Latest Cretaceous – earliest Paleogene vegetation and climate change at the high southern latitudes: palynological evidence from Seymour Island, Antarctic Peninsula Palaeogeography, Palaeoclimatology, Palaeoecology

Fluctuations in Late Cretaceous climate were already in ﬂ uencing biotic change prior to the environmental up- heaval at the Cretaceous – Paleogene (K – Pg) boundary, but their general nature, magnitude and timing remain controversial.Ahigh-resolutiondatasetonterrestrially-derivedpalynomorphsispresentedfromthehighsouth- ern palaeolatitudes that unlocks details of small-scale climate variability throughout this period of signi ﬁ cant global change. Speci ﬁ cally, this isa quantitative spore and pollen analysisof anexpanded uppermost Cretaceous to lowermost Paleogene(Maastrichtian – earliest Danian)shallow marine sedimentary succession fromSeymour Island,offthenortheasterntipoftheAntarcticPeninsula,then(asnow)locatedat~65°S.Usingnearestlivingrel- atives the ﬁ rst detailed vegetation, habitatand climate reconstruction is presentedfor the emergent volcanicarc atthistime.Onthecoastallowlands,acooltowarmtemperaterainforestisenvisagedgrowinginariverineland- scape,withbothwet(rivermargin,pond)andrelativelydry(inter ﬂ uve,canopygap)habitats.Diversepodocarps and southern beech trees grew alongside angiosperm herbs and shrubs in mean annual temperatures of ~10 – 15 °C. Higher altitude araucarian forests gave way to open ericaceous heathland, beyond the tree line, in subalpine to alpine conditions with mean annual temperatures of a cold ~5 – 8 °C. There is no exact modern bo- tanical equivalent, but the closest modern ﬂ ora is that of the Andes of southern Chile and Argentina. Maastrichtian climate is shown to have ﬂ uctuated from cool, humid conditions, through a rapid warming ~2 million years prior to the K – Pg transition, followed by cooling during the earliest Danian, a trend supported by previous work on this interval. © 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/3.0/).


Introduction
Major climatic fluctuations during the latest Cretaceous were already influencing global biotic change immediately prior to the Cretaceous-Paleogene (K-Pg) boundary at 66 million years ago (Ma) (Keller, 2001;O'Dea et al., 2011). However, details of small-scale climate variability, particularly from the climatically-sensitive high palaeolatitudes, have been lacking due to a paucity of suitable Upper Cretaceous sedimentary successions. This paper presents a detailed quantitative analysis of terrestrially-derived palynomorphs (spores and pollen) from a continuous and superbly-exposed succession on Seymour Island, off the northeast tip of the Antarctic Peninsula (Fig. 1). Seymour Island represents the most expanded and best exposed uppermost Cretaceous to lowermost Paleogene succession from the high southern latitudes.
Furthermore, it has remained at approximately its present latitude (65°S) since the Cretaceous (Hathway, 2000).
It is well established that climates were cool and dynamic during the Late Cretaceous and there is indirect evidence for ephemeral ice sheets on Antarctica (Barrera, 1994;Barrera and Savin, 1999;Li et al., 2000;Miller et al., 2005;Davies et al., 2009;Thibault et al., 2010;Bowman et al., 2013a). Global climate change at this time has been associated with episodes of outgassing from major volcanic events (Courtillot et al., 1986;Hansen et al., 1996;Chenet et al., 2009), orbital cyclicity (Miller et al., 2003;Stüben et al., 2003;Gallagher et al., 2008) and tectonism (Frank and Arthur, 1999), before culminating at 66 Ma in the cataclysmic upheaval caused by a large bolide impact at Chicxulub, on the Yucatán Peninsula, Mexico (Schulte et al., 2010).
This study aims at providing the first high-resolution vegetation and palaeoclimatic reconstruction for the latest Cretaceous and earliest Danian of the high southern palaeolatitudes. Well-preserved spore, pollen and other terrestrial palynomorphs (e.g. algal spores and fungal remains) are documented from the López de Bertodano Formation of Seymour Island, and assessments made on their likely analogues by comparison with modern palynomorphs. The observed fossil ranges and quantitative abundance data are then used to reconstruct vegetation types within various habitats on the eastern flank of the Antarctic Peninsula. Palaeoclimatic trends inferred from changes in these plant associations through time are consistent with cool to warm temperate climates and humid to super-humid conditions on Seymour Island and the surrounding areas during the Maastrichtian, across the K-Pg boundary and into the earliest Danian.

Geological setting
The shallow marine López de Bertodano Formation of the Marambio Group crops out on southern Seymour Island in the James Ross Basin, Antarctic Peninsula, unconformably intercalated between the Snow Hill Island Formation beneath and the Sobral Formation above ( Fig. 1; Rinaldi et al., 1978, revised by Crame et al., 2004Olivero et al., 2008). The first comprehensive studies of the geological setting, palaeontology and age of the López de Bertodano Formation were carried out by authors in Feldmann and Woodburne (1988), and are summarised by Bowman et al. (2012).
During the Maastrichtian to earliest Danian, southern Seymour Island was submerged in an inner to mid shelf setting within a back-arc basin (the Larsen Basin). The Antarctic Peninsula may have been joined to southern South America by a narrow land bridge to the distant north (Hathway, 2000;Leppe et al., 2012), and the Larsen Basin opened into the Weddell Sea to the east. There, huge volumes of fine-grained sediment accumulated, having been eroded from the Antarctic Peninsula volcanic arc immediately to the west (Hathway, 2000;Crame et al., 2004;Olivero et al., 2008). Partial inversion of the northern Larsen Basin during the Late Cretaceous led to shallow tilting to the southeast, and the basinward progradation of shallow marine facies (Hathway, 2000). Sediment provenance studies suggest significant volcanic input from the emergent arc, but the intensity of volcanic activity apparently decreased during the deposition of the Marambio Group (Pirrie, 1991). No evidence for any discrete ash deposits within the López de Bertodano Formation was found during sedimentary logging. It seems therefore unlikely that significant tectonic upheavals or volcanic events occurred in this region during the Maastrichtian and earliest Danian that could have affected the near-offshore palynomorph record. The palaeogeography of the Antarctic Peninsula region at this time suggests that the eastern flank of the emergent arc, perhaps tens of kilometres to the west of what is now Seymour Island, was the only potential source area for the terrestrially-derived palynomorphs recovered in this study.
The López de Bertodano Formation comprises~1000 m of unconsolidated silty-clays and clayey-silts and contains no sedimentological evidence of significant hiatuses (Fig. 2). Dinoflagellate cyst biostratigraphy, macropalaeontology, magnetostratigraphy, strontium isotope stratigraphy and a minor iridium anomaly at the K-Pg boundary confidently date this major unit as Maastrichtian to Danian (Fig. 2;Elliot et al., 1994;McArthur et al., 1998;Bowman et al., 2012;Tobin et al., 2012). The López de Bertodano Formation of southern Seymour Island is hence one of the most expanded successions of this age in the world.

Materials and methods
A composite section (D5.251) 1100 m thick was measured perpendicular to strike across southern Seymour Island from the coast immediately north of Cape Lamb, to the southeast coast near Penguin Point during the austral summer of 2006 (Fig. 1). All necessary permits were obtained from the British Antarctic Survey (Cambridge, United Kingdom) for the described study, which complied with all relevant regulations.
The base of the section at Cape Lamb is within the uppermost Haslum Crag Member of the Snow Hill Island Formation, which passes unconformably into the López de Bertodano Formation at 66.5 m in section D5.251, then unconformably again into the lowermost Sobral Formation at 1074.5 m (Fig. 2). The K-Pg boundary is within a glauconitic horizon at 1007.5 m and was identified between samples at 1004 and 1008 m using the sudden appearance of an acme of the dinoflagellate cyst Senegalinium obscurum after Elliot et al. (1994). Elliot et al. (1994) collected samples at 10 cm intervals and also used the appearance of dinoflagellate cyst marker species to identify the K-Pg boundary.
A robust age model for section D5.251 has been constructed by correlating magnetic polarity zonal boundaries (Tobin et al., 2012) and strontium isotope ratios (McArthur et al., 1998) from previous studies on Seymour Island. The correlation was achieved using stratigraphic height from the K-Pg boundary datum and assuming planar bedding (Fig. 2). The timescale is based on Gradstein et al. (2012) and has been added to the age model using linear interpolation between the known ages of reversal boundaries. Strontium isotope stratigraphy for the Maastrichtian (Vonhof et al., 2011) supports the age model with the addition of McArthur et al.'s (1998) strontium isotope ratios measured from macrofossils (Fig. 2). No discrete ashes have been found within the López de Bertodano Formation, but available biostratigraphical analyses further support the age model (e.g. Huber et al., 1983;Harwood, 1988;Olivero and Medina, 2000;Bowman et al., 2012). Sediment samples were collected at intervals ranging between 0.5 and 2 m depending on sedimentological features throughout section D5.251. Selected samples were processed using standard quantitative procedures for extracting and concentrating palynomorphs, and the residues were permanently mounted on slides following staining with Safranin-O (Wood et al., 1996;Bowman et al., 2012;Supplementary Dataset). The British Antarctic Survey curates the remaining bulk sediment and all slides.
Quantitative palynological analyses were undertaken on 81 samples at 20 m intervals throughout the lower 700 m, and every 10 m or less for the remainder of the section ( Fig. 2; Supplementary Dataset). Most slides were analysed using a Leitz Ortholux transmitted light microscope (D5.601.1A and D5.621.1A were analysed on a transmitted light Leica DM750P). All the digital images in Figs. 3 and 4 were taken using a Leica DM750P photomicroscope and processed with Leica image stacking software. The spore and pollen nomenclature primarily follows that of Raine et al. (2011).
The palynomorph counts were all made across predefined, regularlyspaced traverses of known dimensions across the coverslip. When at least 300 marine and terrestrially-derived palynomorphs had been recorded that last traverse was completed resulting in final counts of contemporaneous spore and pollen grains, per sample, reaching up to 450 (sample D5.1061.1A, Supplementary Dataset). Overall, 87% of slides had counts over 200 contemporaneous spore and pollen grains, and 45% over 300 (Supplementary Dataset). A known quantity of residue (representing 0.5 g of sediment) was mounted on each cover slip, and palynomorphs per gramme of sediment were calculated (Supplementary Dataset). The remainder of each slide was then scanned to ensure all taxa present were recorded. A judgement was made on each palynomorph encountered as to whether or not it was likely to have originated from a plant that was growing on the eastern flank of the Antarctic Peninsula geologically contemporaneous with sediment deposition. Consideration was also given to whether it was obviously reworked based on the level of thermal maturity compared with others on the slide, its state of preservation (whether significantly damaged) and its taxonomic identification (i.e. known Permian-Triassic palynomorphs). Unless specifically noted as obviously reworked, all mention of palynomorphs hereafter in this paper refer to those that are considered to be approximately the same age as their surrounding sediment. Most palynomorphs encountered were included in the Fig. 2. Sedimentology, marine and terrestrial palynobiostratigraphy and sampled horizons, section D5.251, southern Seymour Island, Antarctic Peninsula. This succession has been dated using several methods: magnetostratigraphy (Tobin et al., 2012), dinoflagellate cyst biostratigraphy and associated minor iridium anomaly K-Pg boundary (the latter not sought in the current study) (Elliot et al., 1994;Askin and Jacobson, 1996;Bowman et al., 2012), terrestrial palynomorph biostratigraphy (Raine, 1984;Helby et al., 1987), macropalaeontology biostratigraphy (Macellari, 1988; not on this figure) and strontium isotope stratigraphy (McArthur et al., 1998). Other microfossil groups suffer from poor recovery and long stratigraphical ranges (Harwood, 1988;Huber, 1988). The open arrows represent macrofossil strontium isotope analyses, which support the age model (McArthur et al., 1998;Vonhof et al., 2011). The placement of absolute ages against section D5.251has been done using Tobin et al. (2012) updated to the timescale of Gradstein et al. (2012). The approximate stratigraphical location of chron boundaries has been correlated between the section of Tobin et al. (2012), slightly to the north on Seymour Island, and section D5.251 using the K-Pg boundary as a datum and assuming approximately planar bedding along strike inbetween. The figure timescale has been calculated using linear interpolation between neighbouring chron boundaries using Gradstein et al. (2012). The K-Pg boundary is identified in this study primarily by the acme of the dinoflagellate cyst Senegalinium obscurum with reference to Elliot et al. (1994) counts, but for some that were consistently fragmented (e.g. Azolla spp., Grapnelispora sp. 1, and fungal spores and hyphae) only presence/absence data were recorded.
Reconstructing past vegetation from dispersed spore and pollen records is essential for understanding past climates. The identification of a modern spore or pollen grain that is comparable with a fossil morphotype, assuming little evolutionary change in morphology or habitat preference during the intervening period, forms the basis of the "nearest living relative" technique (e.g. Wolfe, 1995), which is used herein. Nearest living relatives (NLRs) of the spore and pollen taxa recorded here are based primarily on Askin (1990a), MacPhail et al. (1994, Gallagher et al. (2008), Truswell and MacPhail (2009) and Raine et al. (2011), and references therein (Table 1). Preferred temperature and precipitation ranges of the NLRs based on their modern geographic distributions have been derived from the Atlas of Living Australia (ALA, http://www.ala.org.au/) and specifically for Araucaria araucana (monkey puzzle tree) from Sanguinetti and Kitzberger (2008) ( Table 2). The climatic data downloaded from the ALA, from the WorldClim database (http://www.worldclim.org, after Hijmans et al., 2005) is averaged for each taxon and includes worldwide (though primarily Australasian and Papua New Guinean) and spatially valid records. Climate parameters are used only for NLRs at species and/or genus level for taxa naturally distributed in Australasia to keep the climate reconstruction Southern Hemisphere-focused and as realistic as possible. Preferred climate types are then defined for each NLR based on mean annual temperature (MAT) ranges within the Holdridge Life-Zone Classification System (Emanuel et al., 1985) ( Table 2).
The validity of these interpretations, based on the NLR technique, depends on the recognition of inherent uncertainties particularly relevant to deep time records. For example, difficulties may arise because a particular fossil spore or pollen type may be affiliated with a modern family, genus or species group with non-uniform ecological tolerances (Martin, 1997). It is also possible that a single fossil taxon may have a conservative morphology and actually represent more than one modern taxon (e.g. Hill, 1990). Similarly, potential extinctions of ecotypes can render interpretations on pre-Quaternary palynofloras somewhat tenuous (Jordan, 1997). Although not directly resolvable from the fossil palynomorph record, there may also have been a variety of extraneous factors influencing vegetation composition on the Antarctic Peninsula during the Maastrichtian and Danian. For example, the dynamics of modern Nothofagus forests in New Zealand have been found to be not only a result of climate-induced change, but may also reflect internal stand processes (Stewart and Rose, 1990). Furthermore, the palynomorph record is of terrestrial material washed or blown into marine sediments and as such, taphonomic processes prior to fossilisation must also be considered. These may include the potential for reworked material, the size and location of the source area, ocean circulation, physical barriers, geochemical conditions and the resistance of sporomorphs (spores and pollen) to degradation through long transport paths. For example, in temperate latitudes wind-pollinated plants producing bisaccate pollen (e.g. Pinus) are over-represented in marine pollen spectra in part due to their resistance to oxidation (e.g. Rossignol-Strick and Paterne, 1999). In addition, interpretations based on abundance patterns of quantitative terrestrial palynomorph counts must consider factors including the relative pollen production of the nearest living relative plants. In the Seymour Island sequence, this may influence the representation of Nothofagidites pollen, which due to high levels of production is known to be transported over long distances, particularly when pollen grains reach higher levels of the atmosphere, also true for bisaccate pollen (e.g. Close et al., 1978;Kappen and Straka, 1988;Gassmann and Pérez, 2006). Nothofagus fusca-type pollen has a particularly high power of pollen dispersal (McQueen and MacPhail, 1983). Askin and Jacobson (1996) discussed further specific factors affecting the marine and terrestrial palynomorph record in the López de Bertodano Formation on Seymour Island. The interpretations herein take all of these factors into account, and provide as realistic a reconstruction of Maastrichtian to earliest Danian vegetation and palaeoclimate as possible from the material available.
To explore the palynological dataset further, a 'Humidity Index' has been generated based on a weighted averaging function using spore and pollen taxa with affinities to known hydrological groups (Table 1). Plants are grouped in order of moisture level preference from Thermo-/Xerophytic (prefer warm and dry, or well-drained conditions) to Hygrophytic (preferring very humid conditions). The resulting Humidity Index is a one-dimensional summary of wetness based on the same principle as the hydroclimatic index used in Holocene plant macrofossil research (e.g. Daley et al., 2012). Thermo-/Xerophytic, Mesophytic, Hygro-mesophytic and Hygrophytic taxa in a sample were given arbitrary scores of 1, 2, 3 and 4 respectively. A weighted average (WA) score was then calculated from the relative percentages of classified taxa within a sample.
The WA score was calculated as follows: x taxon percentage abundance w taxon weight i a given stratigraphic height in the section.
Climatic and ecological interpretations based on NLR analysis of the palynological dataset are necessarily cautious within the constraints already discussed. However, the high stratigraphic resolution of the samples, excellent preservation of the fossils and quantitative laboratory analyses also allow an exploration of the results using multivariate numerical methods, although this is not the focus of this study. To assess potential ecological gradients, and provide support to our qualitative ecological interpretations based on NLR techniques, Non-metric Multidimensional Scaling (NMDS) using the Bray-Curtis dissimilarity distance analysis was employed (Supplementary Dataset 2).

Results
Samples throughout the López de Bertodano Formation yielded well-preserved terrestrial plant-derived and freshwater palynomorphs (Figs. 3 and 4). Marine palynomorphs from the same sequence have been described in Askin (1988), Thorn et al. (2009) and Bowman et al. (2012, 2013a. Forty-four spore and pollen species from thirty-two genera were recorded throughout the section, in addition to further specimens identified to generic level, rare freshwater algal spores and fungal remains (Appendix 1, Supplementary Dataset). A single testate amoeba specimen was tentatively identified from 567 m , although this may also represent a fungal spore. Most taxa are present throughout hundreds of metres of the section with a slight increase in the frequency of appearance of new taxa in the latest Maastrichtian (Fig. 5A). There appears to be no evidence from this section of significant extirpation of land plant taxa across the K-Pg transition in this region (Fig. 5B, Supplementary Dataset). Rare adherent clusters of same-taxon pollen grains were recorded  implying that transport to the depositional site (and subsequent laboratory processing) involved little energy and confirms close proximity to the source vegetation. Fragments of cuticle, conifer tracheid and abundant inertinite are also present. Spore and pollen recovery was variable but largely good with between 1142 and 15,281 palynomorphs, including significantly reworked grains, per gramme of sediment recorded (Fig. 6).
It is considered that only minor proportions of the terrestrial palynoflora are obviously reworked (average 1.4%, Supplementary Dataset), either from the Permian-Triassic or from earlier in the Cretaceous (Fig. 4-21 to -27). Askin and Elliot (1982) also noted Permian-Triassic and Early Cretaceous reworking within Seymour Island sediment.
The relative abundances of the most common palynomorphs are presented in Fig. 6, plotted against stratigraphical depth and grouped by NLR. To facilitate discussion, this terrestrial palynomorph record is divided into six phases delimited by notable changes in spore and pollen abundance (Figs. 5 and 6). The palynoflora is dominated throughout by the bisaccate pollen grains Phyllocladidites mawsonii and Podocarpidites spp., comprising up to 34% and 40% of the palynoflora respectively (Fig. 6). Nothofagidites spp. are also abundant throughout, reaching their highest levels in the lowermost and uppermost parts of the section.
Phyllocladidites mawsonii varies in abundance up-section; it exhibits the most striking pattern of all the pollen taxa. This species increases in abundance gradually from the base to 746 m, through phases 1 and 2, from 3% to 17% of the terrestrial palynoflora (Fig. 6). Above 746 m, it rises rapidly to a peak of 33% at 816 m during phase 3 (Fig. 6). An abrupt drop in abundance to a low of 4% at 833 m, perhaps exaggerated by the lower spore and pollen recovery at this horizon, is followed by a further gradual increase during phase 4 (Fig. 6). Phyllocladidites mawsonii then suddenly decreases in abundance over the K-Pg transition, followed by a rapid rise and fall during phase 5. Here it attains its highest level in the section at 1024 m (34%), before decreasing to 2% at the top of the section (phase 6, Fig. 6).
Podocarpidites spp. comprises a larger component of the palynoflora than Phyllocladidites mawsonii. Although variable, it exhibits a subtle increase in abundance throughout phase 1 and stays at high levels for most of phase 2 (Fig. 6). During phase 3, the abundance of Podocarpidites spp. varies in inverse proportions to that of Phyllocladidites mawsonii with a peak up to 38% at the base, dropping to 16% by the top, followed by a rapid increase to 36% at the start of phase 4 ( Fig. 6). Podocarpidites spp. are relatively common throughout phase 4, decreasing briefly to 9% across the K-Pg boundary (Fig. 6). Podocarpidites spp. increase rapidly again thereafter to a peak of 31% at 1014 m, immediately prior to that of the post K-Pg Phyllocladidites mawsonii peak at 1024 m. Podocarpidites spp. then remain relatively abundant to the top of the section.
Nothofagidites spp. represent the third major component of the terrestrial palynoflora, and appear to subtly vary in inverse proportions to the abundance of Phyllocladidites mawsonii (Fig. 6). Although variable in abundance in the lowermost part of the section, Nothofagidites spp. decrease in abundance throughout phase 2 from 32% to 11% and increases to approximately 20% during phase 3 (Fig. 6). During phase 3, slight changes in abundance indicate a decrease throughout down to 12%, followed by a rapid increase to a minor peak of 22% within phase 4 ( Fig. 6). Thereafter a variable but gradual decrease in abundance occurs throughout phase 4, followed by a rapid increase in abundance (despite relatively low recovery) across the K-Pg boundary, peaking at 29% in the lowermost Danian sample at 1008 m.
Additional taxa that comprise a significant proportion of the terrestrial palynoflora include Laevigatosporites ovatus, Peninsulapollis gillii and Stereisporites antiquasporites (Fig. 6). Laevigatosporites ovatus occurs consistently throughout, and exhibits no particular abundance trend. The broad peak in abundance of this species across the K-Pg transition is unremarkable, and well within the range of variability observed throughout. The abundance pattern of P. gillii is similarly unremarkable throughout most of the section, except for a peak of 11% (its highest level in the section) post K-Pg at 1014 m, coincident with a post K-Pg peak in Podocarpidites spp. (Fig. 6). S. antiquasporites occurs in low numbers (≤3%) throughout phases 1 and 2, after which it begins to increase in abundance, reaching a peak of 16% at 1054 m. There is a slight drop in the abundance of S. antiquasporites at the K-Pg boundary, but this appears to be within the variation for that part of the section.
Other spore and pollen taxa that occur in minor proportions and consistently throughout are Cyathidites minor, Osmundacidites wellmanii, Proteacidites spp. and Retitriletes austroclavatidites (Fig. 6, Supplementary Dataset). These forms all show subtle variability in abundance throughout, however are slightly more abundant during phase 4. Araucariacites australis exhibits an interesting abundance pattern. It occurs rarely (≤ 3%) below the K-Pg boundary, then peaks abruptly up to 6% in the lowermost Danian sample at 1008 m, coincident with an isolated peak in Nothofagidites spp. Thereafter it returns to low levels during the remainder of phase 5, increases again within phase 6, and peaks at its highest level in the section at 15% in the uppermost sample at 1084 m ( Fig. 6).

Terrestrial palynomorph biostratigraphy
All, except Tricolpites lilliei, of the spore and pollen species from the López de Bertodano Formation have long stratigraphical ranges that include the Late Cretaceous and Danian of the Southern Hemisphere (e.g. Harris, 1965Harris, , 1972Stover and Partridge, 1973;Raine et al., 2011), thereby confirming the age of this succession. However, due to these long stratigraphical ranges of the cosmopolitan forms, and the high proportion of endemic angiosperms (Askin, 1989) within this section, further refinement of the terrestrial palynobiostratigraphy within this interval is limited. The shorter stratigraphical ranges of the marine palynomorphs from the same section have, however, been useful for establishing a regional dinoflagellate cyst biozonation scheme (Bowman et al., 2012).
The only biostratigraphically significant terrestrial palynomorph species from the López de Bertodano Formation is Tricolpites lilliei, which is known to have its range top at the K-Pg boundary in Australia and New Zealand (Vajda et al., 2001;Ocampo et al., 2006;Ferrow et al., 2011). This species is restricted to preliminary biozone PM2 in New Zealand (upper Lower Haumurian to Upper Haumurian = middle Campanian to Maastrichtian) (Raine, 1984;Crampton et al., 2004;Hollis et al., 2010). In Australia, it is restricted to the Tricolporites (al. Tricolpites) lilliei and Tricolpites longus Zones of the Proteacidites Superzone, correlated with the lower to middle part of New Zealand zone PM2 (Helby et al., 1987). The ages of the bases of these two zones are considered to be 77 and 73 Ma respectively using Schiøler et al. (2002). In the López de Bertodano Formation on southern Seymour Island, Tricolpites lilliei is present close to the base, ranging up to 984 m (Taxon 48 in Fig. 5, Supplementary Dataset),~24 m below the K-Pg boundary (Bowman et al., 2012).
The base of zone PM2 in New Zealand is marked by the first occurrence of Tricolpites lilliei, which occurs in older strata on Vega Island in the James Ross Basin (Dettmann and Thomson, 1987), suggesting that (based solely on the terrestrial palynomorph record) the lowermost strata exposed on southern Seymour Island can be no older than middle Campanian. The dinoflagellate cyst record implies that the base of the section is within the Maastrichtian (Bowman et al., 2012). The top of zone PM2 is marked by the first occurrence of Tricolpites secarius in New Zealand, which overlaps slightly with the range of T. lilliei, and is considered to correlate approximately with the K-Pg boundary (Raine, 1984). Tricolpites secarius was not recorded from the López de Bertodano Formation on Seymour Island. The New Zealand biozone PM3 extends into the Waipawan (Early Ypresian) (Raine, 1984), beyond the top of section D5.251.  Table 1 Botanical affinities, and their ecological types, of terrestrial palynomorph taxa recorded from the Maastrichtian to lowermost Danian López de Bertodano Formation, Seymour Island. Botanical affinities and nearest living relatives after Askin (1990a), MacPhail et al. (1994, Gallagher et al. (2008), Truswell and MacPhail (2009) and Raine et al. (2011), and references therein. Additional references are noted with the relevant taxon. Ecological type after Wang et al. (2005) Raine et al., 2011(refer also to Sampson, 1996 T/X

Vegetational composition and 'nearest living relatives'
The overall composition of the terrestrial palynomorph assemblage is relatively stable throughout the entire López de Bertodano Formation, with the only significant perturbation in the proportions of the major groups being evident near the K-Pg boundary (summary plot, Fig. 6; Supplementary Dataset 2). Abundant Podocarpidites spp., Nothofagidites spp. and Phyllocladidites mawsonii, despite some likely overrepresentation in the palynoflora (Section 3) suggest a lowland to mid altitude vegetation of mixed podocarp-southern beech temperate rainforest, with abundant available moisture, growing on the eastern flank of the Antarctic Peninsula arc throughout the Maastrichtian and earliest Danian (Table 1, Fig. 7). The presence of Araucariacites australis, Ericipites scabratus, Microcachryidites antarcticus and Proteacidites parvus implies a component of the palynoflora that also reflects a higher altitude, more open, heath-like vegetation at the margin of montane araucarian forests (Table 1, Fig. 8).

Lowland to mid altitude mixed temperate forests
The arborescent canopy was comprised primarily of diverse podocarps (e.g. Podocarpus, Dacrydium) and southern beech with rare ?Cupressaceae/?Taxodiaceae trees and possibly mistletoe (Elytranthe). Elytranthe is locally common in New Zealand today, especially on Nothofagus (McQueen and MacPhail, 1983). It was a relatively open canopy, at least locally, to allow the growth of a diverse understory.
Mixed beech forests in New Zealand today are distributed between podocarp-broadleaf lowland forests and pure beech forests of montane and subalpine zones (Ogden et al., 1996). In Australia, Podocarpus grow where MATs are~15°C in mainly southern temperate climates ( Table 2). The ecology and climate of Nothofagus, the southern beech, is described in detail by Veblen et al. (1996), and grows in cool temperate environments ( Table 2). The co-occurrence of southern beech and podocarps occur today in New Zealand and Tasmania where there is high moisture availability, although natural differences in shade tolerance means that these taxa may have been in forest stands dominated by one or the other on the Antarctic Peninsula lowlands (Poole, 1987;Veblen et al., 1996;Gallagher et al., 2008).
The nearest living relative of Phyllocladidites mawsonii is Lagarostrobus franklinii (the Tasmanian Huon Pine), and is of particular interest due to the distinctive abundance pattern of this pollen type throughout the López de Bertodano Formation (Fig. 6). Lagarostrobus franklinii trees are evergreen, large (up to 30 m tall), long-lived (N220 years), slow-growing and are closely related to Dacrydium (Quinn, 1982;Mabberley, 1997). Lagarostrobus franklinii seeds are buoyant, and they remain viable after floating for one to two months (Shapcott, 1991). This helps their dispersal in freshwater, and their entrainment into the fossil record (Hill, 1990;Molloy, 1995). Their terrestrial dispersal rate is otherwise slow, commonly via bird transport and gravity (Molloy, 1995;. However, they are capable of vegetative regrowth, dominating canopy gaps even before fastgrowing species . Their modern habitat is closely linked with flowing water, and today they are indigenous to riverine margins in wet, southwest Tasmania, occasionally reaching up wet hillsides to 750 m altitude (Molloy, 1995;Farjon, 2010). Rainfall in this region is N 1200 mm/year (Colhoun et al., 1988). Rarely they reach subalpine habitats near the tree line up to 1040 m, tolerating low temperatures down to − 12.6°C (Anker et al., 2001;Bannister and Neuner, 2001), but generally live in cool temperate conditions with a MAT of 10.5°C (Table 2). Pollen attributed to Lagarostrobus franklinii is common (often dominant) in the Paleogene/Neogene sediments of Australia, but may have been produced by a much larger group of species in the past, with most becoming extinct (Hill, 1990). With this caveat, L. franklinii has retained ecological stability over a long period of geological time, always requiring high humidity levels (Martin, 1997), which makes it a useful modern analogue for palaeoenvironmental reconstructions.
Warmth-loving ferns (Dicranopteris/Gleichenia, Schizaeaceae) and angiosperms (?Chloranthaceae, ?Juglandaceae, Gunnera macrophylla and proteas) may have occupied north-facing, sheltered niches, probably associated with gaps in the canopy (Tables 1 and 2). Modern Gleichenia prefers lowland to subalpine heaths and boggy ground in warm temperate climates (Allan, 1961;Mabberley, 1997; Table 2). Gleicheniidites senonicus, affiliated with Gleichenia, occurs rarely throughout the section on Seymour Island, but implies a generally humid climate with sheltered habitat niches throughout the Maastrichtian (Supplementary Dataset). Clavamonocolpites polygonalis (possibly affiliated with the Chloranthaceae) occurs in the latest Maastrichtian, interpreted by Bowman et al. (2013a) as a warm climatic phase. Modern Chloranthaceae are distributed in tropical and warm temperate latitudes (Askin, 1994;Mabberley, 1997). Further, Tricolpites reticulatus is a rare, but distinctive, pollen grain that also occurs during warm phases of the Maastrichtian (Bowman et al., 2013a), and has been compared with the pollen of G. macrophylla, a modern Gunnera that occupies tropical and south temperate super-humid environments (Jarzen, 1980;Wanntorp et al., 2004).
In slightly more elevated, open and well-drained habitats, perhaps on interfluves, monocotyledonous plants (probable Liliaceae) and dicotyledonous herbs (Josephiniasesame family), proteaceous shrubs or small trees (Adenanthos) may have been growing. Proteacidites adenanthoides and Proteacidites sp. cf. Proteacidites crassus may have a natural affinity with the protea Adenanthos, native to warm temperate climates of south and west Australia (Tables 1 and 2; Mabberley, 1997). Although rare in the section, these distinctive proteaceous pollen grains also provide an indication of warm, sheltered habitats in the Antarctic Peninsula forest at this time. In addition, Proteacidites pseudomoides has two possible living relatives within the proteas: Carnarvonia and Lomatia. Carnarvonia is a modern genus of a single Table 2 Temperature data, and derived climate type, of nearest living relatives (NLRs) of the terrestrial palynomorph taxa, Maastrichtian to lowermost Danian López de Bertodano Formation, Seymour Island. Cross-reference to Table 1 and see Section 5.3 for discussion. Averaged temperature parameters for each taxon (except Araucaria araucana), WorldClim data (after Hijmans et al., 2005) from the Atlas of Living Australia (ALA, http:// www.ala.org.au/), mostly records from Australasia and Papua New Guinea: MAT, mean annual temperature; CQMT, coldest quarter mean temperature; WQMT, warmest quarter mean temperature. Araucaria araucana climatic data after Sanguinetti and Kitzberger (2008): MJanT, mean January temperature; MJulT, mean July temperature. Climate type after Emanuel et al. (1985): STr, Subtropical MAT 17-24°C; WT, warm temperate MAT 12-17°C; CT, cool temperate MAT 6-12°C; Sn, Snow (Boreal) MAT 3-6°C. species (Carnarvonia araliifolia) growing into large trees (Hyland, 1995). It is endemic to the wet northeastern corner of Australia (Mabberley, 1997) where the MAT is 21.2°C (Table 2). Lomatia grows in cooler, but still warm temperate climates (MAT 12.5°C, Table 2) as prostrate shrubs and small trees. It is known from the marine palynological record that both cold and warm phases occur in the Antarctic Peninsula region during the latest Maastrichtian (Bowman et al., 2013a) and P. pseudomoides is recorded throughout the López de Bertodano Formation. This suggests that it is likely that this fossil taxon does represent more than one living relative with differing habitat preferences. Proteaceae were a diverse and important part of the Antarctic Peninsula flora, in both rainforest understory to more open areas, although, as discussed by Askin and Baldoni (1998), they are likely under-represented in the spore and pollen counts. They are not prodigious producers of pollen like the wind-pollinated Nothofagus and conifers, due to their different pollination mechanisms. Thus they do not appear in the graphs of more commonly occurring species (Fig. 6), with the exception of Peninsulapollis gillii, which is of uncertain ancestral proteaceous affiliation. Throughout the López de Bertodano Formation, the varied species typically occur as single to rare specimens in any one sample, and the family as a whole (excluding P. gillii in this study) rarely reaches 2% of the terrestrial count (Askin, unpublished data).
In low-lying areas, freshwater ponds were present with aquatic ferns (Salviniaceae), algae and perhaps testate amoebae on the boggy margins. Possible zygospores of the Zygnemataceae (green filamentous algae) indicate that the ponds were clean, oxygen-rich, probably still, shallow and warm for at least part of the year. Furthermore, the presence of these spores indicates seasonality in order to stimulate zygospore formation (Jarzen, 1979). Fungal palynomorphs were Fig. 5. Observed stratigraphical ranges of terrestrial palynomorph taxa from section D5.251, Seymour Island, Antarctic Peninsula. Important note: Observed stratigraphical range bases and tops from this section are not intended to imply speciations and extinctions, respectively. Data presented in this figure is only from section D5.251 of this study. However, data from other parallel sections suggest 10 of the 14 taxa here with apparent range tops during the latest Maastrichtian actually extend to the top of the López de Bertodano Formation (Askin, 1990a,b, 1994 andunpublished data). Taxa considered to be entirely reworked are not included. A, Plotted in stratigraphical order of their first appearance datums. B, Plotted in stratigraphical order of their last appearance datums. Horizontal tick marks indicate observed occurrence of each taxon throughout their range. K, Cretaceous. Pg, Paleogene. Climate phases are discussed in Section 5.3 and also shown in Fig. 6 recorded throughout the section and reflect the saprophytic degradation of terrestrial biomass in a humid palaeoenvironment.

High altitude Araucaria forest and heathland
Some elements of the terrestrial palynomorph assemblages suggest a well-drained, higher altitude habitat for part of this flora, with marginal forest environments typified by Araucaria trees (see Kershaw and Wagstaff, 2001) and open heath-like vegetation with Ericaceae (Ericipites scabratus) and Microcachrys (Microcachryidites antarcticus) on infertile soils (MacPhail et al., 1999) (Fig. 8). Alete pollen grains comparable with Araucariacites australis are produced today by Araucaria araucana (the monkey puzzle tree), a distinctive araucarian tree (Heusser, 1971) that extends to the tree line in modern southern Chile and Argentina. Today, the monkey puzzle tree grows in cool temperate conditions where mean July temperatures (coldest month) are 7-8°C, and mean January temperatures are 17-19°C (Table 2; Sanguinetti and Kitzberger, 2008). High abundances of araucarian pollen, as recorded above 1044 m in the Seymour Island section, are indicative of dense, emergent cover of Araucaria in relatively dry conditions .
The trisaccate pollen Microcachryidites antarcticus is affiliated to the modern creeping shrub Microcachrys tetragona (the sole species of that genus), which is endemic to western Tasmania and grows at high altitude in a boreal-type (snow) climate, where MATs are~5-8°C (Table 2; Silba, 1984;Biffin et al., 2011). Microcachryidites antarcticus occurs consistently throughout the Maastrichtian part of section D5.251 implying that the subalpine/alpine zones of the Antarctic Peninsula arc retained some vegetation throughout this period. It is possible that at these altitudes Microcachrys grew alongside other low-growing shrubs such as the protea Bellendena montana (the NLR of Proteacidites parvus, Table 1), which today also occupies high altitudes in Tasmania with similar climatic tolerances ( Table 2).

Late Cretaceous south polar forests and climate
Mixed forests of similar composition to those observed herein, with both closed and more open canopies, were growing across the high southern latitude landmasses during the Late Cretaceous (Specht et al., 1992). Dettmann and Jarzen (1990) discussed a similar co-occurrence of rainforest and sclerophyllous Proteaceae in southern Australia during the Late Cretaceous and invoked the lack of a nearest living relative, which is certainly the case for complex forests in a polar light regime (Hill and Dettmann, 1996). However, these Late Cretaceous lowland to mid latitude rainforests of the eastern Antarctic Peninsula do have similarities with the modern Valdivian rainforests of Chile, which have an angiosperm-dominated canopy with subordinate gymnosperms, and an The stratigraphic occurrence of Clavamonocolpites polygonalis (circles) and Tricolpites reticulatus (stars) are also plotted (although too rare to plot as abundance curves) because of their thermophilic nearest living relatives (?Chloranthaceae and Gunnera macrophylla respectively, Section 5.3). The summary abundance plot shows how the proportions of each plant type vary throughout the section (spores and pollen grouped by modern botanical affinity, see Table 1). The Humidity Index is a one-dimensional summary of ecological groupings using a weighted average (WA) model based on habitat moisture level preference (Table 1). It is an expression of how wet the palaeoclimate was through the Maastrichtian to earliest Danian at this locality. The palynofloral trends are divided into six climatic phases for ease of discussion, which record both temperature and humidity fluctuations throughout this time period.   Fig. 8). All fauna are known from the Cretaceous fossil record of the James Ross Basin (Chatterjee, 1989(Chatterjee, , 2002Case et al., 2000Case et al., , 2003Case et al., , 2007Clarke et al., 2005;Salgado and Gasparini, 2006;Cerda et al., 2012;Coria et al., 2013). Artist: James McKay, University of Leeds. understory of other angiospermous plants (e.g. Liliaceae) and ferns (Leppe et al., 2012).
The mixture of terrestrial taxa from many plant groups, as seen in the López de Bertodano Formation palynoflora, most likely represents a mosaic of habitats and vegetation types from different altitudes growing on the eastern flank of the Antarctic Peninsula during the Maastrichtian and earliest Danian. These habitats include open and closed canopy forest, river margins, freshwater ponds, well-drained interfluves, montane araucarian-dominated forests and high altitude, infertile heath. Based on the climatic preferences of the NLRs, the lowland to mid altitude eastern flanks of the Antarctic Peninsula mountains probably existed in a very humid, generally cool to warm temperate climate during the Maastrichtian with MAT~10-15°C (Table 2). Such average temperatures would not have precluded freezing conditions in the winter, particularly at the cooler end of this range. In sheltered, perhaps north-facing niches, temperatures may have risen higher supporting diverse proteas in particular ( Table 2). At high altitude, the montane forests grew at the cooler end of this range, and subalpine/ alpine heathlands grew in cold, boreal-type conditions of~5-8°C MAT ( Table 2). This diverse mixture of vegetation types was also clearly capable of surviving the extreme polar light regime together with a fauna that included dinosaurs (in the Maastrichtian) and birds (e.g. Clarke et al., 2005;Cerda et al., 2012) (Figs. 7 and 8).

Vegetational change and climate history throughout the deposition of the López de Bertodano Formation
The terrestrial palynomorph record throughout section D5.251 is divided into six phases (Figs. 5 and 6). Throughout phase 1 there is a gradual, but consistent incoming of new species with Nothofagidites spp. and Podocarpidites spp. dominating the assemblage (Figs. 5 and 6). Abundant Nothofagidites spp., consistent moss and terrestrial fern spores, the presence of aquatic fern spores (Azolla spp., Grapnelispora sp. 1) and rare freshwater algal spores suggest this was a cool and relatively humid period (Read et al., 2010). The Humidity Index suggests moisture levels remained relatively stable throughout this phase with WA scores mainly between 2.0 and 2.5 (Fig. 6). All the extant subgenera of Nothofagus grow within high rainfall regions in the high southern latitudes (Hill and Dettmann, 1996), with subgenus Brassospora (fossil pollen "brassii group") limited today to warmer though upland areas of New Caledonia and New Guinea. In this study, differentiation of the different pollen types of Nothofagidites has not been done, though it should be noted that Nothofagidites "menziesii group" (subgenus Lophozonia) is always very rare and often absent, and that the pollen "fusca" (subgenera Fuscospora and Nothofagus) and "brassii" groups vary in relative abundance though typically occur in similar frequencies, including near the K-Pg boundary (Askin, unpublished data). Further work on these pollen types may provide additional insights on the Maastrichtian vegetation and climate dynamics.
Phase 1 could be interpreted as the final stage in the development of the lowland forests of the eastern Antarctic Peninsula towards a compositional state in equilibrium with the "new" cool and humid climate, following the mid Cretaceous greenhouse. Minor climatic fluctuations are apparent from variations in the terrestrial palynomorph abundances during this phase (Fig. 6). The best example of this is at~350 m where rare Tricolpites reticulatus (Gunneraceae) occur together with a fall in the abundance of Nothofagidites spp. and a rise in Podocarpidites spp., suggestive of a brief warm spell (Fig. 6).
At 407 m, a period of relative stasis in vegetation composition began with only one pollen taxon (Taxodiaceaepollenites hiatus) appearing for the first time within~340 m of sediment (phase 2, Fig. 5). Within this phase, the only significant change in vegetation composition is that Nothofagidites spp. decreased in abundance, suggesting that the southern beech became increasingly limited by gradually warming temperatures. Rare, possible Palmae pollen appeared near the base of phase 2 in a parallel section (Askin, 1994), consistent with a warming trend. In addition, Phyllocladidites mawsonii increased in abundance implying a gradual increase in humidity (Fig. 6). The Humidity Index expresses this gradual change in moisture levels with WA score values reaching 2.5 by the top of this phase (Fig. 6).
During the relatively short phase 3, rapid environmental changes occurred with synchronous falls in abundance of both Podocarpidites spp. and Nothofagidites spp., coincident with a steep rise in abundance of Phyllocladidites mawsonii to its highest level prior to the K-Pg boundary (phase 3, Fig. 6). These abundance trends suggest that southern beech is again limited by warming temperatures, while Lagarostrobus became highly successful in a wet palaeoclimate. The Humidity Index expresses a sudden shift from relatively low to high moisture levels during this phase, followed by a sharp decrease again into phase 4 (Fig. 6). Modern Nothofagus tend to outcompete podocarps only where the climate or soil is suboptimal (McGlone et al., 1984;Case, 1988). However, with both of these in decline towards the top of this phase, one explanation could be that podocarps were limited not by temperature or soil conditions, but by the super-humid climate favoured by Lagarostrobus.
Bryophytes began to increase slightly in abundance during phase 3, a trend that continued upwards. Also at the beginning of this phase (between 746 and 776 m), there was an increase in the number of presumed thermophilic species (including Clavamonocolpites polygonalis, ? Chloranthaceae and Tricolpites reticulatus, Gunneraceae), with five new taxa appearing over 30 m in this section (Fig. 5). Note, however, that three of those five taxa (C. polygonalis, Liliacidites sp. cf. Liliacidites variegatus and Proteacidites subpalisadus) occur lower in the López de Bertodano Formation in other parallel sections on Seymour Island (Askin, 1990a(Askin, ,b, 1994. The presence of Ericipites scabratus pollen at 776 m may represent ericaceous plants from the higher and drier hinterland of the arc, but still within a climate of potentially warm summers (e.g. Bowman et al., 2013a). The terrestrial palynomorph record during phase 3, in only 84 m of the section, suggests there may have been a sudden period of increasing humidity possibly associated with a culmination of the warming trend that began at the start of phase 2.
At 830 m the abundance of Phyllocladidites mawsonii pollen decreases dramatically, together with a significant increase in other podocarp pollen and a minor increase in Nothofagidites spp., suggestive of slight but abrupt warming and drying, corroborated by a significant decrease in the Humidity Index (Fig. 6). Phase 4 is characterised throughout by relatively high abundances of Podocarpidites spp. and relatively low levels of Nothofagidites spp., with many of the taxa present affiliated with modern thermophilic plants (e.g. Clavamonocolpites polygonalis-?Chloranthaceae, and Tricolpites reticulatus-Gunneraceae; Bowman et al., 2013a). In this part of other parallel sections, Askin (1989) records the distinctive pollen Anacolosidites sectus (Olacaceae), about 28 m below the K-Pg boundary and Bombacaceae about 44 m below the boundary. Other thermophilic types were reported sporadically throughout the López de Bertodano Formation, such as Sapindaceae (Cupanieae tribe; Askin, 1989Askin, , 1990a, with increased occurrences of these and possible Palmae pollen (Askin, 1994), near the K-Pg boundary, and appearances of two species of Beauprea-type Proteaceae (Askin, 1989) through phase 4.
Bryophyte abundances are relatively high, and the fern spore Laevigatosporites ovatus reached its highest level in the section. Within phase 4, Phyllocladidites mawsonii recovers almost immediately (by 836 m), only to drop again followed by a gradual recovery almost to the K-Pg boundary, which is also reflected in the fossil wood record of Phyllocladoxylon . Gallagher et al. (2008) also noted an increase in abundance of P. mawsonii towards the K-Pg boundary in southeast Australia. Relative abundance of P. mawsonii can also be high (up to 22%) in the latest Cretaceous (Haumurian) of New Zealand (Pole and Vajda, 2009). The spore and pollen assemblage and abundance trends in this part of the section are corroborated by relatively high Humidity Index values suggesting that this phase was warm and very humid (Fig. 6). The marine palynomorph record from the same samples supports this interpretation (Bowman et al., 2013a). The recovery of the other podocarps during phase 4 implies their growth may no longer have been limited by the super-humid conditions favoured by Lagarstrobus and they grew well in this warmer climate immediately prior to the K-Pg transition. The coincident increase in bryophyte abundance supports a humid climate at this time suggesting one possible scenario of Sphagnum segregation in wetland habitats between more well-drained areas supporting podocarp-dominated forest. However, interpretation of such fine-scale climate fluctuations and habitat mapping must necessarily remain speculative considering the many controlling factors on the palynomorph assemblage recorded. These factors, for example potential changes in the source sediments with climate change and disturbance, will have shifted through time and are considered further in the Methods and materials section above.
Phase 5 (the K-Pg boundary to 1044 m) is characterised by a rapidly changing sequence of abundance peaks of different taxa, which may indicate a successional turnover in forest composition as the landscape and vegetation adjusted to the post-boundary conditions. Detailed analysis of this interval across the K-Pg boundary is being undertaken at much higher resolution to resolve the specifics of the vegetational change both prior to and following the K-Pg impact. By 1044 m, the composition of the lowland vegetation (i.e. mixed temperate rainforest) appears to have largely returned to its state during phase 1 (presumed early Maastrichtian), with low abundances of Phyllocladidites mawsonii and relatively high abundances of Podocarpidites spp. and Nothofagidites spp. The duration of phase 5 is approximately 190 kyr using the present age model for the López de Bertodano Formation.
Throughout phase 6 at the top of the section, across the unconformity into the Sobral Formation, the apparent recovery of the palynoflora suggests a return to cool climatic conditions. High abundances of Araucariacites australis and Nothofagidites at the top of the section are coincident with low Phyllocladidites mawsonii abundances, and a brief spike in Sphagnum spores.
The abundance pattern of Araucariacites australis is of particular interest. This species is rare throughout most of the section, but increased significantly above 1044 m in the Danian (Fig. 6). This pattern suggests that araucarian trees grew in low numbers on the Antarctic Peninsula throughout both the cool and warm phases of the Maastrichtian. Araucariaceae were capable of surviving long periods of adverse climatic conditions during the Early Pleistocene, but most modern araucarians have subtropical to mesothermal climatic preferences Wagstaff et al., 2001). The exception is the Chilean monkey puzzle tree (Araucaria araucana), which prefers cooler conditions . The ability to grow in cool conditions supports the suggested affinity of A. australis with a montane taxon such as the monkey puzzle tree that could retreat higher up into the hinterland during the warmer interval of the latest Maastrichtian (phases 3 and 4). Above 1044 m (phase 6, Fig. 6), the sudden increase in abundance of A. australis may reflect cooling, which caused an expansion in the range of montane araucarians in the hinterland with a lowering of the tree line. This apparent expansion of araucarian forest at this time was also noted in the fossil wood flora from Antarctica, which shows an increase in abundance of araucarian wood across the K-Pg boundary into the Early Paleocene . The lack of these montane araucarians during the cool climate phases 1 and 2 is enigmatic. One explanation for increased abundance of araucarians at the top of the section may be that the upland terrestrial palynomorph signature became more prominent in the offshore record following the disruption to lowland vegetation communities after the K-Pg event. This event may have effectively cleared the pollen transport path resulting in rapid outwashing from more open and well-drained upland habitats in the Antarctic Peninsula hinterland. This may explain the significant decrease in the Humidity Index into this uppermost phase (Fig. 6). Interestingly, numbers of Sphagnum also increase, perhaps reflecting additional transport factors that concentrated some of the lowland wetland spores.

The broader context
This reconstruction of Antarctic Peninsula palaeoclimates in the lead-up to, and immediately beyond, the K-Pg catastrophe provides a general picture of a relatively cool early Maastrichtian and a warm, humid late Maastrichtian, possibly cooling again following the K-Pg event. This is consistent with interpretations of the marine palynomorph record from the same section (Bowman et al., 2013a). Other fossil plant and palynomorph records from the Upper Cretaceous of the Antarctic Peninsula region support this interpretation, although such records are sparse and qualitative. These include terrestrial deposits from the forearc (Birkenmajer and Zastawniak, 1989;Cao, 1992;Shen, 1994) and back-arc (Dettmann and Thomson, 1987;Dettmann and Jarzen, 1988;Askin, 1990a,b;Dutra and Batten, 2000;Césari et al., 2001;Salzmann et al., 2011) regions of the Antarctic Peninsula. Preliminary analysis of dispersed plant cuticles from the López de Bertodano Formation, Seymour Island, suggest MATs of 8 to 15°C for the latest Maastrichtian, which corroborate the MAT estimates from the terrestrial palynology (Upchurch and Askin, 1989;Upchurch, personal communication (1990) cited in Askin, 1992). All of the palynomorph assemblages known from the region, including those from the López de Bertodano Formation in this study, have similar components but differ in their constituent proportions, which are likely to be related to differences in source habitat. They are all interpreted as representing rainforest vegetation growing on both sides of the Antarctic Peninsula arc during this time within a cool to warm temperate and humid palaeoclimate (Askin, 1992), consistent with the results of this study.
In addition to leaves and palynomorphs, fossil wood has also been analysed from Upper Cretaceous sediments in the Antarctic Peninsula region (Francis and Poole, 2002;Cantrill and Poole, 2005). Tree ring analysis on fossil wood from the López de Bertodano Formation on Seymour Island (although sample resolution is low compared with the current study) suggests a cool mean annual temperature (MAT) of 7°C for the Maastrichtian samples (Francis and Poole, 2002), close to the lowest estimate of 10°C for lowland to mid altitude forests suggested herein. A cooling into the Paleocene, as seen in this study, is supported by a decrease in MAT to 4-8°C across the K-Pg boundary interpreted from fossil wood tree ring analysis from Seymour Island (Francis and Poole, 2002).
Global benthic foraminifera carbon and oxygen isotope data suggest cool ocean temperatures during the early Maastrichtian (Friedrich et al., 2012). This was followed by a further decline in ocean temperatures throughout the Maastrichtian in the James Ross Basin and at Maud Rise (Barrera et al., 1987;Barrera and Huber, 1990;Pirrie and Marshall, 1990;Ditchfield et al., 1994). Terrestrial palynofloras from the Gippsland Basin in southeastern Australia confirm cool Maastrichtian temperatures (Gallagher et al., 2008). Evidence from this study from Seymour Island of a warm interval during the latest Maastrichtian, and increasingly wet conditions at this time, supports previous interpretations from the Antarctic fossil plant and palynomorph record (Askin, 1989(Askin, , 1990aDettmann, 1989;Francis, 1991;Bowman et al., 2013a), δ 18 O interpretations from planktonic foraminifera  and modelled Maastrichtian climates (Parrish et al., 1982;Hunter et al., 2008). Similarly, consistent with the data presented herein, Stott and Kennett (1990) and Barrera and Savin (1999) also suggested a cooling into the Paleocene based on oxygen isotope records.
Global sea level compilations for the Maastrichtian indicate a fall, followed by a rise within the final~2 million years of the Cretaceous, exemplified by estimates based on the sequence stratigraphy of the New Jersey passive margin (Haq et al., 1987;Hallam and Wignall, 1999;Miller et al., 2005;Cramer et al., 2011). Palaeoclimate interpretations from this study support the concept of a cool early Maastrichtian, possibly cool enough for limited ice on continental Antarctica, which would contribute to a lowering of global sea level (see also Bowman et al., 2013a). Despite the lack of a direct modern analogue, it is considered possible for a complex vegetation to have existed at these latitudes during the partial glaciation of Antarctica because even the most sensitive of plants, the Salviniaceae (e.g. Azolla), could have survived subzero temperatures (Janes, 1998). The latest Maastrichtian rise in sea level, and the main eruption of Deccan Traps volcanism (Chenet et al., 2009), coincided with the warm interval interpreted from this study immediately prior to the K-Pg boundary. However, there is no sedimentological evidence for major sea level change throughout the Seymour Island section.

Summary
This paper presents the first quantitative terrestrially-derived palynological record from the high southern palaeolatitudes throughout the Maastrichtian and earliest Danian. The data offers an unprecedented high-resolution record of polar palaeoclimate through this interval of dramatic global change, comparable in detail to oceanic oxygen isotope analyses, but providing a direct record of atmospheric conditions from the terrestrial realm.
Using a nearest living relative approach, a diverse mixed rainforest has been reconstructed that grew on riverine lowlands on the eastern flank of the Antarctic Peninsula. The canopy would have been relatively open in places supporting flowering (and perhaps aromatic) herbs and shrubs. This lowland to mid altitude forest comprised tall trees of diverse podocarps and southern beech trees, which hosted mistletoe and epiphytic ferns. Huon Pine lined the riverbanks and duckweed ferns covered freshwater ponds. Diverse mosses, shade-tolerant ground and tree ferns, together with proteaceous shrubs grew on the forest floor and saprophytic fungi degraded the leaf litter. Associated with canopy gaps, particularly during warmer time intervals, thermophilic ferns and angiosperms colonised these sheltered habitats. Higher ground, between catchments, supported a relatively open shrubland with lilies, members of the sesame family and proteaceous shrubs or small trees. Mean annual temperatures in these lowland to mid altitude forests may have been~10-15°C, with the higher end of the range reflecting the sheltered habitat niches, and perhaps the warmer climatic phase prior to the K-Pg boundary. The cooler end of the range would not have precluded freezing temperatures during the winters. At higher altitudes, near the tree line, monkey puzzle trees dominated the forest margins opening on to well-drained, ericaceous heathland. In this subalpine to alpine zone, mean annual temperatures may have been a cold~5-8°C. Moisture availability seems to be the most important ecological factor affecting the vegetation throughout this landscape. The closest modern analogue of the Antarctic Peninsula rainforests would be the Valdivian forests of the southern Andes in Chile, although the Antarctic rainforests were apparently more protea-rich. The upland araucarian forest was similar to modern high-altitude monkey puzzle forests from southern Chile and Argentina.
Changing vegetation patterns at the margin of Antarctica suggest that, following mid Cretaceous greenhouse temperatures, the early Maastrichtian climate was relatively cool and humid, gradually warming and increasing in humidity during the mid Maastrichtian. A rapid warming~2 million years prior to the K-Pg boundary, consistent with evidence from the marine palynomorph record, coincided with a super-humid interval, before a return to cool conditions in the earliest Danian. These results do not preclude small ice caps on the Antarctic continent at this time because even the most sensitive of plants, the Salviniaceae, could have survived sub-zero temperatures at the margins. Causative mechanisms for this palaeoclimatic trend prior to the K-Pg impact remain speculative, but could have been related to changing orbital parameters and the sudden influx of greenhouse gases associated with the main eruption of the Deccan Traps. The nature of vegetational change in the south polar region suggests that terrestrial ecosystems were already responding to relatively rapid climate change prior to the K-Pg catastrophe.

Appendix 1. Taxonomic notes
This section comprises an alphabetical list of terrestrial palynomorph taxa recorded in this study with taxonomic and biostratigraphical notes where relevant. Most author references linked to the taxa listed below can be found in Raine et al. (2011); additional citations that are not included in Raine et al. (2011) can be found in the reference list herein. Nearest living relatives of the contemporaneous sporomorphs are listed in Table 1. * Reworked.

Bryophyta
Stereisporites antiquasporites (Wilson and Webster 1946) Dettmann, 1963 [Fig. 3-1] Stereisporites regium (Drozhastichich in Samoilovitch andMtchedlishvili 1961) Drugg 1967 [Fig. 3-5] Triporoletes reticulatus (Pocock 1962) Playford 1971 Lycopodiophyta Camarozonosporites ambigens (Fradkina, 1967) Playford 1971 Camarozonosporites ohaiensis (Couper, 1953) Dettmann and Playford 1968 [Fig. 3-4] Ceratosporites equalis Cookson and Dettmann, 1958 [Fig. 3-8] Lycopodiacidites bullerensis Couper, 1953 [Fig. 3-6] Taxonomic note -As noted by Askin (1990b, as Retitriletes sp. cf. R. eminulus), the specimens observed in this study are comparable to the type material of Lycopodiacidites bullerensis except the former have a more triangular amb and more irregular distal reticulum. They are also slightly smaller than the type material. Raine et al. (2011) considered the Retitriletes sp. cf. R. eminulus specimen illustrated by Askin (1990b) from Seymour Island to be within the species concept of the senior synonym Lycopodiacidites bullerensis, which is used herein. *Perotriletes majus (Cookson andDettmann, 1958) Evans 1970 [Fig. 4-21] Taxonomic note -The distinctive zonate spore Perotriletes majus is the only consistently occurring taxon recorded throughout the section that is considered reworked (Supplementary Dataset). The specimens recorded include some that are slightly smaller than the type material (for example, Fig. 4-21). The specimen illustrated also does not have the serrated edge or conical spines described in Cookson and Dettmann (1958), but this may be within intraspecific variation. This species had a wide geographical distribution during the mid to Late Cretaceous, with the latest published being from the Maastrichtian of Seymour Island (Askin, 1990a;Raine, 2005). It has since been recorded from the Paleocene Sobral and Cross Valley formations from Seymour Island (unpublished data). It remains uncertain as to whether this species is contemporaneous or reworked in the current section as its exine colour and preservation are variable. Here it is considered questionably to be reworked due to its rare occurrence within any one sample, its dark coloured exine compared with definitely contemporaneous terrestrial palynomorphs and its frequently folded and fragmented preservation.
Retitriletes austroclavatidites (Cookson 1953) Doring et al. in Krutzsch 1963 Pteridophyta Azolla Lamarck, 1783 spp.  Taxonomic note -Azolla spp. occurred as fragments of massula, often with glochidia attached, and as isolated glochidia. Askin (1990b) recorded three morphotypes of Azolla glochidia from earlier studies on the López de Bertodano Formation on Seymour Island, but stated that the different sizes of two of these forms may relate to maturity. Occasional specimens of yet another type encountered in this study exhibited multi-barbed glochidia (for example, Fig. 3-12), but can still be attributed to Azolla sp. (Margaret Collinson, personal communication, 2013). No attempt has been made here to systematically differentiate the fragments of Azolla recorded, which require further detailed taxonomic analysis. *Cyathidites sp. cf. C. australis Couper, 1953 [Fig. 4-23] Taxonomic note -This specimen is compared with Cyathidites australis Couper, 1953 due to its large size (110 μm diameter). The type material is 54-77 μm diameter (Couper, 1953) and those illustrated by Dettmann (1963) are 50-73 μm diameter.

Gymnospermae
Araucariacites australis Cookson 1947 [Fig. 3-17] Taxonomic note -In this study, pollen grains that match the description of Araucariacites australis include those with a slightly coarser sculpture that is comparable with the description of Dilwynites granulatus Harris 1965. Here these grains are treated as end members of the same species in the absence of more detailed taxonomic analysis.
Dacrydiumites praecupressinoides (Couper, 1953) Truswell, 1983 Taxonomic note -The specimens recorded are slightly larger than the type material, but within the size range recorded by Truswell (1983).
Liliacidites sp. cf. L. variegatus Couper, 1953 [Fig. 3-28] Taxonomic note -The specimens observed in this study have similar variability to those recorded by Askin (1994). They compare closely to the description of Liliacidites variegatus, but vary in their form (elliptical to subspherical), commonly with a finer mesh than the type material. Couper, 1953sp. 1 of Askin (1994 Liliacidites Couper, 1953 spp.

Liliacidites
Myricipites sp. cf. M. cavilloratus Drugg 1967 [Fig. 3-37] Taxonomic note -This single specimen is comparable to the type material of Myricipites cavilloratus, except for the presence of two pore plugs. Drugg (1967) observed similar specimens with plugged pores referring them only questionably to Myricipites, but stated they were probably "aberrant examples of M. cavilloratus".

-41]
Taxonomic note -The morphological variety and commonly folded nature of Nothofagidites spp. grains restricted identification of most specimens to generic level within the scope of this study. Nothofagidites spp. grains were all consequently counted to generic rather than species level.
Polycolpites sp. cf. P. langstonii Stover in Stover and Partridge, 1973 [Fig. 4-1] Taxonomic note -The specimens recorded compare well to the description of the type material, but are larger than the type material at up to 78 μm diameter (for example, Fig. 4-1).
Propylipollis subscabratus (Couper, 1960 Proteacidites adenanthoides Cookson 1950 [Fig. 4-2] Taxonomic note -The two specimens recorded in this study (and that of Proteacidites sp. cf. P. crassus) in the López de Bertodano Formation extends the known palaeobiogeographical distribution of this type of Proteacidites spp. pollen onto the Antarctic Peninsula where it was previously thought to be absent (Askin and Baldoni, 1998).
Proteacidites angulatus Stover in Stover and Partridge, 1973 [Fig. 4-3] Biostratigraphical note -Known from Australia only from the Paleogene (e.g. Stover and Partridge, 1973;Hill, 1994), the rare occurrence of this taxon below the K-Pg boundary in this study extends its known stratigraphical range back into the Maastrichtian.
Proteacidites sp. cf. P. crassus Cookson 1950 [Fig. 4-4] Taxonomic note -This single specimen is closely comparable with the type description of Proteacidites crassus (but it is smaller) and is similar to Proteacidites adenanthoides except possessing a coarser reticulum.
Proteacidites parvus Cookson 1950 [Fig. 4-5] Taxonomic note -These specimens compare well to the description of the type material, but include smaller specimens down to 25 μm diameter (for example, Fig. 4 -5).
Proteacidites subpalisadus Couper, 1953 [Fig. 4-10] Proteacidites Cookson 1950ex Couper, 1953 Taxonomic note -These specimens are similar to specimens classified as Proteacidites sp. 3 by Truswell (1983), except for possessing a less prominent ornamentation. Triporopollenites ambiguous Stover in Stover and Partridge, 1973 has a similar ornamentation, but has a slightly more angular amb. Further detailed taxonomic work is required to definitively classify this morphotaxon as either Proteacidites Cookson 1950ex Couper, 1953 or Triporopollenites Pflug & Thomson in Thomson & Pflug 1953. The similarity of the specimens from this study to Truswell's (1983) Proteacidites sp. 3, and the placement of Propylipollis ambiguus Partridge, 1973) Dettmann &Jarzen 1996 in the Proteaceae allow that, for the present, the designation Proteacidites sp. 1 is retained.
Stellidiopollis annulatus Dettmann and Hedlund 1988 [Fig. 4-7] Tetracolpites Vimal 1952 ex Srivastava 1967 sp. 1 (aberrant tricolpate pollen)  Taxonomic note -A single scabrate tetracolpate pollen grain was recorded. It has a quadrangular amb, is symmetrical and angulaperaturate. It has equatorial colpi (but the margins are not thickened), which are incised up to 40% of the grain diameter. Tetracolpites sp. of Mildenhall (1994), from the Cretaceous of the Chatham Islands, is similar, but the specimen from this study has a more orthogonal amb and straighter sides. It is also possible that, as only one specimen of this morphotype was recorded, that this is an aberrant tricolpate pollen grain.
?Nothofagidites sp. 1 (tetra-aperturate pollen)  Taxonomic note -Several echinate tetra-aperturate pollen grains were recorded. They have a quadrangular amb, are symmetrical and angulaperaturate. They have equatorial apertures, which have slightly thickened margins and incised up to~15% of the grain diameter. This morphotype is distinguished from Tetracolpites sp. 1 by its smaller size and echinate sculpture. These may be aberrant grains of Nothofagidites spp. and further detailed taxonomic analysis is required.
?Zygnemataceae spore B  Taxonomic note: These specimens have a subspherical form with regular, low rounded verrucae and no visible laesurae or other apertures. They occur throughout a similar range in section D5.251 to ? Zygnemataceae spore A and Azolla spp. These specimens may be related to the Zygnemataceae.