347

Journal of Paleolimnology 17: 347–367, 1997. c 1997 Kluwer Academic Publishers. Printed in Belgium.

Climatic and non-climatic lake-level changes inferred from a Plio-Pleistocene lacustrine complex of Catalonia (Spain): palynology of the Tres Pins sequences S. A. G. Leroy Centre of Palaeoecology, Queen’s University of Belfast, Belfast BT7 1NN, N. Ireland, UK (e-mail: [email protected]) Received 11 September 1995; accepted 12 September 1996

Key words: carbonate content, pollen, lake level, seasonality, climate, Spain, Plio-Pleistocene

Abstract A 27-m sequence of deposits from the Plio-Pleistocene Banyoles-Besal´u lacustrine complex, at Tres Pins, N.E. Spain, shows lithological, carbonate, pollen and spores, and ostracode-gastropod evidence of climatic and lake-level change. Upland pollen taxa from the lowermost zones 1, 3 and 5, show that the area was forested and indicate a progressive deterioration of the climate. Zone 2 (steppe vegetation) corresponds to a global glacial stage, with mild temperatures, if the comparison to modern analogues is valid. Only minor climate fluctuations occurred subsequently. Aquatic vegetation and micritic sediment facies show a pelagic environment during pollen zones 1 and 3 and a littoral one during pollen zone 5. A short-term lowering of the water level (sandy algal micrite and slightly brackish waters) occurred in zone 2, as a result of severe droughts. Subzone 2c records progressive recovery of the forest. The coarse lithology and the high carbonate content, however, continue to indicate shallow waters. The fluctuation marked by zone 4 (extensive marsh vegetation and spring waters) may be due to a lake-level decrease caused by karst activity, or by lower precipitation with only slight cooling, or more probably by a relative lake-level decrease caused by natural infilling. The sediment of pollen zone 4, a sandy algal micrite, indicates the development of a littoral bench at the core site. From pollen zone 4 to pollen zone 5 an evolution from lakeward to landward position occurred. In zone 2, droughts existed at least during the spring growing season, and probably also during the rest of the year. In subzone 2c, a shift to spring precipitation occurred. In contrast, during pollen zone 4, if there was a decrease in precipitation, it did not take place during the summer growing season. Introduction The Mediterranean area is a transitional zone between Europe, which has undergone temperature changes linked to extensive movements of the ice cap, and North Africa, where the deserts varied in extent over several millions of years. In the Mediterranean area, the roles of temperature and precipitation as limiting ecological factors are intermingled. Under climate stress the arboreal vegetation has often given way to steppe caused by low rainfall. However, the temperatures associated with a steppe vegetation (‘warm’ versus ‘cold’ steppes) are hard to determine from pollen spectra. Pollen analyses of the site of Tres Pins, Spain, can contribute to these frequently raised questions.

*121782

The history of seasonality in the Mediterranean area is relatively well known. Already during the Miocene in Catalonia and S. France, the subtropical climate varied seasonally (Bessedik, 1985). Suc (1984) showed that summer drought, characterising the modern Mediterranean climate, was already developed around 3.2 Ma ago for at least some lengths of time. As a response to Plio-Pleistocene glacial-interglacial cycles, the Tres Pins site has known shift in the season when the rainfall was high enough to sustain vegetation, in the same way as happened for the last climate cycle (Prentice et al., 1992). Lake-level fluctuations can be used to reconstruct palaeohydrological and palaeoclimatic variations when several lakes show synchronous changes

Article: jopl 357 Pips nr 121782 BIO2KAP jopl357.tex; 13/05/1997; 11:59; v.5; p.1

348

Figure 1a. Map showing the location of Tres Pins in Catalonia, Spain.

at the regional level. Past lake-level fluctuations caused by climate indicate changes in precipitation, net radiation and temperature, related to variations in atmospheric circulation (Gaillard & Digerfeldt, 1991; Harrison & Digerfeldt, 1993). Distinguishing between climate and non-climate driven changes of lake waterdepth is difficult. Only multidisciplinary studies and the accumulation of information from several independent drainage basins can be satisfactory. For a PlioPleistocene lake in a region where there is only one known lake remnant of that age, it becomes a challenge. Analyses of the Tres Pins sequences based on lithology, carbonate content, pollen analyses (upland and aquatic vegetations) and some information from ostracode-gastropod assemblages show both climatic and non-climatic environmental changes. Continental Plio-Pleistocene deposits are especially difficult to date because they are out of the range of many radiometric methods such as U/Th and radiocarbon, and also because the available methods (relative or absolute) often give a too broad confidence interval if one wishes to correlate the record to oceanic isotopic stages.

Figure 1b. Simplified geological map showing the location of Tres Pins and B`obila Ordis. 1: lake 3, drillholes BO II and BO III; 2: lake 3, drillhole BO I; 3: lake 1: drillhole BO IV; 4: lake 2: outcrops BOC III and BOC IV; 5: drillhole TP I; 6: drillhole TP II; and 7: water well of the Tres Pins bar.

An additional complication linked to old sediments in karstic areas emerges from a cumulative story of post-infilling events, such as differential compaction, faulting, erosion, etc. (Løvlie & Leroy, 1995). Owing to the problems in dating short continental Plio-Pleistocene sections, it has always been difficult to estimate sedimentation rates, and therefore durations of environmental changes. In consequence, Milankovitch forcing was demonstrated on long continuous marine sections. Once the cyclostratigraphy was established and the influence of astronomical forcing demonstrated for vegetation changes obtained from marine sequences (Leroy & Dupont, 1994; Combourieu-Nebout, 1993), the method could be applied to continental sections (Leroy & Seret, 1992; Leroy & Roiron, in press; Leroy & Løvlie, 1995; Leroy et al., 1994). Previous work in southwest Europe includes several isolated, relatively poorly dated, and often short continental sections (Elha¨ı, 1966, 1969; Ablin, 1985, 1991; Remy, 1958; Br´enac, 1984). Other continen-

jopl357.tex; 13/05/1997; 11:59; v.5; p.2

349 tal sites are better dated, either by convergence of chronological information, e.g. at Nogaret and Bernasso, S. France (Leroy & Seret, 1992; Leroy & Roiron, in press), or by palaeomagnetism and the presence of several consecutive cycles, e. g. at Leffe, N. Italy (Ravazzi & Rossignol-Strick, 1995). In Portugal (Diniz, 1984), two boreholes made in littoral marshes have been tentatively dated by correlation of flora and climate changes to other European palynostratigraphies. Longer and better dated sections have been obtained from marine sediments either on-shore (Combourieu-Nebout, 1993, 1995) or off-shore (Suc & Cravatte, 1982), where dating relies mostly on foraminifera assemblages, and for the more recent work, on correlation to marine isotope stages. The aims of this study are therefore to contribute to the vegetation history of the area, and to help establish criteria to distinguish between climatic and nonclimatic changes, as well as to determine temperature during glacial periods and in which season of the year rain falls, a major limiting factor for vegetation growth in the Mediterranean basin.

Study area and site description TP I and TP II sediment cores were drilled at Tres Pins, 3 km north-west of Banyoles, Catalonia, Spain (42080 4500 N., 2 430 5000 E., alt. 210 m a.s.l.), within a few hundred meters of the B`obila (i.e. brickyard in Catalan) Ordis (190 m a.s.l.) (Figure 1a and b; Figure 2). The present Lake Banyoles (altitude 173 m a.s.l.; Figure 2) is the remnant of a much larger PlioPleistocene lacustrine complex reaching northwards as far as Besal´u and Crespi`a, 4 km wide west-east by 6 km long north-south. The form of the present lake is that of joined subcircular depressions with, in most, a central deep and narrow channel. In some of these, springs are active. Each sinkhole is surrounded by shallow water shelves. Lake Banyoles is endorheic with warm sublacustrine springs feeding the lake through the channels. For example, a 40-m deep channel has a 18  C spring, and a salinity of ca 1.2‰. The temperatures of the artesian springs around the lake range from 19  to 16  C, and those of the springs in the nearby city of Banyoles from 7.6  to 9.6  C. In the whole area, the temperature of the aquifer is around 18  C (Juli`a, 1980). The karst is very active as seen by the recent formation of new small lakes, collapsing of roads, etc. (Figures 2 and 3). In this context, non climatic lake-level changes can be caused by a variety of local factors,

such as karst, natural infilling, damming by travertine, etc. The mean annual air temperature measured at Banyoles is 15.5  C and the mean annual rainfall is 750 mm. The climate is characterised as humid Mediterranean. Pliocene and present-day vegetation shows a clear climate limit near the latitude of Barcelona (Suc et al., 1995). Several Plio-Pleistocene palaeoclimatological studies have been made in the Banyoles-Besal´u basin: pollen and ostracode assemblages in the seasonal rythmites at Mol´ı Vell (6 km north of Tres Pins) (De Deckker et al., 1979), macroflora and macrofauna at Crespi`a (6 km to the north-north-east) (Roiron, 1983; Villalta & Vicente, 1972) as well as palynological investigations in B`obila Ordis (amongst others Juli`a & Suc, 1980; Leroy, 1987, 1988 and 1990; Suc et al., 1992; Løvlie & Leroy, 1995). These and other ancient deposits are now dislocated and collapsed due to continuous karstic activity in the underlying Eocene gypsum (Figures 1b and 3). Clays worked out at B`obila Ordis are thought to represent the infilling of a lake in a karstic sinkhole and the sequence drilled at Tres Pins most probably has a similar origin, though is not necessarily of the same age. B`obila Ordis had been put forward as a keysection for the Mediterranean equivalent (Pl. III-zone) of the Waalian interglacial (Suc, 1984; Zagwijn & Suc, 1984). Further work from new outcrops and drillings revealed the existence of three nested lakes, numbered from the oldest to the youngest (Leroy, 1990). The clays of the brickyard are attributed to the youngest of the three lakes: lake 3. The pollen diagram of lake 2 represents a steppe (Leroy, 1988). New rodent tooth discoveries, identification of the palaeomagnetic Cobb Mountain sub-chron, and the introduction of the cyclopalynostratigraphy concept (Leroy, 1990; Leroy & Dupont, 1994; Løvlie & Leroy, 1995) allowed revision of the age of the oldest known lake deposit in B`obila Ordis (lake 1): one of the interglacial periods around 1.2 Ma, possibly oxygen isotope Stage 35. Despite the proximity of B`obila Ordis and Tres Pins and the presence of outcrops along the road, there is no possibility of correlating the infilling of these two sinkholes, separated by the Usall limestone plateau (Figure 1b).

jopl357.tex; 13/05/1997; 11:59; v.5; p.3

350

Figure 2. Aerial photograph of Lake Banyoles reaching northwards to B`obila Ordis (lower arrow) and to Tres Pins (upper arrow) (July 1988, S. Leroy). White travertines in the foreground surround one of the main circular depressions (25 m deep) open to karstic upwelling.

Figure 3. Collapse of the road 150, 200 m north of Lake Banyoles, by normal concentric faults at Can Guixeres (June 1989, S. Leroy).

jopl357.tex; 13/05/1997; 11:59; v.5; p.4

351 Methods A 2720-cm long and 40-mm wide hole (TP II) was made in 1985 using a drilling rig with a gouge auger made by the Laboratoire de Pal´eontologie et Pal´eog´eographie of Universit´e Catholique de Louvain, Belgium. The subsampling of the auger in 3 to 10 cmlong sections had to be done in the field because of the absence of a liner. A parallel hole, TP I (1940 cm deep), was made at ca 20 m north from hole TP II in 1983 (Figure 1b), but was not studied in detail because of its accidental destruction. A brief on-site description of a water well at 25 m north-north-west of the TP II hole is also available (Juli`a, 1980). The carbonate content was measuredon 50 dry samples weighed before and after digestion in 10% HCl (Figure 4) (Collet, 1987; pers. commun.). A survey of the ostracode assemblages was completed (Coen in: Leroy, 1990). Due to irregular numbers of valves, no detailed diagram can be presented. Therefore occurrences are used to reconstruct palaeoenvironments. Six test samples have been analysed for their gastropod content (Magnin in: Leroy, 1990; Magnin, pers. commun.). Only four contained enough material to allow for autecological interpretations (Magnin, 1991; Glo¨er and Meier-Brook, 1994). At first, 87 palynological samples were taken between 2712 and 1017 cm depth (Leroy, 1987) and prepared by the method described in Dricot & Leroy (1989). An additional 25 samples were later added in order to refine the climate reconstruction (Leroy, 1990). Five samples are devoid of pollen in the upper 1000 cm of the drillcores, as a consequence of oxidisation, with the exception of one relatively poor sample not illustrated here. Two samples lack pollen at the base of the sequence at the contact with the substratum. Detailed diagrams based on 105 pollen rich samples, and 54 662 pollen grains and spores) are presented in Figures 5 and 6. The average number of pollen and spores counted is 520 per sample, with a minimum of 100 grains counted in addition to Pinus, Cyperaceae and spores. The gymnosperms are plotted first, then the other trees, followed by the herbaceous elements in Figure 5. Figure 6 shows the aquatic elements including the Cyperaceae, the spores and the varia (unknown and indeterminable grains). The total sum (called sum 1) is used for the percentages of the aquatics, spores and varia in the aquatic diagram (Figure 6). When the taxa of the last 3 groups (aquatics, spores and varia) are subtracted from this sum, it is called sum 2, the basic sum chosen for the

Figure 4. Lithological log (left) and carbonate content curve (right) of the TP II sequence. Lithological symbols: horizontal dashed lines: micritic sediment; dots: sandy algal micrite.

percentages of the upland pollen diagram (Figure 5). On the whole, the grains are well preserved with only ca 4% indeterminable grains, without any zone of especially poor preservation. The concentrations have been calculated for each sample (Figure 7). In total, over a hundred taxa were identified. Quercus ilex type is rare and only present in ca one third of the samples at less than 1%. Determination of a Cyperaceae-type and a non-psilate monolete spore showed that they are

jopl357.tex; 13/05/1997; 11:59; v.5; p.5

Figure 5. Detailed upland pollen diagram from hole TP II. Sum 2 (aquatics, spores and varia subtracted from the total), the basic sum chosen for the percentages of the upland pollen diagram is used for the percentages. A ten times exaggeration curve has been plotted. Pollen zones were visually determined. The Psimpoll program 2.27 (Bennett, 1995) was used to draw the pollen diagrams.

352

jopl357.tex; 13/05/1997; 11:59; v.5; p.6

Figure 5. Continued.

353

jopl357.tex; 13/05/1997; 11:59; v.5; p.7

Figure 6. Detailed aquatic diagram for aquatics, spores and varia. The sum is the total number of pollen grains and spores (sum 1). A ten times exaggeration curve has been plotted. Pollen zones were visually determined. The Psimpoll program 2.27 (Bennett, 1995) was used to draw the pollen diagrams.

354

jopl357.tex; 13/05/1997; 11:59; v.5; p.8

355

Figure 7. Palynological concentration curves in number of pollen grains and spores per g of dry sediment (gr. g

respectively Cladium cf. mariscus and cf. Thelypteris palustris, illustrated by Leroy (1987: Pl. 1, Figures 2, 5 and 11). Most of the Cyperaceae are Cladium cf. mariscus, although some other types are present in zone 4. Most of the non-psilate monoletes are cf. Thelypteris palustris. It can be assumed that most of the psilate monoletes have been produced by the same fern taxon because they are often associated with sporangia. Diatoms occur only rarely, probably owing to dissolution caused by the sediment alkalinity. A visual zonation of the palynological diagram has been made by the author (Figure 8B). The five zones are in good agreement with the lithology, the carbonate content curve and the fauna. Statistical zonations based on the constrained cluster analysis with a dissimilarity coefficient using Euclidian distance (CONISS), is provided by the Psimpoll program 2.27. It was first applied to the upland diagram, and then to the aquatic, spores and varia diagram, providing two additional zonations (Figures 8A and C). Figure 8 shows zonation with agglomeration from 14 zones up to one. In addition a dashed line indicates the 5-zone level.

1 ).

Lithology, ostracodes and gastropods In TP II drillhole, grey marls, attributed to the regional Eocene formations (Juli`a, pers. commun.), were reached at 2720 cm depth. From 2720 to 1000 cm, the sediment is a micrite with two sandy algal (Characeae) micrite interruptions from 2412 to 2290 cm and from 1577 to 1472 cm (Figure 4). The last 1000 cm are oxidised. The preliminary description of TP I hole indicates a coarser sediment with the same last 10 m of oxidised material. The water well went through ca 4000 cm of organic rich micrite, with some peaty horizons (e.g. at 3000 cm depth) (Figure 1b) (Juli`a, 1980 and pers. commun.). Within a relatively small area (less than 20  20 m), the three sequences show different lithofacies. Most of the sediments recovered at Tres Pins are marls with an average carbonate content of about 40% (Figure 4). Sections with higher carbonate values (>50%) are found at 2560–2290 cm and 1550–1440 cm depth. This high carbonate content is explained by the abundance of calcareous algal remains. SEM and binocular magnifying glass observations show that the sand-size fraction is mainly made up of encrustation on charophyte stems, with numer-

jopl357.tex; 13/05/1997; 11:59; v.5; p.9

356

Figure 8. Statistical and visual zonations for 105 samples of TP II drillhole. The statistical zonation is based on a dissimilarity coefficient which is the Euclidian distance (CONISS). A: upland taxa on sum 2, minimum value for inclusion of a taxon is 0.04, 15 taxa. B: visual zonation. C: aquatics, spores and varia on sum 1, minimum value for inclusion of a taxon is 0.02, 5 taxa.

of vegetation overgrowth (De Deckker, 1979; Robinson, 1980). As pointed out by Hiller (1972) from recent occurrences in northern Germany, true freshwater is clearly a more constraining factor. The first gastropod sample from 2697 to 2687 cm depth (zone 1) contains only aquatic elements: Bithynia tentaculata (L.) and unidentified aquatic snails. The second one, from 2462 to 2450 cm depth (zone 2), also contains a large assemblage of freshwater malacofauna suggesting stagnant waters with a rich aquatic vegetation: Gyraulus crista (L.) predominant, G. laevis (Alder), Radix cf. auriculata (L.), Pisidium sp. and Bithynia tentaculata (L.). A littoral environment is suggested by G. crista. Some species indicate an environment that occasionally dries up. The salinity reaches a maximum of 3.5‰. The penultimate sample, from 1922 to 1905 cm depth (zone 3), is rich in freshwater gastropods: G. crista (L.), Islamia sp., R. cf. auriculata (L.) and B. tentaculata (L.), assemblage characteristic of standing hard waters. Islamia is frequent in karstic springs. The last sample (zone 5) between 1370 to 1345 cm depth has an extremely fragmented land snail fauna indicating dense, humid and cool temperate forest and marshes with Discus sp., Carychium sp. and Vertigo angustior (Jeffreys). There are no more freshwater species.

Palynology of drillhole TP II ous gastropod faecal pellets. SEM examination reveals only calcite (most probably low-Mg calcite, according to the crystal form) crystallites, and a few flakes of a clay mineral (M. Coen, pers. commun.). The ostracode frequency distribution roughly follows the same pattern as the carbonate content curve, i.e. 5 zones. Whereas zone 1 does not contain any ostracodes, the assemblages of the other zones are dominated all through by Cyclocypris sp. Candona neglecta and Darwinula cylindrica are frequent. Potamocypris fallax, present in zone 2 but mostly in zone 4, is a cold stenotherm species known to live in springs and waters flowing from springs (Meisch, 1984). The waters of the present Lake Banyoles are warm and Potamocypris is not found in the present lake. Heterocypris salina (salinity from 0.5 up to 20‰, with an optimum at 5-6‰) together with populations of mostly noded Cyprideis torosa (salinity 5 to 6‰) indicates slightly brackish water. They are both present in zones 2 and 4, but mostly in zone 2. Metacypris cordata (zones 3 and 5) has been considered as an indicator

Upland pollen diagram Description Pollen zone 1 (2712 to 2616 cm) The total concentration of pollen grains and spores is low (Figure 7). The pollen percentages of Quercus, Carpinus, Carya and Ulmus-Zelkova indicate that they probably were the four main components of a deciduous forest, besides Pinus. Less important are Ericaceae, Pterocarya, Tilia and Fagus. The arboreal taxa are well diversified. The herbaceous pollen are dominated by Gramineae. Pollen zone 2 (2616 to 2286 cm) The concentration curve fluctuates and shows two groups of samples with lower values from 2607 to 2535 cm depth and from 2422 to 2362 cm depth. In this zone, the pollen spectra are very different. Pinus pollen grains dominate throughout, rising to 88%

jopl357.tex; 13/05/1997; 11:59; v.5; p.10

357 (Figure 5). The herbs are well represented too, especially Gramineae and Compositae, mostly Liguliflorae. However, Artemisia pollen percentages remain low. Ericaceae decrease above 2535 cm depth. Populus, Pinus haplostellate type and Ephedra are well represented. The beginning and the end of this zone show peaks of conifer pollen (Picea and Abies). Pollen subzone 2b, from 2456 to 2404.5 cm depth (4 samples), displays high percentages of Pinus, maxima of Liguliflorae (>5%), low percentages of Carpinus, Quercus, Carya and Abies, and minimum of UlmusZelkova. The Gramineae are temporarily less well represented. Already in pollen subzone 2c, Quercus values increase in parallel with a rise in the pollen grains and spore concentration. Pollen zone 3 (2286 to 1633 cm) On the whole, concentrations are moderate. Pollen concentration is lower at 1770 cm. After a peak of Carpinus (28%), an assemblage characteristic of a deciduous forest again develops, with the same four dominant taxa as in zone 1, but the diversity of the arboreal taxa is slightly less. Parrotia cf. persica becomes continuously present. Pinus percentages first decrease (subzone 3a), then peak again in subzone 3b (from 1996 to 1977 cm, 4 samples) together with Abies and Picea. In pollen subzone 3b, the deciduous trees, except Carya, recede as do Artemisia and Gramineae. Alnus is absent. After subzone 3b, Carya percentages decrease, while Pterocarya and Artemisia increase. Three taxa have their maxima in subzone 3c: Myrica, Artemisia, and Rosaceae. There are regular occurrences of Betula. The representation of the conifers, other than Pinus, is poor. Pollen zone 4 (1633 to 1419 cm) Maximum total concentrations are reached during this pollen zone with >70 000 gr g 1 . However, low arboreal concentrations (excluding Pinus) are attained from 1530 to 1472 cm depth. Besides Pinus, Carpinus dominates the arboreal grains. Quercus clearly decreases below 10% (1555 to 1485 cm). The other trees (Ulmus-Zelkova, Alnus, Pterocarya), except Carya, are only slightly affected. There is a reduction in the most warm-loving trees, such as Parrotia cf. persica, Pterocarya, Parthenocissus, Eucommia, etc. Populus grains are present. Abies and Picea increase as well, up to 14 and 6.8% respectively. One of the few grains of Tsuga has been observed at 1542 cm depth, one of the levels with high Abies and Picea. In the same

samples, the percentages of pollen of deciduous forest species decrease. The herbaceous taxa do not develop more than in adjoining zones. Compositae (with the exception of Artemisia) slightly increase. Calystegia (a vine; >2% at 1555 and 1542 cm depth) and Salix (a pioneer genus) are characteristic elements. Pollen zone 5 (1419 to 1017 cm) The percentages of deciduous tree pollen slightly increase and the dominant forest taxa are slightly more diversified than in zone 3: Carpinus, Quercus, Carya and Ulmus-Zelkova. Eucommia grains become rare. A new taxon, Corylus, is now present throughout and reaches maximum values of 4.3%. Abies is well represented during the first half of this zone. Non-arboreal values, lower than in pollen zones 1 and 3, are dominated by Gramineae and Artemisia. The AP-Pinus concentration is on the average the highest of the diagram. Interpretation, reconstruction of the terrestrial vegetation Pollen zone 1 This dense and diversified warm temperate deciduous forest presents – with the exception of Carya now absent from Europe and western Asia – some similarities with the forests from the South Caspian and Colchis mid-altitude mountains (Zohary, 1973). The climate of those regions is warm temperate with a mean annual temperature of 13–17  C, very humid with 850– 2300 mm rainfall and no summer drought. The climate has a steep altitudinal gradient. Pollen zone 2 Such a xeric association corresponds to a steppe with Pinus stands. Quercus and Ulmus-Zelkova are the first trees to recover at the end of the zone. In surface samples, steppe and forest steppe spectra contain relatively low Pinus percentages (Bottema & Barkoudah, 1979; Wright et al., 1967; Van Zeist et al., 1970; Saadi & Bernard, 1989). I nevertheless attribute most of the pollen spectra of this pollen zone to a steppe or a forest steppe because the arboreal taxa mostly consist of Pinaceae undiff. and other conifers. In Plio-Pleistocene and recent lacustrine and marine sediments, a strong over-representation of this taxon is common (Heusser, 1988; Suc et al., 1995). This can be partly explained by morphology and by taphonomy. For Plio-Pleistocene pollen diagrams, it is

jopl357.tex; 13/05/1997; 11:59; v.5; p.11

358 therefore often the case to exclude Pinus and Pinaceae undiff. from the sum (Ablin, 1991; CombourieuNebout, 1993, 1995). Because the indeterminable pollen grains in the TP II pollen diagram are low at the Liguliflorae-rich levels, no over-representation owing to differential resistance to corrosion, as often observed by Bottema (1975), can be advocated. A climatic interpretation is therefore proposed for the pine steppe association characterised by low arboreal taxa, high Pinaceae and Liguliflorae percentages. Other Liguliflorae abundance in a similar context to Tres Pins are known from lake 3 (Juli`a & Suc, 1980: succession 3, phases 2 and 3) and lake 2 (in this case up to 31%) (Leroy, 1988: BOC III, and BOC IV 1, 3 and 4) of B`obila Ordis, next to Tres Pins (Figure 1b). However the ostracode assemblages in lake 2 (Potamocypris sp. and H. salina) indicate, at levels that are rich in Liguliflorae, spring and brackish waters and therefore give no direct climatic information (Coen in: Leroy, 1990; Coen, pers. commun.). Pollen spectra with 5% Liguliflorae are also known from the W¨urm Lateglacial at Olot, 19 km west of Tres Pins, Spain (P´erez-Obiol, 1988), but the composition of the spectra are very different from the one of Tres Pins. Liguliflorae are observed there with a mixed forest of Pinus sylvestris and Betula. In the Pliocene of the Alboran Sea, percentages of 15% Liguliflorae are found until 2.4 Ma (Suc et al., 1995). These high percentages illustrate the dry climate during the Pliocene of southern Spain and northern Africa. The fossil pollen record of the Lower Pliocene of Tunisia shows at least 20% to 30% of Liguliflorae (Suc et al., 1995). In a Late Pliocene deep-sea core off shore of N.-W. Africa, Liguliflorae are mostly present from 3.48 to 3.24 Ma when they exceed 10%. This has been interpreted in parallel with changes in oxygen isotopes as a global glaciation with local extension of the savannah and aridification of the Sahara (Leroy & Dupont, 1994). The comparison to modern analogues brings information from present-day steppes. Some surface samples from south-east Europe and the Middle East steppes produce high percentages of Liguliflorae (10– 30%) (Bottema, 1975). However they are always associated with Liguliflorae plants growing at the sampling point. Based on an analysis made by J. Guiot (pers. commun.) of surface samples, it appears that high percentages of Liguliflorae are found today under dry and warm conditions in Morocco and Greece. High percentages of Liguliflorae (from 5 to ca 40%) are always

found above 12  C mean annual temperature (with highest percentages between 12  C and 17  C) and mostly between 0 mm and 250 mm of precipitation. The value of such a comparison is however questionable as it is based on only one taxon, because the modern vegetation has been transformed by humans, and because the transfer function method based on the analogues showed to have poor success for glacial periods, as it is difficult to find modern steppe analogues for dry and cold conditions (only the inference of dry conditions is reliable) (Guiot et al., 1993). In the Late Pliocene of north-west Africa (Leroy & Dupont, 1994) and in the Last Glacial of New South Wales, Australia (Dodson & Wright, 1989), high percentages of Liguliflorae have also been found that have no equivalent in the modern vegetation of those regions and their neighbours. In conclusion, for zone 2, the pollen analyses indicate low rainfall, especially during the growing season (spring), hindering the development of a deciduous forest. The landscape is very open with nearly no trees. The pollen spectra here provide poor information on temperature, which might be similar to the present. It has been well documented that in the north-west and central Mediterranean and in north-west African regions, local dry climatic conditions correspond to a global glacial period (Suc & Zagwijn, 1983; Combourieu-Nebout, 1993; Leroy & Dupont, 1994). Subzone 2c gives signs of some climatic improvement as Quercus stands develop. However the species is unknown and some are drought-resistant. Pollen zone 3 Recovery of the forest continues with expansion of Carpinus, Carya, Abies and Picea before a stabilisation of the vegetation. If climate-controlled, subzone 3b fluctuation could be connected to decreased temperatures and maintained humidity. Such minor climate fluctuation might have been evidenced by higher time resolution at those depths. It is not impossible that similar phenomena occurred undetected elsewhere in the sequence. Nevertheless, pollen subzone 3b is at the boundary of two different pollen subzones: 3a and 3c. The climate of subzone 3c might have been slightly cooler as shown by taxa such as Betula and Artemisia. However the departure from subzone 3a remains short. Zone 3 indicates humid conditions on the whole with some fluctuations of temperature. Subzone 3b is the coolest, subzone 3a is the warmest. Zone 3 diversity has been slightly impoverished in comparison to

jopl357.tex; 13/05/1997; 11:59; v.5; p.12

359 zone 1, most probably owing to the extreme climatic deterioration of zone 2. Pollen zone 4 Spectra based on sum 2 (Figure 5) show that the deciduous forest seems to retreat only slightly. Further, by excluding Pinus, the deciduous forest displays values similar to previous and following zones. However in absolute values (Figure 7), the AP-Pinus percentages and concentrations decreases, especially from 1530 to 1472 cm depth. There are two complementary ways to explain the higher representation of conifers of cool and humid environments. First, it could be caused by an increase of the distant pollen input from surrounding hills and from the Pyrenees. However, these pollen grains are usually poorly transported. Secondly, a decrease in temperature might have caused a lower altitudinal conifer belt and closer proximity of the conifer forest. The presence of Calystegia and Salix both indicate open patches of land and possibly also cooler conditions. The climate interpretation is a phase of sufficient rainfall for sustained deciduous forest during the growing season, i.e. summer (not much different from zones 3 and 5) and probably a short-term cooler episode. Pollen zone 5 The presence of pollen grains of Corylus, a heliophilous shrub or small tree (Reille, 1990) of deciduous forests, indicates openings in the forest. The near disappearance of Eucommia, a taxon that seems to have completely disappeared from the European flora before the Upper Pleistocene, could be explained by cooling in zone 4 or by further growing distance. The higher upland tree concentration values could have been caused by lower sedimentation rates and by the proximity of the forest. Climate conditions of pollen zone 5 were probably only slightly cooler than in pollen subzone 3a. Aquatic, spores and varia diagram Description Pollen zone 1 Monolete and Pteris spores are well represented (Figure 6). Typhaceae and Cyperaceae are present.

Pollen zone 2 Typha-Sparganium values reach up to 4%. Potamogeton pollen has frequent occurrences, especially in subzones 2a and 2b. The psilate monolete spore values are lower from 2412 cm depth to the end of pollen zone 2. At 2535 cm depth, non-psilate monolete spore percentages are strongly reduced whereas Cyperaceae pollen increases up to 5.3%. In subzone 2b the Cyperaceae values are relatively high: 36.4%. In subzone 2c the proportion of the aquatics, spores and varia is the smallest, often less than 10% of the total counted. Pollen zone 3 Aquatic taxa like Typhaceae and Potamogeton are frequent. Cyperaceae and monolete spores are continuously present each with ca 3%. Nymphaea grains have been observed in the middle of pollen zone 3. In subzone 3b, there is a brief peak of Cyperaceae. In subzone 3c, maxima of Typha-Sparganium and occurrences of Typha cf. latifolia have been observed. Pollen zone 4 The first half of zone 4 is characterised by the extraordinary development of Cladium cf. mariscus and, in the second half of the zone, of cf. Thelypteris palustris. Haloragaceae grains are present, whereas Typhaceae and Potamogeton nearly completely disappear. The concentration indicates a first peak at 1600 cm. Afterwards, values are frequently very high, often higher than the upland taxa. They do not decrease from 1530 to 1472 cm. Pollen zone 5 Values of Typhaceae are moderate. Cyperaceae have values similar to pollen zone 3, whereas both monolete spore types are intermittently peaking. Concentrations are relatively high. Interpretation: reconstruction of the aquatic vegetation During the period represented by pollen zones 1 and 3, the lake was surrounded by a narrow belt of aquatic plants. During zone 2, a marsh developed and aquatic vegetation was maintained. Information about the temperature of this zone comes from the presence of Cladium mariscus which is a warmth-loving plant (Godwin, 1975).

jopl357.tex; 13/05/1997; 11:59; v.5; p.13

360 Pollen zone 4 These high percentages of Cladium and Thelypteris are not encountered in 25 surface samples from Lake Banyoles (Leroy, 1990 and unpublished work) despite their occurrence in the present flora of the region. Even the flora of the small marsh at the northern end of Lake Banyoles is poorly illustrated in the pollen spectra of samples taken inside the marsh itself, and mainly by only slightly higher percentages of Gramineae. Kershaw (1979) has shown that maximum pollen deposition of aquatic plant pollen and spores is near the aquatic and marshy vegetation producing them. Therefore during TP II zone 4, the spectra are dominated by purely local vegetation. Moreover, the increase in the aquatic taxa concentration (from about 3000 up to 50 000 gr g 1 ) attests to a local production of Cladium and Thelypteris, confirmed by the frequent occurrences of clusters of spores or pollen, sporangia and stamens, as well as the abundance of spores with a well-preserved perispore. Both plants indicate a marshy environment. The vegetation reconstruction is as follows: locally the vegetation was open (spreading of vines, marshy vegetation, freshwater aquatic plants), due to much enlarged lake shores and shallow waters. The bench platform around the depression provided flat surfaces for the extension of the marsh. The distant input of pollen grains probably increased. Pollen zone 5 Deciduous forest might have (re)-colonised large parts of the emergent lake shores. The replacement of Cladium by Thelypteris is explained by the growth of this fern in carrs. In a Holocene pollen diagram from Lobsigensee, Switzerland, the sediment change from littoral (lake marl) to terrestrial (peat) is paralleled by steep rise in Thelypteris palustris spore values (Amman, 1989). Statistical zonation In the upland diagram, whereas most of the visual zones and subzones (Figure 8B) find a more or less good equivalent in the statistical zonation (Figure 8A), CONISS does not distinguish pollen zone 4 from pollen zone 5. This could indicate that the changes occurring after the transition from pollen zone 3 to pollen zone 4 are of lesser importance for the upland vegetation than, for example, the short events such as subzones 2a and 3b.

In the aquatics, spores and varia diagram, most of the zones are still concentrated around pollen zone 4 (Figure 8C). This is due to highly fluctuating percentages of Cyperaceae pollen grains and monolete spores, either linked to fluctuating lake level, or more probably to the presence of spore clusters. Changes around subzone 2b are recognised by CONISS.

Palynology of drillhole TP I and of a peat level from the water well Due to early destruction of the TP I cores, only very preliminary palynological results could be obtained (Table 1). Five of the seven treated samples provided enough grains for analyses. Between 1935 and 1080 cm depth, the base sum for the percentages excludes the spores (Table 1). The grains are relatively poorly preserved (5 to 14% of indeterminable grains). Pinus values are dominant: 67% to 96%. Samples 1 and 2 have high percentages of cf. Lycopodium. Sample 1 has higher Ericaceae, Ulmus-Zelkova and Gramineae values; sample 4 has higher Alnus, Quercus and Carya values; and finally sample 5 has higher Elaeagnus values. The two upper samples are pollen sterile because they come from the upper oxidised levels. Abundant other microfossils were found: Spiniferites, Concentricystes, fungi, reworked pollen grains, etc. but only rare diatoms. The reworked pollen grains and the dinoflagellate cysts indicate an input from marine sediments, probably via reworking from the Eocene substratum. The presence of Concentricystes suggests run-off. Not only is it hard to reconstruct the vegetation from these spectra, but it is also difficult to link this sequence with drillhole TP II by horizontal correlation. However, the TP I flora, similar to TP II, could belong to the same deposit. A peat sample was collected during the digging of the well from ca 3000 cm depth by R. Juli`a (Table 1). The desiccation of the sample caused poor preservation of the palynomorphs. Only 127 grains have been counted. The spectrum is largely dominated by Pinus. Quercus and indeterminable grains are abundant. The flora is similar to the one of TP I and TP II drillholes.

Lake level, precipitation and temperatures Important facies changes from a more organic sediment to a nearly pure Characeae deposit, including also

jopl357.tex; 13/05/1997; 11:59; v.5; p.14

361 Table 1. Pollen spectra of TP I drillhole and the water well in percentages Site Sample Depth in cm Pinaceae Abies Picea Cupressaceae Alnus Carpinus Elaeagnus Ericaceae Quercus Q. ilex t. Parrotia cf. persica Carya Pterocarya Tilia Ulmus–Zelkova Caryophyllaceae Artemisia Compositae tubulifl. Compositae ligulifl. Cruciferae Cyperaceae Gramineae Plantago Typha-Sparganium Typha latifolia t. Urticaceae Sum Monoletes psilate Monol. non psil. Triletes psilate Tril. non psil. cf. Lycopodium Osmunda Pteris Selaginella Anthoceros Indeterminated Indeterminable Reworked Total sum

TP I 5 1080

TP I 4 1313

85.58

67.27

3.13

2.91

TP I 3 1770 96.28

TP I 2 1930

TP I 1 1935

Water well 3000

87.50

67.82

70.80 0.88 3.54

0.74

1.15

1.47 1.47

1.15

1.47

8.05 1.15 1.15

1.77 13.27 0.88

1.47

1.15

3.54

0.53

0.74

4.60

3.54

0.53

0.74 0.74 0.74

1.15

1.06 4.73 2.55 1.57 0.31

2.18 5.45 0.36

0.53

0.74 4.39 0.31 0.63

0.94 2.19

0.94

5.82 1.09 0.36 1.09 0.36 1.45 0.36

0.36 1.82

1.06

0.74 1.47

2.30 1.15 1.15 6.90 1.15

0.88 0.88

0.36 1.09 319 1.87 0.33

0.36 275 1.65 0.45 0.45

188

136 0.40

87 2.90

1.45 0.72 32.14

29.71

113 1.59

0.33 1.45 0.72 10.96 1.87 374

2.31 4.62 303

fine-grained profundal facies, have occurred. Therefore lithology and carbonate content contribute to the reconstruction of the history of the water-depth and water-level variations. Palynology, in general, gives regional terrestrial information, sometimes also local

13.64 220

1.19 12.30 252

0.79 7.94 138

126

ones, via the aquatic vegetation. The seasonality of the precipitation may sometimes be reconstructed on the following basis. Under interglacial conditions in middle latitude western Europe, trees grow during late spring and summer. Under dry conditions with

jopl357.tex; 13/05/1997; 11:59; v.5; p.15

362

Zone 1 lithology (fine-grained sediment) and fauna (no ostracodes and only aquatic gastropods) indicates a profundal facies. Palynological data suggest temperature conditions similar or higher than today, whereas precipitation was higher, even possibly three times higher.

waters and the gastropods of occasionally drying environments (at least at the transition from subzone 2a to 2b) are explained by shallower water in a littoral environment. Prentice et al. (1992) suggested that, during the last glacial maximum, water levels were high, because precipitation occurring in winter maintained the water body and the aquatic vegetation. The dry summers and rainy winters account for the apparently contradictory steppe conditions displayed by the vegetation. In the TP II sequence, droughts certainly occurred during the growing season in most of the zone but possibly also during the rest of the year during subzone 2a and b, although sufficient rainfall must have occurred to maintain some aquatic vegetation. In subzone 2c, enough rain during the growing season (least evapotranspiration in spring) allowed the re-establishment of trees, possibly drought resistant. A seasonal shift of the rainfall occurred at the transition between subzones 2b and 2c, from a season when it is not available to the plants in spring. There also seems to be a delay between the terrestrial vegetation change caused by better rainfall and the change to coarse grain size and the lowest aquatic percentages, both still indicating low water levels. The temperature of subzone 2b, the Liguliflorae subzone, might have been mild according to the comparison with modern spectra. In the second half, before the forest re-increase, the rare occurrences of Potamocypris fallax could indicate a slight cooling of the water caused by a different origin of the water, possibly only surficial. This is possible but not necessary, because there also is a climate change. For zone 2, if there is a departure from the present temperatures, it probably is a minor one towards lower temperatures. In brief, in zone 2 there is general agreement suggesting low precipitation and strong evaporation throughout the year. However, temperatures might not have changed drastically and might have remained similar to the preceding zone. Pollen subzone 2c, with a slight re-increase of the precipitation during the growing season, experiences, however, the lowest water level.

Zone 2: short-term lake-level lowering due to severe droughts

Zone 3: slightly warmer and more humid than today, but less than zone 1

In zone 2, especially in subzone 2c, a lowering of the lake level occurred. The high carbonate percentages (growth of Characeae usually in shallow oligotrophic waters), the ostracode assemblage of slightly brackish

The lithological facies shows again a high water level. Aquatic plants from open waters are present. Zone 3 is a warm temperate and humid period as in zone 1. The new forest assemblage shows a tendency towards less

Figure 9. Synthesis of changes in temperature, precipitation, and absolute lake level, and their interpretation.

mild temperature leading to a ‘warm’ steppe formation, the season with the least evapotranspiration is winter. Therefore rain falling in winter is unavailable for most trees which grow in spring. Ostracode assemblages provide local information on water depth, salinity, temperature, substratum, etc. Gastropod assemblages are especially informative when they contain terrestrial forms: i.e. from marshes and land, because they indicate variations in lake levels and give information on the type of land vegetation. Past variations of lake level, precipitation and temperature are discussed below zone by zone, clearly distinguished by lithology, carbonate content, palynology and ostracode-gastropod assemblages (Figure 9). Zone 1: more humid and slightly warmer than today

jopl357.tex; 13/05/1997; 11:59; v.5; p.16

363 favourable conditions. A shift towards cooler temperatures occurs at subzone 3b. Zone 4: lake-level lowering or stable due to climatic or more likely non-climatic causes In TP II, shallow, non turbid and oxygenated waters favoured the growth of calcareous algae. The environment reverted to littoral conditions. The climate was, according to the upland pollen diagram, humid and only slightly and briefly colder. The ostracodes (peak of P. fallax at 1480 cm) indicate a higher activity of underwater springs bringing cold waters. From the 1530 to 1472 cm depth, there is a maximum in carbonate content and lower upland tree concentrations, caused either by lower pollen production or by increased sedimentation rate. This information is strictly local and it does not necessarily imply a vegetation/climate change.The ostracode assemblages show a patchwork environment with zones of spring water and zones of shallow brackish waters. The existence of shallow waters not only near the drilling site but also in the surroundings is deduced from continued summer rainfall, presence of open spaces, and marsh and freshwater plants. The increased shallow water surface was favoured by exposure and extension of the platform bench. It is probable that the shelf progressed enough by algal growth to reach the drillhole location. The presence of some open water aquatic plants could be explained by steep bench slopes nearby, separating the sequence from the main depression. The TP II sequence would therefore be located on the lakeward side of the bench platform. Higher lake levels are not likely because of the lithology, the ostracode assemblages and the shallowwater plant representation. Only a stable water level with a natural infilling or a lowering of the water depth are possible. Zone 4 may already show a preliminary natural infilling, as seen from the aquatic vegetation belts, when the water level remained stable and only a relative change of water depth occurred. A lowering of the lake level of the Tres Pins water body could have been caused by karstic activity, but a climate change cannot be excluded. Similar water-level fluctuations occurred at Tigalmamine in the Middle Atlas Mountains, Morocco, according to evidence from mineralogy, diatoms, and ostracode shell geochemistry (Lamb et al., 1995). The pollen data there showed that the forest was not affected by Holocene short dry events (duration of 150 to 400 years) because they were caused by reduced winter precipitation. Lamb

et al. (1995) suggest an association with cooler sea surface temperatures in the N. Atlantic Ocean. Phase 4 of Tres Pins might be like the dry events of Tigalmamine because only tenuous changes in the forest cover have been observed. At Tres Pins, however, the cause of the lake-level lowering may be winter droughts or, more likely, karst activity. A climate change is hard to demonstrate for an isolated Plio-Pleistocene site. In the case of Tres Pins, not only is there a lack of proxies (absence of diatoms, and no geochemical analyses such as Sr/Ca on ostracode shells or stable isotopes), but also only one site is available for the time period considered. When it is possible to examine several lakes in one area, and to detect synchronous changes in different drainage basins, then it becomes possible to assume a climate forcing. In brief, if there is a climate-induced change, it may have been caused by a cooling, with high summer precipitation, and possible droughts in the winter, which resulted in a lake-level lowering. If the change is not climatic, which is more probable in the light of zone 5, it could be caused by karstic movements affecting the water table followed by lake-level lowering, or more probably by natural infilling with a stable lake level. Zone 5: lake infilling, slightly warmer and more humid than today, but less than subzone 3a Although the sediment becomes fine again and has a lower carbonate content in zone 5, the presence of only terrestrial gastropods from cool, humid and forested surroundings (F. Magnin, pers. commun. and in Leroy, 1990) and the high AP/NAP ratios illustrate the continued infilling of the lake. Ostracodes typical of plant overgrowth (Metacypris cordata) are more frequent. The location of the sequence is now clearly on the landward side of the shelf. The infilling of the lake may have already started in zone 4. The AP/NAP ratio is similar to subzone 3a, indicating warm temperate and humid periods, but the composition of the forest indicates more open conditions than in zone 3 (Figure 5).

Discontinuity of the record The Tres Pins deposit has been affected by karst, faulting and synsedimentary erosion. The sedimentation is, therefore, not continuous.

jopl357.tex; 13/05/1997; 11:59; v.5; p.17

364 Within the TP II sequence, small erosion phases in the lake are suggested by hiati altering the continuous record of the past vegetation successions. They are expressed by rather abrupt changes from one pollen zone to the other, such as at 2245 and 1977 cm depth. Progressive vegetation changes have only been recorded at the transition between pollen zones 2 and 3. It is symptomatic that this part of the record occurred during a lake level increase, allowing for continuous sedimentation. The dissimilarity between TP I spectra, the TP II pollen diagram, and the water well sample suggests not only a strong lateral facies variation, but also the existence of subvertical faults separating the three sequences.

sometimes up to one meter per ka (Juli`a, 1980; P´erezObiol & Juli`a, 1994). The second reason is the presence of short hiati of unknown duration. The time represented by the sediment from the bottom to 1000 cm depth, excluding the hiati, could be in the order of less than 30 ka. Indeed the climate cycles are forced by obliquity during Late Pliocene and Lower Pleistocene. Pollen zone 2 belongs to a glacial period, whereas the other zones are interglacial ones. In order to provide bases for regional correlation by palynostratigraphy, pollen zone 2 is named Glacial of Santa Eugenia and pollen zones 3 to 5 are part of an interglacial named Interglacial of Tres Pins (Figure 9).

Conclusions Age and duration of the deposit The age of the deposit remains uncertain. At the time of deposition of the Tres Pins sediment, regional volcanic activity (Donville, 1976) has been very slight, providing no possibility for any Ar/Ar dating. No rodent teeth have been found which would have allowed at least a correlation with other sites in the area (Agust´ı et al., 1987). However, comparison of the palynological data with those of other sections in the same area that are better dated (Suc et al., 1992) makes it possible to propose lower and upper age limits with the necessary caution when using palynostratigraphy as a chronological tool. The whole of the sedimentation of TP II could be slightly older than lake 1 of B`obila Ordis (1.2 Ma) because of the better representation of pollen taxa, which later disappear from Catalonia (Pterocarya, Carya, Parrotia cf. persica, etc.). On the other hand, such an association is unlikely to be older than the first glacial period, around 2.6 Ma, because of the absence of some typical ‘Tertiary’ elements. Amongst the ostracodes, a fossil species is present: Darwinula cylindrica (Staub, 1952; M. Coen, pers. commun.). The exact timing of its disappearance during the PlioPleistocene is still unknown. The flora (typical of the period from 2.6 to ca 1.2 Ma) and the fauna (extinct species) indicate that the deposit could have formed during the Late Pliocene-Lower Pleistocene. The duration of TP II climatic changes is difficult to estimate. The first reason is the varying sedimentation rate with the different lithologies. The Characeae-rich sediment must have had a high sedimentation rate if compared to recent well-dated sediment of the area:

The interpretation of this Late Pliocene-Lower Pleistocene lacustrine sequence, with a duration of less than 30 ka and with synsedimentary hiati, leads to the following conclusions. 1. Vegetation history. Pollen analyses of zones 1, 3, and 5 indicate forested environments under warmer and much wetter conditions than at present. The glacial period of pollen zone 2 caused aridity and the consecutive development of a steppe with pine stands. The marsh vegetation of pollen zone 4 indicated by pollen of Cladium cf. mariscus and spores of cf. Thelypteris palustris has no modern equivalent south of the Pyrenees. Vegetation successions are poorly expressed in the diagram. Only one has been observed, at the transition from subzones 2c to 3, because of continuous sedimentation resulting from a progressive increase in water level. The recovery of the arboreal vegetation after a glacial period is the following: Quercus, then Carpinus and finally the group Carya, Ulmus-Zelkova, Parrotia. 2. Temperature during the glacial period. Pollen zone 2, and especially subzone 2b, illustrates a steppe period rich in Liguliflorae, similar to environments that occurred in southern Europe and North Africa during the glacial stages of the Late Pliocene and Early Pleistocene. The search for modern analogues and the presence of temperature indicators in the aquatic vegetation indicate that the climate is certainly dry. The temperature might have been mild if the comparison to modern analogues is valid. Zone 2 is a global climate event linked to a glacial period.

jopl357.tex; 13/05/1997; 11:59; v.5; p.18

365 3. Causes for the observed palaeohydrological changes. Several factors appear to have influenced lake levels in the TP II sequence. In a karstic environment, subterranean processes are either distant consequences of the climate, or completely independent of it. Drought during the spring growing season was the main influence on vegetation represented by pollen zone 2. Subzone 2c shows the lowest lake level. For pollen zone 4, a global atmospheric circulation change (droughts during the non-growing season, i.e. winter) or more probably local non-climatic factors (karstic collapse, changes in the path of the sublacustrine water inflow, natural infilling, etc.) are possible reasons for a stable or decreasing water level. However the last reason, a local limnological change, is clear at the drilling site during zone 5. It probably already began in pollen zone 4. The sediment facies is profundal in zones 1 and 3, and littoral in zones 2 and 4. Palynology indicates that the drilling site is in a landward position on the bench platform for pollen zone 5. The origin of the springs feeding the Tres Pins lake could have changed from deep phreatic waters in pollen zone 1, 3 and 5, to more surficial and cooler waters in zones 2 and 4. 4. Changes in the rainfall seasonality. Palynological interpretation and lake-level reconstruction allow the detection of seasonal shifts in precipitation. In subzones 2a and 2b, droughts are important during the growing season (spring), but also possibly during the rest of the year. Subzone 2c shows the beginning of recovery of the forest vegetation, possibly linked to a shift of rainfall to the spring growing season. The presence of forest and the extension of marsh in zone 4 testifies to significant precipitation during the growing season (summer) at least, and probably also during the winter. Acknowledgments G. Seret and his team (Universit´e Catholique de Louvain (UCL), Belgium) have been in charge of the drilling, which could not have succeeded without local assistance and support by R. Juli`a (CSIC, Barcelona, Spain) and A. Coll, the bar owner. I am grateful to G. Seret, J.-P. Suc (CNRS, Lyon, France), J. Guiot (CNRS, Marseilles, France), F. Magnin (CNRS, Aixen-Provence, France) and M. Coen (UCL) for fruitful comments during this research and for reviewing this manuscript. I wish to acknowledge B. Collet (UCL)

who made the initial ostracode and carbonate studies. F. Magnin has provided the gastropod data. M. Coen has been in charge of a second, more detailed, ostracode survey, and kindly allowed me to make use of his results. R. Juli`a organised the flight over Lake Banyoles. Thanks are extended to J. Guiot for access to modern pollen spectra stored in the European Pollen Database, to V. Hall (Queen’s University of Belfast, UK) for improving the English of the manuscript and to the two reviewers for constructive comments, one of them being H. Lamb. Technical help has been provided by UCL (pollen treatments) and QUB (drawings). References Ablin, D., 1985. Analyse pollinique des d´epˆots lacustres du Ceyssac (Plio-Pl´eistoc`ene du Velay, Massif Central franc¸ais): flore, v´eg´etation et climat. Th`ese 3`eme cycle, Paris VI, Mus´ee National d’Histoire Naturelle et Univ. P. et M. Curie, 125 pp. Ablin, D., 1991. Analyse pollinique des d´epˆots lacustres de Ceyssac, Plio-Pl´eistoc`ene du Velay (Massif Central, France). Cahiers de micropal´eontologie 6, 1: 21–38. Agust´ı, J., S. Moy`a-Sol`a & J. Pons-Moy`a, 1987. La sucesi´on de Mam´ıferos en el Pleistoceno inferior de Europa: proposicion de una nueva escala bioestratigr`afica. Paleont. i Evol., Mem. Esp. 1: 287–295. Amman, B., 1989. Late-Quaternary palynology at Lobsigensee, regional vegetation history and local lake development. Dissertationes Botanicae, J. Cramer, Berlin-Stuttgart 137: 157 pp. Bennett, K., 1995. Manual for Psimpoll 2.27 and Pscomb 1.02. Cambridge documentation. INQUA boutique. Bessedik, M., 1985. Reconstitution des environnements mioc`enes des r´egions nord-ouest m´editerran´eennes a` partir de la palynologie. Th`ese d’Etat, Univ. Sc. et Techn. du Languedoc, Montpellier, 162 pp. Bottema, S., 1975. The interpretation of pollen spectra from prehistoric settlements (with special attention to Liguliflorae). Palaeohistoria 17: 15–35. Br´enac, P., 1984. V´eg´etation et climat de la Campanie du Sud (Italie) au Plioc`ene final d’apr`es l’analyse pollinique des d´epˆots de Camerota. Ecologia Mediterranea 10, 3/4: 207–216. Collet, B., 1987. Contribution a` l’´etude s´edimentologique du d´epˆot lacustre pl´eistoc`ene de TP II (Banyoles, Catalogne). M´emoire de licence, Universit´e Catholique de Louvain, Belgium: 74 pp. Combourieu-Nebout, N., 1993. Vegetation response to Upper Pliocene glacial/interglacial cyclicity in the Central Mediterranean. Quat. Res. 40: 228–236. Combourieu-Nebout, N., 1995. R´eponse de la v´eg´etation de l’Italie m´eridionale au seuil climatique de la fin du Plioc`ene d’apr`es l’analyse pollinique haute r´esolution de la section de Semaforo (2.46 a` 2.1 Ma). Comptes Rendus de l’Acad´emie des Sciences de Paris, ser. IIa, 321: 659–665. De Deckker, P., 1979. The Middle Pleistocene ostracod fauna of the West Runton freshwater bed, Norfolk. Palaeontology 22, 2: 293–316. De Deckker, P., M. A. Geurts & R. Juli`a, 1979. Seasonal rythmites from a Lower Pleistocene lake in Northeastern Spain. Palaeogeography, Palaeobiology, Palaeoecology 26: 43–71. Diniz, F., 1984. Apports de la palynologie a` la connaissance du Plioc`ene portugais. Rio Major: un bassin de r´ef´erence pour l’his-

jopl357.tex; 13/05/1997; 11:59; v.5; p.19

366 toire de la flore, de la v´eg´etation et du climat de la fac¸ade atlantique de l’Europe m´eridionale. Univ. Sc. et Techn. du Languedoc, Montpellier, 230 pp. Dodson, J. R. & R. V. S. Wright, 1989. Humid to arid to subhumid vegetation shifts on Pilliga Sandstone, Ulungra Springs, New South Wales. Quat. Res. 32: 182–192. Donville, B., 1976. G´eologie n´eog`ene de la Catalogne orientale. Bull. B. R. G. M., 2`eme s´erie, 4, 3: 177–210. Dricot, E. & S. Leroy, 1989. Peptization and sieving for palynological purposes. Geobound, Bruxelles 2: 114–126. Elha¨ı, H., 1966. Deux gisements du Quaternaire moyen. Bruges (Sud-Ouest de France). Ba˜nolas (Catalogne). Bull. Assoc¸. Franc. Et. Quat. 6: 69–78. Elha¨ı, H., 1969. La flore sporo-pollinique du gisement villafranchien de Sen`eze (Massif-Central – France). Pollen et Spores 11, 1: 127– 140. Gaillard, M. & G. Digerfeldt, 1991. Palaeohydrological studies and their contribution to palaeoecological and palaeoclimatic reconstructions. In B. E. Berglund (ed.). The cultural landscape during the last 6000 years in southern Sweden – the Ystad Project. Ecological Bulletins 41: 275–282. Glo¨er, P. & C. Meier-Brook, 1994. S¨usswassermollusken. Deutschen Jugendbund f¨ur Naturbeobachtung, Hamburg, 136 pp. Godwin, H., 1975. History of the British Flora. A factual basis for phytogeography. Cambridge Science Classics. CUP, 541 pp. Guiot, J., J.-L. de Beaulieu, R. Cheddadi, F. David, P. Ponel & M. Reille, 1993. The climate in western Europe during the last Glacial/Interglacial cycle derived from pollen and insects remains. Palaeogeogr., Palaeoclimat., Palaeoecol. 103: 73–94. Harrison, S. P. & G. Digerfeldt, 1993. European lakes as palaeohydrological and palaeoclimatic indicators. Quat. Sc. Rev. 12: 233–248. Heusser, L. E., 1988. Pollen distribution in marine sediments on the continental margin off northern California. Mar. Geol. 80: 131–147. ¨ Hiller D., 1972. Untersuchungen zur Biologie und zur Okologie limnischer Ostracoden aus der Umgebung von Hamburg. Arch. Hydrobiol., Suppl. 40: 400–497. Juli`a, R., 1980. La conca lacustre de Banyoles-Besal´u. Monografies del Centre d’Estudis Comarcals de Banyoles, 188 pp. Juli`a, R. & J.-P. Suc, 1980. Analyse pollinique des d´epˆots lacustres du Pl´eistoc`ene inf´erieur de Banyoles (Ba˜nolas, site de la B`obila Ordis-Espagne): un e´ l´ement nouveau dans la reconstitution de l’histoire pal´eoclimatique des r´egions m´editerran´eennes d’Europe occidentale. Geobios 13, 1: 5–19. Kershaw, A. P., 1979. Local pollen deposition in aquatic sediments on the Atherton Tableland, North-Eastern Australia. Australian Journal of Ecology 4: 253–263. Lamb, H., F. Gasse, A. Benkaddour, N. El Hamouti, S. van der Kaars, W. T. Perkins, N. J. Pearce & C. N. Roberts, 1995. Relation between century-scale Holocene arid intervals in tropical and temperate zones. Nature 373: 143–137. Leroy, S., 1987. Analyse palynologique de deux nouvelles sections du Pl´eistoc`ene inf´erieur de Banyoles (Catalogne). In Civis Llovera, J. & M. F. Valle Hern´andez (eds), Actas del VI Simposio de Palinolog´ıa, A.P.L.E., Salamanca: 307–313. Leroy, S., 1988. Image pollinique d’une steppe du Plioc`ene sup´erieur a` B`obila Ordis, Banyoles (Catalogne). Inst. fr. Pondich´ery, trav. sec. sci. tech. 25: 197–207. Leroy, S., 1990. Pal´eoclimats plio-pl´eistoc`enes en Catalogne et Languedoc d’apr`es la palynologie de formations lacustres. Th`ese de l’Universit´e Catholique de Louvain. Vol. 1: 319 pp; vol. 2: 193 pp.

Leroy, S., P. Ambert & J.-P. Suc, 1994. Pollen record of the SaintMacaire maar (H´erault, southern France): a Lower Pleistocene glacial phase in the Languedoc coastal plain. Rev. Palaeobot. Palyn. 80: 149–157. Leroy, S. & L. Dupont, 1994. Development of vegetation and continental aridity in northwestern Africa during the Late Pliocene: the pollen record of ODP Site 658. Palaeogeog., Palaeoclimat., Palaeoecol. 109: 295–316. Leroy, S. & P. Roiron, in press. Latest Pliocene pollen and leaf floras from Bernasso palaeolake (Escandorgue Massif, H´erault, France). Review of Palaebot. and Palynol. Leroy, S. & G. Seret, 1992. Duration and vegetation dynamic of the Nogaret Interglacial (1.9 Ma, S. France). Tentative correlation with stage 75. In Kukla, G. & E. Went (eds), Start of a Glacial, Proceedings of the Mallorca NATO ARW, NATO ASI Series, Springer Verlag, Heidelberg I, 3: 113–125. Løvlie, R. & S. Leroy, 1995. Magnetostratigraphy of Lower Pleistocene Banyoles palaeolake carbonate sediments from Catalonia, NE-Spain: evidence for relocation of the Cobb Mountain subchron. Quat. Sc. Rev. 14: 473–485. Magnin, F., 1991. Mollusques continentaux et histoire quaternaire des milieux m´editerran´eens (Sud-Est de la France, Catalogne). Unpublished thesis, Universit´e d’Aix-en-Provence, 340 pp. Meisch, Cl., 1984. Revision of the recent Western Europe species of the genus Potamocypris (Crustacea, Ostracoda), part I. Trav. Sc. Mus. Hist. Nat. Luxembourg 3: 1–55. Neale, J. W., 1988. Ostracods and palaeosalinity reconstruction. In De Deckker P., J. P. Colin & J. P. Peypouquet (eds), Ostracoda in the Earth Sciences. Elsevier: 125–155. P´erez-Obiol, R., 1988. Histoire tardiglaciaire et holoc`ene de la v´eg´etation de la r´egion volcanique d’Olot (NE P´eninsule Ib´erique). Pollen et Spores 30, 2: 189–202. P´erez-Obiol, R. & R. Juli`a, 1994. Climatic change on the Iberian peninsula recorded in a 30 000 yr pollen record from Lake Banyoles. Quat. Res. 41: 91–98. Prentice, I. C., J. Guiot & S. P. Harrison, 1992. Mediterranean vegetation, lake levels, and palaeoclimate at the Last Glacial Maximum. Nature 360: 658–660. Ravazzi, C. & M. Rossignol-Strick, 1995. Vegetation change in a climatic cycle of Early Pleistocene age in the Leffe Basin (Northern Italy). Palaeogeog., Palaeoclimat., Palaeoecol. 117: 105–122. Reille, M., 1990. Lec¸ons de palynologie. Editions du CNRS, 206 pp. Remy von, H., 1958. Zur Flora und Fauna des Villafranca-Schichten von Villaroya, Prov. Logro˜no, Spanien. Eiszeitalter und Gegenwart 9: 83–103. Robinson, E., 1980. The ostracod fauna of the interglacial deposits at Sugworth, Oxfordshire. Phil. Trans. r. Soc., Lond. B289: 99–106. Roiron, P., 1983. Nouvelle e´ tude de la macroflore plio-pl´eistoc`ene de Crespi`a (Catalogne, Espagne). Geobios 16, 6: 687–715. Staub, E. W., 1952. Mikropal¨aontologische Untersuchungen im Terti¨ar zwischen Ehingen und Ulm a. d. Donau. Geol. Jb. 66: 433– 524. Suc, J.-P., 1984. Origin and evolution of the Mediterranean vegetation and climate in Europe. Nature 307: 429–432. Suc, J.-P., G. Clauzon, M. Bessedik, S. Leroy, Z. Zheng, A. Drivaliari, P. Roiron, P. Ambert, J. Martinell, R. Domenech, I. Matias, R. Juli`a & R. Anglada, 1992. Neogene and Lower Pleistocene in Southern France and Northeastern Spain. Mediterranean environments and climate. 8th Intern. Palynol. Congress. Cahiers de Micropal´eontologie 7, 1/2: 165–186. Suc, J.-P. & J. Cravatte, 1982. Etude palynologique du Plioc`ene de Catalogne (nord-est de l’Espagne). Pal´eobiol. contin. 13, 1: 1–31. Suc, J.-P. & W. Zagwijn, 1983. Plio-Pleistocene correlations between the northwestern Mediterranean region and northwestern Europe

jopl357.tex; 13/05/1997; 11:59; v.5; p.20

367 according to recent biostratigraphic and palaeoclimatic data. Boreas 12: 153–166. Suc, J.-P., A. Bertini, N. Combourieu-Nebout, F. Diniz, S. Leroy, E. Russo-Ermolli, Z. Zheng, E. Bessais & J. Ferrier, 1995. Structure of West Mediterranean vegetation and climate since 5.3 Ma. Acta Zool. Cracov. 38, 1: 3–16. Villalta, J. F. de & J. Vicente, 1972. Una flora del Cuaternario antiguo en las cercanias de Crespi`a. Acta Geol. Hisp. 7, 4: 120–128.

Zagwijn, W. & J.-P. Suc, 1984. Palynostratigraphie du PlioPl´eistoc`ene d’Europe et de M´editerran´ee nord-occidentales: corr´elations chronostratigraphiques, histoire de la v´eg´etation et du climat. Pal´eobiol. contin., Montpellier 14, 2: 475–483. Zohary, M., 1973. Geobotanical foundations of the Middle East. Gustav Fisher Verlag, Stuttgart, 739 pp.

jopl357.tex; 13/05/1997; 11:59; v.5; p.21