Vegetation dynamics and drivers of change in the Central Highlands of Madagascar during the last 6300 years: Pre- and post-human settlement

Abstract

The Central Highlands region of Madagascar is currently dominated by open and mosaic ecosystems, and the region is targeted for afforestation projects. At present, little is known about the origin of these ecosystems and whether or not they are ancient or anthropogenically derived mosaics. This paper provides new insights into the environmental history and drivers of change in the Central Highlands of Madagascar during the last 6300 years through palaeoecological records from Lake Dangovavy. Our results suggested that open and mosaic ecosystems have been present in the landscape for at least 6000 years that is, 4 cal. ka BP before human settlement (2 cal. ka BP). Between 6.3 and 2 cal. ka BP landscapes were characterized by a mosaic ecosystem dominated by a matrix of open montane grassland and ericoid shrubland with extent forest patches, associated with seasonal variation of rainfall, and fire. A matrix of ericoid shrubland remained abundant in the landscape despite an increase of fire regime and herbivory activities between 2 and 1.1 cal. ka BP. Major changes in the vegetation were recorded during the last millennium with a shift towards a more open C 4 -dominated grassland associated with high fire frequency and herbivory activities, most likely linked to human influence. Our findings support the notion of the natural presence of open ecosystems on the islands. These shed light on the landscape history which should be considered when setting conservation goals with a particular focus on areas targeted for tree plantations and fire bans like the Central highlands of Madagascar.

Full text

https://doi.org/10.1177/09596836241307295
The Holocene
1 –11
© The Author(s) 2024
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/09596836241307295
journals.sagepub.com/home/hol
Introduction
Open and mosaic ecosystems harbour a wealth of biodiversity yet
are often undervalued in biodiversity conservation (Bond and
Parr, 2010; Veldman et al., 2015). These ecosystems are often
assumed to be secondary degraded forests, which makes them tar-
gets for afforestation and other carbon storage schemes. However,
there is growing evidence showing that many open and mosaic
ecosystems are natural and existed long before any human inter-
ventions, supporting high biodiversity richness, as well as soil
carbon stocks (Bond et al., 2019; Fernandes et al., 2016; Tölgyesi
et al., 2022; Veldman et al., 2015). Distinguishing ancient from
derived ecosystems has important management and conservation
implications. Misclassification of open and mosaic ecosystems as
degraded forests can lead to neglect of their conservation and sub-
sequent loss of unique biodiversity (Bond et al., 2019).
Paleovegetation studies of open and mosaic ecosystems in the
tropics and subtropics remain scarce and it is therefore difficult to
distinguish ancient from anthropogenically derived mosaics.
Focussing on Madagascar, a tropical world biodiversity hotspot,
mosaic and open ecosystems, particularly grasslands, are cur-
rently abundant, but their origin and history are debated. One area
of particular controversy is the Central Highlands of Madagascar,
which is dominated by grasslands with localized forest patches
mainly confined to ravines (Vorontsova et al., 2016). The nature
and origin of the landscape are not clear, and it is still debated
whether these grasslands are natural or anthropogenically derived
(Bond et al., 2008, 2023; Lowry et al., 1997; Solofondranohatra
et al., 2018; Vieilledent et al., 2018; Vorontsova et al., 2016).
Some authors argue that open ecosystems were naturally main-
tained by natural fire and herbivores (Hansford and Turvey, 2022;
Solofondranohatra et al., 2020), while others argue that they are
the result of anthropogenic activities associated with the use of
fire and pastoralism (Erdmann, 2003; Vieilledent et al., 2018).
Vegetation dynamics and drivers of
change in the Central Highlands of
Madagascar during the last 6300 years:
Pre- and post-human settlement
Andriantsilavo HI Razafimanantsoa,1,2 William J Bond3
and Lindsey Gillson1,4
Abstract
The Central Highlands region of Madagascar is currently dominated by open and mosaic ecosystems, and the region is targeted for afforestation projects.
At present, little is known about the origin of these ecosystems and whether or not they are ancient or anthropogenically derived mosaics. This paper
provides new insights into the environmental history and drivers of change in the Central Highlands of Madagascar during the last 6300 years through
palaeoecological records from Lake Dangovavy. Our results suggested that open and mosaic ecosystems have been present in the landscape for at least
6000 years that is, 4 cal. ka BP before human settlement (2 cal. ka BP). Between 6.3 and 2 cal. ka BP landscapes were characterized by a mosaic ecosystem
dominated by a matrix of open montane grassland and ericoid shrubland with extent forest patches, associated with seasonal variation of rainfall, and
fire. A matrix of ericoid shrubland remained abundant in the landscape despite an increase of fire regime and herbivory activities between 2 and 1.1 cal.
ka BP. Major changes in the vegetation were recorded during the last millennium with a shift towards a more open C4-dominated grassland associated
with high fire frequency and herbivory activities, most likely linked to human influence. Our findings support the notion of the natural presence of open
ecosystems on the islands. These shed light on the landscape history which should be considered when setting conservation goals with a particular focus
on areas targeted for tree plantations and fire bans like the Central highlands of Madagascar.
Keywords
fire history, herbivory activities, Madagascar, Mid-Holocene, open and mosaic ecosystems, vegetation dynamics
Received 6 July 2023; revised manuscript accepted 18 November 2024
1 Plant Conservation Unit, Department of Biological Sciences, University
of Cape Town, South Africa
2
Human Evolution Research Institute, Department of Geological
Sciences, University of Cape Town, South Africa
3 Department of Biological Sciences, University of Cape Town, South
Africa
4 Leverhulme Centre for Anthropocene Biodiversity, Department of
Biology, University of York, UK
Corresponding author:
Andriantsilavo HI Razafimanantsoa, Plant Conservation Unit,
Department of Biological Sciences, University of Cape Town, Private
Bag X3, Cape Town, Western Cape 7701, South Africa.
Email: tsilavo.razafimanantsoa@uct.ac.za
1307295HOL0010.1177/09596836241307295The HoloceneRazafimanantsoa et al.
research-article2024
Research Paper

2 The Holocene 00(0)
Humans may well have existed in Madagascar since
10,000 years ago (Hansford et al., 2018); but the timing of human
settlement was only recorded at ca. 2 cal. ka BP (Burney et al.,
2004; Douglass et al., 2019; Pierron et al., 2017; Radimilahy and
Crossland, 2015). The few existing palaeoecological records from
the highlands have demonstrated the abundance of ericaceous
heathland in the last glacial (Samonds et al., 2019) and open eri-
coid grassland during the Early Holocene (Burney, 1987c; Straka,
1996). Herbivore fossils were abundant in the region, as recorded
at the Tsaramody site (Samonds et al., 2019), and fire was sug-
gested to be more frequent during the Early Holocene (Burney,
1987b). The complex interplay between climate, and fire and their
impact on open ecosystems is still not well understood (Burney,
1987b; Gasse and Van Campo, 1998, 2001). The interaction
between fire, herbivores, climate, human activity and shifting
vegetation patterns remains tentative in the region. More informa-
tion on the origin and history of the landscapes of the highlands,
as well as their drivers of change, particularly prior to human
settlement, is therefore urgently needed in order to distinguish
between natural and anthropogenically derived ecosystems and to
guide the management of this area.
Compiling high-resolution evidence that would allow recon-
struction of environmental change in the Central Highlands of
Madagascar is critical to understanding the nature and drivers of
change of the current open and mosaic ecosystems and therefore
their conservation management. Here, we present a high-resolu-
tion reconstruction of Holocene environmental changes in the
ecosystems of the Central Highlands of Madagascar using a sedi-
ment core from Lake Dangovavy in the Vakinankaratra region.
Specifically, we analyse multiple proxies to reconstruct past veg-
etation, fire history and herbivore activities using fossil pollen
and stable carbon isotopes, microscopic charcoal and dung fun-
gal spores, respectively. The following research questions are
addressed in this paper:
(1) How dynamic were ecosystems around Lake Dangovavy
from the Mid-Holocene to the present?
(2) How did fire and herbivory activities change over time
prior to and after human settlement?
(3) How did vegetation respond to changing fire regimes, and
herbivory activities before and after human settlement?
(4) What implications do the findings have for conservation
and management?
Methods
Study area
Lake Dangovavy is a small lake of approximately 600 m wide
and 4 m deep situated at −19.287569, 46.857446, and 1700 m
a.s.l. near Faratsiho in the Vakinankaratra region of the Central
Highlands of Madagascar (Figure 1). The region is character-
ized by a subhumid climate alternately affected by cool and dry
winters, and warm and wet summers (Burgess et al., 2004; Raje-
riarison and Faramalala, 1999). It has a mean annual tempera-
ture of about 20°C and receives a mean annual rainfall between
1200 and 2000 mm (Razafindratsima, 2019). The substrate is
characterized by ferralitic and hydromorphic soils (Sourisseau
et al., 2016). Modern vegetation in the region is characterized by
the dominance (ca. 65%) of grasslands with some fragmented
patches of woodlands and subhumid forests at high and mid-
elevation (Burgess et al., 2004; Rabarivola et al., 2019) forming
a wooded grassland-bushland mosaic (Moat and Smith, 2007)
(see Figure 1). Surrounding Lake Dangovavy, the vegetation is
dominated by grassland (particularly, the C4 grass Aristida rufe-
scens, Poaceae) with some shrubs and vestige of trees. Shrubs
were represented by Elephantopus scaber, Helichrysum sp.
(Asteraceae), Agauria polyphylla, Erica goudotiana, Vaccinium
secundiflorum (Ericaceae) and Weinmannnia sp. (Cunoniaceae).
Figure 1. Location of Lake Dangovavy. Ecoregion adapted from Nusbaumer et al. (2010); and vegetation cover based on Moat and Smith
(2007).

Razafimanantsoa et al. 3
The trees are represented by Ilex mitis (Aquifoliaceae), and
Aphloia theiformis (Aphloiaceae), Kotschya perrieri (Faba-
ceae), Syzygium jambos, Syzygium sp. (Myrtaceae), Maesa lan-
ceolata (Primulaceae) and Tina striata (Sapindaceae). Individual
native Uapaca cf. bojeri trees are found in the upland, as well as
plantations of exotic trees, such as Acacia spp., Eucalyptus spp.
and Pinus spp.
Sampling and dating
Using a Russian corer, a 110 cm long sediment core was collected
150 m from the edge of Lake Dangovavy at an approximate 1.4 m
depth of water (Aaby and Digerfeldt, 1986; De Vleeschouwer
et al., 2010). The lithology of the sediment core was described
using a simplified version of Troels-Smith (1955) and the Mun-
sell soil colour charts (Munsell Colour Company, 1954). Subsam-
pling was conducted at 4 cm intervals for pollen, charcoal, and
dung fungal spore analyses.
Accelerator mass spectrometry (AMS) 14C dating was con-
ducted in seven (07) bulk sediment samples at different depths to
describe the chronology of the sediment core (Table 1). The
choice to date the bulk sediment was driven by its strong correla-
tion with the pollen deposit, which serves as the primary proxy in
this research, even though macrofossil plants were also present.
Samples were pre-treated with acid for carbonate removal and
sent to the iThemba LABS facility in Johannesburg (South Africa)
and Beta Analytic Inc., Miami (United States of America). The
dates were calibrated to years before the present (cal. years BP)
using the Southern Hemisphere calibration curve ‘SHCal20’
(Hogg et al., 2020) within the open-source statistical software
programme R v.4.3.0 (R Core Team, 2023). The age-depth mod-
els were conducted using the package ‘rbacon’ v. 2.5.8 (Blaauw
and Christen, 2011) in R v.4.3.0 (R Core Team, 2023). Bacon age
models help to find outliers in the data by reconstructing the accu-
mulation histories of sedimentary deposits. The chronology was
used to identify sediment accumulation change over time and
outliers.
Pollen, stable carbon isotopes (δ13C), stable carbon
and nitrogen isotope composition (C/N), charcoal
and dung fungal spore analyses
Pollen and stable carbon isotope (δ13C) analyses were used to
understand vegetation dynamics in the area over time. Pollen anal-
ysis was based on the standard procedures using HCl, NaOH, HF
and acetolysis (Bennett and Willis, 2001; Ferguson et al., 1976).
Lycopodium tablets with a known concentration (9666 per tablet)
were added to each sample prior to the analysis to estimate con-
centration and influx (Stockmarr, 1971). A minimum standard
count of 300 pollen grains (excluding Cyperaceae and other
aquatic types) was made (Maher, 1972). Pollen value was
expressed in the percentage of total terrestrial pollen counted. Pol-
len identification was based on the modern pollen reference slides,
photographs and theses from the University of Antananarivo in
addition to published pollen atlases like Gosling et al. (2013) and
Schüler and Hemp (2016). Samples for δ13C and δ15N isotope
analysis were pre-treated with HCl and analysed at the Stable
Light Isotope Laboratory at the Department of Archaeology, Uni-
versity of Cape Town. The purpose of the analysis was to identify
the variation of C3 and C4 plants in sediment deposits (δ13C)
(McCarroll and Loader, 2004). Stable carbon isotope and nitrogen
composition (C/N ratio) were used to determine the origin of
organic matter (Cloern et al., 2002) whether it came from aquatic
plants with low C/N ratio (<12) or terrestrial plants with a high
ratio, that is, >12 (Bonn and Rounds, 2010).
Charcoal analysis was used to reconstruct fire frequency in the
area over time. Microcharcoal (<150 μm) is supposed to repre-
sent a regional signal, while macrocharcoal (>150 μm), being
heavier, is supposed to reflect local fire (Carcaillet et al., 2001).
Microcharcoal was identified and counted on pollen slides while
macrocharcoal was counted under a stereomicroscope in a petri-
dish. Charcoal is expressed in units of concentration that is, cm2.
cm−3 for microcharcoal and particles.cm−3 for macrocharcoal.
Charcoal accumulation rate was not used considering that influx
is related to the sediment accumulation rate, which is often com-
plex in the context of tropical ecosystems.
Dung fungal spore analysis was used to determine herbivory
activities in the area over time. In this paper, they are represented
by Sporormiella, Sordariaceae, Coniochaeta and Podospora taxa
assumed to be coprophilous and used in previous studies of mega-
fauna extinction and introduction of livestock in Madagascar
(Baker et al., 2013; Burney et al., 2003). Dung spores were identi-
fied and counted on pollen slides and are expressed as concentra-
tion (spores.cm−3).
Numerical analyses
Cluster analysis based on the Euclidean distance was performed
using the function ‘chclust’ in the R package ‘Analogue’ to iden-
tify the significant stratigraphic changes in the pollen assem-
blage of the Dangovavy sediment core (Birks and Gordon, 1985;
Simpson, 2007).
Palynological richness, evenness-detrended palynological
richness and β-diversity were used to evaluate biodiversity
changes in Lake Dangovavy during the last 6300 years. Palyno-
logical richness is the number of pollen types per sample with a
constant total pollen count and is estimated using rarefaction
analysis (Birks and Line, 1992). It has been used by several stud-
ies to evaluate change in past plant diversity (e.g. Colombaroli
et al., 2013; Giesecke et al., 2012; Meltsov et al., 2011). How-
ever, it can be influenced by pollen production and dispersion,
and therefore evenness-detrended palynological richness was
also considered in this paper. The index of evenness is indepen-
dent of the number of taxa (Smith and Wilson, 1996) and consid-
ers equitably pollen types highlighting rare taxa (Giesecke et al.,
2014). In addition, β-diversity was calculated to analyse changes
in pollen assemblage composition in Lake Dangovavy and sur-
rounds during the study period.
Table 1. Uncalibrated and calibrated radiocarbon dates from Lake Dangovavy.
Depth (cm) Dated material Laboratory ID Uncalibrated 14C-dates (BP) Calibrated age range (BP)
3–5 Organic sediment IT–C–1591 820 ± 50 505–830
20–22 Organic sediment Beta–554608 1310 ± 30 1176–1304
36–37 Organic sediment Beta–525649 2620 ± 30 2608–2695
55–57 Organic sediment IT–C–1710 5050 ± 41a5711–5906
67–69 Organic sediment Beta–554609 1820 ± 30a1695–1793
87–89 Organic sediment IT–C–1589 4870 ± 47 5510–5592
106–107 Organic sediment Beta–525650 5400 ± 30 6159–6202
aIndicates dates that are excluded from the age-depth model and considered as outliers.

4 The Holocene 00(0)
Ordination analysis was used to identify gradients in the veg-
etation composition during the mid-Holocene around Lake Dan-
govavy as already recognized and applied in past studies. First,
the untransformed pollen data were analysed to extract the gradi-
ent length of the first axis of the Detrended Correspondence Anal-
ysis (DCA) (Birks and Gordon, 1985). Since the gradient was
short from Lake Dangovavy (1.3 SD), a linear-based method, that
is, principal component analysis (PCA) with square-root transfor-
mation of the pollen percentage data was used in this paper.
Open-source statistical software programme R v.4.3.0 (R Core
Team, 2023) with the package ‘vegan’ (Oksanen et al., 2022) was
used for all numerical analyses.
Results
Lithology and chronology
Based on a modified version of Troels-Smith (1955), the sediment
core was composed of 10 different stratigraphic layers (SLs),
named SL-I to SL-X from the bottom (110 cm) to the top (0 cm).
The sediment mostly consists of brown organic clay mud with plant
fragments, pieces of charcoal, coarse fine sand, and gravel in some
layers (Figure 2). SL-I (110–93.5 cm) is brown and clayey with
some plant fragments. SL-II (93.5–86 cm) is grey and clayey with
some fine sands in addition to plant fragments. SL-III (86–82.5 cm)
is still grey and clayey with an abundance of coarse sand and
gravel. SL-IV (82.5–60.5 cm) is darker grey and clayey with root-
lets, fine plant fragments and coarse sand. SL-V (60.5–47.5 cm) is
similar to the first layer (SL-I), and SL-VI (47.5–38 cm) to the third
layer (SL-III). SL-VII (38–29.5 cm) is calcareous grey with some
plant fragments, while SL-VIII (29.5–18 cm) remains calcareous
grey but less plant fragments. From 18 cm to the top, the sediment
is darker clayey with the presence of plant fragments. SL-IX (18–
4.5 cm) and SL-X (4.5–0 cm) differed by the presence and absence
of fine sands, respectively.
The bottom of the Dangovavy sediment core was dated at ca.
6.3 cal. ka BP. The age-depth model based on the seven (07) AMS
radiocarbon dates suggests a mean sediment accumulation rate of
59.3 ± 7.3 years.cm−1 throughout the core but a large decrease of
163 years.cm−1 was noted from 4 cm to the top of the core. Two
outlier dates, Beta–554609 (68 cm) and IT–C–1710 (56 cm), were
excluded automatically by ‘rBacon’ as they represented reversals
in the sequence. This may indicate contamination, sediment mix-
ing (e.g. due to bioturbation or root penetration), reworking of
sediment, or deposition inconsistencies, such as hiatuses.
Pollen, stable carbon isotopes, charcoal and dung
fungal spores
Based on the cluster analysis, five (05) statistically significant
pollen assemblage zones were identified, representing major veg-
etation shifts in Lake Dangovavy over the last ca. 6.3 cal. ka BP.
These pollen zones will form the basis for the discussion of the
pollen, stable carbon isotopes, charcoal and dung fungal spore
results. Dates of the transitions between zones are tentative due to
uncertainties in the age-depth model. Figure 3 represents the pol-
len diagram of the abundant terrestrial pollen taxa (greater than
Figure 2. Lithology and age-depth model of the Lake Dangovavy sediment core, based on seven calibrated radiocarbon dates.

Razafimanantsoa et al. 5
2%) alongside the measure of δ13C and concentrations of charcoal
and dung fungal spores during the Mid-Holocene.
Pollen zone 1 (ca. 6.3–5.5 cal. ka BP). The pollen assemblage
was dominated by grassland and ericoid shrubland taxa charac-
terized by the abundance of Poaceae (29.2 ± 1.4%) and Erica-
ceae (22.7 ± 1.8%) pollen, respectively. An abundance of
riverine forest taxa was also recorded during this period repre-
sented by Anthocleista cf. amplexicaulis (12.7 ± 1.4%), and
Entada cf. chrysostachys (5.1 ± 1.1%). Pollen from high-eleva-
tion forest taxa represented by Ilex mitis (4.2 ± 0.9%) and Vitex
type (2.6 ± 0.6%) was slightly abundant during this period. Pol-
len from mid-elevation forests and open woodland taxa were
scarce (ca. 2%). On the other hand, stable carbon isotope (δ13C)
results showed a mean value of −23.7 ± 0.4‰ during the period,
while the C/N ratio value was low (11.7 ± 1.1). The charcoal
concentration was very low with a mean value of 15.3 ± 1.2
charcoal particles.cm−3 for macrocharcoal and 43.8 ± 1.6 cm2.
cm−3 charcoal area for microcharcoal. Dung fungal spores were
also very low during this period with a mean value of
0.2 ± 0.1 ×103 spores.cm−3.
Pollen zone 2 (ca. 5.5–4.3 cal. ka BP). This second pollen zone
was characterized by an increased abundance of ericoid shru-
bland, compositae and mid-elevation forest taxa. Ericaceae pollen
showed a mean value of 27.2 ± 3% and Asteraceae a value of
9 ± 2.3%. Mid-elevation forest taxa were represented by Celtis
type (8.5 ± 1%), Eugenia type (9.4 ± 1.8%) and Trema orientalis
(5.6 ± 1%). A slight increase of open woodland taxa represented
by Uapaca cf. bojeri (2.4 ± 0.3%) was also noted. However,
Poaceae decreased to half of the previous period with a mean
value of 15.7 ± 0.7%. Riverine and high-elevation forest taxa also
decreased with an abundance of less than 2%. The δ13C isotope
values were slightly lighter than in the previous period with a
mean value of −25.3 ± 0.6‰ and this was accompanied by a high
stable C/N ratio value (28.7 ± 4.4). Charcoal concentration
increased more than tenfold during this period whereby macro-
charcoal and microcharcoal increased to a mean value of
366 ± 122 charcoal particles.cm−3 and 623.4 ± 95.7 cm2.cm−3
charcoal area, respectively. An increase of dung fungal spores
was also recorded with a mean value of 2 ± 0.6 ×103 spores.cm−3.
Pollen zone 3 (ca. 4.3–3 cal. ka BP). The third pollen zone was
characterized by dominance of grassland with peaks of Poaceae
pollen at ca. 4 cal. ka BP (39.3%) and 3.5 cal. ka BP (42.2%). Eri-
coid shrubland and mid-elevation forest taxa decreased during this
period compared to the previous zones. Ericaceae, Celtis type,
Eugenia type and Trema orientalis pollen decreased to 16 ± 1%,
6 ± 1.5%, 4.3 ± 1% and 4.2 ± 0.7% respectively. Pollen from
high-elevation forest taxa remained low. The relative abundance of
pollen from riverine forests increased to 8.8 ± 3.6% with a peak of
Anthocleista cf. amplexicaulis (17.7%) and Entada cf. chrys-
ostachys (5.6%) at the end (ca. 3 cal. ka BP). The open woodland
remained stable with a dominance of Uapaca cf. bojeri (2.3%).
δ13C isotope results were slightly more positive with a mean value
of −23.4 ± 0.3‰ while C/N ratio was low (12 ± 0.8). Charcoal
concentration (macro and microcharcoal) was 5- to 10-fold higher
than zone 1 but lower than zone 2 with a mean value of 100 ± 68.4
charcoal particles.cm−2 and 205.2 ± 64.4 cm2.cm−3 charcoal area
for macrocharcoal and microcharcoal, respectively. Macro and
microcharcoal concentration were higher at the beginning of the
period and dropped at the end. Dung fungal spores dropped to a
mean value of 0.5 ± 0.1 ×103 spores.cm−3 during this period.
Pollen zone 4 (ca. 3–1.1 cal. ka BP). This pollen zone was marked
by a return abundance of ericoid shrubland, Compositae, open
woodland and mid-elevation forest taxa. Ericaceae and Astera-
ceae pollen showed a mean value of 21.5 ± 2.2% and 11.9 ± 2.7%,
respectively. Both taxa peaked at ca. 2.6 cal. ka BP with a maxi-
mum value of 30.6% and 9.6% for Ericaceae and Asteraceae,
respectively. Uapaca cf. bojeri was still dominant in the open
woodland with a mean value of 2.5 ± 0.4%. Mid-elevation forest
taxa remained characterized by Celtis type (8.5 ± 1%), Eugenia
type (7.5 ± 2) and Trema orientalis (6.5 ± 1.2%). During this
period, Poaceae pollen was low with a mean value of 16.7 ± 1.5%.
High-elevation forest taxa remained low (<2%) and riverine for-
est taxa were almost absent (<0.5%). δ13C isotope results were
more negative (−24.4 ± 0.6‰) while the C/N ratio increased with
a mean value of 25.7 ± 2.6. Charcoal concentration was higher
than zone 3 with a mean value of 311 ± 115 charcoal particles.
cm−3 and 298.5 ± 74.8 cm2.cm−3 charcoal area for macrocharcoal
and microcharcoal, respectively. Dung fungal spores increased
during this period with a mean value of 5 ± 1.3 × 103 spores.cm−3.
Figure 3. Pollen diagram of Lake Dangovavy grouped in their ecological affiliations (pollen taxa ⩾ 2%) characterized by five (05) statistically
significant pollen assemblage zones: Grassland taxa (yellow), shrubland taxa (orange), open woodland taxa (white), riverine forest taxa (grey),
high-elevation forest taxa (lawn green), mid-elevation forest taxa (dark green). The line graphs indicate the results of δ13C stable isotope (circle
symbol) and C/N ratio (cross symbol) analyses, while dung fungal spores, macrocharcoal, and microcharcoal concentrations are displayed as
brown, purple, and black silhouettes.

6 The Holocene 00(0)
Pollen zone 5 (ca. 1.1 cal. ka BP to present). This last period was
characterized by a large increase in the abundance of grassland
taxa with Poaceae pollen reaching a mean value of 55.1 ± 4.6%.
A peak of Poaceae pollen was recorded at ca. 900 cal. years BP
with a value of 61.8%. On the other hand, other taxa decreased
significantly in the pollen diagram. The ericoid shrubland and
Asteraceae shrub decreased to a mean relative abundance of
6.2 ± 1.2% and 6.7 ± 1%, respectively. The Uapaca woodland,
which was more stable in the previous periods, dropped to a mean
relative abundance of 1.8 ± 0.5%. The high-elevation forest taxa
and riverine forest taxa were very low to almost absent. Mid-ele-
vation forest taxa were also low except Trema orientalis
(5.6 ± 1.5%), which was more or less stable compared to the pre-
ceding period. The stable carbon isotope (δ13C) value during this
last period tended to a more positive value (−20.9 ± 0.3‰). The
ratio of C/N was moderately high compared to the previous period
with a mean value of 16.9 ± 0.5. Charcoal concentration was very
high both for macrocharcoal and microcharcoal with a mean
value of 883 ± 165 charcoal particles.cm−3 and 896.3 ± 115.5 cm2.
cm−3 charcoal area, respectively. Peaks were recorded at ca.
900 cal years BP with 1356 charcoal particles.cm−3 and 1321 cm2.
cm−3 charcoal area for macrocharcoal and microcharcoal, respec-
tively. Dung fungal spores were also very high during this period
with a mean value of 19.3 ± 4.2 × 103 spores.cm−3 with a peak of
30.5 × 103 spores.cm−3 at ca. 900 cal. years BP.
Rarefied pollen richness, pollen evenness and
compositional change
Rarefied pollen richness in the Dangovavy sediment core ranged
between 24 pollen types (ca. 5.9 cal. ka BP) to 41 pollen types
(ca. 1.1 cal. ka BP) with a mean value of 33 ± 2 pollen types in
the entire period (Figure 4a). Between ca. 6.3–4.3 cal. ka BP
(Pollen zone 1–2), pollen richness showed a mean value of
30 ± 2 pollen types. It increased (36 ± 1 pollen types) between
ca. 4.3 and 1.1 cal. ka BP (Pollen zone 3–4). A peak of 41 pollen
types was recorded consecutively at ca. 2.9 and 1.1 cal. ka BP.
From ca. 1.1 cal. ka BP to the present period (Pollen zone 5), a
slight decrease in pollen richness was recorded with a mean
value of 34 ± 2 pollen types.
In parallel, between 6.3 and 1.1 cal. ka BP (Pollen zone 1–4), no
long-term trends were recorded in the pollen evenness (Figure 4b).
It remained fairly constant with a value of ca. 0.8 SD. However,
based on the β-diversity calculation, the taxonomic composition dif-
fered at ca. 5.5 cal. ka BP (between Pollen zone 1 and 2); 4 cal. ka BP
(within Pollen zone 3); 3 cal. ka BP (between Pollen zone 3 and 4)
and at ca. 1.1 cal. ka BP (between Pollen zone 4 and 5, Figure 4c).
From ca. 1.1 cal. ka BP to the present period (Pollen zone 5), a slight
decrease in pollen evenness was recorded with a value of 0.6 SD.
This variation was associated with the last change in taxonomic
composition based on β-diversity calculations.
Ordination
PCA axis 1 and PCA axis 2 (noted as PCA1 and PCA2 hereafter)
explained 57.8% of the variation in the distribution of pollen taxa in
the Dangovavy sediment core (Figure 5). Poaceae, Anthocleista,
Figure 4. Biodiversity changes: (a) Rarified pollen richness, (b)
evenness and (c) compositional change (Beta diversity) in Lake
Dangovavy over the last ca. 6300 cal. years BP.
Figure 5. A Principal Component Analysis (PCA) determining the distribution of abundant and characteristic pollen taxa in a sediment
core sampled from Lake Dangovavy. The colour gradient represents the contribution of each taxa to the axis components (lowest
contribution = black; highest contribution = orange).

Razafimanantsoa et al. 7
Entada, Ericaceae, Celtis and Syzygium pollen taxa contributed
most to the axis components. PCA1 explained 31.2% of the vari-
ance in the pollen data. It was related to dryness where riverine
forest taxa (Anthocleista and Entada) and Poaceae had a high axis
1 score, whereas upland taxa represented by Celtis, Ericaceae,
Trema, Syzygium and Weinmannia, had a low axis 1 score. In addi-
tion, PCA1 showed an elevation gradient where high-elevation for-
est taxa such as Faurea, Ilex and Vitex had high axis scores, while
mid-elevation forest taxa such as Celtis, Syzygium and Trema ori-
entalis had low axis scores. PCA2 explained 26.6% of the distribu-
tion of the pollen data. It was related to fire regime where high fire
regime taxa such as the grassland taxa (Poaceae) had a high axis 2
score, whereas low fire regime taxa such as the ericoid shrubland
(Ericaceae), forest (Celtis and Syzygium) and riverine forest taxa
(Anthocleista, and Entada) had a negative axis score.
Discussion
The vegetation cover and its nature in the Central Highlands of
Madagascar has been highly debated over the last two decades.
Reconstructions have been reported on vegetation in the region;
however, these records had some chronological issues or with
low-resolution reconstruction (Burney, 1987c; Samonds et al.,
2019; Straka, 1996). Here, we present the first high-resolution
reconstruction of Holocene environmental changes in the ecosys-
tems of the Central Highlands of Madagascar, based on a sedi-
ment core from Lake Dangovavy in the Vakinankaratra region.
The sediment core from Lake Dangovavy is dated for about ca.
6.3 cal. ka BP based on radiocarbon dates of the bulk sediments.
Despite some uncertainties, complexity in the age-depth model
and the presence of hiatus (e.g. Broothaerts et al., 2023; Razanat-
soa et al., 2021; Voarintsoa et al., 2017b) which is common in
tropical and sub-tropical environments, the proxies used have
allowed to identify the major changes in the vegetation and retrace
the environmental history of the areas around Lake Dangovavy
prior and after human settlement 2 cal. ka BP.
Environmental change in Lake Dangovavy, Central
Highlands Madagascar, during the last ca. 6.3 cal. ka BP
Prior to human settlement (at least from 6.3 to 2 cal. ka BP). Our
palaeoecological record shows that open and mosaic ecosystems
were already established around Lake Dangovavy and surrounds
at least in the last 6.3 cal. ka BP, more than 4000 years before
human settlement. The vegetation was dynamic and was charac-
terized by mosaic ecosystems comprising forest patches of vari-
able extent in a matrix of open grassland and ericoid shrubland,
where grassland increased significantly with fire and herbivory
activities in the past 1000 years (Figure 6).
Between ca. 6.3 and 5.5 cal. ka BP, the pollen record suggests a
mosaic ecosystem, where open grassy and ericoid vegetation co-
existed alongside pollen from riverine and higher elevation forests
(Figure 6a). The more negative value of δ13C isotope is consistent
with the presence of C3 plants, which could be shrubs, woody
plants (i.e. the forest elements mentioned above), or aquatics plants
as justified by the low C/N ratio (Bonn and Rounds, 2010). The
presence of C3 grass in the area was also possible but this needs
further analysis on the measurements of grass-pollen δ13C to distin-
guish between C3 and C4 grass pollen (e.g. Nelson et al., 2008). Our
new record shows the importance of the co-existence of grass and
ericoid pollen, suggesting an open landscape with grassland and
heathland elements as described in the Madagascar pollen rain
studies (Burney, 1988; Razafimanantsoa and Razanatsoa, 2024). It
seems likely that forest types were restricted to riverine and higher
elevations, with grassland and heathland dominating the landscape.
Charcoal (macro and microcharcoal) and dung fungal spores data
(Figures 6d and 6e) indicate that fire regimes and herbivory
activities were very low in the area during this earliest period. The
restriction of forests might be associated with seasonally dry condi-
tions in the area as suggested by the positive value of the PCA1
during this period (Figure 6f). Based on the literature, the Central
Highlands of Madagascar experienced drier conditions in the
Early- Mid-Holocene as justified by the sediment cores collected in
Lake Tritrivakely that suggested an ephemeral swamp through
mineral-magnetic properties, pollen, and diatom analysis (Gasse
et al., 1994; Gasse and Van Campo, 2001; Williamson et al., 1998).
This drier period was in parallel with the southern Madagascan and
Southeast African climates showing an anti-phase climate relation-
ship to northwestern Madagascar and Eastern Africa during the
Mid-Holocene (Chevalier et al., 2017; Wang et al., 2019).
The period between ca. 5.5 and 4.3 cal. ka BP is marked by a
compositional change in vegetation cover (Figure 6b). Ericoid
shrubland and mid-elevation forest taxa expanded along with an
increase in woodland taxa characterized by Uapaca cf. bojeri (see
previous chapter). Riverine forests almost disappeared, while mon-
tane grassland was reduced to half its former value compared to the
previous period. Expansion of forest taxa was associated with the
slightly lighter value (−25.4‰) of δ13C which is justified from ter-
restrial sources based on the high C/N ratio during the period (Fig-
ure 6c). This change was associated with an increase in fire regimes
and herbivory activities reflected by the charcoal (macro and
microcharcoal) and dung fungal spore records (Figures 6d and 6e).
Despite the regional increase of fire, forests and shrubs might per-
sist in fire-free refugia and wetter areas, and maintain themselves
through fire-vegetation feedback as commonly observed in savan-
nas elsewhere (Huntley, 2023; Staver et al., 2011).
With an abundance of ericoid shrubland, woodland and mid-
forest taxa, montane grasses decreased, which was intensified by
increased grazing in the area during this period. Together, the abun-
dance of shrubs and woody elements account for the more negative
Figure 6. Summary of environmental change in Lake Dangovavy
during the last ca. 6.3 cal. ka BP. (a) Relative abundance of grassland
and ericoid shrubland (Poaceae, Ericaceae), and forest taxa (high-
and mid-elevation forest), (b) Biodiversity changes, (c) Carbon
and nitrogen isotopes, (d) Dung fungal spores concentration, (e)
Charcoal concentrations, and (f) Principal Component Analysis
(PCA) axes. Period of human settlement at ca. 2 cal. ka BP is
indicated by a grey bar.

8 The Holocene 00(0)
δ13C isotope value during this period. This expansion of shrubs and
forests is associated with a negative value of the PCA1 (Figure 6f)
and could be also associated with an increasing humidity in the
region during this period. In fact, around 5 cal. ka BP the Central
Highlands and Southwest of Madagascar supposedly shifted to
more humid conditions (Burney, 1987c, 1993, 1996). This was at
the end of African Humid period that is, the climate began to dry
which might confirm the anti-phase climate relationship of Central
Highlands Madagascar to East Africa previously mentioned. How-
ever, more palaeoclimatic studies are still needed in the region to
confirm this theory. A more humid environment may have increased
biomass production, explaining the increase in fire at the regional
scale, as indicated by the microcharcoal values.
A second compositional change was recorded between 4.3 and
3 cal. ka BP (Figure 6b). Expansion of grassland in the mosaic
ecosystem occurred at this time, with the highest value of Poaceae
pollen at ca. 4 cal. ka BP. Forest and ericoid shrubland decreased
compared to the previous period (Figure 6a). It was associated
with a slight trend towards a more C4 signal from the stable car-
bon results. These might suggest the onset of C4 grass expansion
in the highlands. Local and regional fire frequency were slightly
higher at the beginning of the period which may have promoted
an increase in C4 grass. Our findings suggest an abundance of C4
grass, which corresponded to a positive value of the PCA1 (Fig-
ure 6f), and was possibly associated with the 4.2 cal. ka BP dry
event in the south east African monsoon and Madagascar (Scrox-
ton et al., 2020), the warmest and driest period on the island
(Gasse and Van Campo, 1998; Virah-Sawmy et al., 2010; Wang
et al., 2019). A climatic and fire threshold might have been
reached when the combination of fire with drought led to an eco-
system shift with a decline in heathland and a shift to a more open
grass-dominated system. Herbivory activities were low during the
period, which might be linked to reduced local animal density and
functional species extinction in the landscape (Gill et al., 2009)
possibly due to the drought event that occurred. No human influ-
ence was recorded in the region during this period.
A resurgence of ericoid shrubland, woodland and forest taxa
was recorded in the area between 3 and 2 cal. ka BP (Figure 6a).
This marked the third compositional change in the vegetation
cover (see Figure 6b). Grassland decreased to half the abundance
compared to the previous period. This was associated with a more
negative value of δ13C isotope and a high value of C/N ratio (>12)
confirming the dominance of C3 terrestrial plants (Figure 6c). Fire
frequency and herbivory were low during this period, possibly
contributing to the increase of shrubs and forests in the area at this
time, as seen in the African savanna (Roques et al., 2001). This
period is associated with a negative value of the PCA1 suggesting
a more humid period (see Figure 6f). In the literature, this period
corresponded to a more cool and humid period in the Central
Highlands of Madagascar as suggested by the analyses of mineral-
magnetic proxies (Williamson et al., 1998). These cool conditions
could also have promoted the abundance of ericoid shrubland as
observed during the glacial periods in the Central Highlands
region (Burney, 1987c; Gasse and Van Campo, 1998, 2001).
Post human settlement (2 cal. ka BP to present period). Between
ca. 2 and 1.1 cal. ka BP, the vegetation around Lake Dangovavy
remained dominated by a mosaic of ericoid shrubland, and mid-
elevation forest taxa (Figure 6a). The negative value of δ13C isotope
suggests an abundance of shrubs and trees within this period. Fire
and herbivory activities started to increase slightly in the area com-
pared to the previous period. However, vegetation in the area was
resilient despite these changes, based on the pollen evenness analy-
sis (Figure 6b). Woodland taxa such as Uapaca cf. bojeri, and
Combretaceae were moderately abundant during the period (see
previous Chapter), taxa which are suggested to be resistant to mod-
erate fire in the area (Burney, 1987a). Therefore, fire was not fre-
quent enough to burn savanna trees and this mosaic landscape
might act to limit fire spread creating fire refugia for the ericoid
shrubland. The cool and humid climate recorded in the region until
ca. 1 cal. ka BP (Gasse and Van Campo, 1998) likely promoted the
expansion of shrubs in the area as recorded in the region during the
glacial period (Burney, 1987c; Gasse and Van Campo, 1998, 2001).
Similar patterns of vegetation were recorded in the region as
recorded surrounding Lake Tritrivakely during this period (Burney,
1987c). Despite the beginning of the increase in fire and herbivores
in the area, suggesting human impact, it is likely that human popu-
lation density was low during that period (Beaujard, 2011).
During the last ca. 1.1 cal. ka BP, more open grassland expanded
into the area surrounding Lake Dangovavy, accompanied by a
massive decrease in ericoid shrubland, and forest taxa (Figure 6a).
This change was associated with a shift to more C4 plants accord-
ing to the carbon isotope results, marking the last compositional
change in the vegetation cover in the area (Figures 6b and 6c). This
was accompanied by a decrease in pollen evenness suggesting vul-
nerability of the vegetation. This period is supposedly more drier
than the previous period as indicated by the positive value of the
PCA1 (Figure 6f). Based on the literature, a regional dry period
was recorded at ca. 950 cal. ka BP as indicated by recorded pollen
and diatom values (Gasse and Van Campo, 1998; Virah-Sawmy
et al., 2010). Fire regime and herbivory activities were very high
compared to the rest of the past ca. 6000 years (Figures 6d and 6e).
This result suggests a local increase of human impact in the area
during the last millennium through fire use and pastoralism, pro-
moting the abundance of grassland. Our findings show similarity
to other sites recorded in the highlands region with an expansion of
grassland around 1 cal. ka BP (Burney, 1987a, 1987c; Straka,
1996). Similar trends have been recorded almost across the entire
island, for instance in the Southwest (Razanatsoa et al., 2022), and
in the Northwest suggesting climate and human activities (Rails-
back et al., 2020; Voarintsoa et al., 2017a) as drivers.
Conservation and management implications. Our data highlights
the presence and importance of open and mosaic ecosystems
characterized by montane grassland, and ericoid shrubland around
Lake Dangovavy from ca. 6.3 cal. ka BP to the last millennium.
This finding supports the hypothesis of the presence and abun-
dance of natural open vegetation with forest patches along rivers
and in fire refugia, rather than continuous forest in the Central
Highlands of Madagascar (Bond et al., 2008, 2023; Solofon-
dranohatra et al., 2018; Vorontsova et al., 2016). Though C4 grass-
lands expanded later in the record during a drier period and again
when human impact increased, it is important to note that open
and mosaic ecosystems dominated by a matrix of grasslands and
ericoid shrubland are a natural part of the highlands landscape.
Conserving these ecosystems is worthwhile, especially the eri-
coid shrubland, which deserves focused research priorities
(Hackel et al., 2024). These habitats should not be assumed to be
degraded and replanted as part of the government’s afforestation
programme, which primarily focuses on the Central Highland’s
open spaces (Lacroix et al., 2016). The decision to afforest the
area needs careful re-evaluation, especially with regard to mono-
culture plantations of exotic trees such as Acacia spp., Eucalyptus
spp. and Pinus spp. Exotic species would harm local biodiversity
and increase biomass fuel, which would intensify fires and pos-
sibly put any remnant forest patches in jeopardy by slowing down
native species recovery (Baohanta et al., 2012).
In addition to curbing the spread of alien plantations, however, it is
also important to distinguish anthropogenic grasslands, which became
more prevalent in Lake Dangovavy and surrounds during the last
1000 years in association with increasing human impact, from the
ancient mosaics and heathlands that pre-dated this period. Hence, to
encourage the restoration of ericoid shrubland in its natural habitat
and to safeguard the remnant forest, careful fire management that will
return burning to its historical range of variability is required in the
area. As natural fire has occurred surrounding Lake Dangovavy and

Razafimanantsoa et al. 9
possibly the entire region for thousands of years before human arrival
(Burney, 1987b), banning fire is not necessary. Instead, new measures
to not exceed the threshold for Ericaceous vegetation and forest
patches should be developed in order to save the remaining fire-
adapted trees, including the Uapaca cf. bojeri. In fact, this species
plays an important role in the region as it is exploited by local people
for textile sources and food (Kull, 2002; Razafimanantsoa et al.,
2013). Therefore, conservation projects in the area should focus on the
restoration of such taxa as occurred in the former Ericaceous heath-
land but not the use of exotic trees currently promoted in the region
(Lacroix et al., 2016). Meanwhile, given that locals rely on fire to sup-
port their way of life (Kull, 2000), fire management should concen-
trate on strategies that are sufficient to preserve the diversity of the
environment and be done in conjunction with the local populace.
Conclusion
This paper provides new evidence of environmental change in the
Central Highlands of Madagascar over the last 6 cal. ka BP from
Lake Dangovavy, which highlights changes before and after
human settlement in the region and informs conservation and res-
toration management. The palaeoecological records presented
here offer a better understanding of the debated origin of the veg-
etation in the region and contributes to explaining the role of fire
and herbivory activities in shaping the current landscape. Through
this multi-proxy investigation, the palaeoecological data showed
that the area has been dominated by open and mosaic ecosystems
comprising forest patches of variable extent in a matrix of open
grassland and ericoid shrubland at least 4000 years before human
settlement. It highlighted the abundance of montane grassland
and heathland in the area and suggests that the forest was restricted
to riverine and higher elevations before 5.5 cal. ka BP. Expansion
of heathland, woodland and forest, and then grassland to the mid-
altitude with the onset of grass expansion, were recorded until 4.3
and 3 cal. ka BP, respectively in the area. Ericoid shrubland and
forest taxa resurged to mid altitude between ca. 3 and 2 cal. ka BP,
and even after human settlement more precisely at ca. 1.1 cal. ka
BP. The vegetation remained resilient within these periods despite
its dynamism and changes in regional climate. In parallel, fire was
naturally present in the landscape at ca. 5 cal. ka BP where no
human activities were recorded in the region. Variation of fire
over time conditioned particularly the relative abundance of mon-
tane grass versus ericoid shrubland with grasses increasing as Eri-
caceae decreased and vice-versa. Modern examples of the
changing dominance of ericoid shrublands versus C4 grasslands
are common, for example at Ambohitantely Special Reserve. The
fire events were possibly associated with climate-vegetation-fire
feedback allowing for the maintenance of the mosaic landscape in
the region. Herbivory activities fluctuated over the period peaking
from ca. 5 and again at 2 cal. ka BP possibly maintaining openness
of the mosaic, with a major surge of dung fungal spores from
1.1 cal. ka BP associated with the introduction of pastoralism and
expansion of grassland. During the last millennium, vegetation in
Lake Dangovavy and surrounds shifted to a more grassland-dom-
inated landscape most likely through the influence of human
activities as reflected by a high concentration of charcoal and
dung fungal spores. Our results highlight the importance of
understanding landscape history prior to establishing conserva-
tion and reforestation/restoration projects and caution against the
current trends of afforestation with alien species. It further con-
tributes to calls worldwide for greater recognition of the value and
history of open and mosaic ecosystems to biodiversity and eco-
system services, including carbon storage and grazing resources.
Acknowledgements
The authors would like to thank the Mention Foresterie et En-
vironnement de l’Ecole Supérieure des Sciences Agronomiques,
Université d’Antananarivo Madagascar for facilitating field work
permits in Madagascar. We acknowledge iThemba Laboratory at
the University of Johannesburg for running free samples for ra-
diocarbon dating and Stable Light Isotope Laboratory at the Uni-
versity of Cape Town who ran the samples for carbon content
analysis. Special thanks go to Dr. Estelle Razanatsoa for her initial
discussions and her assistance with proofreading the manuscript,
as well as to Mrs. Samantha Venter for her thorough review.
Author contribution(s)
A. H. I. Razafimanantsoa: Conceptualization; Data curation;
Formal analysis; Investigation; Methodology; Resources; Soft-
ware; Validation; Visualization; Writing – original draft; Writing
– review & editing.
W. J. Bond: Conceptualization; Supervision; Validation; Writing
– review & editing.
L. Gillson: Conceptualization; Funding acquisition; Project ad-
ministration; Supervision; Validation; Writing – review & editing.
Funding
The author(s) disclosed receipt of the following financial support
for the research, authorship, and/or publication of this article:
This research was supported by the Applied Centre for Climate
and Earth Systems Science (ACCESS NRF) [grant number UID
98018]; The Palaeontological Scientific Trust (PAST), Johannes-
burg, South Africa and the NRF/African Origins Platform [grant
number 117666].
ORCID iDs
Andriantsilavo HI Razafimanantsoa https://orcid.org/0000-
0002-9933-6991
William J Bond https://orcid.org/0000-0002-3441-2084
Lindsey Gillson https://orcid.org/0000-0001-9607-6760
References
Aaby B and Digerfeldt G (1986) Sampling techniques for lakes
and bogs. In: Berglund BE (ed) Handbook of Holocene Pal-
aeoecology and Palaeohydrology. New York: John Wiley &
Sons, pp.181–194.
Baker AG, Bhagwat SA and Willis KJ (2013) Do dung fungal
spores make a good proxy for past distribution of large herbi-
vores? Quaternary Science Reviews 62: 21–31.
Baohanta R, Thioulouse J, Ramanankierana H et al. (2012)
Restoring native forest ecosystems after exotic tree plantation
in Madagascar: Combination of the local ectotrophic species
Leptolena bojeriana and uapaca bojeri mitigates the negative
influence of the exotic species Eucalyptus camaldulensis and
pinus patula. Biological Invasions 14(11): 2407–2421.
Beaujard P (2011) The first migrants to Madagascar and their
introduction of plants: Linguistic and ethnological evidence.
Azania 46(2): 169–189.
Bennett K and Willis J (2001) Pollen. In: Smol J, Birks H and Last
W (eds) Tracking Environmental Change Using Lake Sedi-
ments. Dordrecht: Kluwer Academic Publishers, pp.5–32.
Birks H and Gordon A (1985) Numerical Methods in Quater-
nary Pollen Analysis. Biometric Society. London: Academic,
p.317.
Birks HJB and Line JM (1992) The use of rarefaction analysis for
estimating palynological richness from Quaternary pollen-
analytical data. Holocene 2(1): 1–10.
Blaauw M and Christen JA (2011) Flexible paleoclimate age-
depth models using an autoregressive gamma process. Bayes-
ian Analysis 6(3): 457–474.
Bond WJ and Parr CL (2010) Beyond the forest edge: Ecology,
diversity and conservation of the grassy biomes. Biological
Conservation 143(10): 2395–2404.
Bond WJ, Silander JA and Ratsirarson J (2023) Madagascar’s
grassy biomes are ancient and there is much to learn about

10 The Holocene 00(0)
their ecology and evolution. Journal of Biogeography 50(3):
614–621.
Bond WJ, Silander JA Jr, Ranaivonasy J et al. (2008) The antiq-
uity of Madagascar’s grasslands and the rise of C4 grassy
biomes. Journal of Biogeography 35(10): 1743–1758.
Bond WJ, Stevens N, Midgley GF et al. (2019) The trouble with
trees: Afforestation plans for Africa. Trends in Ecology &
Evolution 34(11): 963–965.
Bonn B and Rounds S (2010) Use of Stable Isotopes of Carbon
and Nitrogen to Identify Sources of Organic Matter to Bed
Sediments of the Tualatin River, Oregon (Oregon. U. S. Geo-
logical Surveyed.). Reston, VA, p.58.
Broothaerts N, Razanamahandry VF, Brosens L et al. (2023) Veg-
etation changes and sediment dynamics in the Lake Alaotra
region, central Madagascar. Holocene 33(4): 459–470.
Burgess N, Hales J, Underwood E et al. (2004) Terrestrial Ecore-
gions of Africa and Madagascar: A Conservation Assessment
(WWFU Statesed). Washington: Island Press.
Burney D (1987c) Pre-settlement vegetation changes at lake
Tritrivakely, Madagascar. Palaeoecology of Africa 18: 357–
381.
Burney D (1996) Climate change and fire ecology as factors in the
quaternary biogeography of Madagascar. In: Lourencxo WR
(ed) Biogéographie de Madagascar. Editions de l’ORSTOM,
Paris, pp.49–58.
Burney DA (1987a) Late Holocene vegetational change in central
Madagascar. Quaternary Research 28(1): 130–143.
Burney DA (1987b) Late Quaternary stratigraphic charcoal records
from Madagascar. Quaternary Research 28(2): 274–280.
Burney DA (1988) Modern pollen spectra from Madagascar. Pal-
aeogeography Palaeoclimatology Palaeoecology 66(1-2):
63–75.
Burney DA (1993) Late holocene environmental changes in
arid southwestern Madagascar. Quaternary Research 40(1):
98–106.
Burney DA, Burney LP, Godfrey LR et al. (2004) A chronology
for late prehistoric Madagascar. Journal of Human Evolution
47(1–2): 25–63.
Burney DA, Robinson GS and Burney LP (2003) Sporormiella
and the late holocene extinctions in Madagascar. Proceedings
of the National Academy of Sciences 100(19): 10800–10805.
Carcaillet C, Bouvier M, Fréchette B et al. (2001) Comparison of
pollen-slide and sieving methods in lacustrine charcoal analyses
for local and regional fire history. Holocene 11(4): 467–476.
Chevalier M, Brewer S and Chase BM (2017) Qualitative assess-
ment of PMIP3 rainfall simulations across the eastern African
monsoon domains during the mid-holocene and the last gla-
cial maximum. Quaternary Science Reviews 156: 107–120.
Cloern JE, Canuel EA and Harris D (2002) Stable carbon and
nitrogen isotope composition of aquatic and terrestrial plants
of the San Francisco Bay estuarine system. Limnology and
Oceanography 47(3): 713–729.
Colombaroli D, Beckmann M, van der Knaap WO et al. (2013)
Changes in biodiversity and vegetation composition in the
central Swiss alps during the transition from pristine forest
to first farming. Diversity and Distributions 19(2): 157–170.
De Vleeschouwer F, Chambers F and Swindles G (2010) Cor-
ing and sub-sampling of peatlands for palaeoenvironmental
research. Mires and Peat 7: 1–10.
Douglass K, Hixon S, Wright HT et al. (2019) A critical review of
radiocarbon dates clarifies the human settlement of Madagas-
car. Quaternary Science Reviews 221: 105878.
Erdmann T (2003) The dilemma of reducing shifting cultivation.
In: Goodman S and Benstead J (eds) The Natural History
of Madagascar. Chicago, IL: University of Chicago Press,
pp.134–139.
Ferguson IK, Faegri K, Iversen J et al. (1976) Textbook of pollen
analysis. Kew Bulletin 31(1): 192.
Fernandes GW, Coelho MS, Machado RB et al. (2016) Afforesta-
tion of savannas: An impending ecological disaster. Natureza
e Conservacao 14(2): 146–151 (Associação Brasileira de
Ciência Ecológica e Conservação).
Gasse F, Cortijo E, Disnar J et al. (1994) A 36 ka environmental
record in the southern tropics : Lake Tritrivakely (Madgas-
car). C.R. Acad. Sci. Paris 318: 1513–1519.
Gasse F and Van Campo E (1998) A 40,000-yr pollen and diatom
record from Lake Tritrivakely, Madagascar, in the Southern
Tropics. Quaternary Research 49(3): 299–311.
Gasse F and Van Campo E (2001) Late quaternary environmental
changes from a pollen and diatom record in the southern trop-
ics (Lake Tritrivakely, Madagascar). Palaeogeography Pal-
aeoclimatology Palaeoecology 167(3–4): 287–308.
Giesecke T, Ammann B and Brande A (2014) Palynological rich-
ness and evenness: Insights from the taxa accumulation curve.
Vegetation History and Archaeobotany 23(3): 217–228.
Giesecke T, Wolters S, Jahns S et al. (2012) Exploring holocene
changes in palynological richness in northern Europe–did
postglacial immigration matter? PLoS One 7(12): e51624.
Gill JL, Williams JW, Jackson ST et al. (2009) Pleistocene mega-
faunal collapse, novel plant communities, and enhanced fire
regimes in North America. Science 326(5956): 1100–1103.
Gosling WD, Miller CS and Livingstone DA (2013) Atlas of the
tropical West African pollen flora. Review of Palaeobotany
and Palynology 199: 1–135.
Hackel J, Razafimanantsoa A, Porcher V et al. (2024) Heath-
ers (Erica, Ericaceae) of Madagascar_ taxonomy, evolution,
ecology and uses. ARPHA Preprints. DOI: 10.3897/arphap-
reprints.e141580.
Hansford J, Wright PC, Rasoamiaramanana A et al. (2018) Early
Holocene human presence in Madagascar evidenced by
exploitation of avian megafauna. Science Advances 4(9): 1–7.
Hansford JP and Turvey ST (2022) Dietary isotopes of Madagas-
car’s extinct megafauna reveal holocene browsing and graz-
ing guilds. Biology Letters 18(4): 20220094.
Hogg AG, Heaton TJ, Hua Q et al. (2020) SHCal20 Southern
Hemisphere calibration, 0–55,000 years cal BP. Radiocarbon
62(4): 759–778.
Huntley B (2023) Biomes and ecoregions of the highlands and
escarpments of Angola and Namibia. Namibian Journal of
Environment 8: 29–41.
Kull CA (2000) Deforestation, erosion, and fire: Degradation
myths in the environmental history of Madagascar. Environ-
ment and History 6(4): 421–450.
Kull CA (2002) The “degraded” Tapia Woodlands of Highland
Madagascar: Rural economy, fire ecology, and forest conser-
vation. Journal of Cultural Geography 19(2): 95–128.
Lacroix E, Carodenuto S, Richter F et al. (2016) Restauration des
paysages forestiers Evaluation des potentialités dans le contexte
des engagements de Bonn 2.0 et de la Déclaration de New York
sur les forêts. Méthodologie et résultats pour Madagascar.
Lowry P, Schatz G and Phillipson P (1997) The classification of
natural and anthropogenic vegetation in Madagascar. Natural
Change and Human Impact in Madagascar 5: 93–123.
Maher LJ (1972) Nomograms for computing 0.95 confidence lim-
its of pollen data. Review of Palaeobotany and Palynology
13(2): 85–93.
McCarroll D and Loader NJ (2004) Stable isotopes in tree rings.
Quaternary Science Reviews 23(7-8): 771–801.
Meltsov V, Poska A, Odgaard BV et al. (2011) Palynological rich-
ness and pollen sample evenness in relation to local floristic
diversity in southern Estonia. Review of Palaeobotany and
Palynology 166(3–4): 344–351.
Moat J and Smith P (2007) Atlas of the Vegetation of Madagascar
(K Royal Botanic Gardensed.).
Munsell Color Company (1954) Munsell color company.
Baltimore.

Razafimanantsoa et al. 11
Nelson DM, Hu FS, Scholes DR et al. (2008) Using SPIRAL
(single pollen isotope ratio AnaLysis) to estimate C3- and
c4-grass abundance in the paleorecord. Earth and Planetary
Science Letters 269(1–2): 11–16.
Nusbaumer L, Gautier L, Chatelain C et al. (2010) Loky-
Manambato: point de rencontre des principales unités phy-
togéographiques de Madagascar. In van der Burgt X, van
Maesen J and Onana J-M (eds) Systématique et Conservation
des Plantes Africaines. Royal Botanical Garden, pp.253–264.
Oksanen J, Simpson G, Blanchet F et al. (2022) Vegan: Com-
munity Ecology Package_. R package version 2.6-4. Avail-
able at: https://cran.r-project.org/package=vegan (accessed
12 December 2022).
Pierron D, Heiske M, Razafindrazaka H et al. (2017) Genomic
landscape of human diversity across Madagascar. Proceed-
ings of the National Academy of Sciences. U.S.A. 114,
E6498–E6506 (32).
Rabarivola L, Razanajatovo H, Razafiniary V et al. (2019) Mada-
gascar Grass Atlas.
Radimilahy CM and Crossland Z (2015) Situating Madagascar:
Indian Ocean dynamics and archaeological histories. Azania
50(4): 495–518.
Railsback LB, Dupont LA, Liang F et al. (2020) Relationships
between climate change, human environmental impact, and
megafaunal extinction inferred from a 4000-year multi-proxy
record from a stalagmite from northwestern Madagascar.
Quaternary Science Reviews 234: 106244.
Rajeriarison and Faramalala (1999) Nomenclature Des Forma-
tions Végétales de Madagascar. Association Nationale pour
la Gestion des Aires Protegees (ANGAP).
Razafimanantsoa AHI and Razanatsoa E (2024) Modern pollen
rain reveals differences across forests, open and mosaic land-
scapes in Madagascar. Plants People Planet 6(3): 729–742.
Razafimanantsoa TM, Raminosoa N, Rakotondrasoa OL et al.
(2013) Silkworm moths inventory in their natural tapia for-
est habitat (Madagascar): Diversity, population dynamics and
host plants. African Entomology 21(1): 137–150.
Razafindratsima A (2019) Caracterisation des sols et analyses
des modes d’usages des terres: cas des districts Antanifotsy,
Antsirabe 1, Betafo, Faratsiho; Region vakinankaratra.
Razanatsoa E, Gillson L, Virah-Sawmy M et al. (2022) Synergy
between climate and human land-use maintained open veg-
etation in southwest Madagascar over the last millennium.
Holocene 32(1–2): 57–69.
Razanatsoa E, Virah-sawmy M, Woodborne S et al. (2021) Adap-
tation of subsistence strategies of the southwestern Malagasy
in the face of climate change. Malagasy Nature 15: 41–55.
R Core Team (2023) R: A Language and Environment for Statisti-
cal Computing. Vienna: R Foundation for statistical Comput-
ing. Available at: https://www.r-project.org/
Roques KG, O’Connor TG and Watkinson AR (2001) Dynam-
ics of shrub encroachment in an African savanna: Relative
influences of fire, herbivory, rainfall and density dependence.
Journal of Applied Ecology 38(2): 268–280.
Samonds KE, Crowley BE, Rasolofomanana TRN et al. (2019)
A new late pleistocene subfossil site (Tsaramody, Sambaina
basin, central Madagascar) with implications for the chronol-
ogy of habitat and megafaunal community change on Mada-
gascar’s Central Highlands. Journal of Quaternary Science
34(6): 379–392.
Schüler L and Hemp A (2016) Atlas of pollen and spores and their
parent taxa of Mt Kilimanjaro and tropical East Africa. Qua-
ternary International 425: 301–386.
Scroxton N, Burns S, McGee D et al. (2020) Possible expression
of the 4.2 kyr event in Madagascar and the south-east African
monsoon. Climate of the Past Discussions 1–25. https://doi.
org/10.5194/cp-2020-137
Simpson GL (2007) Analogue methods in palaeoecology: Using
the analogue package. Journal of Statistical Software 22(2):
1–29.
Smith B and Wilson JB (1996) A consumer’s guide to evenness
indices. Oikos 76(1): 70–82.
Solofondranohatra CL, Vorontsova MS, Hackel J et al. (2018)
Grass functional traits differentiate forest and savanna in the
Madagascar central highlands. Frontiers in Ecology and Evo-
lution 6: 1–14.
Solofondranohatra CL, Vorontsova MS, Hempson GP et al. (2020)
Fire and grazing determined grasslands of central Madagas-
car represent ancient assemblages: Grasslands are shaped by
disturbance. Proceedings of The Royal Society B Biological
Sciences 287(1927): 20200598.
Sourisseau JM, Rasolofo P, Belieres JF et al. (2016) Diagnostic
Territorial de la Région du Vakinankaratra à Madagascar.
Staver AC, Archibald S and Levin SA (2011) The global extent
and determinants of savanna and forest as alternative biome
states. Science 334(6053): 230–232.
Stockmarr J (1971) Tables with spores used in absolute pollen
analysis. Pollen et Spores 13(1): 615–621.
Straka H (1996) Histoire de la végétation de madagascaf oriental
dans les derniers 100 millenaires.pdf. 37–47.
Tölgyesi C, Buisson E, Helm A et al. (2022) Urgent need for updat-
ing the slogan of global climate actions from “tree planting” to
“restore native vegetation. Restoration Ecology 30(3): 2–5.
Troels-Smith J (1955) Karakterisering af lose jordarter [Charac-
terization of unconsolidated sediments]. Danmarks Geolo-
giske Undersogelse. IV(10): 1–73.
Veldman JW, Buisson E, Durigan G et al. (2015) Toward an old-
growth concept for grasslands, savannas, and woodlands.
Frontiers in Ecology and the Environment 13(3): 154–162.
Vieilledent G, Grinand C, Rakotomalala FA et al. (2018) Combin-
ing global tree cover loss data with historical national forest
cover maps to look at six decades of deforestation and forest
fragmentation in Madagascar. Biological Conservation 222:
189–197.
Virah-Sawmy M, Willis KJ and Gillson L (2010) Evidence for
drought and forest declines during the recent megafaunal extinc-
tions in Madagascar. Journal of Biogeography 37(3): 506–519.
Voarintsoa NRG, Railsback LB, Brook GA et al. (2017b) Three
distinct holocene intervals of stalagmite deposition and non-
deposition revealed in NW Madagascar, and their paleocli-
mate implications. Climate of the Past 13(12): 1771–1790.
Voarintsoa NRG, Wang L, Railsback LB et al. (2017a) Multiple
proxy analyses of a U/Th-dated stalagmite to reconstruct
paleoenvironmental changes in northwestern Madagascar
between 370 CE and 1300 CE. Palaeogeography Palaeocli-
matology Palaeoecology 469: 138–155.
Vorontsova MS, Besnard G, Forest F et al. (2016) Madagas-
car’s grasses and grasslands: Anthropogenic or natural? Pro-
ceedings of the Royal Society B 283(1823). DOI: 10.1098/
rspb.2015.2262.
Wang L, Brook GA, Burney DA et al. (2019) The African humid
period, rapid climate change events, the timing of human col-
onization, and megafaunal extinctions in Madagascar during
the holocene: Evidence from a 2m Anjohibe Cave stalagmite.
Quaternary Science Reviews 210: 136–153.
Williamson D, Jelinowska A, Kissel C et al. (1998) Mineral-mag-
netic proxies of erosion/oxidation cycles in tropical maar-lake
sediments (Lake Tritrivakely, Madagascar): Paleoenviron-
mental implications. Earth and Planetary Science Letters
155(3–4): 205–219.