Central African biomes and forest succession stages derived from modern pollen data and plant functional types

. New detailed vegetation reconstructions are proposed in Atlantic Central Africa from a modern pollen data set derived from 199 sites (Cameroon, Gabon and Congo) including 131 new sites. In this study, the concept of plant functional classiﬁcation is improved with new and more detailed plant functional types (PFTs) and new aggregations of pollen taxa. Using the biomisation method, we reconstructed (1) modern potential biomes and (2) potential succession stages of forest regeneration, a new approach in Atlantic Central African vegetation dynamics and ecosystem functioning reconstruction. When compared to local vegetation, potential biomes are correctly


Introduction
Plant functional classifications were first proposed by a core project of IGBP, "Global Change and Terrestrial Ecosystems", in the early-mid 1990's as a tool to model vegetation dynamics and ecosystem functioning in response to climate and CO 2 . Such classifications appeared as an ecological alternative to traditional taxonomic entities for the simplification of floristic complexity in global vegetation models (e.g. Prentice et al., 1992;Woodward and Cramer, 1996;Haxeltine and Prentice, 1996;Cramer, 1997;Leemans, 1997;Smith et al., 1997) and for mapping vegetation patterns at key periods in the past (Prentice and Webb, 1998;Prentice et al., 2000). Since then, extensive research has been carried out to identify plant functional types (PFTs) which are sets of plant species exhibiting similar responses to environmental conditions and grouped on the basis of structural and functional characters (e.g. Noble and Gitay, 1996;Diaz and Cabido, 1997;Diaz Barradas et al., 1999). These PFTs are characterised by a set of common biological attributes correlated with their behaviour (Gitay and Noble, 1997;Lavorel et al., 1997;Lavorel and Garnier, 2002;Lavorel et al., 2007).
In Africa, the notion of plant functional types by the means of pollen data was developed for the first time by Jolly et al. (1998a, b), then by Elenga et al. (2000a), Peyron et al. (2000Peyron et al. ( , 2006, Ngomanda (2005) and Vincens et al. (2006) to reconstruct biomes and climate at key periods (Present, 6000 BP and 18 000 BP) or along quaternary pollen sequences. Skarpe (1996), Jolly and Haxeltine (1997) and Hély et al. (2006) have also used such a classification to test the sensitivity of African biomes to changes in climatic parameters (e.g. the precipitation regime or the temperature). According to the high species diversity of the African flora,  plant functional classifications have proved to be a good tool for understanding the present, past and future functioning of the African ecosystems.
The aim of this paper is to apply this concept of classification on a large modern pollen data set (199 pollen assemblages, 272 taxa) from Atlantic Central Africa. In this region the main ecosystems, particularly the forest ones, were not always previously accurately identified by their pollen rain due to their low sampling representation, providing difficult a comparison of the data with vegetation models. Plant functional classifications, associated with the biomisation method , were used following two ways: (1) the reconstruction of modern potential biomes and (2) the reconstruction of potential succession stages of forest regeneration, a new approach in Atlantic Central African vegetation dynamics and ecosystem functioning reconstruction. New PFTs and assignment of pollen taxa in these plant functional types are proposed.

Botanical environmental setting
The study area, covering the southern Cameroon, the Gabon and the southern Congo, is located between latitude 5 • N and 5 • S and between longitude 10 • and 15 • E ( Fig. 1; Table 1). It is mainly occupied by lowlands, with an average altitude of 400 m.
This area is floristically located in the Guineo-Congolian centre of endemism (White, 1983). In this region three main types of vegetation occurring on well-drained soils are differentiated by this author.
(ii) The mixed moist semi-evergreen rain forest ("Congolian forest", Letouzey, 1968Letouzey, , 1985, well developed in southeastern Cameroon (Letouzey, 1968(Letouzey, , 1985 and the eastern half Gabon (Nicolas, 1977) is characterized by a mixture of evergreen and some semi-deciduous species becoming more Clim. Past, 5, 403-429, 2009 www.clim-past.net/5/403/2009/  Lebamba et al., 2009; (c) Vincens et al., 2000; (d) Jolly et al., 1996; (e) Ngomanda, unpublished; (f) Elenga et al., 2000b) important in the canopy. The Guineo-Congolian wet rain forest represents the climax ecosystem of the central part of the study area. The upper stratum of this formation is generally 35-45 m high and is well-distributed in diameter classes. The high canopy density precludes the development of an herbaceous strata and favours epiphytes. The thermal gradient in the canopy is very marked, while the atmospheric moisture is permanently high. This type of forest does not show any noticeable seasonal behaviour (Mayaux et al., 1997(Mayaux et al., , 1999.

The drier types of Guineo-Congolian rain forest ("semi-deciduous rain forest", Letouzey, 1968, 1985)
These forests, also called dry peripheral semi-evergreen rain forest by White (1983), are located at the border of the wet rain forest and are floristically richer than the previous ones. More individuals of the common largest tree species are deciduous (up to 70% in the upper stratum) and lose their leaves during the dry season allowing the development of a continuous shrub stratum. The diameter classes' distribution of the upper layer is irregular. The thermal gradient is less marked than in the previous type, while the seasonality is more marked in mesological conditions (Mayaux et al., 1997(Mayaux et al., , 1999. Secondary forest, occurring on past cultivated areas, is widespread in the region, and corresponds to various stages of forest regrowth in which light-demanding species and pioneers are abundant (e.g. Kahn, 1982;Catinot et al., 1983;White, 1983;Mayaux et al., 1997Mayaux et al., , 1999. The upper layer of the secondary formations is continuous and homogenous and often characterised by a monospecific composition in its earliest stages, with heliophytic and fast height growth pioneer species.

The mosaic of rain forest and secondary grassland
Much of the rain forest at the northern and southern limits of the Guineo-Congolian region has been destroyed by cultivation and fire and replaced by secondary grassland which often occurs in mosaic with small, usually severely degraded, patches of the original forest.
Inside the Guineo-Congolian domain appear vast savannas, either as large patches surrounding the forest massif, or as small islands enclosed within the forest. The trees and shrubs of these savannas are sparse while grasses form a www.clim-past.net/5/403/2009/ Clim. Past, 5, 403-429, 2009 continuous and high stratum. They find their origin in soil conditions (poorly developed, sandy or lateritic soils), in past and present human activity (settlements, fire) or in past climatic changes (e.g. Robyns, 1936;Richards, 1952;Schwartz et al., 2000).

The modern pollen data set
A total of 199 modern pollen spectra were compiled in a data set. They all have been exclusively extracted from surface soil (in savanna) or litter (in forest) samples collected following the Wright method (1967) widely used in African modern pollen studies (eg. Jolly et al., 1996;Lézine and Edorh, 1991;Bonnefille et al., 1993 ;Vincens et al., 1997Vincens et al., , 2000Elenga et al., 2000b). The location of the study modern pollen samples is given in Fig. 1 and Table 1. Seventy three samples were collected in southern Cameroon (Vincens et al., 2000 and unpublished data;Lebamba et al., 2009), eighty three in Gabon Lebamba et al., 2009;Ngomanda, unpublished data) and forty three in southern Congo (Elenga et al., 2000b). All samples come from vegetation formations occurring on well-drained soils excluding riparian and swampy formations since these formations are not directly linked to climate but rather to local hydrological conditions. They cover the three main White's vegetation types described above. Local vegetation at each site, extracted from field observations or detailed inventories, is given in Table 1. The pollen data set comprises a total of 272 pollen taxa which nomenclature was standardized following the list of taxa available in Vincens et al. (2007) and on the African Pollen Data base web site (2008). This list refers to the botanical nomenclature proposed by Stork (1991-1997).

The biomisation method and its application to the modern pollen data set
The biomisation method classifies the plant taxa represented in the pollen assemblages into a number of plant functional types (PFTs) which are broad classes of plants defined by life form (e.g. tree/shrub/lianas/herbs), leaf form (e.g. broadleaved/needle-leaved), phenology (e.g. evergreen/deciduous) and bioclimatic factors. Pollen taxa are assigned to one or more PFTs, then affinity scores are calculated for each biome in turn based on its list of characteristic PFTs. The pollen sample is assigned to the biome to which it has the highest affinity. This method, initially developed for Europe and now used worldwide, was described in detail by .
In this paper the biomisation procedure has been applied on our modern pollen data set from Atlantic Central Africa, following two ways: (1) the reconstruction of modern potential biomes (see 4.1) and (2) the reconstruction of potential succession stages of forest regeneration (see 4.2) Comparisons with local or more regional botanical data were performed in the aim to test the level of confidence of our reconstructions, and particularly of our taxa-PFT assignments.
Compared to previous works undertaken in Africa, we propose in this paper the creation of new PFTs taking into account: (1) a more precise definition of the life form of plants which produce the pollen taxa (trees or shrubs, lianas and herbs) and (2) the place these plants occupy in the different central African ecosystems (e.g. tropical wet rain forest, tropical dry rain forest forest. . . ), mainly linked to bioclimatic factors (e.g. rainfall, temperature, cloud cover, atmospheric humidity. . . ). We have differentiated the trees and shrubs from the lianas and herbs. In the PFTs Te1 (wet tropical Clim. Past, 5, 403-429, 2009 www.clim-past.net/5/403/2009/ Table 3. Allocation of the pollen taxa derived from all sites listed in Table 1 to the plant functional types used for the biomes reconstructions. Family Taxa  Te1 Te2 Tr1 Tr2 Tr3 TLw TLd THw THd g (2000) and Vincens et al. (2006), grasses (Poaceae) represent a particular PFT (g). Thus, a total of 10 PFTs is used in this work ( Table 2). The 245 pollen taxa have been allocated to one or more of these PFTs (Table 3). When a taxa is assigned to more than one PFT, this is generally due to its low level of identification (family or genus). Thus, it can include several species with different biology (e.g. Rubiaceae, Danieilla, Parinari. . . .) or it can comprise species that can adopt different life forms in different environments (e.g. Acacia a tree in savanna and a liana in forest. . . .). The taxa-PFT allocation has been adapted to the study area and so shows many differences compared to the works of Jolly et al. (1998b), Peyron et al. (2000), Vincens et al. (2006) or Lézine et al. (2009). The corresponding plant life form and habitat of pollen taxa have been determined using West and Central African botanical literature (e.g. Flore du Congo Belge et du Ruanda-Urundi, 1948; Flore du Congo, du Rwanda et du Burundi, 1967Burundi, -1971Flore d'Afrique Centrale (Zaïre, Rwanda, Burundi), 1972-2004Hutchinson andDalziel, 1954-1972;Flore du Gabon, 1961-2004Flore du Cameroun, 1963-2001Letouzey, 1968Letouzey, , 1985Stork, 2003, 2006;Tchouto Mbatchou, 2004 (Table 4).

Potential succession stage reconstructions
The same numerical procedure than the one used for potential biome reconstructions has been applied for the reconstruction of the succession stages of forest regeneration. The pollen data set comprises 250 taxa, including here Elaeis guineensis, Musanga, Anthocleista, Vismia guineensis and Polyscias fulva, pioneer taxa which play an important role in the regeneration of the forest in its youngest stages.
The matrix taxa-PFTs comprises 14 PFTs including 13 new ones whose definition is based: (1) on the life form of the plants as for biome reconstructions and (2) on the place they occupy in the forest succession in function of their behaviour and growth strategies (savanna, regrowth, young secondary, old secondary and mature stages) (Richards, 1952;Descoings, 1969;Letouzey, 1968Letouzey, , 1985Schell, 1976;Kahn, 1982;Catinot et al., 1983;White, 1983;White and Abernethy, 1996;Moutsamboté et al., 2000;Lebrun et Stork, 2003Tchouto Mbatchou, 2004;de Namur, unpublished) (Table 5). Such succession status classification was already successfully applied for the aggregation of tropical tree species of the Sabah's lowland rain forests in Malaysia by Köhler et al. (2000) to suit for applications with process-based rain forest growth models. As above, the 250 pollen taxa have been allocated to one or more of these PFTs ( Table 6).
Instead of "biomes" we have created dynamic "stages" of succession of forest regenerartion which are from the youngest to the oldest one: -SAVA: corresponding to the grass herbaceous to semiwoody stage; -TRFE: corresponding to the forest woody regrowth stage. In this stage the dominant shrubs and small trees, mainly heliophilous, such as Albizia, Anthocleista, Harungana, Tetrorchidium, Trema and Vernonia conferta are mixed with many coarse herbs (e.g. Zingiberaceae), soft woody shrubs and small climbers (e.g. Dioscorea).
-TYSF: characteristically this young secondary stage is dominated by the fast growing heliophilous Musanga cecropioides which is the most abundant and characteristic secondary forest tree in tropical Africa, associated with Myrianthus, Macaranga or Albizia for the most abundant trees. The herbaceous and shrubby layer is dense and lianas are abundant (e.g. Apocynaceae).
-TMFO: This is the ultimate stage, or climacic mature stage, of forest regeneration. The floristic composition of this forest stage, the presence of shrub and herbaceous strata depends on the status of the forest: semi-deciduous or mixed semi-evergreen.
The final matrix involving the allocation of plant functional types to the 5 main succession stages is shown in Table 7.
The results

Comparison between reconstructed biomes and vegetation at each sampled site
The results of the comparison (Table 8 and for detailed results refer to Table 1) show that among the 199 pollen sites considered in this study, 87 are correctly reconstructed as a potential biome Tropical Rain Forest (TRFO). They correspond to 74 sites of rain forest from Gabon and to 13 sites from the Cameroon littoral rain forest (8) and from the Dja forest area at the wet/dry rain forest transition (5). For the 43 other sites, all originating in the Mayombe forest massif in southern Congo, a potential TRFO biome is reconstructed. This result arises the problem of the status of this forest in Atlantic Central Africa mapped by White (1983) as Guineo-Congolian dry rain forest (Fig. 1), i.e. such as the Letouzey's semi-deciduous forest of southern Cameroon, north of 4 • N (see specific discussion below, in Sect. 6.1.). A potential biome Tropical Seasonal Forest (TSFO) is reconstructed at 49 sites. Among them, 45 sites from Cameroon, located in dry rain forest and 3 in the Dja forest area at the wet/dry rain forest transition are correctly reconstructed. Three other sites inside the wet rain forest of Gabon (2) and Cameroon (1) are clearly incorrectly reconstructed as potential TSFO biome such as one savanna in Cameroon.
All the 20 remaining sites are reconstructed as a potential biome Savanna (SAVA). Among them, 19 are really savanna, and one is from a forest regrowth, but largely disturbed by Man for cultivation, leading to high frequencies of Poaceae in the pollen spectra.

Comparison between reconstructed biomes and White's Central Atlantic vegetation types
The results of the comparison are given in Fig. 1a and Table 9. Among the 199 sites, 65 sites from the wetter types of rain forest (i.e. Cameroon coastal evergreen forest and the Gabon mixed moist semi-evergreen rain forest) (vegetation type 1a, White, 1983) are correctly reconstructed as potential TRFO biome.
In the Dja area in southern Cameroon, where occurs the transition between the mixed moist semi-evergreen forest (type 1a) and the dry rain forest (type 2) (indicated in Table 9 as 1a/2 transition) the 9 sites can be considered as correctly reconstructed with 5 sites as potential TRFO biome and 4 sites as potential TSFO biome.
Inside the White's dry rain forest (type 2), only 8 Cameroon sites are correctly reconstructed as potential TSFO biome. One site is reconstructed as a SAVA but, corresponding locally to an enclosed savanna inside the forest, this reconstruction can be considered as correct. The 43 remaining sites are all the forest sites from the Congolese Mayombe massif which are reconstructed as potential TRFO biome and not as potential TSFO biome as it could be expected according to White's vegetation map.
The last White's vegetation type occurring in Central Africa (of drier type in southern Cameroon, and wetter type in central and southern Gabon) -mainly developed at the border of the forest massif -is the mosaic of rain forest and secondary grassland (type 11a). Inside this vegetation type, 17 Gabon sites are reconstructed as TRFO biome (Lopé area), 34 Cameroon sites as TSFO biome (Kandara area) and 19 Gabon and Cameroon sites as SAVA biome (Lopé area and southern Gabon, Kandara area) indicating well the mosaic character of the vegetation. For each site (total of 70), when the potential biome proposed is compared to the local vegetation as defined in the field (Table 1), it appears that all sites are well reconstructed. This shows the importance to have a minimum of botanical information at each sampling site.

Reconstruction in terms of forest dynamics
In a first step, we have considered all the succession stages which can be observed in a dynamics of reconstruction of the forest, from the younger herbaceous stage (SAVA) to the Tropical Mature FOrest stage (TMFO), including successively the Tropical Forest REgrowth (TFRE), the Tropical Young Secondary Forest (TYSF) then the Tropical Old Secondary Forest (TOSF). The results of the comparison between potential reconstructed stages and local vegetation show that only 3 stages can be considered as correctly reconstructed, with a number of correct assignments exceeding the number of incorrect ones (Table 10a and for details at each site refer to Table 1). These are the TMFO (148 sites), TOSF (6 sites) and SAVA (19 sites) stages. The youngest stages of arboreal recolonisation are poorly (TYSF) or totally uncorrected (TFRE) reconstructed.
According to these results we have re-arranged these stages into only 3 main stages: TMFO, TSFE (Tropical Secondary Forest) grouping all secondary succession stages, and SAVA (Table 10b and Fig. 1b). In this way, the potential stages correctly reconstructed are of 97.4% (148 sites), 66.6% (18 sites) and 95% (21) of confidence, respectively.

Discussion
Our reconstructions in terms of biomes or succession stages of forest regeneration arise some questions concerning (1) the status of the Congolese Mayombe forest inside the Guineo-Congolian forest massif and (2) the boundary features between two biomes or two succession stages.

The status of the Congolese Mayombe forest
The potential biomes reconstructed at all the sites from the Congolese Mayombe forest show discrepancies compared with the White's Central Atlantic vegetation type locations. Indeed, all these sites are located in the dry rain forest (type 2) and are reconstructed as Tropical Rain Forest biome (TRFO) such as the Gabon forested sites and not as Tropical seasonal Forest biome (TSFO) as it could be expected ( Table 1).
In the light of these botanical information, we have analysed part of our pollen data set -i.e. considering only forest sites (176 and 272 taxa) -using hierarchical cluster analysis (Ward, 1963) and correspondence analysis (CA) (Benzécri, 1973) to identify the possible affinities between pollen assemblages from Congo and those from Gabon and Cameroon.
The dendrogram of the cluster analysis ( Fig. 2) shows, at a first level of division (1), a clear differentiation between the Cameroon pollen spectra of Kandara and Mayos located in the northern dry peripheral rain forest (or semi-deciduous forest) and all the others. The next division (2) separates the spectra from Makokou and Belinga in Gabon and from Mayombe (Congo) from the others. At the third level of division (3a) spectra from Makokou and Belinga are well differentiated from those from the Mayombe and (3b) spectra from the Lopé (Gabon) are well separated from the other Cameroon and Gabon spectra. The Correspondence analysis (Fig. 3) displays the same features than the cluster analysis showing along the first axis a clear separation of pollen spectra from the dry peripheral rain forest of Cameroon (Kandara and Mayos) from the others, and along the axis 2 better affinities between the Congolese Mayombe spectra and the Makokou and Belinga ones than with all the others.
These results show that the Mayombe forest, though under lower mean annual rainfall (1400 mm/year) and a longer dry season of 4 months, has more floristic affinities with the wet inland evergreen semi-evergreen rain forest occurring in Gabon (mean of 1600 mm/year, a dry season of 3 months) as previously proposed by Doumenge (1992)   mapped by de Namur (1990), than with the dry Cameroon semi-deciduous rain forest (1600 mm/year, a dry season of 3 months) as mapped by White (1983). This would confirm the hypothesis expressed by Lebamba et al. (2009) suggesting the importance of the role played by the cloud cover and the relative atmospheric humidity during the dry season in the floristic composition of forests north and south of the meteorological equator, rather than the annual rainfall amount and/or the length of the dry season. High values of these two climatic parameters, linked to monsoon influences from the Gulf of Guinea in Congo and Gabon, are measured during June, July and August versus low values in Cameroon linked to influences of continental trade winds (Harmattan) during December, January and February (Nicholson, 2000;FAO Web LocClim, 2008).

The boundaries between biomes or succession stages
While the boundary between grassland and rain forest is clearly identified in the field and on aerial or satellite photographs, by their floristic composition, structure and physiognomy, as well as in pollen assemblages by the abundance of the Poaceae versus arboreal taxa, inside the forest massif the limit between the wet and the dry rain forest, and between the old secondary and the mature forest, is often difficult to establish. The difference between the wet and the dry forest types is chiefly floristic and this floristic difference does not produce a fundamentally different physiognomy. The transition from one community to the other is usually gradual; one passing into the other imperceptibly except where a concomitant change of soil and climate occurs, it may then be more abrupt (Aubreville, 1938;Richard, 1952). The gradual character of this transition, but also its mosaic character (Tchouto Mbatchou, 2004) is well evidenced in our biome reconstructions in sites of southern Cameroon in the Dja forest area (Nkoul and Cyrie sites) where some sites are reconstructed as TRFO potential biome and others as TSFO potential biome, so they have all been considered as correctly reconstructed (Tables 1  and 8).
The same problem is observed at the boundary of old secondary forest and mature forest. Richards (1952) estimates that it is almost impossible to differentiate these two types of forest formation. In tropical central American forest Budowski (1970) used 3 criteria to differentiate the old secondary forest: (1) a high concentration of deciduous species in the evergreen domain, (2) abundance of lianas of low diameter and (3) the paucity in epiphytes, which are also the most characteristic criteria of the west and central old African secondary forest (Kahn, 1982). Other authors base the boundary of these two forest types in central Africa on the presence of markers such as Elaeis guineensis only in the old secondary forest (Kahn, 1982) or species of Rinorea only in the mature one (Achoundong, 1996 or in dynamic term, on the acquisition of the mechanism of regeneration by windfall that get forest species in position to regenerate in mature forest (Kahn, 1982).
But, all these criteria used by botanists in the field are very difficult to consider in our palynological work due to the lack of detailed botanical inventories at each pollen site and to the level of identification of our pollen taxa, more generally at the genus or family levels. Only the site of Kandara can illustrate this transition between old secondary forest and mature forest, pollen samples being collected along 2 contin-uous transects (Vincens et al., 2000 and unpublished data). Along these transects the boundary between the two formations has been placed following botanical inventories and mainly the presence of Rinorea species only in the mature forest . Unfortunately, pollen grains of Rinorea have not been identified in our samples due probably to the entomophilous character of the pollination of this plant. This boundary is not clearly identified by our pollen assemblages since some samples from the old secondary forest are reconstructed as a TMFO potential stage and some samples from the mature forest as a TOSF potential stage (Table 1). But often the calculated scores for these two stages in a same sample are very close. There is probably not a well defined boundary between these two types of forest such as between dry and wet forest; instead we suggest a transitional zone with a mosaic character. Moreover, our work being based on pollen analysis, part of pollen grains produced by the plants growing on a plot (mainly great pollen producers with well dispersed pollen grains) can be transported in a contiguous plot and so affect the local pollen rain. This could also explain why the regrowth (TRFE) and the young secondary forest (TYSF) stages are not well reconstructed by pollen assemblages, probably associated also with a high anthropogenic impact in the northern Kandara transect.

Conclusions
The application of biomisation to Atlantic Central African pollen data using a larger data set, a new and more precise classification of plant functional types and new allocations of pollen taxa to these PFTs adapted to the study area than in previous works, demonstrates that this objective and quantitative method is able to accurately predict the potential vegetation in a tropical forest region with high taxonomic diversity. In the majority of the cases, the results are comparable to site-specific descriptions of the vegetation. Savannas (SAVA potential biome), tropical rain forests (TRFO potential biome) and tropical seasonal forests (TSFO potential biome) are correctly reconstructed at 97.5% of the sites, such as the main succession stages of forest regeneration (93% of the sites): savanna (SAVA potential stage), tropical secondary forest (TSFE potential stage) and tropical mature forest (TMFO potential stage). But inside the secondary forest the young and the pioneer stages are not clearly identified due probably to their low sampling representation. Some reconstructions can remain questionable mainly at sites located at the boundary of two forest biomes (wet rain forest and dry rain forest) or two forest stages (old secondary forest and mature forest) due to the lack of precise botanical inventories on these sites and to the fact that it does not occur a clear boundary between these forest types but a gradual and transitional zone, with also a probable mosaic character.
This work evidences the ability of modern pollen data to predict accurately the present potential vegetation in tropical Clim. Past, 5, 403-429, 2009 www.clim-past.net/5/403/2009/ African forest ecosystems in terms of biomes but also for the first time in Africa in terms of vegetation dynamics. These positive results open the possibility to use now with confidence fossil pollen data to reconstruct more precisely potential vegetation and its dynamics in Atlantic Central Africa during the Late Quaternary from lacustrine pollen sequences and also to refine climate past reconstructions in this region for a more accurate comparison data/modelling.