Gain and loss: Human and environmental wellbeing – drivers of Kilimanjaro’s decreasing biodiversity

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 1 / 23 OPEN ACCESS Citation: Hemp A, Miyazawa M, Hurskainen P (2025) Gain and loss: Human and environmental wellbeing – drivers of Kilimanjaro's decreasing biodiversity. PLoS One 20(10): e0334184. https://doi.org/10.1371/ journal.pone.0334184 Editor: Mattias Gaglio, University of Ferrara, ITALY Received: June 13, 2025 Accepted: September 23, 2025 Published: October 29, 2025 Copyright: © 2025 Hemp et al . This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data availability statement: The data are deposited on https://doi.pangaea. de/10.1594/PANGAEA.942806,10.5281/zeno- do.13929135,10.5281/zenodo.13929113, and 10.5281/zenodo.13929084. Census data are available from National Bureau of Statistics, Dar es Salaam: National Bureau of Statistics RESEARCH ARTICLE Gain and loss: Human and environmental wellbeing – drivers of Kilimanjaro’s decreasing biodiversity Andreas Hemp 1,2,3*, Mieko Miyazawa4, Pekka Hurskainen 5,6 1 Department of Plant Systematics, University of Bayreuth, Bayreuth, Germany, 2 Waldkundeinstitut Eberswalde, Eberswalde, Germany, 3 Global Mountain Biodiversity Assessment, University of Bern, Bern, Switzerland, 4 Department of International Studies, Graduate School of Frontier Sciences, The University of Tokyo, Kashiwa, Chiba, Japan, 5 Finnish Environment Institute, Helsinki, Finland, 6 Department of

Geosciences and Geography, Earth Change Observation Laboratory, University of Helsinki, Helsinki, Finland * andreas.hemp@uni-bayreuth.de Abstract Tropical mountains such as Kilimanjaro are biodiversity hotspots providing ecosystem services for millions of people, but many are under great pressure. Effective policies to halt biodiversity loss require an understanding of which anthropogenic factors are the main direct causes. While previous research focused mainly on climate change and on the effects rather than the causes, we investigated the effects of multiple drivers on biodiversity. The focus is on floristic and vegetation diversity, as vegeta- tion is closely related to the diversity of other taxa and plays a fundamental role in ecosystem functioning. We show that land-use change caused by rapid population growth was the main direct driver on Kilimanjaro between 1911 and 2022, when 75% of natural species per km2 disappeared from the lower slopes. Climate change, on the other hand, had no apparent influence on the observed trends in biodiversity. The significant increase in traditional and diverse agroforestry and the establishment of protected areas show possibilities for mitigation. Kilimanjaro is thus an example of the challenges of global change, but also of the prospects and opportunities for other tropical regions. Introduction Globally, human population growth and increasing per-capita demands for natural resources and energy are escalating human impacts on nature [1]. These indirect

factors — such as economic development, institutions, and governance — contribute to biodiversity loss and influence direct anthropogenic drivers, which affect biodiver- sity at multiple levels (within species, between species, and across ecosystems) [2]. The most notable direct drivers include land-use change, climate change, natural resource exploitation, pollution, and invasive species [3]. Human well-being relies

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 2 / 23 on the biodiversity of ecological systems and the benefits they provide to humans (known as nature’s contribution to people, NCP, formerly “ecosystem services”) [4–6], making anthropogenic impacts on nature a growing concern [7,8]. While climate change has attracted much attention as one of the drivers of biodiversity loss [8–10], it is not the only one [11–13], necessitating further examination of which drivers have the greatest effects on biodiversity [14]. Additionally, research has mainly focused on impacts rather than the drivers of global environmental change [15,16]. Kilimanjaro, the world’s highest free-standing mountain, features diverse ecosys- tems, high species diversity, and distinct land use systems [17]. The NCP provided by this tropical volcano includes climate and water regulation, food and timber provision and the spillover of crop pollination services into adjacent agro-ecosystems [18,19]. Due to increasing human pressures, Kilimanjaro’s ecosystems face severe threats [20,21], classifying it as a biodiversity hotspot [22]. It exhibits typical aspects of envi- ronmental change in tropical Africa while combining various features and challenges on one mountain, making it an excellent model for the investigation of the relationship between nature and humans. The objective of this study is to investigate how the main direct and indirect anthropogenic factors (with the exception of pollution, for which no data are available)

affect biodiversity at Kilimanjaro and to assess the role of conservation in relation to biodiversity. The focus is on floristic and vegetation diversity, as vegetation is closely related to the diversity of other taxa and plays a fundamental role in ecosystem func- tioning and thus NCP. Our biodiversity assessment is restricted to richness of plant species (overall species numbers and numbers of indicator species). First, we investigate the spatial and temporal relationship between population growth and economic development (as indirect drivers) and land use (as a direct driver) by relating changes in population density and economic development to changes in land use and vegetation cover. For this, we use historical German maps and census data from the beginning of the 20th century to get an overview of the most important changes between 1911 and 1976. A more detailed insight into the sit- uation since then is provided by the availability of satellite imagery. We focus on the inhabited areas of Kilimanjaro below the national park where no people live. Further- more, we convert these changes in land cover into changes in biodiversity by using the number of species of different plant groups as an indicator. To do this, we use a large dataset of 1600 vegetation plots with almost 3000 plant species that have been established throughout the mountain range since 1996. With this approach we want to answer the following questions: (i) Which of the anthropogenic factors addressed have the greatest impact on

biodiversity and how have the impacts of these anthropogenic factors changed in recent decades? (ii) How does climate change affect biodiversity? (iii) How is biodiversity changing? (iv) Can nature conservation mitigate the negative effects? Tanzania. 1978. Population and Housing Census 2002. Bureau of Statistics. Available: https://catalog.ihsn.org/catalog/2876/get-mi- crodata, National Bureau of Statistics Tanzania. 1988. Population Census. Analytical report. Bureau of Statistics, Planning Commission, President’s Office, Dar es Salaam, Tanzania. Available: https://microdata.nbs.go.tz/index. php/catalog/23/get-microdata National Bureau of Statistics Tanzania. 2003. Population and Housing Census 2002. National Bureau of Statistics, Planning Commission, President’s Office, Dar es Salaam, Tanzania. Available: https://microdata.nbs.go.tz/index.php/cat- alog/24/get-microdata National Bureau of Statistics Tanzania. 2013. 2012 Population and Housing Census. National Bureau of Statistics, Ministry of Finance, Dar es Salaam and Office of Chief Government Statistician, President’s Office, Finance, Economy and Development Planning, Zanzibar. Available: https://microdata. nbs.go.tz/index.php/catalog/44/get-microdata National Bureau of Statistics and President’s Office. 2022 The 2022 Population and Housing Census: Initial Results. – Ministry of Finance and Planning, Office of the Chief Government Statistician, Zanzibar. Dodoma, Tanzania. Available: https://microdata.nbs.go.tz/index.php/ catalog/45/get-microdata. Funding: German Research Foundation (DFG),

(HE 2719/14-1), Dr. Andreas Hemp. Competing interests: The authors have declared that no competing interests exist.

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 3 / 23 We hypothesize the following: (i) Population growth is the main indirect driver of biodiversity change, while land use is the main direct driver. Demo- graphic change differs between rural and urban areas: The population in urban areas is growing faster than in rural areas, and the population is more stable in mountainous areas than in lowland areas. Therefore, land cover changes are more pronounced in the lowlands than in the highlands. (ii) Climate change affects biodiversity mainly through changes in land cover after catastrophic events (e.g., fires after droughts or landslides after heavy rainfall), while the direct effects are minor. (iii) Species diversity is generally decreasing, but differently for different plant groups. (iv) Direct anthropogenic impacts inside the nature reserves have decreased, while they are increasing outside the reserves. Materials and methods Study area Kilimanjaro, located in North-Eastern Tanzania, is a UNESCO World Heritage site, boasting approximately 3000 vas- cular plant species [23], representing about one-third of Tanzania’s approximately 10,000 vascular plant species [24]. The study area encompasses the inhabited lower slopes and foothills below Kilimanjaro National Park (KINAPA) with 1688 km2 within five districts of Kilimanjaro Region: Moshi Municipal, Moshi Rural, Hai, Rombo, and Siha, covering 3282 km2 (Fig 1). Surrounding the mountain between 700 and 1100m is a dry and hot colline savanna zone. This zone is characterized

by a heterogeneous mosaic of disturbed savanna vegetation, croplands, and built-up areas (Fig 2). The remaining natural vegetation faces significant pressure from human activities such as agricultural expansion and intensification, urban- ization, firewood collection, brick manufacturing, and grazing [21,25–27]. The predominant natural vegetation consists of savanna woodlands with species of the genera Acacia, Terminalia, Combretum, Ozoroa and Sclerocarya in the tree and shrub layer, but the woodlands have degraded typically into grasslands dominated by Themeda, Eragrostis, Hetero- pogon, Chloris and Cenchrus species. Remnants of tall forest canopy exist along the river courses with fig trees (mainly Ficus sycomorus), Sorindeia, Trilepisium and Mimusops and in some areas with high groundwater table closed albeit heavily disturbed lowland forests still exist. Here, Oxystigma msoo a near-endemic trees species occurs together with Khaya anthotheca, Trichilia emetica and Milicia excelsa. Rushes and Sedges (Typha domingensis, Cyperus papyrus and Schoenoplectus corymbosus) dominate the few still existing swamps. Alkaline-resistant grasses (Sporobolus spicatus and S. robustus) and shrubs (Suaeda monoica, Salvadora persica, Megalochlamys revoluta and Dobera loranthifolia) char- acterize the quite widespread dry vegetation on alkaline soils, while table palms (Phoenix reclinata and Hyphaene com- pressa) dominate on alkaline soils with higher groundwater. Smallholder farms cultivate extensive areas with maize, sunflower, beans, and millet. South of Moshi, large rice paddy

fields and sugar cane plantations exist (Fig 3), while large wheat farms are present in the western part. During fallow periods, these fields are covered with ruderal vegetation consisting of mainly introduced species such as Malvastrum cor- omandelianum, Acanthospermum hispidum, Bidens pilosa, Commelina benghalensis and Hyptis suaveolens. A detailed physiographic map by Hemp et al. [28] illustrates the different land use and land cover types (LULC) of Kilimanjaro includ- ing our study area. In the highlands within the submontane zone below KINAPA only small remnants of forest remain in the deepest val- leys, where Newtonia buchananii, Entandrophragma excelsum (the tallest tree species of Africa [29]) and Garcinia tanza- niensis (a critically endangered tree species) grow. Everywhere else, the forests have mostly been converted into Chagga

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 4 / 23 homegardens, a unique form of agroforestry prevalent on the southern and southeastern slopes. These homegardens feature four vegetation layers: Under a sparse tree layer (mainly Albizia schimperiana) bananas are grown, under the bananas coffee trees and on the ground vegetatables, maximizing land use efficiency and biodiversity, with over 500 plant species (most of them forest species) and endemic arthropods [30,31]. Steep, dry slopes in the valleys of the submontane cultivation zone are covered by species-rich mown Hyparrhenia grasslands, consisting of Hyparrhenia rufa, Satureia abyssinica, Crepis carbonaria and Microchloa kunthii. A detailed description of these land use and vegetation types is presented by [17,23,31–33]). According to the Köppen and Troll/Pfaffen climate classification system (in [34]), the study area experiences a seasonal dry tropical climate influenced by the Inter-Tropical Convergence Zone. Rainfall occurs bi-modally, with rainy seasons from March to May and November. Rainfall and temperature vary with altitude and exposure to the dominant winds from the Indian Ocean [23]. Mean annual temperatures range between 15 and 25°C , while annual precipitation ranges from 400 to 2800 mm. Southern and eastern slopes receive more rainfall due to wind exposure from the Indian Ocean compared to the northern and western slopes [23,35]).

Fig 1. Study area. Study area (outer thick grey line) with delineation of the wards, situation of 2022 with 94 wards (grey lines), five districts (red lines; MM: Moshi Municipal), the border of KINAPA (green line), international border (dashed/dotted line), location of the map in Fig 10 (yellow line), Namalok nature reserve (orange line) and the available vegetation plots (red dots). Map source: Reprinted from World Imagery Basemap (Esri, Maxar, Earthstar Geographics, and the GIS User Community) under a CC BY license, with permission from Esri Deutschland GmbH, original copyright 2024. https://doi.org/10.1371/journal.pone.0334184.g001

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 5 / 23 Climate data For the temperature analysis of the past decades, we used data from the NOAA National Climatic Data Centre (NCDC) (https://www.ncdc.noaa.gov/). The temperature data are much more limited than the precipitation data, as only a few weather stations have collected long-term temperature data. A station with useful data was Kilimanjaro Airport in the southwestern corner of the study area at 850 m a.s.l. The average daily temperature was determined from 4 to 24 mea- surements/day. For precipitation data, we used the published data from Otte et al. [36]. We used a generalised least squares (GLS) regression model to test the statistical significance of the long-term trends. Data of population and economic growth The scale and time frame of our study were determined by the available census and remote sensing data: The smallest spatial unit for which census data were available was the ward level, an administrative structure for several villages, one Fig 2. The nine land cover/land use (LULC) classes used for the study area. The first two lines represent anthropogenic LULC classes. GRA: sub- montane grassland (class 4); AFO: agroforestry (class 2), a: commercial coffee plantation, b: traditional agroforestry system of a Chagga homegarden; FPL: forest plantation (class 7); BTU: built-up area (class 8); AGR: agriculture (class 1), a: rice field, b: wheat field, c: maize field. The last two lines rep-

resent natural LULC classes. FOR: forests (class 3), a: groundwater forest, b: Ficus riverine gallery forest, c: submontane Newtonia ravine forest; REG: forest regeneration (class 5); WAT: waterbody of Lake Chala (class 9); SAV a and b: savanna grasslands and woodlands, c: palm shrubland on alkaline soil with high groundwater table (class 6). https://doi.org/10.1371/journal.pone.0334184.g002

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 6 / 23 single town or portion of a bigger town (Fig 1). The first available Landsat image for Kilimanjaro was from 1976, and the first eligible national population census was from 1978. The other censuses were from 1988, 1998, 2012 and 2022. Due to these limitations (including the availability of cloud-free satellite imagery), we analysed the years 1976 (using the 1978 census data), 2002 and 2022. As corresponding basis for analysing the historical maps from 1911 we used the 1913 census by mission statistics [37]. We used this data to calculate population density (number of inhabitants per km2) for analyses. In addition to population, we applied house building activity as a proxy for economic growth, as no data on per capita income or other indicators of economic development, such as ownership of superior goods (e.g., the number of motor vehicles), were available for the entire period. We used a global satellite-derived built-up area statistic called GHS-BUILT-S from the Global Human Settlement Layer (GHSL) project led by the Joint Research Centre of the European Commission [38]. In parallel, we used our land cover classification to delineate built-up areas (Table 1). For visualisation of these trends we digitized groups of buildings using time series of Google Earth images and the four 1:50,000 Ordnance Survey topographic sheets from the early 1990s (Sheet No: Moshi 56/4, Ol Molog 42/3, Rombo 57/1, Himo 57/3). Land cover

We used the LULC maps for 1976 and 2000 of Hemp [20,23] based on Landsat MSS imagery from 1976, and Landsat ETM+ imagery from 2000. To update the 2000 map to correspond to the 2002 census year, we made adjustments by visually inspecting various very high resolution (VHR) satellite imagery available at Google Earth and ESRI World Imagery platforms, correcting boundaries of agriculture/savanna and other classes. Similarly, we updated the physiographic map by Hemp et al. [28] to reflect the situation as of 2022. Due to the lower spatial, spectral, and radiometric resolution of the Multi Spectral Scanner (MSS) on Landsat 1 satellite compared to the more advanced ETM+ and operational land imager (OLI) sensors on later Landsat satellites, the number of LULC classes had to be reduced to match what was realistically achievable to map from the 1976 Landsat 1 image. This led to a very rough categorisation of the main land cover and ecosystem types (Table 1): Fig 3. Cultivated lowlands on the southern slope of Kilimanjaro. In the foreground, the sugar cane plantations of TPC, surrounded by smallholder fields and – further up – rice fields, March 2025. https://doi.org/10.1371/journal.pone.0334184.g003

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 7 / 23 Since we could not delineate built-up areas on the Landsat image from 1976, we used the oldest available topo- graphic maps 1:50.000 from 1982. At that time, only Moshi existed as an urban centre. For the years 2002 and 2022, we used VHR satellite images in Google Earth to delineate the urban centres with built-up areas. The LULC maps were prepared using ArcMap 10.8.2. For more specifications of mapping these LULC classes see Supplementary Information S1 Text. To quantify the change flows, or transitions, between LULC classes and the three time steps (1976, 2002 and 2022), we used the functions in OpenLand R library [39]. We visualized the change flows with a Sankey diagram, which illus- trates relative transitions between the classes for the three time steps, as well as the net and gross changes with a stacked barplot. In order to obtain an overview of the main changes in the landscape since the first census on Kilimanjaro in 1913 [37], we also analysed historical maps from the beginning of the 20th century. The map of Hans Meyer’s Ostafrika-Expedition 1911 [40] at a scale of 1: 300,000 provides very detailed and botanically sound descriptions of the vegetation, especially in the savanna areas of Kilimanjaro. For final delineation of the LULC types (see also S1 Text) we used additional histori- cal information from the map and vegetation description by Meyer (1890) [41] together with information in Volkens (1897) [42], Jaeger (1909) [43] and Klute (1912) [44]. Biodiversity

We used a subset of the 54 land cover types identified by Hemp [23,29,32] for the entire mountain range based on 1,600 releves. We combined the occurring plant communities to fit into our proposed nine classes to create species lists (see Supplementary Information S1 Text for more details). The species numbers resulting from the analysis of the releves are listed in Table 2. The status of endemics (species with a restricted occurrence) follows a list by Gereau, Missouri Botanical Garden (pers. comm./unpublished data). We considered species in the following IUCN categories as Red List species (reviewed at https://www.iucnredlist.org/): Criti- cally Endangered, Endangered, Vulnerable, Near Threatened. The status of neophytes (introduced species) was checked in FTEA (1952–2012) and Kew Plants of the World Online (https://powo.science.kew.org/). Using these figures in order to calculate the number of species per km2 and to compare it with the human population density per km2 required two further steps. We first had to take into account the fact that the number of species increases Table 1. The Land Use and Land Cover classification schema (cp. Fig 2). LULC/ecosystem type Description (1) Agriculture Maize, sunflower and millet fields of smallholder farmers, commercial farms with wheat fields on the western and northern slopes, sugarcane plantations and rice fields south of Moshi (2) Agroforestry Agroforestry including traditional Chagga home gardens and commercial coffee plantations with a sparse tree layer of Albizia schimperiana, Rauvolfia caffra and Cordia africana

(3) Forest Moist forests including riverine and groundwater forests (with Ficus sycomorus, Oxystigma msoo and Khaya anthotheca) and submontane ravine forests with Newtonia buchananii, Entandrophragma excelsum and Mitragyna rubrostipulata in the tree layer (4) Grassland Mosaic of anthropogenic grasslands with Hyparrhenia rufa, Bulbostylis densa and Sporobolus africanus and agricultural fields in the submontane and lower montane zone, especially on steep slopes along the major river valleys and near KINAPA (5) Regeneration Forest regeneration with vegetation recovering after previous disturbances, mainly shrubland with Lantana camara, Carissa edulis and Ochna insculpta (6) Savanna Savanna with grasslands (Cyperus niveus, Themeda triandra and Heteropogon contortus), woodlands and dry forests with Acacia nilotica, Combretum molle and Sclerocarya birrea (7) Forest plantation Forest plantations with exotic tree species (Pinus patula and Cupressus lusitanica) on the northwestern and northern slopes (8) Built-UP Continuous impervious surfaces with infrastructure (buildings, roads, pavements, yards) (9) Water body Lakes https://doi.org/10.1371/journal.pone.0334184.t001

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 8 / 23 with area. This species-area relationship (SAR) is one of the few very well documented patterns of species richness [45,46]. This model is still the most commonly used and is considered the best for describing species-area calculations [47]. We used the calculated SAR for the 13 major land use and vegetation classes of Kilimanjaro of Hemp et al. [48] to scale up the species counts of the nine classes to 1 km2. Second, these classes show a certain overlap of species. For example, many species of open agricultural land also occur in agroforestry systems. Based on the analysis of the releves, we calculated this overlap which varied between the different classes and depended on the number of classes within a given area, in our case a ward. In one ward with all nine classes, the reduction factor increased to 0.51, meaning that the species numbers of each class had to be halved (see S1 Table with the species numbers and reduction factors for all combinations). Using the same approach, we calculated the number of endemic species, IUCN Red List species and neophytes, from the 772 releves (Table 2). In a final step, we used these reduced species numbers to calculate the final species numbers per km2 by relating them to the proportion of each LULC in the different wards. We used the detailed LULC classification by Hurskainen et al. [75] from the southern slope of Kilimanjaro to demon- strate the correlation between different LULC and species richness on a finer scale. Data analysis

We related changes in human population density at the ward level, to changes in land use and vegetation cover in the respective wards. We then analysed the effects of land cover changes on biodiversity using the number of species of vascular plants as an indicator. The total number of species and the number of different plant groups (natural vegetation plants, endemic and Red List species, neophytes) were used as indicators. To analyse trends in species numbers, we fitted linear functions using R2 as the measure of fit. The bivariate correlation between biodiversity data as response vari- able and population and land use data as explanatory variable was performed with Pearson’s correlation coefficient using SPSS 29.0.2.0. For critical remarks on the methodology see Supplementary Information S1 Text. Permits, inclusivity in global research We obtained the necessary permits to access the sites and conduct fieldwork from the Tanzania Commission for Science and Technology (COSTECH), Tanzania National Parks (TANAPA), and the Tanzania Wildlife Research Institute (TAWIRI). Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information (S1 Checklist). Table 2. Species numbers in the study area based on the 772 releves. LULC/ ecosystem type Number of releves Species number Number endemics Number IUCN Red List species Number neophytes Agriculture* 136 700 2 1 100 Agroforestry 76 581 4 0 138 Forest 101 846 25 16 62 Grassland 163 509 6 0 46

Regeneration 24 300 0 0 15 Savanna 272 1059 9 10 43 Water body 0 0 0 0 0 All classes 772 2057 35 26 180 *Including classes forest plantation and built-up (see Supplementary Information S1 Text) https://doi.org/10.1371/journal.pone.0334184.t002

extent of savanna and agriculture were nearly identical, but agriculture then expanded to become the dominant LULC class. Fig 4. Human population growth, reduction of natural vegetation and increase of town area on Kilimanjaro. Human population growth during the last 133 years (red dots with polynomial regression of degree 2), starting in 1889 ([37,41], Bureau of Statistics 1968, 1998, 2003, 2013, 2022 [49]); reduction of natural vegetation (green dots with polynomial regression of degree 2) and increase of town area (black dots with exponential regres- sion); correlation (Pearson’s r) of natural vegetation vs population (1911−2022) −0.985, p < 0.01 and vs town area (1976−2022) −0.915, ns. https://doi.org/10.1371/journal.pone.0334184.g004

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 10 / 23 Similarly, agroforestry expanded to 26%, surpassing savanna, which declined to 18%, becoming the second-largest land-use class in 2022. Apart from savanna, the other two natural vegetation types (forest and regeneration), also showed a steady decline. Forest plantations exhibited only minor changes in their extent, while built-up areas increased twelvefold between 1976 and 2022 (Fig 4). The same patterns but from a different perspective can be seen in Fig 6B, which shows the net and gross changes across the period 1976–2022. From this figure, two key insights arise: Agriculture and savanna were the most dynamic classes. Furthermore, all anthropogenic LULC classes experienced a net gain, while natural vegetation experienced a net loss across all categories. The Sankey diagram (Fig 6C) provides further insights into LULC transitions by illustrating the relative changes between LULC classes across three time periods (1976, 2002, and 2022). The diagram highlights which LULC classes Fig 5. Maps of land cover/land use (LULC) on Kilimanjaro between 1911 and 2022. (A-D) LULC maps of 1911, 1976, 2002 and 2022. AGR. agricul- ture, AFO: agroforestry, BTU: built-up area, FPL: forest plantation, FOR: forest, GRA: anthropogenic grassland, REG: forest regeneration, SAV: savanna, WAT: water body. Reprinted from World Imagery Basemap (Esri, Maxar, Earthstar Geographics, and the GIS User Community) under a CC BY license, with permission from Esri Deutschland GmbH, original copyright 2024.

https://doi.org/10.1371/journal.pone.0334184.g005

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 11 / 23 contributed to the expansion of specific categories. Notably, savanna conversion was the primary driver of agricultural expansion, with smaller contributions from other LULC classes. Biodiversity trends The number of all plant species per km2 (mean at ward level) exhibited only a slight decrease of approximately 10%, between 1976 and 2022 (Table 4). The decline in plant species of natural habitats, including forest, regeneration, and savanna, was much more pronounced, with a decrease of 46%. Conversely, there was an increase of approximately 25% in neophytes during the same period. This means that not only is species richness changing, but also species composi- tion, i.e., there is an increase in partially invasive and widespread species at the expense of native species with partially restricted distribution. Discussion Demography and economic development as main drivers of land use changes World population is projected to continue growing throughout the century, reaching between 9.6 billion and 12.3 billion at the end of the century, with Africa expected to experience at least a 3.5-fold increase in population. The current population of Africa, approximately 1.5 billion people in 2024, is projected to reach approximately 2.2 billion in 2054 [50]. Tanzania’s population growth rate stands at about 3.0% annually. If this trend persists, Tanzania’s population is estimated to increase

from about 65 million in 2022 to approximately 95 million in 2050 and over 300 million by the end of the century [49,51]. The human population on Kilimanjaro, comprising Moshi Urban, Moshi Rural, Rombo, Siha, and Hai Districts, has multiplied 28 times over 133 years from about 50,000 in 1889 [41] and 100,000 in 1913 [37] to over 1.4 million in 2022 ([49], Fig 4). As a result, population density increased from 15 people/km2 in 1889–30 in 1913, 50 in 1948 [52], and 211 in 1978, eventually reaching 430 people/km2 in 2022. The local population density (outside the cities) rose to unprecedented levels of up to 1500 people/km2 in the most fertile highland area of the Chagga homegardens. This population growth resulted in a severe shortage of cultivable land and the subdivision of land into fragments too small to support a family and too numerous to allow for further expansion [25,53]). Consequently, many young people migrated from the highlands to urban centers such as Moshi, Arusha, and Dar es Salaam, or settled on the foothills of Kilimanjaro. This emigration from the highlands explains why the demographic growth rate on Kilimanjaro is lower than the national average in Tanza- nia and why the population growth rate in the lowlands (2.9%) was much higher than in the highland homegarden areas Table 3. Land cover changes 1911-2022. LULC/ ecosystem type 1911 area (km2) 1911 area (%) 1976 area (km2) 1976 area (%) 2002 area (km2) 2002 area (%) 2022 area (km2) 2022 area (%) Agriculture 66 2 1062 32 1406 43 1486 45 Agroforestry 268 8 395 12 650 20 849 26

Forest 319 (500*) 10 201 6 119 4 60 2 Grassland NA NA 134 4 32 1 28 1 Regeneration 180 5 281 9 87 3 43 1 Savanna 2449 75 1090 33 795 24 588 18 Forest plantat. 0 0 116 4 170 5 150 5 Built-up 0 0 6 0 24 1 74 2 Water body 2 0 2 0 2 0 2 0 (*) estimated number including not covered riverine and ravine forests; NA: anthropogenic grasslands could not be delineated for 1911 (see Supplemen- tary Information S1 Text). https://doi.org/10.1371/journal.pone.0334184.t003

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 12 / 23

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 13 / 23 (0.9%) between 1976 and 2022. In 1976, approximately 30% of the population resided in the lowlands, a proportion that increased to about 50% by 2022. In the causal chain of factors affecting biodiversity, demography acts as an indirect driver via the direct drivers [7], at Kilimanjaro by controlling land use change as direct driver. In other regions of Tanzania, it was also the population growth that led to the expansion of cultivated land [54]. The observed growing population density results in the overall reduction of natural habitats since 1911 (Fig 4), with increasing annual reduction rates from 1.0%, 1.9% and 2.4% between 1911 and 1976, 2002 and 2022. This can be shown in more detail for the period 1976–2022 at ward level (Fig 7). In addition to the clear correlation between increasing population density and decreasing natural vegetation, Fig 7 shows that not only the mean share of natural vegetation was higher in 1976 than in 2002 and 2022, as also seen in Fig 4, but also, that at a given population density, e.g., of 1000 people/km2, the natural vegetation share was Fig 7. Changing share of natural habitats over time in correlation to population density per km2 at ward level on Kilimanjaro. Blue symbols 1976, red symbols 2002, green symbols 2022 with logarithmic regressions. Pearson’s correlation coefficient r 1976: −0.594, p < 0.001; 2002: −0.475, p < 0.001; 2022: −0.407, p < 0.001. https://doi.org/10.1371/journal.pone.0334184.g007

Fig 6. Changes of land cover/land use (LULC) on Kilimanjaro between 1976 and 2022. (A) Extent of LULC classes in 1976, 2002 and 2022. (B) Net and gross changes of LULC between 1976 and 2022. Anthropogenic LULC types increased on expense of natural vegetation. (C) Sankey diagram, illustrating relative LULC change flows between 1976 (left), 2002 (middle) and 2022 (right). https://doi.org/10.1371/journal.pone.0334184.g006 Table 4. Changes in species numbers 1976-2022 (mean at ward level). 1976 2002 2022 Pearson’s r All plants 704 640 636 −0.892 ns Natural plants 340 250 184 −0.996* Savanna plants 162 132 101 −1.000** Neophytes 90 105 120 1.000** Pearson correlation with population density: ** p < 0.01, * p < 0.05, ns: non significant. https://doi.org/10.1371/journal.pone.0334184.t004

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 14 / 23 approximately 18% in 1976, compared to 6% in 2002 and only 1% in 2022. This decline may be due to rapid economic development at an even higher rate than population growth increasing per capita demand for resources, e.g., via higher availability of machinery or pesticides, with this increasing land use intensity and reducing natural areas. Fig 3 gives an impression of the extent to which the vegetation at the southern foot of Kilimanjaro has been reshaped by humans. For the entire Kilimanjaro region, including the five study districts and the Pare Mountains, Gross Domestic Product (GDP) grew about 5.7 times from 1980 to 2020 [49]. During this time, GDP per capita in Tanzania increased approxi- mately 1.8 times [55], indicating that economic growth was much more rapid in Kilimanjaro. In our study area, reflected through construction activities as indicators of economic growth, Fig 4 shows an annual growth rate of built-up areas at 5.9% between 1976 and 2022, and 6.1% from 2002 to 2022, surpassing population growth. This aligns with global satellite-derived built-up area statistics named GHS-BUILT-S from the Global Human Settlement Layer (GHSL) project (see Supplementary Information S1 Text; Table 5). Trends post-2000 indicate significant expansion of built-up areas com- pared to the 1975–2000 period. This is also in line with trends on a continental level: Africa’s urban areas will more than double until 2025. The growth

in urban area expansion is projected to outpace urban population growth, at 3.2% compared to 2.3% per year. The total urban footprint is projected to increase from 175 000 square kilometres to 450 000 square kilometres between 2020 and 2050 with an increasing number of urban agglomerations from 9,000 to over 11,000 [56]. The trend of house building activities and the development of new town centers between 1990 and 2016, particularly in the lowlands of Hai and Siha districts on the southwestern side of Kilimanjaro, is illustrated in Fig 8. This aligns with GHS- BUILT-S (Table 5). The role of KINAPA as a barrier against encroachment by settlements is notable in Fig 8, highlighting the importance of governance through governmental protection of nature that delivers ecosystem services to Tanzania’s people. Demographic and socio-economic factors have driven many land-use changes at Kilimanjaro. By 2000, a corridor of submontane forest that once connected Kilimanjaro and Mt. Meru had been largely replaced by human settlements and agriculture (Fig 5B). This forest corridor was crucial for the dispersal of forest animals [21], but Kilimanjaro is now increas- ingly isolated, impacting biodiversity and endemism. On a positive note, the growth of Chagga homegardens – a sustainable land-use practice supporting various natural species [31,53,57] – has helped mitigate biodiversity loss, especially as these gardens have expanded at the expense of agricultural fields (Fig 5C). From 1976 to 2022, the area of homegardens more than doubled to 849 km2.

Climate change was not an important driver of the observed biodiversity trends A reanalysis of meteorological data for East Africa shows an approximate surface temperature increase of 0.95 K over the past 60 years (specifically between 2.5° to 5.0° South and 37.5° to 40.0° East). Additionally, the RCP6.0 scenario projects a further increase of 3 K by 2100 [58]. However, such modeled climate data for tropical mountains is sometimes very unre- liable [35]. Beside increasing trends of temperature no consistently significant rainfall trends over the period 1973–2014 Table 5. Built-up surface (km2) from GHS-BUILT-S. District 1975 2000 2020 1975 → 2000 2000 → 2020 Siha 3.34 4.97 10.70 + 49% + 115% Hai 7.82 12.08 20.85 + 54% + 73% Moshi Rural 23.26 32.02 42.93 + 38% + 34% Rombo 19.79 21.85 25.03 + 10% + 15% Moshi Municipal 6.32 9.82 11.04 + 55% + 12% https://doi.org/10.1371/journal.pone.0334184.t005

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 15 / 23 could be found for the lowlands of Kilimanjaro [36]. Consequently, we conclude that climate change did not significantly influence observed biodiversity trends or land use in the lowlands. If it had been a primary driver of land use change, we would expect to see negative impacts on subsistence agriculture and agroforestry, yet these categories showed significant gains. However, this differs from the situation above the cultivation belt within the national park. There, signs such as glacier retreat [59] indicate a gradual decline in regional humidity and precipitation, and an interplay between increasing human activities and decreasing humidity leads to an increasing frequency and intensity of forest fires, affecting biodiversity espe- cially in the upper montane forest zone and the subalpine ericaceous belt [20]. Forest fires are also frequent in the humid lower montane forest zone near cultivated areas in the vicinity of larger anthropogenic clearings due to a changing micro- climate with increasing air temperature and decreasing humidity [60]. In the last 120 years, Kilimanjaro has lost about 50% of its forest area [23]. This not only affects biodiversity, but also has far-reaching consequences for the regional climate. Deforestation leads to a significant increase in maximum air temperature and cloud base height [61], which on Kilimanjaro primarily affects the upper regions within the national park and even the glaciers [62] and impairs the area’s ability to pro-

vide vital NCPs, especially water. This interaction demonstrates that climate change is not always a challenge originating from outside but also a “homemade” problem caused by regional land-use change and boosting together biodiversity loss. However, such changes (e.g., the increase in wild fires) have not been observed in the lowlands of Kilimanjaro. Land use change: Main driver of biodiversity loss Land use change and intensification are the main drivers of the decline of biodiversity and ecosystem services globally [3,5,63,64], exceeding the impact of climate change, particularly in tropical areas [11,14,65–67]. For Kilimanjaro, Fig 9 illustrates the effects of population-induced natural habitat loss on species diversity. Species den- sity of natural species (including endemics and Red List species) correlates closely with Fig 7, indicating that land-use change, notably ecosystem loss, is the primary cause of reduced species density (Fig 9A). Fig 8. House building activities on Kilimanjao between 1990 and 2016. Left: Buildings (black dots) 1990; Right: New buildings (red dots) 2016. Blue lines: town centres; green line: KINAPA (National Park Boundary). Buildings digitized by Ethan Oleson based on time-series Google Earth images and the 1:50,000 Ordnance Survey topographic sheets. Reprinted from World Imagery Basemap (Esri, Maxar, Earthstar Geographics, and the GIS User Community) under a CC BY license, with permission from Esri Deutschland GmbH, original copyright 2024. https://doi.org/10.1371/journal.pone.0334184.g008

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 16 / 23 The loss of about 40% of natural savanna habitats (76% since 1911) significantly threatens Kilimanjaro’s biodiver- sity, as savanna ecosystems are crucial for plant species diversity [17,68]. This decline also affects traditional medic- inal plants, primarily found in Kilimanjaro’s lowlands [69]. This trend aligns with global patterns; over the last century, tropical savanna ecosystems have decreased by over 50% due to agricultural conversion [4]. In Sub-Saharan Africa, around 20% of former savanna land is now cropland, with much of the remaining area used for livestock grazing [70]. The FAO predicts further encroachment of subsistence agriculture into former savanna regions due to population growth [71]. Forests ecosystems have also significantly declined, from over 500 km2 in 1911–60 km2 in 2022 (Table 3), particularly impacting endemic species (Fig 9B). This decrease varied over time and between wards. Between 1911 and 1976, many natural forest ecosystems on the western and northern slopes were converted to exotic tree plantations, while substantial Fig 9. Biodiversity trends 1976-2022. (A) natural species, (B) endemics, (C) IUCN Red List species, (C) neophytes. Blue symbols 1976, red symbols 2002, green symbols 2022. Species numbers per km2 in correlation to population density at ward level with logarithmic regressions. Correlation indices are presented in Table 6. https://doi.org/10.1371/journal.pone.0334184.g009

Table 6. Pearson correlation coefficient for densities of plant groups vs population (Fig 9). Natural Endemic Redlist Neophytes 1976 (n = 82) −0.673*** −0.399*** −0.717*** 0.696*** 2002 (n = 82) −0.527*** −0.315** −0.589*** 0.459*** 2022 (n = 94) −0.488*** −0.341*** −0.419*** 0.137 ns n: number of wards, which changed after the division of several wards after 2002 (see Supplementary Information S1 Text). ***p < 0.001, **p < 0.01, ns: non significant. https://doi.org/10.1371/journal.pone.0334184.t006

PLOS One | https://doi.org/10.1371/journal.pone.0334184 October 29, 2025 17 / 23 forest areas on the southern slope became homegardens, and between 1976 and 2002 the Meru-Kilimanjaro forest corri- dor was destroyed (Fig 5). In parallel to the decreasing area of natural habitats, Fig 9A shows not only a similar negative trend over the study period, but also a similar decrease in the number of natural species at a given population density. For example, at an average population density of 1000 people/km2, 245 species per km2 were recorded in 1976, decreasing to 180 by 2003 and 115 in 2022. This decline seems linked to increasing land use intensity driven by economic development, similar to trends in natural vegetation (Fig 7). Peters et al. [68] identified land use intensity as a key factor influencing bio- diversity on Kilimanjaro. Intensified land use, particularly pesticide application, significantly harms susceptible plant spe- cies, such as epiphytes in Kilimanjaro’s agroforestry systems. Chagga homegardens showed higher numbers of epiphytic mosses compared to intensively managed commercial coffee plantations [57]. Like other high mountains in tropical East Africa and Ethiopia, Kilimanjaro’s alpine zone has a high proportion of endemic species, with numbers decreasing at lower elevations [72]. Our dataset identified 35 endemic species, making up only 2% of the total species in Kilimanjaro’s lowlands. Fig 9B shows the declining trends of endemics in relation to population density over time.