Maua1a (3000–800 cal yr BP)

During the early Maua record, ericaceous and montane forest pollen types decreased slightly, while Poaceae and Cyperaceae increased (Figs. 4 and 7). Charcoal also increased slightly throughout Maua1a and positively correlated with variations in Cyperaceae pollen. This correlation suggests Cyperaceae made much of the fuel source for the locally created charcoal and that there were increased abundances in Cyperaceae in the wetland during Maua1a. Peak pollen abundances of Podocarpus in the Maua record occur from 2300 to 2000 cal yr BP, concomitant with peak abundances at Lake Rutundu, Mount Kenya (3140 m asl; Coetzee, Reference Coetzee1967; Hamilton, Reference Hamilton1982a), and with increasing relative abundances from 2600 to 1600 cal yr BP at Oblong Tarn, Mount Kenya (Courtney Mustaphi et al., Reference Courtney Mustaphi, Gajewski, Marchant and Rosqvist2017). A previously studied wetland sediment pollen record from Kilimanjaro (2650 m asl; Coetzee, Reference Coetzee1967) shows an increase in Podocarpus from ~4000 to 1800 cal yr BP, although the sampling resolution is rather low, and the relative change is due to large variations in Ilex in that record. It is challenging to examine variability in anemophilous pollen taxa in mountainous areas (Birks and Birks, Reference Birks, Birks, Birks and Birks1980; Markgraf, Reference Markgraf1980; Solomon and Silkworth, Reference Solomon and Silkworth1986), but these coincidental variations in Podocarpus could be related to the expansion of montane Podocarpus-dominated forests driven by hydroclimatic changes or changing fire disturbance regimes.

Figure 7.

Cumulative relative abundances for pollen groups, rate of change (turnover, RoC; Urrego et al., Reference Urrego, Bush, Silman, Correa-Metrio, Ledru, Mayle, Valencia, Vimieux, Sylvestre and Khodri2009; Correa-Metrio et al., Reference Correa-Metrio, Urrego, Cabrera and Bush2011), and principal components analysis axes 1 and 2 are shown with the pollen-based CONISS zonation (horizontal solid and dashed lines, see Fig. 4). The ratio of the relative pollen abundances Poaceae:Cyperaceae (log transformed for visualization) and the total relative abundance Poaceae phytolith morphotypes from Maua are shown (Fig. 5). Additional palaeoenvironmental datasets are plotted for compairson: including an oxygen isotope record from the Kilimanjaro south ice field (black curve); and three proxy records from Lake Challa: a smoothed mixed composite record (red line) (Thompson et al., Reference Thompson, Mosley-Thompson, Davis, Henderson, Brecher, Zagorodnov and Mashiotta2002); macroscopic charcoal (>180 μm; Nelson et al., Reference Nelson, Verschuren, Urban and Hu2012); a branched versus isoprenoid tetraether (BIT) index of monsoon-promoted sediment transport into the lake (low BIT = drier conditions; Buckles et al., Reference Buckles, Verschuren, Weijers, Cocquyt, Blaauw and Sinninghe Damsté2016); and a stable carbon isotope record derived from sediment total organic carbon (Moernaut et al., Reference Moernaut, Verschuren, Charlet, Kristen, Fagot and De Batist2010; Blaauw et al., Reference Blaauw, van Geel, Kristen, Plessen, Lyaruu, Engstrom and van der Plicht2011; Sinninghe Damsté et al., Reference Sinninghe Damsté, Verschuren, Ossebaar, Blokker, van Houten, van der Meer and Plessen2011). A mid-elevation temperature reconstruction derived from the sediments of Lake Rutundu, Mount Kenya (Loomis et al., Reference Loomis, Russell, Verschuren, Morrill, De Cort, Damsté and Olago2017) and a charcoal record from Kwasebuge, North Pare Mountains (Finch et al., Reference Finch, Marchant and Courtney Mustaphi2017). Land-use modeled outputs for what is now the United Republic of Tanzania: human population density estimates (Kaplan and Krumhardt, Reference Kaplan and Krumhardt2011; Kaplan et al., Reference Kaplan, Krumhardt, Ellis, Ruddiman, Lemmen and Klein Goldewijk2011) and total cropland area (Klein Goldwijck et al., Reference Klein Goldewijk, Beusen and Janssen2010) are also shown to compare general patterns in land-use changes. Livelihood strategies on and around Kilimanjaro have also been qualitatively summarized alongside the published radiocarbon dates from archaeological sites (Odner, Reference Odner1971; Mturi, Reference Mturi1986; Tagseth, Reference Tagseth2008; Stump and Tagseth, Reference Stump, Tagseth and Clack2009; Chuhila, Reference Chuhila2016; de Bont, Reference de Bont2018; see Supplementary Table 1). HG = hunter-gatherer; P = pastoralism (and agropastoral); Ag = agricultural; Irr = furrow irrigation for agriculture on the lower slopes of Kilimanjaro; the blue dotted line shows potential regional onset of terraced agriculture (Sutton, Reference Sutton, Widgren and Sutton2004; Stump, Reference Stump2006); M = mercantile (Euro. = increased European connections and colonialism). Calibrated radiocarbon age ranges using OxCal 4.2 (n = 14 dates, 95.4% confidence intervals shown, SHCal13; Bronk Ramsey, Reference Bronk Ramsey2009; Hogg et al., Reference Hogg, Hua, Blackwell, Buck, Guilderson, Heaton and Niu2013) and converted to cal yr BP are associated with iron production and iron working in the Mwanga region (<1400 m asl), North Pare (Iles et al., Reference Iles, Stump, Heckmann, Lang and Lane2018; Iles, Reference Iles2020). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The wetland was probably ice free by the Holocene transition (Shanahan and Zreda, Reference Shanahan and Zreda2000; Mark and Osmaston, Reference Mark and Osmaston2008), and glacial ice on Kilimanjaro was greatly reduced and varied slightly during the late Holocene before experiencing significant reductions over the past century (Hastenrath and Greischar, Reference Hastenrath and Greischar1997; Thompson et al., Reference Thompson, Mosley-Thompson, Davis, Henderson, Brecher, Zagorodnov and Mashiotta2002). The wetland was dominated by Cyperaceae-Poaceae vegetation throughout the Maua1a zone. Increased abundances of charcoal and Cyperaceae appear to be correlated with wetter intervals (less negative δ¹⁸O values) associated with the composite of isotope records of the Kilimanjaro ice fields (Fig. 7; Thompson et al., Reference Thompson, Mosley-Thompson, Davis, Henderson, Brecher, Zagorodnov and Mashiotta2002); this potentially supports the argument that local wetland biomass and burning was influenced by high-elevation hydroclimate during this interval. The charcoal was most likely locally created in the wetland and its watershed, although it was not assessed if charcoal is transported uphill by convection from lower elevation fires (Pisaric, Reference Pisaric2002; Tinner et al., Reference Tinner, Hofstetter, Zeugin, Conedera, Wohlgemuth, Zimmermann and Zweifel2006; Adolf et al., Reference Adolf, Wunderle, Colombaroli, Weber, Gobet, Heiri and van Leeuwen2018), highlighting a topic for further research on tropical mountains (Vachula and Richter, Reference Vachula and Richter2018).

The pollen record from Small Momela Lake at 1440 m asl on Mount Meru shows centennial-scale variability in the relative abundances of Poaceae and Cyperaceae from 3000 to 1200 cal yr BP (Bolick, Reference Bolick1991). The increased charcoal accumulation at Maua from 110 to 800 cal yr BP coincided with increased charcoal accumulation from 1300 to 900 cal yr BP at Lake Challa (Nelson et al., Reference Nelson, Verschuren, Urban and Hu2012), large increases in Poaceae from 1400 to 800 cal yr BP at the Namelok wetland (Rucina et al., Reference Rucina, Muiruri, Downton and Marchant2010), and increased charcoal at the Enkongu wetland, Amboseli (Githumbi et al., Reference Githumbi, Kariuki, Shoemaker, Courtney Mustaphi, Chuhila, Richer and Lane2018b). All these sites are located in the lowlands surrounding Kilimanjaro (Fig. 7), and it is difficult to disentangle regional climate and anthropogenic drivers of environmental change to explain these variations. The varying relative importance of fire regime drivers at each site may explain some of the spatiotemporal variability (Gillson, Reference Gillson2004). Ericaceous pollen is relatively low during this initial burning interval (1200–800 cal yr BP), while Hagenia, Podocarpus, Cyperaceae, and cf. Ehretia abundances are relatively high (Fig. 4).

Phytolith data between 3000 and 2500 cal yr BP represent local input consisting largely of Cyperaceae morphotypes, indicating the wetland was relatively moist during this interval. Between 2500 and 800 cal yr BP, the phytolith assemblage reflects both local and watershed-area inputs derived from woody dicots, particularly the spheroid morphotypes (Fig. 5). There is a notable presence of the Pooideae morphotypes, mainly derived from the tussock grass types common on hummocky alpine periglacial ecosystems. This suggests that the water level in the wetland was generally very low, and the absence of aquatic vegetation pollen types supports the interpretation that the wetland was covered by terrestrial vegetation.

People practicing hunting and gathering were almost certainly living on Kilimanjaro throughout the Holocene and may have even preferred higher-elevation areas as the savanna-montane forest ecotonal zone retreated at the termination of the African Humid Period (Ambrose and Sikes, Reference Ambrose and Sikes1991). The oldest archaeological remains suggestive of livestock on Kilimanjaro were deposited ~4000 cal yr BP on the western side of the mountain, just below 2000 m asl, in the lower montane forest zone (Mturi, Reference Mturi1986). Material remains affiliated with the Savanna Pastoral Neolithic archaeological group, and thus specialized pastoralism, are also present here and date to between 3000 and 1500 cal yr BP. Faunal remains from these sites are predominantly of cattle and sheep or goat, though wild animals are also present, suggesting a mixed livelihood of pastoral activity and subsistence hunting. Indeed, regional archaeological records indicate that hunting and gathering modes of production would have overlapped and persisted with pastoralism and agriculture over the last three millennia, rather than having been displaced through time (Lane, Reference Lane2004). The possibility of intensive “wild” resource management by hunter-gatherer and livestock-herding populations on Kilimanjaro during and prior to the last three millennia also merits further consideration, as there is clear evidence for the careful management of morphologically undomesticated plant species in other regions of sub-Saharan Africa (Lane and Shoemaker, Reference Lane, Shoemaker and Spear2017). Yet despite Kilimanjaro being one of eastern Africa's “islands of intensive agriculture” (Widgren, Reference Widgren, Widgren and Sutton2004, p.1), the early history of human-vegetation relationships remains almost entirely unknown. Farming in the wider region is typically thought to have become established after 2000 cal yr BP, however, accompanied by Bantu language speakers, agropastoralism, and iron working (see Crowther et al., Reference Crowther, Prendergast, Fuller and Boivin2018 for a review). This date is probably best considered a terminus ante quem for the origins of farming in the area because finger millet (Eleusine coracana) was very likely cultivated in eastern Africa before the so-called Bantu expansion (Fuller, Reference Fuller, Neumann and Butler2003; Fuller and Boivin, Reference Fuller and Boivin2009). Some have suggested that Asian domesticates, such as taro (Colocasia esculenta), chicken (Gallus gallus domesticus), and banana (Musa spp.), were also being farmed in eastern Africa prior to 2000 cal yr BP (Chami, Reference Chami, Chami, Pwiti and Radimilahy2001; Mbida et al., Reference Mbida, Van Neer, Doutrelepont and Vrydaghs2000, Reference Mbida, Doutrelepont, Vrydaghs, Swennen, Swennen, Beeckman and De Langhe2001, Reference Mbida, Doutrelepont, Vrydaghs and Beeckman2004, Reference Mbida, De Langhe, Vrydaghs, Doutrelepont, Swennen, Van Neer, de Maret, Zeder and Bradley2006; Lejju et al., Reference Lejju, Taylor and Robertshaw2005, Reference Lejju, Robertshaw and Taylor2006; De Langhe, Reference De Langhe, Denham, Iriate and Vrydaghs2007), and Kilimanjaro may have been an early locus for banana cultivation, though the evidence is inconclusive (Sinclair, Reference Sinclair, Starkey, Starkey and Wilkinson2007; Neumann and Hildebrand, Reference Neumann and Hildebrand2009; Shipton et al., Reference Shipton, Crowther, Prendergast, Kourampas, Horton, Douka and Schwenninger2016). Coastal archaeological records could suggest that the widespread adoption of Asian domesticates by eastern African farmers did not occur until the last millennium (Walshaw, Reference Walshaw2015; Crowther et al., Reference Crowther, Prendergast, Fuller and Boivin2018). Iron smelting may have also occurred on Kilimanjaro over the last two thousand years, though archaeologists have yet to document iron-working remains, suggesting it was extremely limited in scope compared to operations underway in the Pare Mountains from the second half of the first millennium (Iles et al., Reference Iles, Stump, Heckmann, Lang and Lane2018; Iles, Reference Iles2020). It is possible that people used high-elevation areas of Kilimanjaro during the Maua1a zone for herding, hunting and gathering, transportation routes, or ceremonial activities, though direct evidence is lacking. Certainly, the surrounding lower-elevation areas were being modified by human activities, including pastoralism, hunting and gathering, and agriculture by 2000 cal yr BP (probably increasingly so).

Maua1b subzone (800–600 cal yr BP)

The pollen assemblage subzone is dynamic and represents an assemblage transition between Maua1a and Maua2, with distinctively low charcoal, high Cyperaceae abundances, increasing turnover rates in the pollen assemblage (Fig. 7), highly variable phytolith assemblages, and negative values in the pollen PCA axis 2 scores (Fig. 7); the samples from this zone also cluster together on the PCA biplot (Supplementary Fig. 1). Low CHARs could be due to several factors: The local hydroclimatic regime could have become too wet and reduced flammability; decreased ignitions (anthropogenic, lightning, spontaneous) may have limited fires; fuel connectivity could have also been interrupted as vegetation became more sparse or increasingly dissected by, for example, animal trails; and lastly, productivity changes could have affected total aboveground biomass (fuel availability). Aboveground biomass varies owing to vegetation productivity and herbivory regimes; the latter has been linked to consumption of the potential fuels for carrying fires and fragmenting the grass canopy (Archibald et al., Reference Archibald, Bond, Stock and Fairbanks2005; Bond and Keeley, Reference Bond and Keeley2005; Hempson et al., Reference Hempson, Archibald, Bond, Ellis, Grant, Kruger and Kruger2015; Donaldson et al., Reference Donaldson, Archibald, Govender, Pollard, Luhdo and Parr2018). There are very few studies of high-elevation herbivory on Afroalpine vegetation by domestic or wild animals (Mulkey et al., Reference Mulkey, Smith and Young1984; for subalpine effects, see Wood, Reference Wood1965b; Kikoti et al., Reference Kikoti, Mligo and Kilemo2015) or information that explains reduced fire ignitions; thus, it remains difficult to explore the equifinality of the observed low CHARs. During this subzone, the wetland transitioned from a Cyperaceae-Poaceae–dominated habitat to a more Cyperaceae-dominated one (see high abundances in Fig. 4) with lower fuel flammability and connectivity and lower grazing desirability. The Maua1b subzone also has concomitant changes in both the montane forest and the ericaceous taxa assemblages (Fig. 4; Supplementary Fig. 1). Several montane forest taxa, such as Nuxia, Hagenia, and Schefflera, increase and then decrease during the subzone, and a conspicuous decrease in Podocarpus abundances is evident (Fig. 4). During Maua1b, Kilimanjaro ice field δ¹⁸O values are markedly more negative, as is the Lake Challa branched versus isoprenoid tetraether (BIT) index and reconstructed mean annual air temperatures at Lake Rutundu (Fig. 7; Sinninghe Damsté et al., Reference Sinninghe Damsté, Verschuren, Ossebaar, Blokker, van Houten, van der Meer and Plessen2011; Loomis et al., Reference Loomis, Russell, Verschuren, Morrill, De Cort, Damsté and Olago2017), suggesting that—at least at middle and lower elevations—the region was cooler and drier.

Maua2 (600 to –59 cal yr BP)

This zone is significantly different from Maua1 and has more positive scores on the PCA axis 1, suggesting further rapid changes in the mountain flora in and around the wetland following the Maua1b transition (Fig. 7). Pollen assemblage turnover remains highest during this most recent interval, and the uppermost samples are characterized by very different pollen assemblages (Fig. 7). The phytolith assemblage shows a trend of decreasing forest indicators and woody dicots. Cyperaceae morphotypes also decrease, though Poaceae morphotypes increase; Chloridoideae C₄ short grasses in particular become more prominent. Overall, the phytolith evidence suggests either drier conditions or clearing of woody plants, possibly both (Fig. 5). From 500 cal yr BP to the present, a general decrease in or total absence of forest indicators, dicots, and Cyperaceae morphotypes is evident, while Chloridoideae C₄ grasses are notably present during Maua2. The charcoal record suggests a very different fire regime of increased burning since the onset of Maua2, indicating more fire-adapted taxa (Poaceae and Ericaceae) with consistently higher Poaceae phytolith morphotypes (GSCPs) and higher local biomass (Figs. 4 and 5). Increased charcoal during Maua2 could be related to increased vegetation biomass available to burn (Courtney Mustaphi and Pisaric, Reference Courtney Mustaphi and Pisaric2014) in response to increased moisture availability in the wetland. The site may have thus transitioned from a Cyperaceae to a Poaceae tussock-dominated wetland, as it remains at present. The data do not make it possible to examine if the wetland itself was ever covered by ericaceous vegetation, as is the case for some high-elevation mires (Salt, Reference Salt1954) or if it remained consistently dominated by Poaceae throughout Maua2. While CHARs decreased during Maua2 to the present, counts are nonetheless predominantly much higher in comparison to previous intervals. On Kilimanjaro at present, herbaceous biomass increases from the mid-montane forest to alpine elevations; this is linked to changing moisture availability (Ensslin et al., Reference Ensslin, Rutten, Pommer, Zimmermann, Hemp and Fischer2015), potentially suggesting a connected between surface moisture, productivity, and fire.

The epiphytic filamentous green algae Spirogyra was observed in very low abundance (three individual remains with a count of 650 pollen) at 393 cal yr BP, which suggests the potential for puddling in the wetland among the vegetation (Ross, Reference Ross1953; Denny et al., Reference Denny, Bowker and Bailey1978). Spirogyra dominated the phytoplankton community in the Lower Kitandara Lake (4020 m asl) and in the Ruwenzori Mountains (Richardson, Reference Richardson1964); it was also found in late Pleistocene peaty sediments at a swamp in the Udzungwa Mountains (Mumbi et al., Reference Mumbi, Marchant, Hooghiemstra and Wooller2008). Spirogyra is frequently present in aquatic ecosystems with high thermal and radiation variability (Gelorini et al., Reference Gelorini, Verbeken, van Geel, Cocquyt and Verschuren2011), diurnal conditions that prevail in the Maua wetland. If occasional puddling in the wetland occurred during the past few hundred years, it suggests that multiannual alternation between wet and dry conditions would be an important characteristic of the climate-vegetation-fire relationship during Maua2, a situation markedly different than during Maua1.

The pollen and charcoal records indicate that local vegetation during this zone becomes predominantly ericaceous with high abundances of Erica, Artemisia, Anthospermum, and Stoebe. Taxa in both the Ericaceae and Poaceae families likely benefited from increased fire, as has been described in other Holocene records from high-elevation Erica heathland sites in eastern Africa (Gil-Romera et al., Reference Gil-Romera, Adolf, Benito, Bittner, Johansson, Grady and Lamb2019). The associations between charcoal, ericaceous taxa, and Poaceae at Maua suggest a more fire-adapted vegetation community at the wetland during a time of regionally drier high-elevation climate (Thompson et al., Reference Thompson, Mosley-Thompson, Davis, Henderson, Brecher, Zagorodnov and Mashiotta2002), decreasing reconstructed daytime relative humidity (Hepp et al., Reference Hepp, Zech, Rozanski, Tuthorn, Glaser, Greule and Keppler2017), and higher ericaceous pollen abundances at the Maundi crater (Schüler et al., Reference Schüler, Hemp, Zech and Behling2012). This suggests a more nuanced study of the complex local ecohydrology is needed to explain these patterns to further understand hydroclimate, vegetation and fire, and grazing interactions.

The lacustrine sediment record from Small Momela Lake does not cover this time interval in detail, but the low elevation sites of Lake Challa (Fig. 7; Sinninghe Damsté et al., Reference Sinninghe Damsté, Verschuren, Ossebaar, Blokker, van Houten, van der Meer and Plessen2011) and Lake Duluti show evidence of change during Maua2. At Lake Duluti, significant changes to diatom assemblages occur in the past 700 years, both in the relative abundances of planktonic and periphytic taxa and with respect to an overall decrease in periphytic diatoms toward the present, suggesting more open water (Öberg et al., Reference Öberg, Norström, Ryner, Holmgren, Westerberg, Risberg and Eddudóttir2013). The stable carbon-isotopic composition of fossil plant leaf waxes from sediments in Lake Challa increase and remain at their highest values (averaging −30‰) from the past 3000 years (average −31‰; Fig. 7; Sinninghe Damsté et al., Reference Sinninghe Damsté, Verschuren, Ossebaar, Blokker, van Houten, van der Meer and Plessen2011), suggesting relatively wet conditions at least at low elevations. Lake Challa is groundwater fed, and it has been noted that there is a negative relationship during the late Holocene between δ¹⁸O values in Lake Challa diatoms (Barker et al., Reference Barker, Hurrell, Leng, Wolff, Cocquyt, Sloane and Verschuren2011). Similar to Maua, an onset of increased fire activity in the recent centuries occurred at Kwasebuge in the North Pare Mountains, though it has been more directly attributed to changing human activities in the forests using pollen from tree taxa that benefit from forest-edge disturbances (Finch et al. (Reference Finch, Marchant and Courtney Mustaphi2017). Evidence of vegetation and erosion changes from colluvium and palustrine deposits in the Pare Mountains suggests increasing land-use intensities, increased erosion, and changing vegetation cover (Heckman, Reference Heckmann2014). It has been hypothesized that this intensification of soil erosion over the last five centuries occurred owing to wood use for increased iron smelting and smithing (Schmidt, Reference Schmidt, Hamilton and Bensted-Smith1989; Håkansson, Reference Håkansson, Cliggett and Pool2007) or to intensification in agricultural production to meet the demands of an increasingly international landscape connectivity as trade caravans became larger and more frequent, particularly in response to global market demand for ivory (Biginagwa, Reference Biginagwa2012; Iles et al., Reference Iles, Stump, Heckmann, Lang and Lane2018). Furthermore, large changes to CHARs are evident in several palustrine records from areas surrounding Kilimanjaro and show high spatial variation in fire histories (Gillson, Reference Gillson2004; Githumbi et al., Reference Githumbi, Courtney Mustaphi, Yun, Muiruri, Rucina and Marchant2018a, Reference Githumbi, Kariuki, Shoemaker, Courtney Mustaphi, Chuhila, Richer and Lane2018b). CHARs increase after 500 yr BP at the Kimana and Enkongu wetlands in Amboseli, although the geochronological uncertainty is high at Kimana (Githumbi et al., Reference Githumbi, Courtney Mustaphi, Yun, Muiruri, Rucina and Marchant2018a), and the Namelok wetland record shows increased charcoal content after 300 yr BP-present (Rucina et al., Reference Rucina, Muiruri, Downton and Marchant2010). Charcoal accumulations are lower from 400 to 500 yr BP to the present at both Kanderi and Tsavo (Gillson, Reference Gillson2004) and at Esambu, Amboseli (Githumbi et al., Reference Githumbi, Kariuki, Shoemaker, Courtney Mustaphi, Chuhila, Richer and Lane2018b). The differing signals in charcoal accumulation highlight local-scale fuel and fire responses, the spatial heterogeneity of savannah fire regimes, the understudied complexities of fire regimes at savannah-wetland edges (Kirby et al., Reference Kirby, Lewis and Sexson1988; Kirkman, Reference Kirkman1995; Casanova and Brock, Reference Casanova and Brock2000; Cassidy, Reference Cassidy2007; Nielsen et al., Reference Nielsen, Podnar, Watts and Wilson2013; Saintilan and Rogers, Reference Saintilan and Rogers2015), and more broadly the role of human activities in and around wetlands (Seki et al., Reference Seki, Shirima, Courtney Mustaphi, Marchant and Munishi2018).

National-scale models estimate that populations and agricultural land use were expanding during this time (Fig. 7), tentatively supporting the hypothesis that anthropogenic drivers of land-cover change intensified (Klein Goldwijck et al., Reference Klein Goldewijk, Beusen and Janssen2010; Kaplan and Krumhardt, Reference Kaplan and Krumhardt2011; Kaplan et al., Reference Kaplan, Krumhardt, Ellis, Ruddiman, Lemmen and Klein Goldewijk2011). Suggestions that intensifications in land use were escalating during the last five centuries should be interpreted with caution, however, as evidence for anthropogenically induced environmental change—in the form of written records, detailed oral histories, and more conspicuous archaeological remains—becomes increasingly accessible toward the present, potentially leading to underestimations of the magnitude and intensity of earlier land use. Nonetheless, several lines of evidence suggest a diversity of human-environment modifications during this time (Fig. 6): remains of terraced irrigation infrastructure at highlands in the region (up to 1200 m asl; Sutton, Reference Sutton, Widgren and Sutton2004; Stump, Reference Stump2006; Lang and Stump, Reference Lang and Stump2017); remains of caves, dugouts, and bolt-holes used to shelter people, cattle, and valuables during conflicts (Clack, Reference Clack2009; Silayo, Reference Silayo2016); oral histories for mfongo (stream-fed) irrigation systems on the slopes of Kilimanjaro (Tagseth, Reference Tagseth2010); cultivation of crops introduced from the Americas (including maize and tobacco; Miracle, Reference Miracle1965; Cherniwchan and Moreno-Cruz, Reference Cherniwchan and Moreno-Cruz2019); Chagga forest gardens (Fernandes et al., Reference Fernandes, Oktingati and Maghembe1985); historical records of elephant hunting, especially using game-pits dug in the upper cultivation zone, at ~2000 m asl (Meyer Reference Meyer1891, p. 126); records of ivory-trade involvement (Coutu et al., Reference Coutu, Lee-Thorp, Collins and Lane2016); historical records of apiary operations in the heathland vegetation zone at, ~2900–3550 m asl (Meyer, Reference Meyer1891, p. 133); limited-scale iron smelting at Uru, Mamba, and Marangu (Wimmelbücker, Reference Wimmelbücker2002, p.62); and oral histories of other mountain forest uses (Stump and Tagseth, Reference Stump, Tagseth and Clack2009). On the lower southeastern slopes of Kilimanjaro, mfongo irrigation systems, together with water-storage ponds (Kimaro et al., Reference Kimaro, Scharsich, Kolb, Huwe and Bogner2019) have likely contributed significantly to modifying local ecologies. Mfongo networks are recognized to be central to the cultural heritage of the Chagga people (Tagseth, Reference Tagseth2010, p. 16). Oral traditions, historical references, and linguistic evidence allow for us to conclude that irrigation features on Kilimanjaro were being built by at least the seventeenth century (Stump and Tagseth, Reference Stump, Tagseth and Clack2009), though irrigation systems had emerged in the wider region from as early as the fourteenth century (Fig. 7; Sutton, Reference Sutton, Widgren and Sutton2004; Lang and Stump, Reference Lang and Stump2017; Kabora, Reference Kabora2018; Thornton-Barnett, Reference Thornton-Barnett2018; Kabora et al., Reference Kabora, Stump and Wainwright2020). The emergence and intensification of irrigation agriculture is linked to societal, ecological, and climatic factors (Boserup, Reference Boserup1965; Börjeson, Reference Börjeson2007; Håkansson, Reference Håkansson, Cliggett and Pool2007; Davies, Reference Davies2015). Furrow expansion and defensive constructions during the nineteenth century (Kusimba et al., Reference Kusimba, Kusimba and Wright2005; Clack, Reference Clack2007, Reference Clack2009) do, however, seem to coincide with a time of political centralization, conflict, slave raiding (Wimmelbücker, Reference Wimmelbücker2002), and intensified long-distance caravan trade (Stump and Tagseth, Reference Stump, Tagseth and Clack2009, p. 111). Kilimanjaro was a major node along multiple caravan trade routes, and high-elevation trails above the forest zone crossed the upper slopes of Kilimanjaro during this time, allowing people to avoid potentially hostile confrontations (Stahl, Reference Stahl1964, p. 344) and perhaps increase the intensity of anthropogenically induced change at elevations closer to Maua.

The uppermost sediments at Maua show moderately high ericaceous pollen, high Poaceae pollen and phytolith relative abundances, and moderately high charcoal (Fig. 7). European descriptions of the mountain emerged during the mid to late nineteenth century (Krapf, Reference Krapf1858; Meyer, Reference Meyer1890, Reference Meyer1891, Reference Meyer1900). Further explorations in the early twentieth century, including ascents of the mountain (e.g., Uhlig, Reference Uhlig1904; Jaeger, Reference Jaeger1909; Klute, Reference Klute1920; Gillman, Reference Gillman1923), contributed to the written documentation of the diversity of human land-use modifications and observations of vegetation, wildlife, and glaciers of Kilimanjaro (Alluaud, Reference Alluaud1908; Sjoestedt, Reference Sjoestedt1910; Engler, Reference Engler1925; Cotton, Reference Cotton1930; Humphries, Reference Humphries1959). Increasingly after 1900, colonialists also annexed large areas of Kilimanjaro for farmland and a forest reserves (Swynnerton, Reference Swynnerton1949; Sunseri, Reference Sunseri2003; Mwakikagile, Reference Mwakikagile2006) and added to land distribution issues. More detailed descriptions of flora and fauna at the upper elevations were not published until the mid-1900s (Salt, Reference Salt1954); systematic air photography also began in the mid-1900s (McGrath, Reference McGrath1976). In the interim years, there had been a steady decline in millet production in favor of newer commercial cultivars, such as maize, wheat, and coffee, and a decrease in investment in irrigation in the uplands concomitant with a major expansion in agricultural activity on the lower slopes of Kilimanjaro (Tagseth, Reference Tagseth, Tvedt, Coopey, Jakobsson and Østigaard2006; Bender, Reference Bender, Tvedt and Oestigaard2016; Chuhila, Reference Chuhila2016). Satellite-observed remote-sensing products available since the 1970s support observations that decadal-scale variability at the ecotonal gradients between montane forest and ericaceous vegetation cover are highly dynamic and are driven by fire (Wood, Reference Wood1965b; Hemp, Reference Hemp2005a, Reference Hemp, Spehn, Liberman and Körner2005b; Chuhila, Reference Chuhila2016; Detsch et al., Reference Detsch, Otte, Appelhans, Hemp and Nauss2016; Kilungu et al., Reference Kilungu, Leemans, Munishi, Nicholls and Amelung2019). Several lines of evidence support rapid (decadal scale) changes to vegetation cover on Kilimanjaro's slopes. Future work should integrate several lines of evidence from paleoecology, archaeology, history, and the social sciences (Courtney Mustaphi et al., Reference Courtney Mustaphi, Capitani, Boles, Kariuki, Newman, Munishi, Marchant and et al2019) to further examine the relative importance of hydroclimate, fire, plant competition strategies, local topography, herbivory, and anthropogenic interactions that influence spatial heterogeneity in the vegetation biogeography on Kilimanjaro.

Summarizing long-term human-environment interactions

Several factors have long been observed as annual-to-decadal scale drivers of spatial complexity in vegetation on the mountain, including interannual rainfall variability, fuel accumulation, and fire frequency, seasonality, and type (Wood, Reference Wood1965b). Abiotic and biotic factors explain the spatiotemporal complexity of vegetation on the mountain, including geology and the development of soils (Little and Lee, Reference Little and Lee2006; Zech et al., Reference Zech, Leiber, Zech, Poetsch and Hemp2011; Montade et al., Reference Montade, Schüler, Hemp, Bremond, Salamanca Duarte and Behling2018), topographic interactions with cold katabatic air masses and humidity regimes, the large and deep radial ravines (Hemp, Reference Hemp2006a), plant competition, and large mammal wallowing and trampling (Wood, Reference Wood1965b). Among the factors that must be considered when examining ecosystem functioning are long-term and legacy effects of anthropogenic activities. For example, plant biogeographers on the mountain have long found it difficult to explain some distribution patterns, such as that of the limited bamboo cover (Moreau, Reference Moreau1944; Hemp, Reference Hemp2006a), and there is some speculation that such distributions were potentially modified by anthropogenic activity (Grimshaw, Reference Grimshaw1999), including selective logging (Bjørndalen, Reference Bjørndalen1992) or sparing of sacred taxa/space, cultivation (Hemp, Reference Hemp2006a), and influences on homogenizing and diversifying fire regimes (Archibald et al., Reference Archibald, Staver and Levin2012; Beale et al., Reference Beale, Courtney Mustaphi, Morrison, Archibald, Anderson, Dobson and Donaldson2018; Coughlan et al., Reference Coughlan, Magi and Derr2018).

Anthropogenic activities, notably grazing, hunting, metallurgy, settlement, agriculture, and burning (e.g., for honey hunting, agricultural field preparation, grazing resource management, animal viewing spaces in parks), significantly modify vegetation patterns (Wood, Reference Wood1965b; Archibald et al., Reference Archibald, Staver and Levin2012; Vehrs and Heller, Reference Vehrs and Heller2017; Boles et al., Reference Boles, Shoemaker, Courtney Mustaphi, Petek, Ekblom and Lane2019). The potential of increased natural or anthropogenic-driven burning (or both) influencing the increased apparent biodiversity in the pollen taxa assemblage during Maua2 is particularly interesting among investigations of relationships between fire and biodiversity (Parr and Andersen, Reference Parr and Andersen2006). Wood (Reference Wood1965b) reports that honey collecting and fire use above the forest line contributed to fire occurrences in ericaceous vegetation cover, although this reference was missed in global syntheses of fire use by hunter-gatherer societies (Binford, Reference Binford2001; Coughlan et al., Reference Coughlan, Magi and Derr2018) and has yet to be corroborated with oral histories. Moreover, fire activity (or specific components of fire regimes), including the development of fire plans in conservation management, have yet to be studied for their role in promoting homogeneity or heterogeneity in high-elevation Afromontane ecosystems (e.g., Kelly et al., Reference Kelly, Bennett, Clarke and McCarthy2015; Trauernicht et al., Reference Trauernicht, Brook, Murphy, Williamson and Bowman2015; Beale et al., Reference Beale, Courtney Mustaphi, Morrison, Archibald, Anderson, Dobson and Donaldson2018). Further research combining paleoecology, archaeology, historical ecology, and ethnobiology would advance knowledge of historical human-environment interactions on the mountain. Human-induced defaunation on the mountain, both of biodiversity and of biomass, cascades and alters plant-animal-soil dynamics (Child, Reference Child1965; Wood, Reference Wood1965b; Lundgren and Lundgren, Reference Lundgren and Lundgren1972; Newmark, Reference Newmark1991, Reference Newmark1996; Grimshaw et al., Reference Grimshaw, Cordeiro and Foley1995). The disruption of persistent land-use practices that have unknown historical provenance and duration remains unassessed in terms of potential negative consequences and beneficial legacy effects on the environment and socioecological systems.

An estimate of the number of people living on Kilimanjaro (Meyer, Reference Meyer1891, p. 114) prior to the disastrous outbreaks of famine, smallpox, and rinderpest beginning in the 1890s suggests that over the last century the population has increased to roughly 30 times its former size. Historically, and still today, there are concerns that human practices such as water extraction (Stump and Tagseth, Reference Stump, Tagseth and Clack2009, p. 117; Bender, Reference Bender, Tvedt and Oestigaard2016), burning (Hemp, Reference Hemp2005a) and logging (Hemp, Reference Hemp2006a) on the mountain are unsustainable and will lead to environmental degradation. Of further concern is the decline in rainfall, indicated in precipitation records and observed by people living on Kilimanjaro's slopes (Ngana, Reference Ngana and Ngana2002; Hemp, Reference Hemp2005a; Tagseth, Reference Tagseth2010, p. 17). The need for appropriate resource management regimes in the face of these challenges warrants research on long-term trends in regional climatic variability and land-use histories.

Holocene chronologies of land-use changes have relatively high temporal uncertainties, high spatial complexity, and patchworked knowledge gaps. The history of Kilimanjaro is no less complex, and there is abundant evidence of diverse livelihood practices and dynamic cultural exchanges occurring throughout the past millennia. The archaeological record (e.g., Lane et al., Reference Lane, Ashley, Seitsonen, Harvey, Mire and Odede2007; Prendergast, Reference Prendergast2008; de Maret, Reference de Maret, Mitchell and Lane2013; Crowther et al., Reference Crowther, Prendergast, Fuller and Boivin2018; Shoemaker, Reference Shoemaker2018) suggests that livestock herding, metallurgy, agriculture, and hunter-gatherer-fisher modes of production were introduced, maintained, abandoned, and reproduced at highly variable temporal and spatial scales (Lane, Reference Lane2004, Reference Lane2009; Kusimba and Kusimba, Reference Kusimba, Kusimba and Stahl2005). In addition, we must recognize that the so-called traditional practices have not been immune to innovation and change (de Bont et al., Reference de Bont, Komakech and Veldwisch2019). The legacies of land-use activities such as honey gathering, hunting, and grazing as well as delineating sacred spaces to preserve them from resource extraction have, however, had influences on the mountain's vegetation cover patterns (Kikoti et al., Reference Kikoti, Mligo and Kilemo2015). More recently, ecotourism has become a major industry in northern Tanzania (United Nations Conference on Trade and Development, 2002; Kilungu et al., Reference Kilungu, Leemans, Munishi, Nicholls and Amelung2019) and is currently the main high-elevation land-use activity on Kilimanjaro, increasing the risk of introducing new exotic and invasive plant species (Hemp, Reference Hemp2008). Evidence from nearby savannah ecosystems suggests that introduced plant species often radiate from roadsides and refuse areas that service the ecotourism industries (Bukombe et al., Reference Bukombe, Kija, Loishooki, Sumay, Mwita, Mwakalebe, Kihwele and Keyyu2015a, Reference Bukombe, Kija, Loishooki, Sumay, Kihwele and Keyyu2015b).

Knowledge summaries of sustainable and unsustainable uses and human-environment impacts need to be investigated holistically to combat misunderstandings and miscommunications about desirable and less desirable resource management practices and to also assess chronocentric biases (Adams and Anderson, Reference Adams and Anderson1988, p. 522; Magliocca et al., Reference Magliocca, Ellis, Allington, De Bremond, Dell'Angelo, Mertz and Messerli2018). The high biodiversity and relatively lower degree of mountain endemism that characterizes the flora of Kilimanjaro is hypothesized to have been driven in part by millennia of anthropogenic activity (Hemp, Reference Hemp2006a), but further archaeological and historical ecology research is required if we wish to elucidate how people have used mountain resources over time and how anthropogenic activities have been modifying Kilimanjaro ecosystems during its long history of human-ecosystem interaction. As for much of eastern Africa, the potential consequences of land use on mountain areas remain unexplored and will require linked and colocated archaeological, historical, ethnographic, and paleoenvironmental inquiry (Marchant et al., Reference Marchant, Richer, Capitani, Courtney Mustaphi, Prendergast, Stump and Boles2018; Capitani et al., Reference Capitani, Garedew, Mitiku, Berecha, Hailu, Heiskanen and Hurskainen2019; Courtney Mustaphi et al., Reference Courtney Mustaphi, Capitani, Boles, Kariuki, Newman, Munishi, Marchant and et al2019; Cuní-Sanchez et al., Reference Cuní-Sanchez, Omeny, Pfeifer, Olaka, Mamo, Marchant and Burgess2019).