Holocene foraminiferal and geochemical records in the coastal karst dolines of Cres Island, Croatia

Numerous karst dolines have been formed along the Croatian coast and many have been submerged during the Late Glacial and Holocene sea level rise. The coastal area of Cres Island in the Northern Adriatic is a typical example of this geomorphological setting, where transitional forms from subaerial to submerged dolines are present. Once dolines are formed they can accumulate soil, water and sediments due to their morphology. Sediments are an especially valuable source of environmental data. This paper presents the results of the study of foraminiferal assemblages and sediment geochemistry, supplemented with grain-size and mineralogical data, from the marine ponds developed in the karst dolines on Cres Island. Obtained data is correlated with the sediment core record from submerged dolines in the present-day embayments along the coastal zone of Cres Island. In total, 3 sediment cores were collected in the marine ponds Marinska, Arcij and Podbrajde, while 2 longer sediment cores have been extracted from the Jaz and Sonte embayments. The Marinska, Arcij and Podbrajde marine ponds have distinct geochemical and mineralogical sediment compositions, with monospecific foraminiferal assemblages and generally differ from each other. The common characteristics are their high N and P concentrations and the algal origin of organic matter. Agglutinated foraminiferal taxa (Haplophragmoides canariensis and Trochammina inflata), typical for intertidal environments, are abundant in the brackish-water Marinska pond, while stress-tolerant species Ammonia tepida has been identified in the Arcij marine pond. Environmental conditions in the Podbrajde marine pond did not facilitate the development of a rich foraminiferal fauna. Results from the present-day marine ponds enabled recognition of similar environments in the sediment cores collected in the Jaz and Sonte embayments that were progressively inundated during the Holocene sea level rise. A palaeo-marine pond existed in the Sonte embayment until 6610 cal BP, when the sea flooded the investigated area. A marine pond in the Jaz embayment was formed at 711 cal BP. Low-diversity foraminiferal assemblages in these palaeo-ponds are similar to those recognized in the present-day Arcij marine pond on Cres Island. However, differences in the geochemical composition of palaeo-marine ponds, in comparison to the present-day ponds, exist. They might be attributed to climate variability over time and variations in the geological setting of each environment. High Mo concentrations and abundant organic matter content are the main sediment characteristics of the recognized palaeo-marine ponds in the Jaz and Sonte embayments.

The specific geomorphological setting of the karstified eastern Adriatic coast (PIKELJ & JURAČIĆ, 2013), enabled development of unique marginal marine environments. Veliko and Malo Jezero on Mljet Island are examples of submerged karst Holocene foraminiferal and geochemical records in the coastal karst dolines of Cres Island, Croatia sitional environments present along the karstified eastern Adriatic coast.
Recent foraminiferal studies along the eastern Adriatic coast have been mostly restricted to investigation of species distribution (VIDOVIĆ, 2010;ĆOSOVIĆ et al., 2011) and to the possible application of foraminifera as indicators of anthropogenic pollution (VIDOVIĆ et al., 2009;POPADIĆ et al., 2013;VIDOVIĆ et al., 2014). Only a few studies have focused on the determination of fauna inhabiting marginal marine environments and their dependence on environmental conditions, as well as their application in palaeoenvironmental reconstructions. However, foraminiferal assemblages in marine lakes on the Mljet Island have been described in detail (VANIČEK et al., 2000;ĆOSOVIĆ et al., 2016), while SHAW et al. (2016 conducted a study of assemblages present in the Jadrtovac and Blace salt-marshes. Results have produced valuable data about sea level variations in the investigated area and proven the utility of salt-marsh foraminifera in transfer functions. FELJA et al. (2015) determined a foraminiferal fauna in sediment cores from the Mirna River valley in Istria in order to decipher the palaeoenvironmental development of the valley.
Examples of the previously conducted research of sediment geochemistry in the marginal marine environments along the Croatian coast of the Adriatic Sea include studies of the Veliko and Malo Jezero (CUCULIĆ et al., 2009;SONDI et al., 2017), Mir (MLAKAR et al., 2015) and Zmajevo oko (MIHELČIĆ et al., 1996) marine lakes. Furthermore, the geochemical composition of sediments from the shallow marine environments in Bakar Bay (CUKROV et al., 2014) and Makirina Cove (ŠPARICA at al., 2005;MIKO et al., 2008;KOMAR et al., 2015) was studied extensively. Most of these studies focused on deciphering natural and anthropogenic metal enrichment and geochemistry was not used in order to reconstruct palaeoenvironments.
Our study offers a new insight into foraminiferal assemblages and sediment geochemistry in the marine ponds along the coastline of Cres Island. These environments have been formed in the karst dolines and should be considered as marginal marine environments due to their proximity to the sea and marine influence through karst. The main aim was to document typical foraminiferal species and their distribution in these unique water bodies since such studies are lacking along the eastern Adriatic coast. Special emphasis was on the determination of the linkages between foraminiferal assemblages and established environmental conditions manifested in the physical and chemical water properties and geochemical, mineralogical and sedimentological properties of the sediment. A further aim was to characterize environments in the Jaz and Sonte embayments, similar to those in the present-day marine ponds on Cres Island, using extracted sediment cores. Obtained results would be indicative of the timing of the existence of Holocene palaeo-marine ponds in these presently submerged karst dolines. We hypothesize that presentday marine ponds on Cres Island represent modern analogs of submerged Holocene palaeo-marine ponds in the Jaz and Sonte embayments.

STUDY AREA
The investigated environments are located along the southwestern coastline of Cres Island, in the northern part of the Adriatic Sea in the Kvarner region ( Fig. 1). This region is characterized by NW-SE elongated islands (Cres, Lošinj, Krk, Rab, Pag), with bays and channels (Rijeka Bay, Kvarnerić Bay, Kvarner Bay, Vinodol-Velebit Channel, Lošinj Channel) located between the Vinodol-Velebit coastline and the Istrian peninsula. Final formation of the steep and rocky eastern Adriatic coast occurred during the Late Glacial and Holocene periods when sea flooded preexisting folded, faulted and karstified relief. Anticlines became island chains while synclines became bays and channels (BENAC & JURAČIĆ, 1998;JURAČIĆ et al., 1999;KELLETAT, 2005;PIKELJ & JURAČIĆ, 2013).
The Mesozoic carbonate rocks and their geological setting on Cres Island have been a target of many detailed studies (MAMUŽIĆ, 1968;MAGAŠ, 1968;HUSINEC et al., 2000;KORBAR et al., 2001;KORBAR & HUSINEC, 2003;FUČEK et al., 2012;FUČEK et al., 2014). Limestones and dolomites of Cretaceous age are predominant sedimentary rock formations. The oldest deposits have been formed during the Lower Cretaceous. Younger Palaeogene strata (foraminiferal limestones and flysch) are sporadically present in the northern part of Cres Island (FUČEK et al., 2014). Quaternary sediments are rather scarce and present in the form of terra rossa outcrops, alluvial deposits and colluvial deposits on the slopes (MAGAŠ, 1968;BENAC & DURN, 1997).
Subaerial exposure of Cretaceous deposits enabled the formation of numerous karst dolines on Cres Island, especially in its southwestern part. Dolines are enclosed karst depressions with variable depth and dimensions (FORD & WILLIAMS, 1989). It is considered that post-Miocene karstification in the Mediterranean occurred in tectonically susceptible areas during periods of sea level lowstands. The most prominent and best preserved evidence of karstification developed during the most recent sea level lowstand (SURIĆ, 2005;MOCOCHAIN et al., 2009;PIKELJ & JURAČIĆ, 2013).
The Marinska, Arcij and Podbrajde marine ponds investigated in this study have variable depths, sill elevations separating the pond from the direct marine influence, distances to the sea and size (Tab. 1). They have been developed in the karst dolines within two lithostratigraphic units (Fig. 1), which are composed of alternations of dolomites and shallow marine limestones assigned to the Upper Albian-Lower Cenomanian boundary (Sis unit) and Cenomanian pelagic limestones (Belej unit) (FUČEK et al., 2014). The karstic nature of Cres Island enables sea-water seepage through the karstified sill separating the ponds from the sea. Dip directions trend towards the northeast (FUČEK et al., 2014) which possibly additionally facilitates this seepage through developed layering, underground fissures and conduits.
The present-day climate on Cres Island is considered to be temperate and humid with hot summers, while at the highest elevations summers are warm (ŠEGOTA & FILIPČIĆ, 2003). Meteorological measurements between 1971-2000 at the nearby me-   , 2008). Relatively high amounts of rainfall have been recorded in the area, with mean annual precipitation of approximately 930.5 mm and the highest precipitation rates in the autumn .
The human presence on Cres Island is documented by numerous archaeological sites (REGAN & NADILO, 2010;DO-NEUS et al., 2017). The channel that separates islands of Cres and Lošinj in the Osor village has been dug artificially and today has a strong impact on water circulation in the Lošinj Channel, with currents causing deepening of the area south of Osor (DO-NEUS et al., 2017).

MATERIALS AND METHODS
Short sediment cores (up to 45.5 cm long) were collected in the Marinska, Arcij and Podbrajde shallow marine ponds, by inserting a plastic pipe by hand into the sediment. Sampling was conducted in September 2015. In total, 3 cores were extracted (LK-5, LK-6 and LK-7) and subsampled in 1 cm sections in the field using the core extruder (Tab. 1). Longer sediment cores (up to 371 cm long) from the Jaz and Sonte embayments were collected using a piston corer and coring platform in April 2014 (Tab. 1). Detailed subsampling of piston cores LK-2 and LK-3 was conducted after splitting the cores lengthwise. Additionally, physical and chemical water parameters such as conductivity, temperature, pH and dissolved oxygen content were determined at each coring location using a portable multiparameter Multi 3430 WTW probe.
Five mollusc samples from sediment cores LK-2 and LK-3 were dated by radiocarbon dating method (AMS 14 C) in order to establish core chronology and to determine the timing of palaeoenvironmental changes in the present-day Jaz and Sonte embayments. Analyses have been conducted at Beta Analytics Laboratory in USA. The radiocarbon data was corrected for the marine reservoir effect determined for the Adriatic Sea (FAIVRE et al., 2015).
Grain-size was measured using a laser diffractometer Shimadzu SALD-2300. Selected samples were treated with 30% hydrogen peroxide (H 2 O 2 ) in order to remove organic matter. Afterwards, samples were centrifuged, decanted and resuspended in distilled water prior to the analysis. In total, grain-size was determined on 6 samples from sediment cores LK-5 and LK-7, 4 samples from core LK-6, while 25 and 73 samples were analysed from sediment cores LK-2 and LK-3, respectively. Statistical analysis of the grain-size results was conducted using the GRADISTAT software (BLOTT & PYE, 2001) and sediments were classified according to FOLK & WARD (1957).
The bulk mineralogical composition of selected samples, as well as their clay mineralogy, was determined using a PANalytical X'Pert Powder X-ray diffractometer. For the measurements, the diffractometer was set at 45 kV and 40 mA, with a step size of 0.02⁰ 2Ө. Different phases present in the samples were identified following MOORE & REYNOLDS (1997). The clay fraction was analysed by mounting oriented aggregates on glass slides, after the removal of carbonates (where needed), using a buffered sodium acetate (NaOAc) solution. Oriented samples were scanned, air-dried, ethylene-glycolated and analysed after being heated to 400°C and 550°C.
Core tops (1 cm thick) of the short sediment cores from marine ponds (LK-5, LK-6 and LK-7) were used for micropalaeontological analysis. Samples were treated with rose Bengal stain and 70% ethanol in the field and left for 14 days before being washed, following the procedure described by the FOBIMO group (SCHÖNFELD et al., 2012). Piston cores LK-2 and LK-3 from the Jaz and Sonte embayments were analysed in more detail. Foraminiferal analysis on these cores focused on the first cm of each core and intervals where geochemical, sedimentological and mineralogical data indicated significant shifts. These samples were not treated with rose Bengal stain.
For determination of the total foraminiferal assemblages (stained+empty tests) selected sediment samples were washed over a 63 µm sieve in the laboratory in order to remove silt and clay particles. According to SCHÖNFELD et al. (2012) >63 µm fraction is appropriate for foraminiferal analyses in palaeoenvironmental studies. Each sample was divided by microsplitter and approximately 300 specimens were picked. In the samples where the total number of foraminifera specimens was less than 300, the whole sample was examined and counted. Genera and species were recognized following the available literature and classifications by CIMERMAN & LANGER (1991), SGARRELLA & MONCHARMONT ZEI (1993) and LOEBLICH & TAPPAN (1987). Calculation of the species richness, Shannon-Wiener index, Fisher α index, Dominance, Evenness and Equitability was conducted using the PAST software (HAMMER et al., 2001). Cluster analysis of the total foraminiferal assemblages was performed using STATISTICA 10 software with Ward's method and Euclidean distances (STATSOFT, 2011). Scanning electron microscope (SEM) images of the common agglutinated and calcareous taxa observed in the samples were obtained using the Jeol 35 CF scanning electron microscope, while the chemical composition of their tests was determined using the energy dispersive spectroscopy (EDS-Oxford X-ACT) and INCA detection unit.
Analyses of the total organic (TOC) and inorganic carbon (TIC) and total nitrogen (N) were performed on 134 samples on a Thermo Fisher Scientific Flash 2000 NC Analyzer. Before analysis, the samples were freeze-dried, finely ground and treated with hydrochloric acid (HCl) for TOC measurements (TUNG & TANNER, 2003). Calculation of TIC is based on the difference between total carbon (TC) (untreated samples) and TOC (samples treated with HCl). C/N ratio was calculated from TOC and N measurements.
Elemental analysis was carried out on ground samples using Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) in the ACME laboratory, Canada. Geochemical data was analysed using multivariate statistical techniques. Discriminant function analysis (DFA) included a selection of geochemical compositions (TOC, TIC, N, Ca, Mg, Fe, K, Al, P, S, Cu, Pb and Mo) and a single ratio (C/N). Results of the geochemical analysis represent a typical example of compositional data (CoDa) where correlations between relative elemental abundances may not be unequivocal in the absence of any other information or assumptions (LOVELL et al., 2015). Discriminant function analysis (DFA) is a traditional multivariate statistical technique which is particularly useful in building the predictive model for the two-or multiple-group discrimination based on the suite of independent (predictor) variables. It is commonly exercised in geo-and environmental sciences when geological logic calls for the use of some autonomous criterion with regards to the variables in the analysed dataset, such as, in this case, the multi-proxy sediment core data analysis. Objectives and principles of DFA are described comprehensively in many statistical textbooks (e.g., DAVIS, 1973;DAVIS, 1986;DILLON & GOLDSTEIN, 1984;ROCK, 1988;REIMANN et al., 2008). Geochemical data were processed using the statistical software package of STATISTICA 10 (STATSOFT, 2011) in the spirit of the CoDa analysis.

Physical and chemical water parameters
The investigated environments showed variability of the measured physical and chemical water parameters. Table 2. shows conductivity (E c ), pH, temperature (t) and dissolved oxygen (O 2 ) concentrations. The E c measurements reveal significant differences, with the lowest E c values in the Marinska marine pond (11.1 mS/cm) and the highest in the Podbrajde marine pond (56.8 mS/ cm). The pH ranged from 8.24 (Sonte embayment) to 9.07 (Podbrajde marine pond). Measured O 2 concentrations varied from 9.11 mg/l in the Sonte embayment up to 15.22 mg/l in the Arcij marine pond. The highest temperature of 28.1⁰C was determined in the Marinska pond.

Sediment core analysis
Since the cores from marine ponds (LK-5, LK-6 and LK-7) are short (up to 45.5 cm) and analysis of the foraminiferal assemblages focused only on the core tops, we did not section them into distinct units (Figs. 2 to 4). Based on the sediment characteristics such as grain-size, TOC, TIC and N content, and the geochemistry of selected major and trace elements several units were dif-

Core chronology
The radiocarbon data obtained from the macrofossils of the piston cores LK-2 and LK-3 is shown in Table 3. The sediment sequence in both cores is of Holocene age and spans >6610 years in core LK-3 and >711 years in core LK-2. Age reversal has been found in sediment core LK-3, with the result at the core depth of 354 cm being younger than the samples at 252 cm and 217 cm.
Due to the fact that results at 252 and 217 cm have been obtained on single mollusc shells each and the result at 354 cm has been obtained on mixed shells of gastropods and bivalves, the former are thought to be more reliable. Therefore, the result at 354 cm has been discarded (Tab. 3).

Grain size analysis
Grain-size analysis on 110 samples from all sediment cores revealed variations in the grain-size both between cores and within particular cores (Figs. 2 to 6, Tab. 4). Core tops are predominantly composed of coarse silt to medium sand. In most cores grain-size slightly decreases downwards with the general predominance of silty material (Figs. 2 to 6).

Mineralogical analysis
Bulk mineralogical analysis of sediment core LK-5 indicates the predominance of calcite and quartz, while aragonite is also abundant. Halite, Mg calcite, aragonite, gypsum, quartz, muscovite/ illite and pyrite are present in sediment core LK-6. Sediment core LK-7 is characterized by the occurrence of halite, quartz, muscovite/illite and calcite. In the sediment cores LK-2 and LK-3, quartz is the dominant mineral phase, while the calcite presence is variable downcore. In the basal part of the LK-2 core quartz, muscovite/illite and kaolinite are abundant (unit LK-2-1). The upper part of the LK-2 core (unit LK-2-2) is characterized by calcite, Mg-calcite, aragonite and halite. Muscovite/illite and plagioclase are present throughout the LK-2 core. The lowermost part of the core LK-3 (unit LK-3-1) shows the presence of quartz, muscovite/illite, chlorite and kaolinite. Calcite and Mg-calcite predominate in the upper part of the core (units LK-3-2, LK-3-3 and LK-3-4). Aragonite, dolomite, halite, plagioclase, muscovite/ illite and kaolinite are less abundant. The presence of halite in all samples is a consequence of its crystallization from the pore water after sediment drying. The same clay minerals (chlorite, illite and kaolinite) characterize all the analysed cores. The only exception is the presence of Mg clay mineral sepiolite in the sediment core LK-5.

Foraminiferal analysis
Overall, 87 different species were identified in the total foraminiferal assemblages (all cores). Samples from cores LK-5, LK-6 and LK-7 were less diverse and encompass 13 species. In the samples from the sediment cores LK-2 and LK-3, 81 foraminiferal species were determined. One sample (LK-2-1, 67-68 cm) did not contain foraminifera, while in the samples from the basal part of LK-3 core (unit LK-3-1) and from LK-7 core foraminifera specimens were scarce and poorly preserved.
Altogether, 52 SEM-EDS analysis were conducted on selected grains of agglutinated taxa from the LK-5 core, while several analysis were performed on calcareous taxa from the first cm of LK-5 and LK-6 sediment cores. The mineralogical composition of agglutinated taxa is variable. Quartz, amphibole, mica (biotite and muscovite) and feldspars (plagioclase and potassium feldspars) are common mineral phases found in agglutinated foraminiferal tests (Fig. 8).

Discriminant function analysis (DFA) of the geochemical data
Discriminant function analysis (DFA) of geochemical data is summarized in the composite Table 6., describing the geochemical exploratory model for the sediment cores. The table comprises both the multivariate test for the overall significance of discrimination and the test of residual roots (discriminant functions). The Wilks' k statistical test is commonly used with the purpose of validating the probability level (p < 0.05) required to proceed safely with computing discriminant functions (DFs). It is used to select both the statistically significant functions as well as explaining the maximum of the within-group variation. The scatterplots of variable loadings and group centroids are constructed for the first three DFs that explain the greatest proportion of the between-group variance, almost 92% (Fig. 10). In the computed model the first discriminant function DF1 plays by far the greatest role in discrimination between the groups, accounting for more than 60% of the total variance (Tab. 6).
In the context of the environmental conditions studied, this multivariate method is particularly helpful in chasing the major sources of between-group differences originating from dissemination of the chemical elements in the sediment cores. In combining the cores from marine ponds (LK-5, LK-6 and LK-7) and    distinguished sediment core units from LK-2 and LK-3 cores, nine groups of cored sediments were created to provide the most effective rapport between their geochemical signature and immediate sedimentary environment. As seen from the scatterplot of variable loadings (Fig. 10), DF1 is highly bipolar separating principally the LK-5 and LK-7 groups from most of the LK-3 sediment core groups (LK-3-2, LK-3-3 and LK-3-4) while other groups remain to a great extent poorly differentiated sticking closer to the point of the axis intersection. The rationale for this separation is founded primarily in the negative association between P and Mo. Geochemically it can be understood as the LK-5 and LK-7 cores are relatively enriched in P and depleted in Mo in comparison to the upper three LK-3 units. The groups positioned in the middle reflect the average composition of all the investigated core units -a "mixed" geochemical signature.
DF2, comprising over 25% of the total variance of geochemical data is also bipolar and can be interpreted as reflecting a negative correlation between the element suite including Cu, Pb, K, Al, Fe against another set of elements led by TIC, Ca, N. This pattern is a geochemical signature dividing the LK-2-1 from LK-6 groups on the basis of the clay/carbonate (inorganic carbon) background. Despite making only a minor contribution to the overall model significance (6%), DF3 as a monopolar function clearly highlights the LK-3-1 and LK-3-2 groups based on their affinity with C/N with a slight TOC association (Fig. 10).
The main sediment and water characteristics of the Marinska, Arcij and Podbrajde marine ponds investigated here are shown in Table 7. At the sampling time, water in the Marinska pond was brackish with a temperature above 28⁰C and oxygen saturation of 15.12 mg/l. Sediments in this marine pond are characterized by their high N and P contents (Figs. 9F and 9G). A high P concentration in the surface sediments probably occurs due to delay in mineralisation of recently settled organic material (HOLTAN   al., 1988). A source of these elements could also be input of bird faeces (guano) into the ponds (BATANERO et al., 2017), considering that the investigated areas are habitats for migratory marsh birds (most commonly mallard-Anas platyrhynchos), as well as woodcock (Scolopax rusticiola) and seagulls (KRALJ, pers.comm.). The mallard and woodcock are also frequently hunted in the area. High nutrient concentrations enabled the development of environmental conditions that facilitated accumulation of organic matter in the sediments (Figs. 9F and 9G). A low C/N ratio implies an algal origin of the preserved organic matter (MEYERS, 1994;MEYERS, 2003;LAMB et al., 2006). The TIC also proved to be an important sediment component (Tab. 5, Fig.  9H). The Mo and S concentrations in the Marinska pond sediments are low (Figs. 9D and 10E; Fig. 11A). Preserved Mo is not an indicator of anoxic environmental conditions and can be linked to the input of terrestrial soil material based on the high correlation of Mo with Al ( Fig. 11C) (GOLDBERG et al., 1996). Terra rossa topsoils from sites in Dalmatia have a mean concentration of 3 mg/kg Mo (MIKO et al., 2007).
Only agglutinated taxa typical of intertidal environments are present in the Marinska marine pond. The species H. canariensis and T. inflata predominate, while E. macrescens and M. fusca are significantly less abundant. Different species of the genus Haplophragmoides have already been recognized in the Adriatic Sea (CIMERMAN & LANGER, 1991;SERANDREI BARBERO et al., 2004;SHAW et al., 2016), while T. inflata appears to be frequent in the marginal marine environments along the Croatian coast of the Adriatic Sea (FELJA et al., 2015;SHAW et al., 2016). Generally, this species has been reported from many restricted, brackish water environments in the Mediterranean region (SE-RANDREI BARBERO et al., 1999;SERANDREI BARBERO et al., 2004;DEBENAY & GUILLOU, 2002;   al., 2011a) and elsewhere (LIDZ & ROSE, 1989;SEN GUPTA et al., 2009). Recognized agglutinated foraminiferal fauna typically occur in the zone between mean sea level and mean high water level in salt-marsh areas of the Adriatic Sea (SERANDREI BAR-BERO et al., 1999;SHAW et al., 2016), indicating that in the investigated environment tidal influence is important. The Marinska marine pond is located 50 m from the sea and the possible tidal influence could be through karst features. Although foraminiferal species richness in the Marinska pond is low, recognized species seem to be well adapted to the restricted environmental conditions due to their abundance. However, a significant number of damaged tests were also observed, possibly indicating conditions unsuitable for their preservation (Plate 1). DEBENAY (2001) observed the presence of T. inflata in marshes with variable salinity and TOC content, while DEBENAY & GUILLOU (2002) reported the preference of this species for muddy sediments in microtidal Mediterranean environments. However, our results proved that although TOC content was high, this species occurred in the brackish water environment with high percentages of sand (up to 75%). Grain-size distribution, as well as the mineralogical composition of the sediment, has important implications for agglutinated taxa (ALLEN et al., 1999;ARMYNOT DU CHÂTELET et al., 2008;ARMYNOT DU CHÂTELET et al., 2013). SEM-EDS analysis showed that identified agglutinated foraminifera from the Marinska pond used variable mineral grains for building their tests (quartz, amphibole, mica and feldspars) (Fig. 8). Some mineral grains were not recognized in the bulk mineralogical composition of the sediment, probably due to their low abundances. This indicates the preference of agglutinated taxa for specific mineral grains. ALLEN et al. (1999), AR-MYNOT DU CHÂTELET et al. (2008), MAKLED & LANGER (2010) and ARMYNOT DU CHÂTELET et al. (2013) in their studies have determined that some species preferentially select certain mineral grains from their surroundings, while significant differences among species also exist. Mineral analysis also revealed the presence of the clay mineral sepiolite, usually found in restricted environments, ponds or shallow brackish lakes with high evaporation rates (ORDÓÑEZ et al., 1991;VELDE, 1995;MEUNIER, 2003;BUSTILLO & ALONSO-ZARZA, 2007). Direct precipitation of sepiolite occurs from solutions with abundant Mg and Si (MAYAYO et al., 1998) in the high pH environments (STARKEY & BLACKMON, 1984). BUSTILLO & ALONSO-ZARZA (2007) indicate that sepiolite can be formed by the transformation of illites or smectites in a vadose alkaline environment with Mg rich groundwater originating from dolomitic aquifers. The presence of sepiolite implies the significant impact of climate conditions (precipitation, evaporation) on the water level in the Marinska pond that has been developed in a generally dolomitic lithological setting. Seasonal climatic variability with periods of high evaporation from the marine pond, could have enabled the development of environmental conditions that favoured the precipitation of sepiolite in the Marinska pond.
Conductivity (E c ) and O 2 measurements in the water in the Arcij marine pond indicate the existence of an oxygenated en- Figure 11. Scatterplots of A) Mo (mg/kg) against S (%) B) N (%) against P (%), C) Mo (mg/kg) against Al (%), D) TOC (%) against C/N. closed marine environment. Regardless of the larger distance of this marine pond from the sea (92 m), a stronger seepage of sea water probably occurs due to the well-developed karstification and fissures system of the ridge separating the pond from the sea. A silty surface sediment is enriched with TOC (7.58%) and TIC (8.29%), while the organic matter is predominantly of algal origin (C/N<12; MEYERS, 1994;MEYERS, 2003;LAMB et al., 2006). Similarly to the Marinska marine pond, the primary productivity indicators (N and P) are abundant. The Mo and S concentrations in the sediment are higher in comparison to the other marine ponds, causing reducing sediment conditions and thus further facilitating the preservation of organic matter. In the Arcij pond, the source of Mo could be different to that of the Marinska pond (Fig. 11C). In many papers (PEDERSEN, 1989;CRU-SIUS et al., 1996;CALVERT & PEDERSEN, 1993;ALGEO & LYONS, 2006;SCHOLZ et al., 2017) the application of Mo as an indicator of redox conditions was emphasized. It is considered that Mo precipitation often occurs in anoxic and organic matterrich silled basins (ALGEO & LYONS, 2006). We suggest enrichment with Mo in the Arcij pond is a consequence of stagnant bottom water and reducing conditions. The S concentrations are also high (Fig. 9E). The presence of hydrogen sulfide (H 2 S) seems to be important for the uptake of Mo, especially in shallow water environments and non-silled basins (PEDERSEN, 1989;ALGEO & LYONS, 2006;SCHOLZ et al., 2017). This can be applicable to the Arcij pond (Fig. 11A), where pyrite is formed. Established environmental conditions do not seem to be a limiting factor for the presence of foraminifera. The most common species occurring in the Arcij pond is A. tepida, with a relative abundance of almost 97% (Tab. 5). This species has been identified in numerous shallow water marine environments, restricted marginal marine environments (lagoons, estuaries, salt-marshes) and inland saline pools due to its tolerance to normal, brackish and hypersaline water conditions (JORISSEN, 1988;DEBENAY, 1990;ALMOGI-LABIN et al., 1992;DEBENAY et al., 2001;DEBE-NAY & GUILLOU, 2002;DEBENAY & GUIRAL, 2006;MUR-RAY, 2006;VIDOVIĆ, 2010;FRONTALINI et al., 2011a). In previously conducted research, the ability of A. tepida to tolerate environmental stress has been emphasized (ALMOGI-LABIN et al., 1992;DEBENAY et al., 2001;DEBENAY & GUILLOU, 2002). Abundance of this species in the organic matter-rich Arcij marine pond further supports this. A significant number of living specimens was observed, implying their adaptation to the environmental conditions in the investigated marine pond. Ammonia specimens with abnormal test morphology have also been recognized, further indicating the existence of environmental stress or possibly genetic or mechanical influences (ALMOGI-LABIN et al., 1992). A relatively small number of living deformed specimens (6 specimens; Tab. 4) was observed in the analysed sample, which does not facilitate explanation of the dominant factor causing the stress in the Arcij pond. Oxygen deficiency and nutrient abundance in the sediment could be possible factors.
In the Podbrajde marine pond, located 136 m from the sea, only 7 foraminifera specimens have been found in the analysed core top (Tab. 5). This general absence of abundant foraminiferal assemblages, in comparison to the other marine ponds, could be explained by the greater distance of this environment from the sea and a more prominent disconnection from a direct marine influence. However, measured E c values indicate the existence of normal marine conditions which makes it difficult to explain the lack of rich foraminiferal assemblages. Geochemical sediment analysis revealed similar conditions to those in the previously de-scribed marine ponds, especially the Marinska pond, with nutrient enrichment and an algal source of organic matter (MEYERS, 1994;MEYERS, 2003;LAMB et al., 2006) (Tab. 4). The Mo in the Podbrajde marine pond also seems to be of terrestrial origin (Fig. 11C) (GOLDBERG et al., 1996). The only notable difference in the geochemical record, in comparison to the previously described ponds, is the high Pb concentration (Fig. 9C). Generally, it is considered that elevated concentrations of Cd, Pb, Cu and Zn are a consequence of anthropogenic activities (CLARK, 2001). Sediment enrichment with Pb could also occur due to the spent shot during hunting (MATEO, 2009;MIGANI et al., 2015). The results from ROMANO et al. (2016) imply that metallic Pb from spent shot during hunting is oxidized and dissolved in wetlands leading to its mobilization and redistribution in wetland sediments. MIGANI et al. (2015) came to similar conclusions for lagoons on the northern Adriatic coast. SUOKHRIE et al. (2017) provided an overview of studies related to foraminifera exposed to different pollutants including heavy metals. Most studies indicate that high Pb concentrations in marine sediments result in low diversity of fauna and the predominance of opportunistic species, as well as an increase in the abnormalities of foraminiferal tests. FRONTALINI et al. (2015) exposed Ammonia parkinsoniana specimens, cultured in mesocosms, to various concentrations of Pb in sediments. Specimens showed cytological modifications that might be related to pollutant-induced stress. The elevated concentrations of Pb could possibly inhibit the development of abundant foraminiferal assemblages within the studied coastal marine pond. If the concentrations of Pb in the studied marine pond have a major influence on the lack of an abundant foraminiferal fauna, then correlation of modern environments with those of the past (pre-flooding) could be somewhat limited.
The presence of foraminifera in environments isolated from a direct marine influence has been attributed to transport by birds (DEBENAY, 1990;ALMOGI-LABIN et al., 1992). The same transport mechanism can explain the foraminiferal presence in the investigated marine ponds. However, seepage through the karstified underground is probably the dominant factor contributing to the foraminiferal dispersal. Sirocco and bora winds, that are common along the eastern Adriatic coast (PANDŽIĆ & LIKSO, 2005;SIGNELL et al., 2010), could also introduce marine fauna into the ponds.

Palaeo-marine ponds and the palaeoenvironmental evolution of the Jaz and Sonte embayments
Differentiated units from sediment cores collected in the submerged dolines in the Jaz and Sonte embayments, correspond to different Holocene palaeoenvironments. Foraminiferal analysis was conducted only in intervals were the existence of a transitional terrestrial to marine environment was assumed. Furthermore, recognized foraminiferal fauna and geochemical data were compared to the data from the Marinska, Podbrajde and Arcij marine ponds in order to possibly detect Holocene analogs of the marginal marine environments that nowadays exist in the coastal zone of Cres Island. Foraminiferal assemblages from the core tops collected in the embayments were also investigated in order to compare typical shallow marine assemblages in the area with the taxa present downcore and in the marine ponds. A similar type of research has already been conducted in the Adriatic Sea in the Venice lagoon (SERANDREI BARBERO et al., 2004).
Two distinct intervals were recognized in the sediment core LK-2 collected in the Jaz embayment. The first interval (unit LK-2-1), comprising the basal part of the core, implies soil accumulation with high metal concentrations (Al, Fe, K, Cu, Pb) ( Fig.  10) in the karst depression. The geochemical and mineralogical signature is typical for terra rossa soils developed in carbonate terrains (DURN et al., 1999;MIKO et al., 2001). Foraminiferal absence is a prominent characteristic of this environment, which further supports the palaeoenvironmental interpretation (Figs. 12A and 12B). The abrupt change in the core data (unit LK-2-2) (Fig. 5) was observed at approximately 711 cal BP evidencing an important environmental transition. A significant rise in the Mo and S concentrations (Fig. 5) probably indicates establishment of an oxygen depleted restricted water body (PEDERSEN, 1989;CRUSIUS et al., 1996;CALVERT & PEDERSEN, 1993;ALGEO & LYONS, 2006;SCHOLZ et al., 2017). The high TOC content supports this, while also implying increased primary productivity (Fig. 5). However, it seems that this environment was not as productive as the present-day marine ponds on Cres Island or these differences could be a consequence of diagenesis (Figs. 9F and 9G; Fig. 11B). The C/N ratio is indicative of a mixed terrestrial and algal organic matter provenance, which also differs in comparison to the present-day marine ponds (Fig. 11D) (MEY-ERS, 1994;MEYERS, 2003;LAMB et al., 2006). At this transition in core LK-2, a more detailed analysis of foraminiferal assemblages was conducted (Fig. 3). The presence of foraminifera proved that the marine influence in the Jaz embayment begun at approximately 711 cal BP. A poorly diversified assemblage was determined, predominantly composed of numerous specimens of foraminifera typical of the brackish or shallow marine conditions usually established in the marginal marine environments (A. tepida, A. parkinsoniana and H. depressula) (MURRAY, 2006;VANIČEK et al., 2000). Differentiated species seem to be well adapted to the reduced oxygen availability in the sediment. The susceptibility of A. tepida and the genus Haynesina to environmental stress has been previously determined (ALMOGI-LABIN et al., 1992;DEBENAY et al., 2000;DEBENAY et al., 2001;DE-BENAY & GUILLOU, 2002;VIDOVIĆ et al., 2009). However, A. parkinsoniana is not considered to be a stress-tolerant species (VIDOVIĆ et al., 2014). Cluster analysis enabled the identification of different core intervals characterized by similar assemblages (Fig. 7). We consider the similarity of samples from the transitional zone of the Jaz core and samples from the Arcij marine pond as evidence of the establishment of similar environments. Therefore, a palaeo-marine pond with a strong marine influence, analogous to the present-day Arcij marine pond, developed at approximately 711 cal BP when the sea-water level approached the depth of the sill at -0.5 m (Fig. 12C). However, statistical analysis of geochemical data does not fully support this interpretation (Fig. 10), possibly due to differences in the geological setting of the Arcij marine pond and the Jaz embayment where siliciclastic input is more prominent (Fig. 10). It is probable that this environment existed for a very short time before it was flooded with sea-water and the Jaz embayment was formed.
The LK-2 core top analysis implies the existence of an oxygen depleted sedimentary environment in the present-day Jaz embayment (Fig. 12D). Geochemical analysis of the surface sediment revealed high Mo concentrations (8.2 mg/kg) and lower TOC and TIC contents in comparison to the investigated inland marine ponds. A significant difference is in the nutrient availability and source of organic matter, with a high C/N ratio indicating mixed algal and terrestrial organic matter origin (MEYERS, 1994;MEYERS, 2003;LAMB et al., 2006). The core top sample proved to contain a highly diversified foraminiferal assemblage in comparison to present-day marine ponds, with 32 recognized species. The most common species A. tepida, A. parkinsoniana, Ammonia sp., B. striatula, E. margaritaceum, E. translucens and H. depressula are typical for organic matter enriched marine environments along the eastern Adriatic coast (VIDOVIĆ, 2010).
Palaeoenvironmental development of the Sonte embayment encompasses a longer time span (Figs. 12A and 12D). The present-day Sonte embayment has a maximum depth of 5 m, while the sill depth is 3 m. Accumulation of the sediment sequence was possible due to the morphology of this depression. Analysed samples from the lowermost part of the core exhibit similarity in geochemical and mineralogical composition to the basal part of the LK-2 core (Fig. 10) and indicate terrestrial soil (DURN et al., 1999;MIKO et al., 2001). Within this interval (unit LK-3-1) poorly preserved specimens of foraminifera occur in low numbers, making it difficult to interpret an established environment. The rare specimens were probably deposited due to transport by waves and/or winds as a consequence of the gradual Mid-Holocene sea level rise on the seaward side of the karstified sill/barrier. Further up-core (unit LK-3-2), a significant rise in the TOC content was observed (Fig. 9I). We suggest development of a stagnant water body at approximately 8000 cal BP (according to the age-depth model) on previously formed soil. Increased nutrient availability could facilitate organic matter production, similar to the Marinska, Podbrajde and Arcij marine ponds (Fig. 11B). However, most of the organic matter has a terrestrial source possibly indicating input from the land under a different climatic setting or the enhanced growth of terrestrial plants (Fig. 11D). This interval could be correlated with the Holocene pluvial period recognized in the Adriatic Sea (WUNSAM et al., 1999;SCHMIDT et al., 2001) and in the lacustrine sediments from karst poljes and lakes located along the eastern Adriatic coast (SCHMIDT et al., 2000;BALBO et al., 2006;ILIJANIĆ et al., 2018).
The established environment was probably poorly oxygenated, which enabled the preservation of organic matter. Relatively high values and covariation of Mo and TOC corroborates this conclusion (CRUSIUS et al., 1996;CALVERT & PEDERSEN, 1993;ALGEO & LYONS, 2006;SCHOLZ et al., 2017). The Mo concentrations are significantly higher in comparison to the present-day concentrations in the marine ponds (Fig. 9D), indicating low oxygen abundance due to the development of a restricted environment with water stratification. The S concentrations are high and similar to those in the Marinska, Arcij and Podbrajde marine ponds on Cres Island, further supporting the development of a restricted and oxygen poor water body (Fig. 9E, Fig. 11A). Dominant taxa, A. tepida, Ammonia beccarii, H. depressula, Asterigerinata mamilla, Porosononion sp., Elphidium fichtelianum and C. gerthi imply the existence of shallow marine to possibly slightly brackish water environmental conditions (MURRAY, 2006). A. tepida and A. beccarii have been frequently observed in the sediments along the eastern coast of the Adriatic Sea with high P and TOC content (VIDOVIĆ et al., 2014). The presence of the genus Haynesina can also be correlated with the enrichment of TOC in sediment (DEBENAY et al., 2001;VIDOVIĆ et al., 2009;VIDOVIĆ et al., 2014). Statistical analysis demonstrated the highest similarity of the foraminiferal assemblages from this unit with the present-day Arcij marine pond and previously described transitional zone in the core from the Jaz embayment. This suggests development of a palaeo-marine pond with normal marine water (Fig. 12A). However, such similarity was not observed in the statistically analysed geochemical data, where this unit correlates well with the topmost part of the Jaz embayment core therefore implying similarity of these environments (Fig. 10). Micropalaeontological data and geochemistry therefore suggest different palaeoenvironments, but generally provide evidence of a significant marine influence during the deposition of sediments from unit LK-3-2.
At 6610 cal BP (unit LK-3-3) a decrease in TOC content and Mo concentrations, followed by an increase in TIC indicate major environmental change in comparison to the previously described unit (Fig. 6). Carbonates become a more important sediment component (Fig. 9H, Fig. 10). Surface sediments deposited along the present-day eastern Adriatic coast are also enriched in carbonates (PIKELJ et al., 2009;PIKELJ, 2010). The newly formed environment was nutrient-deprived, which prevented organic matter accumulation (Fig. 11B). However, preserved organic matter has mixed algal-terrestrial origin (Tab. 4, Fig. 6) (MEY-ERS, 1994;MEYERS, 2003;LAMB et al., 2006). Foraminiferal assemblages became highly diversified and dominated by A. tep-ida, E. translucens, A. beccarii, Ammonia sp., C. gerthi, E. fichtelianum, Porosononion sp.1, Q. parvula and Q. seminula specimens. Dominant species are similar to the previously described interval. However, species richness and diversity increased significantly. Increases in the relative abundance of miliolids were also observed. This can suggest the existence of an enclosed environment (DEBENAY & GUILLOU, 2002;LIDZ & ROSE, 1989) or hypersaline lagoon (DEBENAY et al., 2001). Genus Quinqueloculina is abundant in the Mljet Lakes (ĆOSO-VIĆ et al., 2016), while DEBENAY & GUILLOU (2002) reported the association of A. tepida and Q. seminula in the subtidal areas, marshes and mudflats with developed seaweeds. According to VIDOVIĆ (2010) miliolids are common in marine environments along the Croatian coast of the Adriatic Sea. The determined assemblage, predominantly comprising the genera Ammonia, Elphidium, Haynesina and Quinqeloculina, can be compared to the previously recognized Haynesina-Ammonia assemblage from the shallow marine environments in the Soline embayment and Nin Bay (VIDOVIĆ, 2010). Cluster analysis grouped samples from this unit into the same subcluster as the surface sample from the Jaz embayment where shallow marine environmental conditions prevail today. Statistical analysis of the geochemical data also indicated this similarity. It is our interpretation that at 6610 cal BP a marine influence became more prominent in the Sonte embayment, possibly with the sea water spilling over the sill (Fig.  12B). This is in general agreement with the published global and regional sea level curves (CORREGGIARI et al., 1996;WAEL-BROEK et al., 2002;LAMBECK et al., 2014;VACCHI et al., 2016).
The topmost part of the core (unit LK-3-4) can be geochemically distinguished from the previously described core intervals (Fig. 6). This is probably indicative of the establishment of a more permanent, fully marine environment (Fig. 12D). A highly diversified foraminiferal assemblage consisting of 41 recognized species was determined in the LK-3 core top. The dominant Sonte embayment foraminiferal fauna (G. praegeri, H. depressula, H. germanica, A. tepida, A. mamilla and Bolivina pseudoplicata) is typical for littoral environments in the Adriatic Sea (JORISSEN, 1988;VIDOVIĆ, 2010).
In the investigated embayment, the determined surface assemblage is present in the sand dominated sediment, with low Mo and S concentrations, low TOC content and high TIC values (Fig. 6). However, although the sediment is oxygenated, measured oxygen concentrations in the water were the lowest among all the investigated environments (Tab. 2), probably due to the greater depth of this environment. Primary productivity in the Sonte embayment is limited due to the nutrient deficiency (Fig.  11B) and therefore organic matter content is low (Fig. 9I). Cluster analysis grouped foraminifera from this sample into the same subcluster as the previously described marine samples (Fig. 7). Statistical analysis of the geochemical data indicated a similarity to other marine units from the LK-3 and LK-2 cores (Fig. 10).

CONCLUSIONS
Marine ponds have been developed in the coastal karst dolines along Cres Island. Sediments preserved in the Marinska, Arcij and Podbrajde marine ponds revealed important micropalaeontological and geochemical data that was used to characterize these marginal marine environments. Detailed analysis of sediment cores from the Jaz and Sonte embayments, developed in now submerged dolines, revealed the complex palaeoenvironmental evolution of the investigated area. The downcore micro-palaeontological and geochemical data enabled detection of the Holocene palaeo-marine ponds. We suggest the development of a palaeo-marine pond, similar to the present-day Arcij pond, at approximately 711 cal BP in the Jaz embayment. A palaeo-marine pond in the Sonte embayment existed up to 6610 cal BP. These palaeo-marine ponds were flooded during the Holocene sea level rise, when the marine environment was established. Therefore, recognition of different environments developed in the karst dolines, located in the coastal zone of Cres Island, was possible. The research of the present-day marginal marine environments along the eastern Adriatic coast could prove to be valuable for palaeoenvironmental studies and especially Holocene sea level change research.