Effects of Holocene climate change, volcanism and mass migration on the ecosystem of a small, dry island (Brava, Cabo Verde)

Palaeoecological data provide an essential long‐term perspective of ecological change and its drivers in oceanic islands. However, analysing the effects of multi‐scalar and potentially co‐occurring disturbances is particularly challenging in dry islands. Here, we aim to identify the ecological consequences of the integrated impacts of a regional drying trend, volcanic eruptions and human mass migrations in a spatially constrained environment—a small, dry oceanic island in Macaronesia.


| INTRODUC TI ON
Oceanic islands worldwide have been identified as frontline sites of major socio-ecological challenges (Baldacchino & Niles, 2011). It is known that they are epicenters of species extinctions and have generally experienced serious loss and fragmentation of habitats since human settlement, providing model systems for understanding human-environment interactions . Dry (climatically arid) island environments can be highly sensitive to disturbances such as an increase in extreme weather events or long periods of drought (Holmgren et al., 2006), and these natural factors can interact with and aggravate human impacts (Lindskog & Delaite, 1996). A comparison of disturbance dynamics before and after human arrival can increase our understanding of how human impacts overprint -or interact with-naturally occurring features of the island environment (Dearing et al., 2006;Paine et al., 1998). It is particularly challenging to develop records of long-term environmental change in dry regions due to poor microfossil preservation and a scarcity of undisturbed sediment archives (Brunelle et al., 2018). Nevertheless, palaeoecological research carried out in dryland soil stratigraphic sequences can provide an essential historical perspective on ecological change in dry islands (e.g. van Leeuwen et al., 2008).

Cabo Verde is a tropical archipelago in the eastern Atlantic
Ocean that has received increasing attention from ecologists and biogeographers in the last decade. Studies in this archipelago are yielding new insights on the diversification and evolutionary mechanisms of Macaronesian fauna and vascular flora (e.g. Romeiras et al., 2019;Vasconcelos et al., 2010) and of the challenges of conserving biodiversity in oceanic islands (e.g. Romeiras et al., 2016). Cabo Verde stands at a crossroads between sub-tropical Macaronesia and the dry Sahel region in terms of biogeography and ecology; species from the arid lowlands show similarities with mainland Africa, while mountainous vegetation show close relationships with the Canary Islands (Beyhl et al., 1990). To further our understanding of Cabo Verde's biodiversity and biogeography, it is vital to assess the composition and natural variability of its ecosystems through time, and the impacts that human settlement brought to different islands. For instance, it is thought that centuries of colonial land mismanagement in Cabo Verde, have led to habitat fragmentation and the soil depletion (Lindskog & Delaite, 1996;Norder et al., 2020). However, the role of volcanic hazards and climate fluctuations in long-term ecological changes and their interplay with anthropic pressures remain to be assessed. Volcanic eruptions can impact human societies by triggering migrations and land use shifts, as well as producing chemical, biological and even climatic alterations (Payne & Egan, 2019). Studying these multi-scalar and potentially co-occurring disturbances and their socio-ecological effects may provide a much-needed historical perspective on landscape degradation and ecosystem variability. This is pivotal to: (i) understanding the effects of regional to local scale environmental change in a spatially constrained ecological/socio-ecological system; (ii) improving biodiversity conservation guidance (Nogué et al., 2017); and (iii) planning responses to future environmental crises.
Cabo Verde comprises 10 volcanic islands and three islets in the African-Sahelian climatic region (Duarte & Romeiras, 2009). Within islands of steep topography (Santo Antão, São Nicolau, Santiago, Fogo and Brava), elevation has a strong influence on temperature (mean annual temperatures vary between 23 and 26°C at sea level and 17-20°C in the highlands) and moisture (Duarte et al., 2008;Rivas-Martinez et al., 2017). Over centuries and millennia, regional climatic fluctuations driven by the oscillations of the West African Monsoon are thought to have shaped natural vegetation composition and distribution (Neto et al., 2020). During the last African Humid Period (AHP: 12,000-5000 yr BP), the African monsoon migrated to the north, leading to higher precipitation in the Saharan region. After this phase, regional conditions became drier, leading to the present extensive Sahara Desert (Pausata et al., 2020).
Near-shore marine records offshore from Senegal reveal sharp decreases in precipitation after 4000 cal yr BP as well as increases in the deposition of Saharan dust after 3500 and after 200 cal yr BP Tierney et al., 2017). After Portuguese settlement (1460 CE), ecosystems in the northern islands of Cabo Verde were transformed by direct and indirect anthropogenic disturbances (Castilla-Beltrán et al., 2019. In recorded history, episodes of socio-ecological distress in Cabo Verde were caused by volcanic eruptions, hurricanes and multi-annual droughts, which are thought to have destabilized colonial socio-ecological systems (Garfield, 2015;Green, 2012;Heckman, 1985;Lindskog & Delaite, 1996;Patterson, 1988).
Over the last two decades, Cabo Verde vascular plants have become increasingly threatened, mostly as a consequence of the increase in exotic species, habitat degradation and human disturbance (Romeiras et al., 2016).
Brava (Portuguese for 'wild') is a small island (64 km 2 ) with a maximum elevation of 976 m asl. It is regarded as one of the most fertile in Cabo Verde, earning it the title of the 'Garden Island'. It is one of the youngest of the islands, having formed c. 3.0 Ma, and it shares its Island c. 1680 CE potentially triggered land use change and intensification, causing a reduction of native vegetation in Brava.

K E Y W O R D S
African Humid Period, Anthropocene, Cape Verde, ecological change, environmental disturbances, Macaronesia, natural hazards, palaeoecology, tropical islands volcanic base with Fogo Island (Madeira et al., 2010). Brava harbours some of the most diverse vegetation of Cabo Verde and the highest total plant species richness, with 239 species, and second-highest endemic floral species richness, with 25 taxa (Duarte et al., 2008), but due to its small size it has received little attention regarding conservation initiatives. Brava's north-facing highlands (700-960 m) are well-placed for the capture of cloud moisture and favoured by humid local climates (Correia, 1996). Currently, it is estimated that 13% of Brava's land area is wooded, much of it comprising introduced taxa such as Prosopis juliflora, Grevillea robusta, and Cupressus lusitanica. Native vegetation is under pressure due to the grazing of cows, donkeys and goats (GEF/UNEP, 2015). Brava's neighbouring islands, Santiago and Fogo, were the first to be significantly settled by Europeans (c. 1460 and 1470 CE, respectively), but it is estimated that Brava remained only marginally inhabited until c. 1680 CE, when 'many families' from Fogo fled to take refuge in Brava after an earthquake and a major volcanic eruption (Correia, 2000;Mitchell-Thomè, 1981;Ribeiro, 1960) (Table S1). Here, we present multiple palaeoecological analyses to show how ecosystems and soils in Brava changed over the last 9700 yr in response to global-to-local and potentially co-occurring environmental changes, including regional drying, within-archipelago volcanism and inter-island human migration.

| MATERIAL S AND ME THODS
Our study of ecological change in the highlands of Brava consists of multiple analyses of palaeoenvironmental information contained in a 220-cm soil profile excavated in a volcanic caldera (810 m asl).
We recorded the changes in the plant, fern and fungal communities using fossil pollen grains, and fern and fungal spores (non-pollen palynomorphs; NPPs). For changes in organic matter content we carried out Loss on Ignition analysis (LOI), for identifying the occurrence of fire we quantified charcoal particles, for shifts in sedimentology we used X-ray fluorescence (XRF) and grain size distribution (GSD) analysis, and to assess the occurrence of significant episodes of deposition of volcanic ash we studied the abundance of microscopic tephra shards.

| Fieldwork and sample collection
In May 2019, we visited Brava Island and carried out excavations and soil-profile sampling in two highland calderas: Cova Galinha (CG, 810 m asl) and Fondo Mato (FM, 750 m asl). We selected the Cova Galinha caldera (14°51′26.2″N, 24°42′09.5″W) for this study based on its superior micro-fossil preservation, that could be linked to finer sediment grain sizes and the absence of gravel-dominated soil horizons. Cova Galinha caldera is currently used for low-intensity agriculture, mostly maize (Zea mays) cultivation and cattle grazing.
We opened a 2.0 × 1.0-m trench in the centre of the caldera and collected a set of samples of about 30 g of sediment every 5 cm (sample set A, taken across horizontally at c. 2 cm thickness), and a set of contiguous 2-cm 3 block samples (sample set B) ( Figure 1). We placed the samples in sealed plastic bags. The samples were transported to the University of Southampton and stored in a cold room at +4°C within the School of Geography and Environmental Science.

| Dating methods
To develop a chronological model of the Cova Galinha site we obtained 10 radiocarbon (RC) dates: nine from bulk sediment samples and one AMS RC date from a macrofossil sample via the Belfast and SUERC Radiocarbon laboratories (Table 1). We also carried out Pb-210 and Cs-137 analysis on the top meter of the core using sample set A, to achieve precise chronological control of sediment deposits in the last 120-150 years. These analyses were undertaken in the GAU-Radioanalytical Laboratories at the National Oceanography Centre (Southampton), using gamma spectrometric-analysis in HPGe well-type detectors (Cundy et al., 2006). We used Bacon in R (Blaauw & Christen, 2013) to calculate the age-depth model based on all dates and defined level 20 cm as the level representing post-CE 1950, based on a sharp increase in Cs-137 and two 'post-bomb' RC dates ( Figure S1).

| Granulometry, elemental composition and Loss on Ignition analyses
For Grain size distribution (GSD), we used sample set A and sample set B for the section 80-40 cm. We used a Mastersizer Hydro (Malvern), programmed to measure soil properties through measurement of non-spherical soil grains. We carried out a minimum of five 20-second measurements per sample. Results were averaged once the standard deviation score of the three size fractions Dx10, Dx50 and Dx90 was equal to or below 0.5, 0.3 and 0.5, respectively.
For elemental composition analysis, we used set A (5 cm intervals). We used a hand-held X-ray Fluorescence (XRF) device (de Lima et al., 2019), model Niton XL3T GOLDD, using a test-stand. In measurements of 160-seconds per sample, we obtained proportion values (%) of elements above a detection limit of 0.001%.
For Loss on Ignition analysis, (LOI) we used sample set B (contiguous 2 cm samples). We used a high-precision scale, and followed the Heiri et al. (2001) protocol, ashing dry samples for 4 h at 550°C. LOI was calculated using the weight value of the dry sample and the ashed sample and used as a proxy for sample organic content.

| Pollen and non-pollen palynomorph analyses
To determine vegetation, fern and fungal community changes, and assess the presence of herbivores, we processed 44 2-cm 3 samples from set A for pollen and NPP analyses following standard procedure that included sieving with 10 µm mesh, processing with hydroflouric acid (HF) to eliminate excess silica, acetolysis (Erdtman, 2013) and adding one tablet of Lycopodium exotic spores for calculating concentrations (batch nr 140119321, avg 19,855 spores per tablet).
We used a high-power microscope to identify palynomorphs at ×400 and ×1000 magnification and consulted African pollen atlases (e.g. Gosling et al., 2013) and our Cabo Verde and Canary Islands reference collections stored at the University of Southampton and the University of La Laguna to identify pollen grains. NPP literature was used to identify fungal and fern spores (van Geel et al., 2003;Gelorini et al., 2011). All pollen grains were included in the calculation of the total pollen sum and categorized as local (endemic and native taxa to Cabo Verde), introduced, derived from long-distance F I G U R E 1 Maps of Cabo Verde with information of documented migrations between southern islands. Topographical map of Brava modified after Danielson and Gesch (2011). Pictures of Cova Galinha caldera and the studied soil profile (modified to show sampling strategy) taken by the authors. [Colour figure can be viewed at wileyonlinelibrary.com] TA B L E 1 Radiocarbon dates of the Cova Galinha site, Brava Island, Cabo Verde, and their calibration. 14 C enrichment value is reported in ages between 1950 CE and the present. transport, or unidentifiable (mostly too damaged to identify). The latter was included in the sum to account for unidentified components of the vegetation. We counted a maximum of four microscope slides per level, reaching a minimum of 250 pollen grains in samples with abundant pollen (55-0 cm), a minimum of 100 grains in much of the rest of the record, but we included counts over 50 pollen grains in the section with scarcer pollen abundance (115-195 cm). The pollen and NPP (ferns and fungal spores) data are presented as percentages over the pollen sum.

| Macro-and micro-charcoal
For macro-charcoal quantification we used 44 2-cm 2 samples from sample set A (taken every 5 cm) and sieved the material through a 180-µm sieve. We used a low-power microscope to count charcoal fragments directly after sieving. We performed micro-charcoal quantification in pollen-slides by counting angular opaque particles between 10 and 180 µm alongside exotic Lycopodium, until reaching a sum over 200 items and then calculating micro-charcoal concentrations (Finsinger & Tinner, 2005).

| Silica structures: phytoliths and tephra
For phytolith and tephra analyses, we used a dry ashing sample preparation (Parr et al., 2001) on sample set B. We sieved the samples through 10 and 180-µm sieves. We tested density separation using sodium polytungstate at density 2.5 sg in 10 samples, which resulted in the flotation of more than 50% of the material. As a costeffective alternative, we followed recommendations of Lentfer and Boyd (1998) for sandy sediments, directly mounting microscope slides using Canada balsam, a method we previously used for diatom and phytolith analysis in Cabo Verde (Castilla-Beltrán et al., 2019).
We used phytolith literature (e.g. Piperno, 2006) for phytolith identification. We spiked the samples with a known quantity of exotic Characterization of tephra geochemistry and comparison of shard morphology with tephra shards from Fogo Island will be carried out in future studies.

| Zonation and ordination analyses
We used Tilia software to establish a zonation by performing stratigraphically constrained CONISS analysis (Grimm, 1993) resulting in six zones (Table S2). We carried out Canonical Correspondence Analysis (CCA) using the R Vegan package (Oksanen et al., 2013) to assess the influence of six environmental drivers on vegetation change (pollen percentage values). 3 | RE SULTS

| Ordination analyses
The

| Taphonomical processes and the interpretation of soil biostratigraphy
Palaeoecological research in dry islands is challenging due to the lack of permanent water mires that preserve microfossils in water-  Figure S5). In these sections, NPPs showed better preservation potential, and they are useful to assess continuity or change in local environmental conditions. The process of incorporation of micro-fossils into the soil matrix can be complex: for instance, Zea mays and agricultural weed phytoliths identified as Commelina benghalensis (Eichhorn et al., 2010), taxa introduced by humans after 435 BP (Moran, 1982), were registered in small percentages in sections dated to before first human arrival.
In the context of the Cova Galinha site, the stratigraphic position of these pollen grains underneath the anthropic agricultural soil horizons (c. 10 cm) supports the interpretation that their presence is due to limited microfossil intrusions through leaching, root percolation or herbivore trampling.
Sediment mixing and redeposition can also affect soil stratigraphic sequences: for instance, our age-depth model shows two  Figure   S6). Instead, it fluctuates with Pb-214 activity, indicating that the vertical distribution of Pb-210 is controlled by input on reworked sediment particles and by variations in sediment composition, rather than by direct atmospheric fallout (Cundy & Stewart, 2004). Even with the constraints of working in a dry environment, the clear-cut biostratigraphic patterns found in Cova Galinha caldera sediments provide unique insights into environmental change in Brava, and our age-depth model successfully provides a chronological framework to discuss natural and human-driven disturbance regimes in Brava.

| Erosion, vegetation and regional climatic change
The

| Vegetation responses to volcanism
Assessing the relationship between changes in vegetation and the consequences of volcanic activity, such as tephra deposition and tephra domination of soils, is challenging. Volcanic eruptions can have indirect influences in vegetation through a change in localand micro-climates (increased humidity and/or precipitation) due to augmented aerosol cover during prolonged episodes of volcanism (Payne & Egan, 2019). According to popular knowledge in Cabo Verde, after volcanic eruptions, good years of rain are to be expected ('Anos bons' in Creole) (Correia, 1996(Correia, , 2000. Close examination of meteorological records from the 20th century does not show a robust link between eruptions and change in local climate (Correia, 2000). This cannot be ruled out in previous centuries, however, as the eruptions of Pico do Fogo in the 20th century were of relatively low intensity, duration and frequency. Vegetation change could also be directly linked to the effects of the deposition of tephra layers, which can only constitute a minor ecological disturbance, and can lead to better water retention and enhanced plant growth (Crisafulli et al., 2015). After the eruption of Mount St. Helens (Washington State, USA), woody species benefitted in some areas from limited tephra falls (Zobel & Antos, 2017). Soils that incorporate volcanic ash are rejuvenated and stocked with essential nutrients, such as phosphorous, that are readily accessible to plants (Schlesinger et al., 1998); this also enhances their potential to sequester carbon (Fiantis et al., 2019). But vegetation responses to ash falls are highly variable: on the Eastern Andean flank forest responses to substantial ash fall varied from almost negligible to the expansion of species with pioneering qualities (Loughlin et al., 2018), while in New Zealand, complete revegetation of forests occurred in 200 years after volcanic disturbance (Wilmshurst & McGlone, 1996).
Abundant deposition of tephra shards took place in Brava, especially between 1800 and 650 cal yr BP. Brava Island has no recorded eruptions in the historical period, but the island's volcanic cones may have erupted in the Late Holocene (Worsley, 2015), meaning that the observed tephra could be of local origin or even re-deposited by local erosion. However, it is also likely that most tephra, if not all, originated from Fogo (highly active in the historical period 450 BPpresent) and was transported by northeast trade winds (Table S1).
On Brava, vegetation changed after these tephra deposition events.

| Colonial settlement, mass migration and human ecological footprint
From 500 BP to the present day, Brava's highlands underwent significant anthropogenic ecological disturbances due to the arrival of European settlers (the first known permanent settlers on the island).  (Ribeiro, 1960). Volcanism was once again a driver of change in southern Cabo Verde, but this time spurring inter-island human migration and resultant land use change in Brava. Historical accounts describe how these catastrophic episodes in Fogo led to the abandonment of sterile lands and damaged settlements, triggering migration between islands (Correia, 2000;Ribeiro, 1960

| CON CLUS IONS
The study of sediments from a volcanic caldera in Brava, the smallest inhabited island of Cabo Verde, provides the first characterization of highland vegetation throughout the Holocene on this island. It also offers a long-term view of local ecological responses to environmental changes and disturbances originating at regional scales and those with more local, within-archipelago and within-island origins. Pollen evidence suggests that taxa such as Brava. Studying long-term environmental change in small islands reveals how major climatic transitions and disturbance episodes impact spatially constrained ecological systems. This is key for the current and future development of guidelines on the protection of island biodiversity and ecosystem services, and for managing future sustainability in islands and archipelagos.

ACK N OWLED G EM ENTS
We thank the Royal Geographic Society for funding our expedition, and NERC for radiocarbon support, which funded a set of six

DATA AVA I L A B I L I T Y S TAT E M E N T
The dataset supporting our findings has been uploaded to the Neotoma Palaeoecology Database (online and open-access: https:// apps.neoto madb.org/explo rer/?datas etid=48891) and can also be requested via email to the corresponding author of this paper.