Hydrobiologia DOI 10.1007/s10750-014-1817-5 PRIMARY RESEARCH PAPER Modern distribution of ostracodes and other limnological indicators in southern Lake Malawi: implications for paleocological studies Margaret Whiting Blome • Andrew S. Cohen Matthew J. Lopez • Received: 19 February 2013 / Revised: 24 January 2014 / Accepted: 27 January 2014 Springer International Publishing Switzerland 2014 Abstract This modern distribution study from the southwest arm of Lake Malawi quantitatively relates variables of the lake environment to surficial assemblages of ostracodes and other paleoenvironmental indicators (molluscs, Botryococcus, fish, and charcoal) from 102 sites, across a gradient of littoral to shallow profundal conditions. The goal of this research is to use the resultant relationships to help quantify paleoecological interpretations of the fossil record from sediment cores. Site locations varied by depth (1–60 m) and adjacent shoreline environment. Thirty-three ostracode species are identified from 54 sites including four new, undescribed species of Cypridopsinae (2) and Limnocythere (2). Ostracodes are extremely abundant between 1 and 25 m water depth, but are rare to absent between 30 and 60 m. This disappearance is probably taphonomically controlled, with carbonate dissolution in the death assemblage since abundant live ostracodes have been found in the lake at greater than 30 m depth, where bottom sediments lack calcium carbonate. Constrained correspondence analysis (CCA) of ostracode species abundance suggests depth and dissolved oxygen (DO) content to be the primary environmental variables affecting their distribution. Additional CCA models using all biological indicators suggest limnologic variables correlated with depth (e.g., bottom water temperature and DO) and adjacent shoreline environment were most significant. Keywords Ostracodes East Africa Lake Malawi Modern distribution study Paleoecology Introduction Handling editor: Jasmine Saros Electronic supplementary material The online version of this article (doi:10.1007/s10750-014-1817-5) contains supplementary material, which is available to authorized users. M. W. Blome (&) BP America, Houston, TX, USA e-mail: mwblom@email.arizona.edu A. S. Cohen University of Arizona, Tucson, AZ, USA M. J. Lopez Tucson High School, Tucson, AZ, USA Ostracodes are small, bivalved crustaceans with shells of low-magnesium calcite, typically 0.5–3 mm long. Although the soft inner tissue is not often preserved in the sedimentary record, the outer calcite carapace preserves well in most alkaline environments (pH [7.5) (Holmes, 1992; Cohen, 2003). The shape or size of the valve, the presence or absence of nodes, pores and/or reticulations on the outer surface, and internal valve characteristics readily distinguish the species commonly found in East Africa, and in Lake Malawi in particular (e.g., Martens, 2002, 2003). The 123 Hydrobiologia Fig. 1 Site locations for the current study within the southeast arm of the extreme southern end of Lake Malawi. The primary depth transects are labeled A–G, as is the location of the SCUBA transect A4/5. Each cross marks the location of a sample site. Exact coordinates for each site can be found in Table 1. The inset also shows the site locations for a previous study of the ostracodes of modern Lake Malawi (Martens, 2002). Satellite imagery from GeoMapApp (http://www.geomapapp. com) environment in which an ostracode lives is determined by the ecological interactions it experiences within a water body and the geology, hydrology, and climate of the watershed (Delorme, 1969). Ostracode assemblages often reflect lake water chemistry variations with strong correlations to total conductivity and alkalinity (Cohen et al., 1983) and solute composition and concentration (Delorme, 1969; Mezquita et al., 2005). In addition, lake level histories and water depth can be inferred based on modern analogs and ecological tolerances (e.g., Mischke et al., 2002; Horne, 2007). 123 The known relationships between ostracodes and their environment increase with each new study, however, it is important to remember that these associations are strictly demonstrating a correlation and not causation between the two groups of data (Fritz, 1996). Lake Malawi, situated between 9 and 14S, is the southern-most lake in the East African Rift System (EARS) (Fig. 1—inset map). It is a large (29,500 km2), deep (706 m), permanently stratified (meromictic), lake filling the Malawi Rift basin of the EARS. Within the lake, there is both a thermal and Hydrobiologia chemical stratification throughout much of the year (Patterson & Kachinjika, 1995), however, the lake is located far enough south to have seasonal variations in temperature, wind, and precipitation (Eccles, 1974). These variations are caused by the migration of the Intertropical Convergence Zone (ITCZ). The ITCZ is a narrow band of atmospheric convergence and monsoonal precipitation that migrates across the tropics following the location of maximum insolation intensity throughout the year. During the cool, windy season at Lake Malawi (May–August), the ITCZ is sufficiently north of the lake to enhance the SE trade winds at the lake. Sample collection for this study occurred in mid-March specifically to avoid the windy season and capture the variation in lake water chemistry with depth while the lake is stratified. Had these measurements been taken later in the year when the surface waters were well mixed, there would have been no chemical gradient with depth within the sampling depth interval (e.g., Patterson & Kachinjika, 1995—Figs. 1.18, 1.29). A modern distribution study quantitatively relates variables of the lake environment (e.g., water chemistry, physical limnology) to the taxa present at a variety of sites, over a wide gradient of environmental conditions (Gasse et al., 1995; Fritz, 1996). This type of study determines the range of ecological tolerance of individual taxa based on where they are found at present, which then allows a paleoecologist to apply the range determined from modern relationships to fossil assemblage data to interpret paleo-lake conditions. Paleoecological reconstruction from lacustrine fossils is not a new technique. It is a well-tested method that has been used for many years with a variety of indicators. However, it has not yet been quantitatively applied to ostracodes from Lake Malawi. Researchers have long made inferences about lake paleoecology using an understanding of the relationship between species assemblages and modern environmental conditions (e.g., Fritz, 1996). In addition, a number of these relationships have been used successfully to quantify paleoenvironmental change (e.g., diatoms, Gasse et al., 1995; pollen, Vincens et al., 2005). The usefulness of ostracodes as paleoecological indicators has long been understood (Delorme, 1969). As a result, ostracodes have been used across the globe in quantitative paleoclimatic reconstructions from China (e.g., Mischke et al., 2002) to Europe (e.g., Horne, 2007), and the United States (e.g., Smith, 1993). Surface shell collections are time-averaged (in this study each sample represents *15–20 years of deposition), and are therefore more appropriate for comparison with sampled intervals in paleoecological studies than a snapshot sample of live ostracodes. In addition to ostracodes, other biota commonly used as paleoenvironmental indicators include lake fly larvae (chaoborids), molluscs (bivalves and gastropods), fish, green algae (Botryococcus), and diatoms. Chaoborids are often found in lakes with an anoxic bottom layer which they need to escape predation by fish (Baker et al., 1985; Dawidowicz et al., 1990). There were no chaoborids found in this study indicating the deepest sampled depths were oxygenated. Mollusc presence indicates littoral–profundal conditions above a permanent oxicline, similar to ostracodes. Fish fossil abundance in Lake Malawi has been interpreted as indicative of shallow lake depths, or widely fluctuating lake levels (Reinthal et al., 2011). The green algae Botryococcus is commonly preserved in lake sediments as a result of the high silica content in its cell walls (Cohen, 2003). Botryococcus has previously been interpreted as representing a wide range of environments from arid/semi-arid to fresh water conditions, and is thought to out-compete other organisms in rapidly changing environments (GuyOhlson, 1992). Diatoms have been widely used in paleoenvironmental studies because many species have very restricted ecological niches and are therefore excellent indicators of past environmental conditions. In this study, a single diatom (Surirella fuellebornii var. constricta) was frequently observed and therefore counted purely out of curiosity. It has previously been documented in two of the in-flowing rivers (Baka and Songwe) and was assumed to be present in Lake Malawi itself (Cocquyt & Jahn, 2007). Finally charred particles provide a sedimentological indicator of fire frequency in the watershed. Sandsized charred particles (macro-charcoal) cannot be readily transported long distances before being deposited in a basin and therefore likely represents the local fire history, whereas micro-charcoal may be windborne, and therefore more indicative of regional fire history (e.g., Thevenon et al., 2003). There are a few limitations to be aware of when conducting a modern distribution study within a single lake basin, and specifically Lake Malawi. First, there is very limited water chemistry variance within the modern lake (Halfman, 1993; Patterson & Kachinjika, 123 Hydrobiologia 1995). Second, the range of conditions in the modern lake at variable depths only bear a partial relationship to a similar depth range under much lower lake stands and more saline/alkaline lake waters. Third, the environmental gradient lengths are short, and fourth, most of the taxa of modern Lake Malawi are endemics (Martens, 2002), which limits our ability to draw comparative conclusions about the lake under past conditions. A modern distribution study using only the one lake may not be ideal, however, since a number of the ostracode species in the paleo-lake were also endemic (Park & Cohen, 2011; Blome et al., in prep.), a training set approach using multiple modern lakes would produce no modern analog for ostracode species assemblages. The environmental variables chosen for this study are ones we hope to be able to reconstruct to interpret past lake conditions. This study focuses on the modern distribution of both ostracodes and other commonly preserved biota within the nearshore (1–60 m) environment of Lake Malawi, in order to identify the relationship between ostracode species and these other, frequently cooccurring biotic indicators, and the ecological range of each of their habitats. The relationships between these bio-indicators and quantifiable variables not observable in sediment cores (e.g., water chemistry, depth), as well as to variables of physical limnology, which are measurable in a sediment core (e.g., grain size, organic and inorganic carbon content), will inform future interpretations of the paleoenvironment of Lake Malawi (e.g., Blome et al., in prep.). Methods Water chemistry and ostracode sample collection Prior to the March 2010 field season, we acquired detailed bathymetric maps of the Cape MacLear region at the southern end of Lake Malawi from the Department of Surveys—Hydrographic Unit in Malawi (Malawi Government, 2003). Using these maps and Google Earth images of the shoreline, we laid out the locations of numerous depth transects (perpendicular to isobaths). Transects were labeled A–G with A south of Malembo, and G north of Chembe Village (Fig. 1). Transect starting locations were chosen for different types of shoreline environments (Fig. 2): high relief (HR) with coarse sandy 123 beach, low relief (LR) near river inflow, and LR near a marshy environment. Surface sediment samples were collected every 5 m depth along each transect beginning at the shore (1, 5, 10, 15, …, 60 m) for the northern-most transects (F, G). Transects A–E were sampled at 5-m intervals from 1 to 30 m and at 10-m intervals between 30 and 60 m due to time constraints in the field. Additional ‘‘in-fill’’ sampling at shallower depths along isobaths (parallel to shore) most commonly occurred at depths of 5 and 20 m (e.g., Transect ZA50 located near Transect A along the 50 m isobath). Samples from Transects A4 and A5 were collected by hand via SCUBA (table in ES15). At each site, depth was confirmed using a handheld depth finder, and surface water temperature was measured using a Hach pH-meter over the side of the boat. In addition, at all sites up to 25 m, bottom temperature and dissolved oxygen (DO) content were measured using a YSI-83 multi-meter (limited to 25 m by the length of the probe cord) (tables in ESM 16, 17). Previous research has shown specific conductivity and major ion composition in the upper 25 m of the modern lake to be relatively invariant (Halfman, 1993; Patterson & Kachinjika, 1995). For this reason, we chose not to measure major ion concentrations, and only measured conductivity at a handful of sites (four) to compare to previously reported measurements (tables in ESM 16, 17). In addition to temperature, DO, and conductivity measurements, surface sediment samples were collected at each location using a PONAR grab-sampler (see Strayer, 2010—Fig. 5b for example). Arms attached to the grab-sampler limited its penetration into the sediment to approximately 2 cm. There were 20 sites where no sample was collected for a variety of reasons, most commonly due to repeated improper closure of the PONAR sampler. These sites are not included in Table 1. In addition, sites 58–77 were rock surface scrapings taken by SCUBA sampling which yielded insufficient sediment for this study, and are therefore also not included in Table 1. The total number of sites presented in this study is 102, with a relatively even spread over the different shoreline environments and across the depths of interest (Table 1). Sample preparation: wet-sieving, laser diffraction granulometry, and loss-on-ignition (LOI) Approximately 16–20 grams of wet sediment from each sample was split into two aliquots of 12–16 and Hydrobiologia Fig. 2 Categorization of different offshore environment transects. Sites are divided into ‘‘High Relief—Sand,’’ ‘‘Low Relief—River,’’ and ‘‘Low Relief—Marsh.’’ The first part of the category describes the topography inland of each transect, and the second part of the label describes the immediate shoreline environment: coarse sandy beach, river inflow, and marshland, respectively. The images to the left typify each subenvironment. The middle panel is a digital elevation model (DEM) map (http://www.geomapapp.com, Ryan et al., 2009) of the study area with the main transects illustrated as solid lines from shore to their end at 60 m water depth. Extending onto land from each sampled transect is a dotted line showing the location of the topographic transects shown in the panel on the right. Vertical exaggeration (VE) is 15 for transects A–D (‘‘Low Relief’’), and VE = 3 for transects E, F, and G (‘‘High Relief’’) 4 g. The 12–16 g split was passed through two wiremesh sieves, the first with a mesh size of 2 mm, the second with a mesh size of 63 lm, yielding sediment splits of [2 mm (gravel), 2 mm–63 lm (sand), and \63 lm (silt and clay). The two larger fractions were dried at low temperatures and weighed to determine the percent of each grain size within a sample. The smallest grain size fraction was further prepared by mixing the sediment with 10 ml sodium hexametaphosphate (25 g/500 ml DI water) to insure complete disaggregation of particles prior to grain size distribution analysis by laser diffraction. This last analysis was used to obtain the fraction of silt and clay-sized particles in each sample. The 4 gram aliquot was subjected to sequential heating to estimate the water, organic, and inorganic carbon content of the sample (Dean, 1974; Heiri et al., 2001). The samples were first heated for 24–48 hours at 60C, then weighed to get a dry weight (DW60) and to estimate water content. This was followed by 4 hour burn at 550C to combust the organic matter in the sample (DW550), followed by a short, high-temperature burn at 1,000C for 2 hours of the residual ash to evolve carbon dioxide from the inorganic carbonate in the sample (DW1000). The following calculations were used to determine (1) percent water content to estimate initial dry weight of the sediment, (2) LOI550, which 123 Hydrobiologia Table 1 Total number of sites sampled by depth (above) and shoreline environment (below), number of sites with more than five total ostracodes, and the percentage of total sites with more than five total ostracodes Depth # sites total # sites [5 cods 1 5 2 5 18 17 10 9 9 % sites [5 cods 40 94.44 100 15 6 5 20 20 17 85 83.33 25 30 4 6 1 0 25 0 35 2 0 0 40 9 0 0 45 2 0 0 50 11 0 0 55 2 0 0 60 8 0 0 Shoreline environment # sites total # sites [5 cods % sites [5 cods HR—sand 32 10 31.25 LR—marsh 32 18 56.25 LR—river 38 23 60.53 Olympus SZH binocular microscope. Ostracode identification followed multiple authors’ work including: Sars (1910, 1924), Klie (1944), Howe (1955), Neale (1979), Van Harten (1979), Kempf (1980), Martens (1986, 1988), Rossetti & Martens (1998), and Park & Cohen (2011), with the bulk of the Lake Malawi endemic species identified using Martens (2002) and (2003). In addition, the wet-sieved residues were counted for absolute abundance of other common fossil remains in this same size range including: charred particles, fish (comprising bones, teeth, and scales), mollusc fragments, and the relative abundance of algae (Botryococcus) measured as a percentage of sample material. As each sample contained a different mass of sediment analyzed for a number of reasons from sampling to analysis, each sample was weighed prior to analysis and the ‘‘weight picked’’ was recorded. In this way, taxa abundance can be compared across samples by their density (number/grams dry sediment) (e.g., Cohen et al., 2007). Fossil identifications were confirmed using a Hitachi 3400 N SEM and compared to previously described modern material. Rarefaction analysis: cross sample comparison of species richness represents the percent weight loss due to combustion of organic matter, and (3) LOI1000, the percent weight loss due to combustion of inorganic carbonate (Heiri et al., 2001). % water ¼ 1 ððInitial wet sediment weightÞ=DW60 Þ; ð1Þ LOI550 ¼ ððDW60 DW550 Þ=DW60 Þ 100; ð2Þ LOI1000 ¼ ððDW550 DW1000 Þ=DW60 Þ 100: ð3Þ Wet-sieved sand-sized residue analysis Wet-sieved residues were counted for total number of ostracodes per sample, up to *300 valves, although only five of the 102 samples contained [300 valves. All other samples found to contain ostracodes were counted in their entirety. Noted in each sample were the species and taphonomic condition (% broken, whole carapace, adult) of each valve using an 123 In order to compare species richness across sites with differing total abundances of ostracodes, we needed to account for rarefaction: as abundance increases, so does the likelihood of finding more species in a given sample. Rarefaction analysis was conducted using the computer program ‘‘psimpoll’’ (Bennett, 2007). This type of analysis estimates the number of taxa (t) one would expect to find in a random sample of individuals (n) taken from a larger collection of individuals (N) containing T taxa (Birks & Line, 1992). N and T are the actual abundances and number of species found in a given sample, respectively, and n is the minimum number of individuals chosen by the researcher and can be based on a number of factors. For our analysis, we chose 25 in order to have more individuals than were found in a single sample and to maximize the number of samples included in the analysis. All samples with fewer than 25 individuals were excluded from this analysis, yielding a total of 35 sites with a combined 33 species. Hydrobiologia Constrained correspondence analysis (CCA) Results We used CCA to better understand the patterns of variability in the biological data that are accounted for by the measured environmental variables. This method is an extension of correspondence analysis (CA); whereas CA finds the best theoretical environmental gradient that maximizes the separation of species with unimodal distributions (ter Braak and Prentice, 1988), CCA constrains that axis to be a linear combination of actual environmental variables (Juggins, 2009; Oksanen, 2004). CCA is considered extremely robust to noisy data and is not very susceptible to outliers (Juggins, 2009). Therefore, assuming the biological assemblages are controlled by the environment, sites that cluster in ordination space will have similar species composition and come from a similar environment, and vice versa (Birks, 2003). A ‘‘full’’ CCA model will analyze the amount of variance constrained using ALL measured environmental variables, however, for reasons of parsimony and practicality it is prudent to find a minimal adequate model. This ‘‘reduced’’ CCA model is the one that explains (both ecologically and statistically) the biological data (almost) as well as the full model (Oksanen, 2004). We simplified our models in two steps; first by testing the significance of each individual environmental variable in the full model, and then by specifying those remaining variables as co-variables and asking whether each remaining variable explains any significant portion of variance once the effect of the other variables has been taken into account (Juggins, 2009). Permutation tests on both the full model and the reduced model were run to test the significance of each. The following constraints were applied to maintain a high level of robustness in our analysis: (1) sites needed to have greater than 5 total ostracodes to be included and (2) species had to be present in more than one site to be included. Following these criteria, seven sites (Nos. 4, 8, 17, 22, 30, 31, and 124) and three species (Cypridopsine sp.A, Limnocythere s.l. sp.8, and Limnocythere sp.B n.sp.) were removed from analysis. Prior to analysis, biological percentage data were arc-sine transformed, count data were log transformed, and environmental data were standardized to normalize the distribution of the data. The CCAs and significance testing were performed using the ‘‘vegan’’ package and the program R, version 2.12.0 (R Development Core Team, 2010; Oksanen et al., 2011). Raw limnological results can be found in the Electronic Supplementary Material and their implications as environmental variables for the distribution of ostracodes and other biota are discussed further in the results of the CCA analysis. Photographs of each species found in this study can be found in figures in ESM 1–8. Coarse residue counts: total abundances of sedimentological and biological indicators (charred particles, ostracodes, fish, molluscs, and Botryococcus) These data are presented in terms of abundance per gram of dry sediment to provide a density of each of these taxa per sample and allow their abundances to be comparable across all samples taken (tables in ESM 16, 17). Sites offshore from the river inlet typically have orders of magnitude greater charred particle abundance than the other two sites between 1 and 30 m water depth (figure in ESM 9). In addition, the highest proportion of sites containing any charred particles (regardless of abundance) is from the offshore river group at most depths. Total abundances of ostracode valves, fish bones, and molluscs, and relative abundance of Botryococcus colonies are shown by depth in table (ESM 17) and Fig. 3. Data are plotted on a logarithmic scale for the absolute abundances as they differ from site to site by orders of magnitude. There is a marked difference in indicator abundance beginning between 20 and 25 m. Above those depths ostracodes, molluscs and to a lesser extent, fish remains are all quite abundant, with ostracodes being the most abundant remains at each shallow depth. At 25 m, both the frequency of occurrence and total abundance of ostracodes decrease precipitously. From 30 to 50 m, ostracodes are found at only four sites, and never in densities greater than three valves/gram sediment. Below 30 m, fish bones become much more abundant. Botryococcus does not occur above 20 m, and gradually increases in relative abundance until 50–60 m where its abundance is high, though with significant variability. Like ostracodes, molluscs are quite abundant, and present at numerous sites in water depths less than 25 m. Below that depth they become less abundant, and occur in fewer 123 Hydrobiologia LOG Total Count/gram sediment Ostracodes Molluscs 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 80 Botryococcus 2.5 70 2.0 60 1.5 50 1.0 40 0.5 30 0.0 20 -0.5 10 Percent (%) LOG Total Count/gram sediment Fishbone 0 -1.0 0 10 20 30 40 50 60 Depth (m) 0 10 20 30 40 50 60 Depth (m) Fig. 3 Biological indicator abundance plotted against depth, with each indicator on its own labeled panel (Ostracodes, Molluscs, Fishbone, and Botryococcus). Absolute abundances of ostracodes, molluscs, and fishbones are plotted as the LOG abundance/gram sediment counted (left-hand axis). The relative abundance of Botryococcus is given as a percentage of the total sample (right-hand axis) samples overall, although they are present at each sampled depth below 25 m, unlike the ostracodes. found at depths greater than 30 m have some of the thickest shells (see figures in ESM 4A–G, 6J–L). Six species had limited depth ranges including Cypridopsine n.gen. sp.A and Limnocythere s.l. sp. 6 (5 m), Limnocythere s.l. sp.3 and Cypriodpsine n.gen. sp.X n.sp. (5–10 m), and Limnocythere s.l. sp.B n.sp. and Gomphocythere irvinei (20 m). Multiple (13) species were quite ubiquitous and found at multiple depths across the entire depth range from 1–5 to 20–25 m and are marked in Fig. 4 with a star. In order to compare species abundances across multiple sites of the same depth and/or offshore environment, average abundances were calculated for each species at each depth and within each environment. This involved calculating the average abundance of each species found within each subdivision of either depth or environment (e.g., (total 9 species at 1 m depth)/(total ostracodes at 1 m depth), then repeat Coarse residue counts: ostracode species abundances A complete list of all species identified in each sample (including their relative and absolute abundances, and the number of species per sample) can be found in table in ESM 18. The depth range of each species found in this study is illustrated in Fig. 4. For taxonomic references see Blome et al., in prep. A list of total number of samples in which each species was recovered and an index of ESM figure number for each species can be found in Table 2. Figure 4 illustrates the precipitous decline in ostracode abundance below 20 m depth, as was alluded to in the previous section. The three species 123 Hydrobiologia Fig. 4 Depth range of each ostracode species found in this study. Species are sorted top to bottom first by median (dark bar across each box plot), then by average. For example, the first nine species all have median of 5 but the first two have the lowest average and the last two have the highest. The width of the box plot is proportional to sample size—those species only found at one or two sites have the smallest width. Species labeled with ‘‘asterisk’’ were quite ubiquitous and found at multiple depths across the entire depth range from 1–5 to 20–25 m Limnocythere sp.8 Cypridopsine sp.A Limnocythere sp.3 Limnocythere sp.6 Cypridopsine sp.1 Gomphocythere huwi Limnocythere sp.A Limnocythere sp.2 *Cyprinotinae sp.2 Cypridopsine sp.N1 *Limnocythere sp.10 *Cypridopsine sp.P Cypridopsine sp.C *Cyprinotinae sp.1 *Cyprinotinae sp.3 *Cypridopsine sp.L *Limnocythere sp.9 Candonopsis sp. *Cypridopsine sp.K Cypridopsine sp.G Cypridopsine sp.D *Zonocypris costata *Alicenula serricaudata *Gomphocythere emyrsi Cypridopsine sp.O Cypridopsine sp.B *Limnocythere sp.1 *Ilyocypris sp.1 Cypridopsine sp.F Limnocythere sp.4 Cypridopsine sp.N Gomphocythere irvinei Limnocythere sp.B Shallower O O O O Intermediate O O Deeper O 0 at 5 m depth, and so on) (Table 3). One sample exceeded the rarity requirements at 25 m depth, but because there were no other samples with which to average it, this sample is not included in the average depth dataset (Nsites = 50, Ncods = 6,328). However, the data from that site are included with the average abundance by environment dataset (Nsites = 51, Ncods = 6,361). Pooling sites of the same depth or environment allow some trends to be more visible than looking at each site individually. Those species that comprised over 5% of the total assemblage of a given depth or environment are highlighted in Fig. 5. At 1 m nearly half of the assemblage (Ncods = 27) is unidentifiable due to valve damage, and although eight total species are identified, only four comprise [5% of the total assemblage. This is the only depth at which G. huwi is found in comparatively large quantities. At 5 m the greatest total number of ostracodes are found, nearly 100 times those found at 1 m (Ncods = 2,716). This depth also has the highest number of total species identified (28), however, O 10 O O 20 O 30 40 50 many of those species were comparatively rare; 22 species had cumulative totals less than 5% of the total assemblage. The percentage of unidentified ostracodes decreases to 11% at 5 m and decreases further to 7% at 10 m (Ncods = 1,706) where valve damage is minimal. There are 25 total species identified at this depth with 6 greater than 5%. 15 m depth had the most even representation of species with 8 of 21 total species individually comprising [5% of the assemblage (Ncods = 569). The portion of unidentifiable valves increases to 16% at this depth and increases further to 22% of the total (Ncods = 1,310) at 20 m. There are 26 species found at this depth with seven [5% of the total. This is the only depth where Cypridopsine n.gen. sp.N contributes significantly to the total assemblage. Interestingly, a number of species repeatedly account for [5% of the assemblage, regardless of depth. These include Cyprinotinae n.gen. sp.2, G. emyrsi, and L. sp. 9 (the latter with the exception of 1 m depth). When comparing species composition by adjacent shoreline environment, HR has the highest percentage 123 Hydrobiologia Table 2 Number of sites in which each species was found Species # sites/ species ESM figure Image Alicenula serricaudata 31 6 A–C Candonopsis sp. 13 6 D–E 1 1 A–C Cypridopsine n.gen. sp.A Cypridopsine n.gen. sp.B 5 1 D–E Cypridopsine n.gen. sp.C 12 1 G–I Cypridopsine n.gen. sp.D 3 1 F Cypridopsine n.gen. sp.F 11 1 J–K Cypridopsine n.gen. sp.G 3 2 A–B Cypridopsine n.gen. sp.K 11 2 C–E Cypridopsine n.gen. sp.L 14 2 F–H Cypridopsine n.gen. sp.N 14 2 I–L Cypridopsine n.gen. sp.O 13 3 A–C Cypridopsine n.gen. sp.P 30 3 D–F Cypridopsine n.gen. sp.X n.sp. 2 3 G–H Cypridopsine n.gen. sp.1 n.sp. Cyprinotinae n.gen. sp.1 5 27 3 4 I–L A–C Cyprinotinae n.gen. sp.2 33 4 D–G Cyprinotinae n.gen. sp.3 36 4 H–J Gomphocythere emyrsi 37 5 A–D Gomphocythere irvinei 2 5 E–G Gomphocythere huwi 7 5 H–J Ilyocypris sp.1 24 6 F–I Limnocythere s.l. sp.1 25 7 A–F Limnocythere s.l. sp.2 10 7 G–H Limnocythere s.l. sp.3 7 7 I–J Limnocythere s.l. sp.4 8 7 K–L Limnocythere s.l. sp.6 5 7 M–N Limnocythere s.l. sp.8 1 7 O Limnocythere s.l. sp.9 37 8 A–C Limnocythere s.l. sp.10 Limnocythere s.l. sp.A n.sp. 20 14 8 8 D–F G–L Limnocythere s.l. sp.B n.sp. 1 8 L 20 6 J–L Zonocypris costata Also included in this list are the references for the ESM figure numbers for each species of unidentifiable valves (24% of Ncods = 508) and also has the least number of total identified species (26), though not by a significant amount. This observation is supported by the rarefaction analysis, which shows that the range of expected number of taxa is similar for all environment types regardless of depth, and that the variance in species richness does not vary significantly 123 with depth between 5 and 20 m (figure in ESM 10). The offshore marsh sites (Ncods = 2,074) comprise 28 total species of which six account for greater than 5% of the total assemblage. The offshore river sites have the highest total number of ostracodes (Ncods = 3,779) and the highest number of species identified (29) with seven accounting individually for [5% of the total. This is the only environment where Ilyocypris sp.1 is found in comparatively high abundance, which may have interesting paleoecological reconstruction ramifications (Blome et al., in prep). A number of species are found in high abundances in all three environments including C. n.gen. sp.2&3, G. emyrsi, and L. s.l. sp.9. This suggests one of two things; either these species would be poor paleoecological indicators of nearshore environment as they have a broad range of suitable ecological niches, or that geographic area is a less significant factor in determining an ostracode’s ecological niche in Lake Malawi, as was suggested by Martens (2002). Constrained correspondence analyses For all further analyses, sites with five or fewer total ostracode valves and species found in fewer than two sites have been removed. Although it is quite informative looking at average abundances of each ostracode species by geographic location and depth, CCA analyzes the abundance of species data at each site in multidimensional space and clusters those sites with similar species assemblages, while using environmental data to constrain the axes (Figs. 6A, B, C). When performing a CCA, there must not be gaps in either the environmental or biological data and the number of sample sites for each must match. As we were limited by equipment to measuring DO and bottom water temperature to 25 m and shallower, we ran three separate CCA models: (1) On ostracode species data alone using all measured environmental variables (Nsites = 37), (2) Using all biological data counted and all measured environmental variables (Nsites = 43—only sites shallower than 30 m could be included), and (3) Using all biological data from all depths but eliminating DO and bottom water temperature as environmental variables (Nsites = 99). We present the results of the three CCA models below, each of which contributes to our understanding of the interaction of ostracode species assemblages with environmental variables and/or other biological data (graphic abbreviations in Table 4). 1.3 Cypridopsine sp.L 2.9 1.4 Cyprinotinae sp.3 8.5 2.9 1.9 2.8 Limnocythere sp.9 Limnocythere sp.10 Limnocythere sp.A Zonocypris costata 7.3 9.3 1.9 2.1 11.3 1.6 1.3 15.6 6.3 0.5 0.7 4.6 7.0 1.2 0.9 5.1 22.4 3.6 0.7 0.7 5.3 0.7 0.3 4.4 12.5 23.7 0.4 0.8 1.4 12.7 0.2 1.0 0.8 0.2 2.0 2.7 0.6 11.4 8.2 10.0 3.7 4.7 1.4 0.6 5.3 0.4 0.4 1.6 0.6 1.6 1.0 2.4 HR Ncods = 508 Nsites = 10 14.5 0.7 3.3 2.4 9.1 0.1 1.0 0.3 2.0 2.3 2.9 0.3 0.4 7.4 9.4 17.6 9.4 0.0 7.5 0.6 0.3 1.7 0.5 0.1 0.3 0.8 1.8 0.3 2.8 LR—marsh Ncods = 2,074 Nsites = 18 10.8 8.0 0.6 1.9 7.1 1.1 0.1 0.8 0.3 1.5 4.8 0.9 0.0 12.1 7.2 14.7 10.9 0.1 3.8 1.1 3.4 1.3 0.7 0.1 1.0 0.8 0.2 0.2 0.3 4.2 LR—river Ncods = 3,779 Nsites = 23 Total number of ostracodes and sites included in each pooled calculation is given in the column heading Ncods and Nsites, respectively. Abundance values of 5% or greater are shown in bold and indicate which species are displayed graphically in Fig. 5 11.1 0.1 Limnocythere sp.6 48.1 0.7 Limnocythere sp.4 Unidentified 0.5 Limnocythere sp.3 0.8 1.8 1.6 3.7 11.1 Limnocythere sp.2 Limnocythere sp.1 Ilyocypris sp. 0.2 9.2 0.7 19.5 2.0 5.4 5.6 1.8 0.7 1.1 17.2 1.6 10.5 6.3 4.6 1.9 8.6 0.3 0.2 0.2 2.2 0.9 0.2 0.2 0.6 8.8 20 m Ncods = 1,310 Nsites = 17 Gomphocythere huwi 5.3 7.4 17.0 9.8 0.2 4.3 0.4 3.9 4.0 0.7 1.2 1.1 1.4 2.8 15 m Ncods = 569 Nsites = 5 Gomphocythere irvinei 7.4 7.4 Cyprinotinae sp.2 Gomphocythere emyrsi 20.1 11.1 Cyprinotinae sp.1 12.8 0.0 12.9 Cypridopsine sp.X 7.3 Cypridopsine sp.P 1.2 0.4 Cypridopsine sp.K 0.1 0.1 0.8 0.1 0.0 Cypridopsine sp.G 0.4 0.3 Cypridopsine sp.F 3.7 0.1 Cypridopsine sp.D 0.7 0.2 Cypridopsine sp.O 0.4 Cypridopsine sp.C 0.6 1.0 10 m Ncods = 1,706 Nsites = 9 Cypridopsine sp.N 0.1 Cypridopsine sp.B 0.1 3.0 1.4 3.7 5m Ncods = 2,716 Nsites = 17 Cypridopsine 1 n.sp. 3.7 Candonopsis sp. 1m Ncods = 27 Nsites = 2 Alicenula serricaudata Species Table 3 Percent abundance of each species by depth and environment Hydrobiologia 123 Hydrobiologia 1m Depth Cyprinotinae 2 11% n = 27 5m Depth Unidentified 11% n = 2716 Cyprinotinae 1 13% Cyprinotinae 3 8% Gomphocythere huwi 7% Unidentified 48% 0 < 5 PCT 22% Cyprinotinae 2 20% 22 species Ilyocypris 11% Gomphocythere emyrsi 5% 4 species Limnocythere 9 9% Cypridopsine P 7% 0 < 5 PCT 15% Unidentified 7% Cyprinotinae 1 10% 0 < 5 PCT 21% 10m Depth n = 1706 15m Depth n = 569 Unidentified 16% Cyprinotinae 3 13% Cyprinotinae 1 6% Cyprinotinae 2 11% Cyprinotinae 2 17% Cypridopsine P 5% 19 species Zonocypris costata 6% 0 < 5 PCT 20% Cyprinotinae 3 8% Gomphocythere emyrsi 17% Zonocypris costata 9% Limnocythere 9 11% 13 species Limnocythere 6 7% Limnocythere 9 5% Ilyocypris 5% Gomphocythere emyrsi 19% Fig. 5 Pie charts showing the most abundant species at each depth and each shoreline environment. Cutoff for display was greater than 5% of the total ostracodes in a zone. All other species are comingled in the wedge labeled ‘‘0 \ 5 PCT.’’ For individual abundances for all species see Table 3 For the ostracode-only CCA, the full model included all 11 measured environmental variables and constrained *40% of the variance with a P value of 0.01 (Table 5). Whereas the reduced model explains only 14% of the variance, only two environmental variables are needed to double the F factor and decrease the P value of the model to 0.005, yielding the minimal adequate model. Figure 6A shows that these significant constraining environmental variables are depth and DO. As would be expected, once fit to the existing constraining axes, bottom water temperature shows an inverse relationship to depth, and increases in the same direction as DO (seen in figure a, b in ESM 11). Many of the ostracodes plotting near the origin of the CCA plot are those that were starred in Fig. 4 for being found across multiple sites at a wide range of depths. Those ‘‘deeper’’ species (Fig. 4) like Cypridopsine n.gen. sp.N and C. sp. F, Limnocythere sp. 1 and L. sp. 4, and Ilyocypris sp.1 plot toward increasing depth, and vice versa for the ‘‘shallower’’ species. Interestingly, although the range of depths investigated was only 1–25 m, depth was still a key explanatory variable. Although dropped from the minimally adequate model, shoreline environment and the other environmental data added post hoc fit the constrained model with P values of better than 0.09, with the exception of surface water temperature (Table 5). 123 Hydrobiologia Cyprinotinae 1 6% 20m Depth High Relief n = 1310 n = 508 Cyprinotinae 2 5% Unidentified 22% Cyprinotinae 2 10% Unidentified 24% Cyprinotinae 3 8% Cypridopsine N 9% Cypridopsine L 5% Limnocythere 9 5% Cypridopsine P 5% Gomphocythere emyrsi 9% 0 < 5 PCT 22% Limnocythere 9 13% 0 < 5 PCT 24% 19 species Ilyocypris 13% 20 species Alicenula 9% Unidentified 15% Cyprinotinae 1 9% Low Relief - Marsh Low Relief - River n = 2074 n = 3779 Unidentified 11% Cyprinotinae 2 18% Gomphocythere emyrsi 11% Cyprinotinae 1 11% Cyprinotinae 2 15% 0 < 5 PCT 24% 0 < 5 PCT 25% 22 species Cyprinotinae 3 7% Cyprinotinae 3 9% 22 species Gomphocythere emyrsi Cypridopsine P 7% 8% Limnocythere 9 9% Ilyocypris 5% Zonocypris costata 8% Gomphocythere Limnocythere 9 7% emyrsi 12% Fig. 5 continued For the shallow-only CCA using all biological indicators, the full model again included all 11 measured environmental variables and constrained *45% of the variance with a P value of 0.005 (Table 6). In this case, the minimal adequate model explains 16% of the variance with two significant environmental variables which nearly doubles the F factor of the full model and keeps the P value at 0.005. Figure 6B shows that these significant constraining environmental variables are bottom water temperature and DO. Unlike the ostracode-only analysis, the minimal range of depths covered here is insufficient to break out depth as a significant factor, likely because much of the difference in depth range between ostracodes and the other biological indicators begins at 25 m (see Fig. 3). However, the two most significant variables are ones which are known to inversely co-vary with depth (Fig. 6; figure a, b in ESM 11). Molluscs (M) plot in ordination space near total ostracode count (Tc) suggesting the environmental range for both indicators overlap. Conversely, Botryococcus (B) and Surirella (S) plot far away from the ostracode cluster. Fish and charcoal plot closer to ostracodes, but may be responding to something other than bottom temperature and DO (see Fig. 6B). Although dropped from the minimally adequate model, the environmental data added post hoc fit the constrained model with P values of better than 0.09 (Table 6). 123 A B 1.5 Gi 1.5 Hydrobiologia B SiltClay 1.0 1.0 CB CG HR A CF CN −0.5 0.5 CCA2 Ge CLRM L4 L10 L9 CO CP Gh CC Zc CT2 Lx CT1 CT3 L2 Gi L6 LRMHR F L4 CT3 M A CT2 CP CO L3 Tc CT1 Lx C L9 L10 CC DO L1 CL CB Cx −1.0 −1.0 CK CD C1 Ge BTemp BTemp Zc L3 LRR Gh STemp Sand L2 I L6 STemp LRR Ch CF I 0.0 DO GrSand CN −0.5 Depth 0.0 CCA2 0.5 CK Cx CD CL L1 Sand GrSand S CG C1 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 CCA1 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 CCA1 C 1.5 C1 Sand 1.0 C 0.5 Gh L3 F Gi Air 0.0 CB GrSand CC HR L1 LRM CO L2 M CP Depth Lx CT3 L9L10 Ge Tc CT2 I OrgC −0.5 CCA2 CL L4 LRR −1.0 S CG CT1 A L6 Ch CK CF B −1.5 CN Zc Cx SiltClay CD −2 −1 0 1 CCA1 Fig. 6 Ostracode species are in red, the most significant environmental variables included in the reduced model are blue and bold, and those that were initially significant are fitted post hoc and are black. Where species labels overlapped, a red dot was placed at the true location, and the labels placed so as to be legible. For index of abbreviations see Table 4. A Ostracode species CCA—depths included 1–25 m, Nsites = 37. Depth and DO were the most significant constraining environmental variables in the reduced CCA model for explaining the distribution of the biological data. Model and variable significance testing results in Table 5. B All biological species CCA—depths included 1–25 m, Nsites = 43. Non-ostracode biological indicators are in red and bold to distinguish them. Bottom water temperature and DO were the most significant constraining environmental variables in the reduced CCA model for explaining the distribution of the biological data. Model and variable significance testing results in Table 6. C All biological species CCA—depths included 1–60, Nsites = 99. Bottom water temperature, DO, and inorganic carbon were excluded from the analysis due to gaps in data. Non-ostracode biological indicators are in red and bold to distinguish them. Depth and shoreline environment were the most significant constraining environmental variables in the reduced CCA model for explaining the distribution of the biological data. Model and variable significance testing results in Table 7 For the ‘‘all-depths’’ CCA using all biological indicators, the full model included seven measured environmental variables and constrained *28% of the variance with a P value of 0.005 (Table 7). In this case, the minimal adequate model explains 15% of the variance with two significant environmental variables which increases the full model F factor and keeps the P value at 0.005. Figure 6C shows that these significant 123 Hydrobiologia Table 4 A listing of all abbreviations used in the CCA plots Table 4 continued CCA Name CCA Name A Alicenula serricaudata HR High relief C Candonopsis sp. C1 CB Cypridopsine 1 n.sp. Cypridopsine sp.B Grouped from top to bottom by ostracode species, other indicators, and environmental variables CC Cypridopsine sp.C CD Cypridopsine sp.D CF Cypridopsine sp.F CG Cypridopsine sp.G CK Cypridopsine sp.K CL Cypridopsine sp.L CN Cypridopsine sp.N CO Cypridopsine sp.O CP Cypridopsine sp.P Cx Cypridopsine sp.X n.sp. CT1 Cyprinotinae sp.1 CT2 Cyprinotinae sp.2 CT3 Cyprinotinae sp.3 Ge Gomphocythere emyrsi Gh Gi Gomphocythere huwi Gomphocythere irvinei I Ilyocypris sp. L1 Limnocythere sp.1 L10 Limnocythere sp.10 L2 Limnocythere sp.2 L3 Limnocythere sp.3 L4 Limnocythere sp.4 L6 Limnocythere sp.6 L9 Limnocythere sp.9 Lx Limnocythere sp.A Zc Zonocypris costata B Botryococcus Ch Charcoal F Fish M S Molluscs Surirella Tc Total ostracodes Air Air temperature Btemp Bottom temperature Stemp Surface temperature OrgC Organic carbon content GrSand Fraction [ sand SiltClay Fraction silt and clay LRR Low-relief river LRM Low-relief marsh constraining environmental variables are depth and shoreline environment. The first CCA axis is likely a depth gradient; with the exception of molluscs, all other biological data (in bold) plot in ordination space well away from ostracodes, with Botryococcus and Surirella the furthest removed. The second CCA axis may reflect more of the characteristics of the offshore environment given the spread along CCA2 of the class variables (HR, LRM, and LRR) and the trend of the ‘‘fitted’’ data. For example, the amount of silt and charcoal both increase significantly in the offshore river environment (figure a in ESM 12, figure in ESM 13), and both plot toward negative CCA2 values as does the LRR categorical variable (Fig. 6B). Although dropped from the minimally adequate model, the environmental data added post hoc fit the constrained model with P values of better than 0.09 (Table 7). Although environmental data measurability forced the running of two separate CCA models to look at the relationships of all collected biological data, the results support one another. Discussion The interaction of numerous ecological variables affects both the presence/absence and the relative species abundances of ostracodes and other biota in modern Lake Malawi. These factors include DO content, water and air temperature, substrate grain size and carbon content (organic and inorganic) measured in this study, as well as other potential factors such as conductivity, local pH, and the habitat preferences of potential predators (e.g., fish—Abdallah, 2003) for which we currently lack data. Figures in ESM 9, 11, and 14 illustrate how each of these factors varied with depth and shoreline environment within the study area. Variables such as bottom water temperature, DO, inorganic carbon content, and fish presence (estimated based on abundance of fishbone in samples—Fig. 3) changed systematically with depth. Others, including silt/clay content, and charcoal 123 Hydrobiologia Table 5 Ostracode-only CCA (N = 37) Ostracode CCA Individual significance Conditional significance Reduced model Variable P value Keep/drop P value Keep/fit Fit P value DO 0.005 Keep 0.025 Keep ** Depth 0.005 Keep 0.08 Keep ** Bottom temp 0.005 Keep 0.29 Fit 0.001 Sand 0.015 Keep 0.14 Fit 0.075 Surface temp 0.03 Keep 0.17 Fit 0.101 [Sand 0.035 Keep 0.36 Fit 0.085 Shore env 0.09 Keep 0.17 Fit 0.067 Silt ? clay 0.25 Drop Organic C 0.25 Drop Inorganic C 0.28 Drop Air temp 0.37 Drop Ostracode CCA Total inertia Constrained inertia % constrained F factor P value Full model 2.4712 0.9997 40.45 1.3588 0.01 Reduced model 2.4712 0.3405 13.78 2.7164 0.005 Data showing the significance testing process for each environmental variable as well as the permutation test statistics for the full and reduced models. The full model includes all variables listed at left. The two-step significance test follows in the next two columns with the statistical P value of each variable and the resultant action taken. The variables with ‘‘**’’ in the final column are the only ones included in the reduced model. The final column evaluates the significance of the fit of each environmental variable to the constrained axes of the reduced model abundance, showed greater variation by environment. Based on the CCA models, three of the four significant environmental variables affecting the distribution of biological indicators in Lake Malawi co-vary with depth, although shoreline environment is also significant when the full biological dataset is used. Ostracode distribution between 25 and 50 m There is a drastic decrease in both the number of ostracode-bearing samples and total ostracode abundance between 25 and 50 m depth. Ostracodes are completely absent from samples deeper than 50 m. An ecological threshold relating to lake water conductivity, temperature, or DO content between 20 and 25 m could cause ostracodes to survive only in the shallowest 25 m of Lake Malawi. The range of conductivities measured at the four sites in this study (235–247.8 lm/cm, table in ESM 16) is well within the range of those previously reported for the upper 100 m of Lake Malawi (Halfman, 1993; Patterson & Kachinjika, 1995), suggesting there is no conductivity threshold crossed in the upper 25 m. The rapidly changing temperature profile in our study area from 29–31 to 123 24–25C over 25 m depth suggests a relatively shallow thermocline at the time of sampling (mid-March), which is supported by 2 years of cruise data showing a similarly shallow thermocline persisting in southern Lake Malawi from December to April (Patterson & Kachinjika, 1995). However, water temperatures below 25C are unlikely an ecological barrier to ostracodes because other researchers sampling Lake Malawi during the same month in 1999 (Martens, 2002) found significant numbers of live ostracodes at depths up to 30 m, where temperatures would presumably have been similar to and likely colder than those measured at 25 m in our study (temperature was not explicitly reported). Although measured DO at 25 m is on average less than that at 20 m, the waters of Lake Malawi do not become anoxic until 200 m (Patterson & Kachinjika, 1995), and therefore too little oxygen is not a likely cause for the decreased abundance of ostracodes beginning at 25 m. For these reasons, it is unlikely that one of these ecological variables is inhibiting the survival of ostracodes at depths below which they were recovered in our study. The other variable that decreases systematically with depth is inorganic carbon in the sediment (figure Hydrobiologia Table 6 All-indicators, shallow-depths CCA (N = 43) Full model: Depth, ShoreEnv, GrSand, Sand, SiltClay, OrgC, InorgC, Air, S.Temp, B.Temp, DO mgL All Bio CCA N = 43 Individual significance Conditional significance Reduced model Variable P value P value Fit P value Keep/drop Keep/fit DO 0.005 Keep 0.02 Keep ** Bottom temp 0.005 Keep 0.03 Keep ** 0.001 Sand 0.005 Keep 0.45 Fit [Sand 0.005 Keep 0.46 Fit 0.001 Silt ? clay 0.005 Keep 0.59 Fit 0.001 Shore Env 0.01 Keep 0.24 Fit 0.003 Surface temp 0.066 Keep 0.17 Fit 0.081 Depth 0.15 Drop Inorganic C 0.2 Drop Organic C 0.29 Drop Air temp 0.46 Drop All Bio CCA N = 43 Total inertia Constrained inertia % constrained F factor P value Full model 1.7398 0.7741 44.49 2.0039 0.005 Reduced model 1.7398 0.2789 16.03 3.8175 0.005 Data showing the significance testing process for each environmental variable as well as the permutation test statistics for the full and reduced models. The full model includes all variables listed at left. The two-step significance test follows in the next two columns with the statistical P value of each variable and the resultant action taken. The variables with ‘‘**’’ in the final column are the only ones included in the reduced model. The final column evaluates the significance of the fit of each environmental variable to the constrained axes of the reduced model b in ESM 14). In this study, with the exception of two sites, samples from depths of 25 m and greater have less than 2 WT% LOI1000. This corresponds to the depth at which pH begins to decrease rapidly from 8.65 between 0 and 25 m depth to 8.0 by 75 m (Patterson & Kachinjika, 1995—Fig 1.26). As carbonate is more soluble in water at lower pH values, it is likely that the disappearance of ostracodes from our record results from a taphonomic bias—dissolution across a lysocline and thus lack of accumulation below the lake’s current carbonate compensation depth (CCD). This would explain why live ostracodes have been recovered from sediment samples taken in 30? meters of water elsewhere in Lake Malawi (Martens, 2002): ostracodes are able to preferentially take in carbonate while living, but once dead, the shells dissolve from the record. This relatively shallow CCD is in contrast to the supersaturated surface waters of Lake Tanganyika (pH 8.7–9.2) where thin-shelled organisms have been recovered from cores taken as deep as the oxycline (*200–300 m), although a CCD exists at greater depth (Alin & Cohen, 2004). This difference in thin-shelled versus thick-shelled organisms may explain why the three ostracode species found at depths greater than 30 m have some of the thickest valves (Fig. 4; figures D–H in ESM 4, J–L in ESM 6), and also why gastropods are still present in the dataset below 30 m, including a single specimen found at 60 m. Smaller, more thinly walled shells (e.g., most ostracode valves) would dissolve faster than larger, thicker shells as the speed of the reaction would be the same for both, making it more likely to find gastropod shells in deeper surface sediments than ostracodes. In summary, the data suggest that the sudden decrease in the number of ostracodes recovered from sites deeper than 25 m may be the result of postdepositional dissolution caused by carbonate undersaturation and decreased pH at depths greater than 30 m (Patterson & Kachinjika, 1995). Therefore, it is likely that our data do not accurately reflect the living assemblage from deeper depths in the modern lake. This also places a clear limit on our ability to extrapolate from these results to times in the past 123 Hydrobiologia Table 7 All-indicators, all-depths CCA (N = 99) Full model: Depth, ShoreEnv, GrSand, Sand, SiltClay, OrgC, Air All Bio CCA N = 99 Individual significance Conditional significance Reduced model Variable P value Keep/drop P value Keep/fit Fit P value DO 0.005 Keep 0.02 Keep ** Depth 0.005 Keep 0.005 Keep ** Shore env. 0.005 Keep 0.005 Keep ** Sand 0.005 Keep 0.26 Fit 0.001 Silt ? clay 0.005 Keep 0.29 Fit 0.001 Air temp. 0.01 Keep 0.31 Fit 0.013 [Sand 0.034 Keep 0.58 Fit 0.003 Organic C 0.07 Keep 0.18 Fit 0.082 All Bio CCA N = 99 Total inertia Constrained inertia % constrained F factor P value Full Model 1.9218 0.5457 28.31 4.4609 0.005 Reduced Model 1.9218 0.2896 15.07 5.6175 0.005 DO, surface, and bottom water variables not included in analysis. Data showing the significance testing process for each environmental variable as well as the permutation test statistics for the full and reduced models. The full model includes all variables listed at left. The two-step significance test follows in the next two columns with the statistical P value of each variable and the resultant action taken. The variables with ‘‘**’’ in the final column are the only ones included in the reduced model. The final column evaluates the significance of the fit of each environmental variable to the constrained axes of the reduced model when the lake was shallower and more alkaline than present (e.g., 110-95 thousand years ago—Cohen et al., 2007). It cannot be assumed that ostracodes would not have lived and been preserved at depths greater than 30 m during such periods of lowered lake levels, although a deeper CCD most likely existed under higher alkalinity conditions than at present. Ostracode distribution between 1 and 25 m Post-depositional dissolution may explain the disappearance of ostracodes below 30 m, however, species assemblage composition changes between 1 and 25 m as well. Based on the ostracode-only CCA (Fig. 6A), species distribution is best explained by two ecological factors: water depth and DO content. This finding supports a similar assertion by Martens (2002), who found that samples from sites at the same depth in widely separate geographic locations had a higher similarity index (more species in common) than samples from the same location but from different water depths (8 and 30 m). The influence of depth on ostracode assemblage is most likely the result of crossing species-specific tolerance thresholds in DO or light penetration (not measured in this study) which is 123 typically inversely correlated with turbidity (Halfman, 1996). In contrast, coastline environment (and its impacts on offshore conditions like substrate grain size and carbon content) is less significant when looking only at ostracode species distribution (Fig. 6A; Table 5). Although standardized species richness did not vary greatly by either depth or environment type in this study, it may be interesting to compare these modern values with those from samples in the paleo-record to see how ostracode diversity has changed with time. Comparison of current study to Martens (2002) live-assemblage distribution The largest difference between our results and those of Martens (2002) is that of the five sites he described, species richness was between 7 and 14 species/site. As can be seen in table in ESM 18, the average number of species per site in our study was 9, and the maximum was 21. However, this is not an unexpected discrepancy for two reasons. First, Martens (2002) specifically chose to describe those sites with the highest ostracode diversity. And second, the Martens’ (2002) report only counted those species that were alive when Hydrobiologia collected, and could therefore have soft-parts of the ostracode preserved for further description. Live collections typically have fewer total species because they represent a snapshot in time, whereas surface shell assemblages have the potential to time average ostracode species present over the course of one or more years (Alin & Cohen, 2004). Historically, typical offshore sediment accumulation rates in Lake Malawi have been between *1 mm/year (Pilskaln & Johnson, 1991) and *1.3 mm/year (Finney & Johnson, 1991), however, a recent study suggests near-shore sedimentation rates are increasing due to enhanced regional soil erosion (Otu et al., 2011). Given the depth of penetration by the grab-sampler in this study of *2 cm, it is likely that our samples averaged at most between 15 and 20 years of deposition. Since the purpose of choosing the five sites reported on by Martens’ (2002) was to maximize overall species diversity (to describe as many new species as were seen in the entire dataset), and the sites chosen were at depths of 7, 8, 10, 11 (‘‘shallow’’), and 30 (‘‘deep’’) meters, it is difficult to draw specific comparisons between the two datasets. However, some general similarities should be highlighted: In Martens (2002) atlas, Cyprinotinae sp.2 and G. emyrsi are found in ‘‘shallow’’ sites at multiple locations, suggesting that they have a geographically broad tolerance; these species are also found in abundance in all three environment categories in our study (Fig. 5; Table 3). In addition, of the eight species found at the ‘‘deep’’ sites in the live assemblage (Martens, 2002), four were observed in our study area: Limnocythere s.l. sp.1, Alicenula serricaudata, Cyprinotinae sp.1, and Cypridopsine n.gen. sp.N. Each of these species was present in the deeper samples (20 m) of our study (Table 3) and three of these four were most abundant at that depth than any other (Fig. 5). A final point concerns the extremely similar distributions of Limnocythere s.l. sp.9 and Limnocythere s.l. sp.10 (see Figs. 6A, B, C). It was suggested by Martens (2002) that these may, in fact, be males and females of the same species, and given their similar distribution and similar valve attributes (figures A–F in ESM 8), our data does not refute this assertion. Near-river sediment characteristics The percentage of silt and abundance of charcoal also change with depth, however, the variation differs further by environment. In sites offshore from river inlets, the values for both variables increase drastically at *20 m depth and remain high until at least 50 m (figures in ESM 9, 12, 13). In addition, increased charcoal and silt/clay plot very close to the low-relief river region in ordination space (Fig. 6C). This covariation between variables suggests two things: (1) grain size distribution in near-shore environments of Lake Malawi is dependent on the subenvironment— offshore channeling (e.g., figure in ESM 13, Transect C) from even a small river may redistribute larger grain sizes (silts) to deeper depths than settling velocity models would suggest (e.g., Gibbs et al., 1971); and (2) that macro-charcoal is overwhelmingly supplied to the lake by rivers. The latter point is of particular significance for paleoenvironmental interpretation of sediment cores from the lake, since it may indicate that changes in local, rather than distant (wind-borne) fire frequency are responsible for the changes observed in these records (e.g., Whitlock & Millspaugh, 1996; Thevenon et al., 2003). Relationship of ostracode abundances to other paleoecological indicators Important biological indicators often preserved as fossils in Lake Malawi paleo-records include ostracodes, fish, and the green alga Botryococcus. For this reason, understanding the distribution of these indicators in the modern lake may inform interpretations about the past lake environment (e.g., Blome et al., in prep.). Since the precipitous decline of ostracodes below *25 m in this study is likely caused by taphonomic bias, its potential abundance at deeper depths in shallower, paleo-lakes cannot be assessed. However, the coincidence of decreased ostracode abundance with increased Botryococcus (Figs. 3, 6A, B) should not necessarily be overlooked as this dichotomous relationship has also been observed between the two indicators in the paleo-record of Lake Malawi (Blome et al., in prep.). There may very likely be an ecological variable that both taxa are responding to but its interpretation in this study is hampered by the taphonomic dissolution of ostracodes. Abundant fish remains may relate to geographic location in addition to depth in this study. In the CCA models, fish abundance plotted near the highrelief environmental category (Fig. 6C). Since 123 Hydrobiologia cichlid fish in modern Lake Malawi are primarily rocky-dwelling littoral communities (Cohen et al., 2007), increased deposition of fish bones would be expected in the offshore HR areas as these often had sandy to rocky shorelines. Fish bones are abundant across all depths sampled in this study. Although total abundance increases with depth, the total variance is similar for all depths (excluding 35, 45, and 55 m depths due to small sample sizes) (Fig. 3). Reinthal et al. (2011) suggest that fish are most abundant at intermediate depths, which may indicate abundance increases with depth until an ‘‘optimal depth,’’ where conditions are ideal, before decreasing in abundance. However, this study did not sample deep enough to observe this theoretical ideal depth and subsequent decreased abundance. Often in the paleo-record discussed in Blome et al., in prep., the highest abundance of fish fossils occurs at transitions between ostracode-rich and Botryococcus-rich intervals which are interpreted as relatively shallow and deep lake environments, respectively. It would be interesting for a future study to sample deeper depths in the modern Lake Malawi to find the depth of highest fish abundance, as this value may be directly applicable to past studies. • • • • Conclusions • Modern ecological data from the southwest arm of Lake Malawi including limnological variables, grain size, and organic and inorganic carbonate content of sediments serve as a data set through which to interpret the paleo-abundance patterns of fish, mollusc, charcoal, Botryococcus, and ostracodes within the study area. The relationship between modern taxa and modern lake ecology may be useful for paleoenvironmental reconstruction. Interpretations of the data from this study allow the following conclusions to be drawn regarding the distribution and abundance of potentially fossilizeable materials in the littoral zone of modern Lake Malawi. • The general absence of ostracodes below 30 m is likely the result of a taphonomic bias caused by shell dissolution in the undersaturated, lower alkalinity waters below that depth. This is likely not a direct reflection of the living ostracode assemblage, although it is probable that ostracodes 123 would be present in lower abundances due to the increased difficulty in calcifying their valves below that depth. It is likely that times in the past when lake level was lower, ostracodes were more abundant and preserved at deeper depths due to increased alkalinity and therefore a deeper CCD. More species per site were found in this study than Martens (2002) in accordance with predicted time averaging differences between live and live/dead assemblages. Constrained ordination of ostracode abundance data suggests that species niches are most significantly affected by depth and DO content rather than substrate grain size or OC content. Based on these results, it is possible that a limnological variable not measured in this study, such as light penetration or turbidity which also co-vary with depth, may influence ostracode distribution in Lake Malawi. As originally suggested by Martens (2002), our analyses support the hypothesis that Limnocythere s.l. sp.9 and Limnocythere s.l. sp.10 may be females and males of the same species, respectively. Even a small inflowing river can significantly affect the particle size distribution of lake sediments, and it can serve as a conduit for macrocharcoal deposition into the basin from the surrounding watershed. The distribution by depth of the commonly fossilizable indicators ostracodes (shallow), fishbone (shallow to intermediate), and Botryococcus (intermediate) in the modern lake suggests that changes in their abundances in the past may reflect relative changes in lake level. Acknowledgments Funding for the field work for this project was provided by the U.S. National Science Foundation-Earth System History Program (EAR-0602350). Funding for sample preparation and sedimentological analyses was provided to author Lopez by the SAGUARO (Southern Arizona Geosciences Union for Academics, Research and Outreach) program at the University of Arizona. SEM imaging was performed on a Hitachi S-3400 N housed in the Geosciences Department at the University of Arizona. Funding for the SEM facility was through the Arizona LaserChron Center Grant NSF EAR-0929777. Lead author Blome would like to thank a number of people for their invaluable assistance during the 2010 field season in Cape MacLear, Malawi including: Linkston Chataka (captain), Bernard Gulo (diver), Reidwel Nyirenda (geologist at Malawi Geological Survey Department), and Jeffery Stone (fellow researcher), as well as Jay Stauffer and Hydrobiologia Leonard Kalindekafe for logistical assistance and Howard and Michelle Massey-Hicks for their hospitality over the course of the trip. References Abdallah, A. M., 2003. Environmental factors controlling the distributions of benthic invertebrates on rocky shores of Lake Malawi, Africa. Journal of Great Lakes Research 29: 202–215. Alin, S. & A. S. Cohen, 2004. The live, the dead, and the very dead: taphonomic calibration of the recent record of paleoecological change in Lake Tanganyika, East Africa. Paleobiology 30(1): 44–81. Baker, A. L., K. K. Baker & P. A. Tyler, 1985. Close interval sampling of migrating Chaoborus larvae across the chemocline of meromictic Lake Fidler, Tasamania. Archiv Fur Hydrobiologie 103(1): 51–59. Bennett, K. D., 2007. psimpoll, 4.27 edn. Birks, H. J. B., 2003. Direct Gradient Analysis. Department of Botany: 1–23. Birks, H. J. B. & J. M. Line, 1992. The use of rarefaction analysis for estimating palynological richness from Quaternary pollen-analytical data. The Holocene 2: 1–10. Cocquyt, C. & R. Jahn, 2007. Surirella fuellebornii (Bacillariophyta) and related taxa: lectotypification and distribution. Systematics and Geography of Plants 77(2): 213–228. Cohen, A. S., 2003. Paleolimnology: The History and Evolution of Lake Systems. Oxford University Press, Oxford. Cohen, A. S., R. Dussinger & J. Richardson, 1983. Lacustrine paleochemical interpretations based on eastern and southern African ostracodes. Palaeogeography, Palaeoclimatology, Palaeoecology 43: 129–151. Cohen, A. S., J. R. Stone, K. R. M. Beuning, L. E. Park, P. N. Reinthal, D. Dettman, C. A. Scholz, T. C. Johnsor, J. W. King, M. R. Talbot, E. T. Brown & S. J. Ivory, 2007. Ecological consequences of early Late Pleistocene megadroughts in tropical Africa. Proceedings of the National Academy of Sciences USA 104: 16422–16427. Dawidowicz, P., J. Pijanowska & K. Ciechomski, 1990. Vertical migration of chaoborus larvae is induced by the presence of fish. Limnology and Oceanography 35(7): 1631–1637. Dean, W. E., 1974. Determination of carbonate and organic matter in calcareous sediments and sedimentary rocks by loss on ignition: comparison with other methods. Journal of Sedimentary Petrology 44: 242–248. Delorme, L. D., 1969. Ostracodes as Quaternary paleoecological indicators. Canadian Journal of Earth Sciences 6: 1471–1476. Eccles, D. H., 1974. An outline of the physical limnology of Lake Malawi (Lake Nyasa). Limnology and Oceanography 19: 730–742. Finney, B. P. & T. C. Johnson, 1991. Sedimentation in Lake Malawi (East Africa) during the past 10,000 years: a continuous paleoclimatic record from the southern tropics. Palaeogeography, Palaeoclimatology, Palaeoecology 85(3–4): 351–366. Fritz, S. C., 1996. Paleolimnological records of climatic change in North America. Limnology and Oceanography 41: 882–889. Gasse, F., S. Juggins & L. B. Khelifa, 1995. Diatom-based transfer functions for inferring past hydrochemical characteristics of African lakes. Palaeogeography, Palaeoclimatology, Palaeoecology 117: 31–54. Gibbs, R. J., M. D. Matthews & D. A. Link, 1971. Relationship between sphere size and settling velocity. Journal of Sedimentary Petrology 41: 7–18. Guy-Ohlson, D., 1992. Botryococcus as an aid in the interpretation of palaeoenvironment and depositional processes. Review of Palaeobotany and Palynology 71(1–4): 1–15. Halfman, J. D., 1993. Water column characteristics from modern CTD data, Lake Malawi, Africa. Journal of Great Lakes Research 19: 512–520. Halfman, J. D., 1996. CTD-Transmissometer Profiles from Lakes Malawi and Turkana. In Johnson, T. C. & E. O. Odada (eds), The Limnology, Climatology and Paleoclimatology of the East African Lakes. Gordon and Breach, Amsterdam: 169–182. Heiri, O., A. F. Lotter & G. Lemcke, 2001. Loss on ignition as a method for estimating organic and carbonate content in sediments: reproducibility and comparability of results. Journal of Paleolimnology 25: 101–110. Holmes, J. A., 1992. Nonmarine ostracods as Quaternary palaeoenvironmental indicators. Progress in Physical Geography 16: 405–431. Horne, D. J., 2007. A mutual temperature range method for Quaternary paleoclimatic analysis using European nonmarine Ostracoda. Quaternary Science Reviews 26: 1398–1415. Howe, H. V., 1955. Handbook of Ostracod Taxonomy. Louisiana State University Press, Baton Rouge. Juggins, S., 2009. Analysis of Environmental Data – Constrained Ordination. School of Geography, Politics & Sociology. Newcastle University. Kempf, E. K., 1980. Index and Bibliography of Nonmarine Ostracoda, Vol. 1. Geologisches Institut Der Universitaet Zu Koeln, Koeln. Klie, W., 1944. Ostracoda Exploration du Parc National Albert: Mission H Damas (1935–1936). Instituut der Nationale Parken van Belgisch Congo, Brussels. Malawi Government, 2003. Nkhudzi Bay to Malembo (map). Department of Surveys – Hydrographic Unit, Malawi. Martens, K., 1986. Taxonomic Revision of the Subfamily Megalocypridinae Rome, 1965 (Crustacea, Ostracoda). Paleis der Academien, Brussels. Martens, K., 1988. Seven new species and two new subspecies of Sclerocypris SARS, 1924 from Africa, with new records of some other Megalocypridinids (Crustacea, Ostracoda). Hydrobiologia 162: 243–273. Martens, K., 2002. Task 3: Taxonomy of Invertebrates. In Irvine, K. (ed.) The Trophic Ecology of the Demersal Fish Community of Lake Malawi/Niassa, Central Africa: 49–59. Martens, K., 2003. On the evolution of Gomphocythere (Crustacea, Ostracoda) in Lake Nyassa/Malawi (East Africa), with the description of 5 new species. Hydrobiologia 497: 121–144. Mezquita, F., J. R. Roca, J. M. Reed & G. Wansard, 2005. Quantifying species–environment relationships in nonmarine Ostracoda for ecological and palaeoecological studies: examples using Iberian data. Palaeogeography, Palaeoclimatology, Palaeoecology 225: 93–117. 123 Hydrobiologia Mischke, S., D. Fuchs, F. Riedel & M. E. Schudack, 2002. Mid to Late Holocene palaeoenvironment of Lake Eastern Juyanze (north-western China) based on ostracods and stable isotopes. Geobios 35: 99–110. Neale, J. W., 1979. On the Genus Cyprinotus and its Interpretation. In Krstic, N. (ed.), Proceedings of VII International Symposium on Ostracodes: Taxonomy, Biostratigraphy and Distribution of Ostracodes. Serbian Geological Society, Beograd, Serbia: 77–85. Oksanen, J., 2004. Ordination and environment. Department of Biology. University of Oulu, Finland: 122–152. Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. Henry, H. Stevens & H. Wagner, 2011. vegan: Community Ecology Package. R package version 2.0-1 edn. Otu, M. K., P. Ramlal, P. Wilkinson, R. I. Hall & R. E. Hecky, 2011. Paleolimnological evidence of the effects of recent cultural eutrophication during the last 200 years in Lake Malawi, East Africa. Journal of Great Lakes Research 37(1): 61–74. Park, L. E. & A. S. Cohen, 2011. Paleoecological response of ostracods to early Late Pleistocene lake-level changes in Lake Malawi, East Africa. Palaeogeography, Palaeoclimatology, Palaeoecology 303: 71–80. Patterson, G. & O. Kachinjika, 1995. Limnology and phytoplankton ecology. In Menz, A. (ed.), The Fishery Potential and Productivity of the Pelagic Zone of Lake Malawi/Niassa. Natural Resource Institute, Chatham: 1–67. Pilskaln, C. H. & T. C. Johnson, 1991. Seasonal signals in Lake Malawi sediments. Limnology and Oceanography 36: 544–557. R Development Core Team, 2010. R: A language and environment for statistical computing, 2.12.0 ed. R Foundation for Statistical Computing, Vienna, Austria. Reinthal, P. N., A. S. Cohen & D. L. Dettman, 2011. Fish fossils as paleo-indicators of ichthyofauna composition and climatic change in Lake Malawi, Africa. Palaeogeography, Palaeoclimatology, Palaeoecology 303: 126–132. Rossetti, G. & K. Martens, 1998. Taxonomic revision of the Recent and Holocene representatives of the Family 123 Darwinulidae (Crustacea, Ostracoda), with a description of three new genera. Biologie 68: 55–110. Ryan, W. B. F., S. M. Carbotte, J. O. Coplan, S. O’Hara, A. Melkonian, R. Arko, R. A. Weissel, V. Ferrini, A. Goodwillie, F. Nitsche, J. Bonczkowski & R. Zemsky, 2009. Global multi-resolution topography synthesis. Geochemistry Geophysics Geosystems 10: Q03014. Sars, G. O., 1910. Zoological Results of the Third Tanganyika Expedition, conducted. Proceedings of the Zoological Society of London: 732–760. Sars, G. O., 1924. The freshwater Entomostraca of the Cape Province (Union of South Africa). Annals of the South African Museum 20: 105–193. Smith, A. J., 1993. Lacustrine ostracodes as hydrochemical indicators in lakes of the north-central United States. Journal of Paleolimnology 8: 121–134. Strayer, D. L., 2010. Benthic Invertebrate Fauna, Lakes and Reservoirs. In Likens, G. E. (ed.), Lake Ecosystem Ecology: A Global Perspective. Elsevier, Amsterdam: 27–41. ter Braak, C. J. F. & I. C. Prentice, 1988. A theory of gradient analysis. Advanced Ecological Research 18: 271–317. Thevenon, F., D. Williamson, A. Vincens, M. Taieb, O. Merdaci, M. Decobert & G. Buchet, 2003. A late-Holocene charcoal record from Lake Masoko, SW Tanzania: climatic and anthropologic implications. Holocene 13(5): 785–792. Van Harten, D., 1979. Some new shell characters to diagnose the species of the Ilyocypris gibba-biplicata-bradyi group and their ecological significance. In Krstic, N. (ed.), Proceedings of VII International Symposium on Ostracodes: Taxonomy, Biostratigraphy and Distribution of Ostracodes. Serbian Geological Society, Geograd, Serbia: 71–76. Vincens, A., G. Buchet, D. Williamson & M. Taieb, 2005. A 23,000 year pollen record from Lake Rukwa (8S, SW Tanzania): New data on vegetation dynamics and climate in Central Eastern Africa. Review of Palaeobotany and Palynology 137: 147–162. Whitlock, C. & S. H. Millspaugh, 1996. Testing the assumptions of fire history studies: An examination of modern charcoal accumulation in Yellowstone National Park, USA. Holocene 6(1): 7–15.