrninita comprises 60% of this fauna and is extremely rare in all other associations (fidelity = 1). The only calcareous taxon in this association is one specimen of Elphidium advenum.
This association is known from only a few, similar, low salinity environments elsewhere in the worldin South Australia, Trinidad and Brazil (Murray, 1991).
This association is only recorded from one station, a high intertidal pool having periodic low salinity caused by freshwater inflow from an adjacent small stream. Nearnormal salinity would be reinstated at every high tide. The association is dominated by agglutinated Trochammina  inflata (89 "/o) and also characterized by fragile, agglutinated Jadammina macrescens (6%), which occurs only rarely in other brackish associations. This association is the common high marsh fauna occuring above mean high water along the mid-latitude coasts of North and South America (Murray, 1991).
This association occurs around the upper reaches of Pauatahanui Inlet in the area where 90% of the inlet's freshwater runoff enters. Thus salinity here, even at high tide would be lower than nearer the mouth. The association extends from just below low tide level, across the extensive intertidal sandy mudflats into the surrounding salt marsh at high tide level.
Miliammina fusca is dominant in each sample (mean abundance 56%) with Haplophragmoides wilberti and Trochammina inflata the secondary dominants. All three have high relative abundance and fidelity values here. The rare agglutinated Ammotium fragile is recorded almost exclusively from salt marsh sample 35 where it comprises 15% of the fauna.
Upper estuarine environments around the world are mostly dominated by Miliammina fusca. Haplophragmoides (often H. wilberti) is abundant in association with Miliammina in some Australian, European, Caribbean and South American estuaries (Murray, 1991).
This association occurs in two mid-tidal stations near the foot of the somewhat shelly, sandy beach that forms the fringe of the inlet along much of its northeastern side. Salinity would be similar to or fractionally higher than association C.
Elphidium excauatum is the dominant species in both samples (mean abundance 63%) with Miliammina fusca the secondary dominant. Trochammina inflata and Ammonia beccarii are also common.
This association is common worldwide in middle estuary environments (Murray, 1991).

Association E -Ammonia beccariilaaynesina depressula
Depth : 3 m to high tide. Sediment : mud to slightly muddy, fine sand and slightly muddy, sandy shell gravel.
This association occurs in a belt around the margins of the outer and middle parts of the inlet and as a wide zone across the inlet between the more saline outer and middle parts and the brackish shallow upper reaches (Fig. 4). Ammonia beccarii is dominant throughout the association (30-90% of the fauna in samples, mean 55%). Haynesina depressula is usually the secondary dominant (mean 12% abundance). Common associated taxa are the dominants of surrounding associations, e.g. Elphidium excavaturn, E. charlottensis, Bolivina cf. translucens, Textularia earlandi.
Associations dominated by Ammonia are common worldwide in similar lower estuary environments (Murray, 1991).
This association occurs in the middle of the inlet (3 stations) and in a backwater to one side of the mouth and entrance channel of the inlet. All samples are in very fine sediment in quiet parts of the inlet, probably with lowered oxygen concentration in the surface sediment.
The characterizing taxa together with Ammonia beccarii are the four dominant species in this association. Other common taxa that mainly occur in this association (high fidelity and relative abundance) are Trochammina sorosa, Astrononion novozealandicum, Siphouvigerina glabra,

Reophax arctica, Rotaliammina bartrami and Fissurina lucida.
This association is unusual. Textulari earlandi i s a common constituent of near normal salinity lower estuarine associations worldwide but seldom in association with Boliuina (Murray, 1991).

Association G -Elphidium charlottensis/PateNinella inconspicualQuinqueloculina seminula
Depth : high tide to 10 m. Sediment : fine sandy mud to shelly, slightly muddy, fine sand. This association occurs in the mouth, entrance channel and central portion of the outer and middle parts of the inlet. It occurs in fine sediment at all depths in both strongly current swept localities and quieter backwaters. All are clustered around the mouth and entrance channel which provide a fresh flush of normal salinity water into the inlet during each tidal cycle.
The characterizing species are E[ph~dium charlottensis, Patellinella inconspicua and Quinqueloculina seminula, mainly because of their high relative abundances in this association. The faunas are dominated (5-15% each) by these three characterizing species, together with Ammonia beccarii, Haynesinu depressula and Elphidium aduenum.
The latter three are codominants in adjacent association E.
Many species that occur commonly in shallow, normal marine environments, are mainly confined to this association in the present study area. the dominants in shallow water normal salinity environments in middle latitudes worldwide, but Patellinella has not been recorded as an additional characterizing species outside of New Zealand (Murray, 1991).

Planktic foraminifera
The tests of planktic forms comprise 0 to 4 YO of the total foraminiferal faunas. Planktic forms are only present in the more saline associations E-G suggesting that they have been carried by tidal currents into Pauatahanui Inlet from more open waters. (Table 2 There is a marked jump up in diversity then to the more saline (and most mixed) associations F and G. The two most saline associations near the inlet mouth (F and G), have the highest evenness scores (E) reflecting the more even spread of dominance of their diverse faunas. Associations E and B have the lowest evenness scores reflecting the high dominance of Ammonia beccarii and Trochamrnina irzfluia, respectively. These overall trends of increasing diversity and increasing evenness from brackish to saline and from intertidal to deeper water are similar to those observed elsewhere in the world and are useful guides in palaeoenvironmental interpretation of extinct shallow or brackish water foraminifera] faunas.

ARE THE ASSOCIATIONS REAL?
Are the seven benthic foraminifera] associations recognized here (Fig. 4), distinctive faunal entities likely to recur many times in nature, or are they artificial divisions applied to a gradually changing faunal pattern? Are these associations recurring combinations of species that have considerable The authors of any study such as this, producing putative associations, must ask whether their suggested associations have any general validity or are they merely artefacts of the particular data set under consideration.
Non-hierarchical clustering techniques attempt to minimize some criterion or function of the distance(s) between points within each cluster. They are iterative, refining the inital clustering and offer no guarantee that their output represent a global minimum of the clustering criterion. Frequently different initial assignments of stations to clusters produces output clusters which also differ. This nondeterminancy can be used to test the robustness of any reported clusters or associations. If many differing initial random subdivisions of the stations are input to the method, the resulting outputs can be compared. It was found that all initial subdivisions produced very similar patterns with only a few stations (near the boundaries of E,F and G) not consistently falling into particular associations. Thus the seven associations are not a purely random subdivision of a continuum, but the most natural subdivision with the most distinctive boundaries that can be determined using the data available. Even so, quick perusal of the characterizing species lists (Table 1) shows that each association is not taxonomically discrete and that adjacent associations often grade into each other, sharing many species in common.
Examination of the individual distributions of the more common species in Pauatahanui Inlet (Fig. 6) shows them to have non-random distributions apparently related to combinations of physical factors. Several patterns are apparent. Some widely distributed species have relatively low background densities but several clumps of much greater abundances (e.g. Ammonia beccarii, Bolivina cf. translucens, Buccella frigida, Elphidium advenum, E. excavatum, Haynesina depressula). Several other widespread species have a single peak area with abundances decreasing gradually away from it (e.g. Elphidium charlottensis, Miliammina fusca, Patellinella inconspicua, Quinqueloculina seminula). Other species with more restricted distributions have one or more high density peaks (e.g. Bolivina subexcavata, Trochammina inpata, Trochamminita irregularis, Textularia earlandi). Other less abundant species have their total distribution restricted to certain parts of the inlet and are never faunal dominants (e.g.

Haplophragmoides wilberti, Jadammina macrescens).
Although the distribution of n o two species exactly coincide, there is considerable approximate coincidence, especially approximate coincidence of peak densities. In places there is also consistent peripheral overlap of species' distributions that appear to occur in similar environmental conditions. It is these coincidences and overlaps that generally produce the recognized associations.
In the computer analysis it was found that the strongest subdivision was two-fold into a brackish group (associations A-D) and a more saline group (E-G). This distinct break is clearly seen in the distribution patterns of individual species (Fig. 6). Five of the more common species are almost exclusively confined to the brackish group of associations; ten are almost exclusively confined to the more saline group and only Elphidium excavatum straddles the boundary. Within the brackish group, associations A and D are each primarily produced by the abundance peak of only their dominant species (Trochamminita irregularis and Elphidium excavatunz respectively) whereas associations B and C are both produced by the coincidences of the abundance peaks of their two characterizing species with highest association scores (B -Trochammina inpata, Jadammina rnacrescens; C -Miliammina fusca, Haplophragmoides wilberti).
Boundaries between the more saline associations (E-G) are more gradational but the associations are defined by the abundance coincidences and overlaps of a greater number of species than in the brackish group. Association E is produced by the abundance peak of Ammonia beccarii and its approximate coincidence with peak abundances of Haynesina depressula. These overlap with the peripheral Thus these associations are distinctive faunal entities, although not discrete as they grade into neighbouring associations. The 7-fold subdivision adopted is the most natural obtainable from the data. It is obvious that the fauna is not distributed randomly nor in a uniformly changing continuum and thus the associations are not purely artificial subdivisions. Recognition of associations does tend to obscure the individual distributions of species and the natural gradations from one association into the next.
Within Pauatahanui Inlet, most recognized associations (except A, B and D) have considerable spatial and mappable extent and their distributions appear to be related to consistent abiotic factors. All associations (except F) are characterized by combinations of species that are known to recur in similar environments elsewhere in New Zealand, e.g. Brook et al., 1981 ( G ) ; Gregory, 1973 (C,E); Hayward, 1981 (B,G); Hayward & Hollis, in press (A-E).
Thus in most respects we consider these associations to be real and their recognition and mapping to be one of the best ways of portraying and understanding the natural faunal pattern.

RELATIONSHIP OF ASSOCIATIONS TO ENVIRONMENTAL FACTORS
The relationship between the sample (via the species) scores and the environmental factors is displayed on the biplots (Figs 7, 8) produced using the Canonical Correspondence Analysis (Ter Braak 1987). The length of the arrow, indicating the axis and direction of increasing values of an environmental factor, is a measure of the correlation between that factor and the faunal pattern. It was found that the three factorsproportion of gravel, sand and mudwere highly correlated. as expected. Consequently the fitted relationships between the species (and hence the sample  (Fig. 3) with the seven associations and the axes of the more highly correlated environmental factors (produced by Canonical Correspondence Analysis) superimposed. The dimensions plotted jointly explain 55% of the variation in the data. Fig. 8. Two-dimensional configuration of samples in the more saline associations (E-G) produced by Detrended Correspondence Analysis with the three associations and the axes of the more highly correlated environmental factors (produced by Canonical Correspondence Analysis) superimposed. The two dimensions plotted jointly explain 62% of the variation in the data. scores) and these variables were unstable. Only the factor, mud, was retained for subsequent analysis. The biplot derived from the full data set is shown in Fig.  7. The strongest associations are with the environmental factors, open sea, exposure and freshwater, and these three factors appear to be largely responsible for producing the pattern of associations. Freshwater (low salinity) and exposure (the proportion of time a station is exposed above the tide level) factors appear to be the two main environmental influences that split the faunas into saline and brackish groups. The more saline associations, E, F and G, have approximately constant values when projected onto the axes for these two factors. These two factors however seem to be responsible for most of the differentiation between the four associations within the brackish group. Association A is the least saline, followed by B and C then D. Associations A and B also differ from C and D in having greater exposure.

Open Sea
The data from the three saline associations (E, F and G) were also analysed separately to determine which environmental factors, if any, differentiate the associations (Fig. 8). The environmental factors having greatest influence appear to be: proximity to open sea, tidal current strength and percentage of mud. Association G is differentiated from E and F by being closest to the open sea, having the highest salinity, the least mud, and generally the strongest tidal currents. Association E is the furthest from the open sea, has the strongest freshwater influence and the weakest currents in this group. Association F appears to be influenced by a combination of high mud content and factors related to greater water depth.
There is considerable residual variation within these saline faunal associations. The biplot (Fig. 8) suggests which environmental factors have the greatest influence on the intra-association variability. It can be seen that faunas in stations 37, 38 and 39 differ considerably from other members of associations E and G. These stations lie in the direction of increasing exposure and are indeed the only intertidal samples in these associations. Other factors that strongly correlate with intra-association faunal variability and may be producing it include mud content (in associations E and F), and tidal current strength (in association G).

POST-MORTEM TRANSPORT OF FORAMINIFERAL TESTS
The strong correlation found between association G and the open sea and tidal current factors suggests that post-mortem transport of foraminifera1 tests is probably a significant factor in determining the composition of the faunas in this association.
Strong tidal currents sweep through the narrow channel at the entrance of Pauatahanui Inlet (Fig. 2c) and have potential to carry in (and out) the tests of foraminifera. Outside the entrance, the combination of strong tidal currents, bottom currents and an exposed coast adjacent to turbulent Cook Strait, makes conditions ideal for the tests of normal salinity foraminifera to be lifted into suspension and transported into the inlet (see Murray et a/., 1982). Association G is best characterized as a mixture of taxa that lived in the area (e.g. Boliuina cf. translucens, Elphidium charlottensis) and those that have been transported in from open sea areas outside the inlet (e.g. Nonionellina flemingi, Cassidulina carinata) and more brackish areas within the inlet (e.g. Ammonia beccarii, Haynesina depressula).
The distinctiveness of the brackish associations and the high fidelity values for the characterizing species in associations A-C (Table 1) suggests that post-mortem transport and faunal mixing is of little consequence around the head of the inlet, where there are low current strengths (Fig. 2c).