Homo heterogenus: Variability in early Pleistocene Homo environments

To understand the ecological dominance of Homo sapiens, we need to investigate the origins of the plasticity that has enabled our colonization of the planet. We can approach this by exploring the variability of habitats to which different hominin populations have adapted over time. In this article, we draw upon and synthesize the current research on habitats of genus Homo during the early Pleistocene. We examined 121 published environmental reconstructions from 74 early Pleistocene sites or site phases to assess the balance of arguments in the research community. We found that, while grasslands and savannahs were prominent features of Homo habitats in the early Pleistocene, current research does not place early Pleistocene Homo, in any single environmental type, but in a wide variety of environments, ranging from open grasslands to forests. Our analysis also suggests that the first known dispersal of Homo out of Africa was accompanied by niche expansion.

the research regarding the variability in suitable environmental conditions.We focus on sites dated to between ∼2.0 and 0.8 Ma.When using the term human, we are referring to any member of the genus Homo.In many cases, human presence is indicated just by archaeological remains, making species identification impossible.However, in this period the human species occupying this site can often be treated as Homo erectus sensu lato.We nevertheless remain agnostic about the taxonomy of Homo in the early Pleistocene and operate at the genus level.

| Understanding the literature
Traditionally, the prevailing hypothesis has been that the first human dispersal out of Africa occurred concurrently with the expansion of savannah-grasslands, 11,12 the habitat in which Homo presumably evolved in eastern Africa.This theory of environmentally constrained expansion is supported by evidence suggesting that at the key early Pleistocene locations in eastern and northern Africa, Levant, and Asia, humans are associated with savannah-like environments.According to this view, these environments extend continuously from eastern Africa to northern China, later expanding to Europe. 12,13Robin Dennell coined the term "Savannahstan" 11,14 to describe this extension of savannah environments.Here, we use the term, "Savannahstan hypothesis", to refer to ideas proposing an intimate link between savannah-like environments and expanding early Pleistocene Homo even if the term was not explicitly used by authors.This hypothesized Savannahstan would have promoted the dispersal of Homo into Eurasia, along with other associated African fauna for example, Theropithecus oswaldi, Megantereon whitei, and Hippopotamus antiquus. 15Under the strict form of the Savannahstan hypothesis, early Pleistocene Homo was so closely linked and adapted to these habitats that its delayed appearance in Europe relative to Asia would have been linked to forests preventing earlier dispersal to the region. 13,16Following this reasoning, Homo would only have been able to disperse into the region when more open environments expanded into Europe towards the end of the early Pleistocene.Similarly, it has been argued that in East and South-East Asia, H. erectus was not able to penetrate the subtropical forests 17 and that the relatively recent extinction of H. erectus in Indonesia was caused by the expansion of forests in the region. 18,19wever, this perspective has also been challenged.1][22][23][24] Moreover, some studies have argued that early Pleistocene Homo would have thrived in non-savannah environments.Belmaker and O'Brien 8 found that at the time of human occupation, the environment at 'Ubeidiya was more humid than today with a woodland-forest vegetation, challenging the notion that hominins would have preferentially occupied open grassland habitats.Similarly, Saarinen et al. 25 have suggested that early Pleistocene humans generally occupied relatively diverse habitats and that the evidence from early Pleistocene archaeological sites of Barranco León and Fuente Nueva 3 in Spain suggests palaeoenvironments where grasses were only a minor component of the vegetation. 25us, the question about the diversity of suitable habitats, or more broadly about the niche size of early Pleistocene Homo, remains unresolved.To resolve this debate one needs to understand how narrow or wide the niche space occupied by Homo was in this period.The geographic distribution of a species can be thought of as a representation of its realized niche.
As such, much of the debate about drivers of the early Pleistocene human range expansion is about niche stability vs niche expansion, and specifically, whether the first range expansion was due to climate and environmental change resulting in Homo being able to expand its distribution as suitable environments became available outside Africa or as a result of biological or cultural novelties allowing niche expansion and consequent exploitation of a wider range of environments.Our main aim, emerging from the abovementioned debate, is to contribute to answering the question of whether, when Homo dispersed out of Africa for the first time, it diversified to exploit new environments or remained within the scope of its African niche.To address this, we need to explore the diversity of environments Homo occupied across its early Pleistocene range in Africa and Eurasia.Therefore, we turned to the research community and assessed the balance of arguments by compiling, analyzing, and synthesizing reconstructions of early Pleistocene Homo's environmental context published between 2000 and 2021 (Figure 1 and Supporting Information Appendix).

| HOW TO LEARN FROM THE LITERATURE?
Descriptions of palaeoenvironments are often contextdependent and specific to the study locality or time period.
When researchers are making broad hypotheses, for example, about early Pleistocene hominin habitats, they often make implicit inferences from a corpus of such descriptions.
To be able to explicitly synthesize varied proxy records and interpretations researchers have derived from them, we had to make the results of individual studies comparable even though in some cases this results in a loss of details and contextual information.
To analyze the reconstructions of early Pleistocene Homo's environmental context we first assembled a corpus of the published literature relevant to our research question. 20, We isted the terms used to describe the vegetation of early Pleistocene Homo environments and returned to the literature to investigate how these terms were used in the corpus.To be able to compare the environments described in the corpus we compressed the diverse language used to give environmental reconstructions under six core categories-Wetland, Grassland, Shrubland, Mixed, Woodland, and Forest.The grassland category also includes environments containing vegetation that is described as open or as steppe, while the mixed category covers savannah, wooded grassland, wooded steppe, and forested steppe.We did not include "savannah" as its own category to ensure as consistent a treatment as possible of environments containing interspersed woody and grassy vegetation.Vegetation described as closed, dense, or closed canopy are included in the forest category.Anything described only as "woody" (or similar) is categorized as woodland.Details of the scoring of the corpus alongside the data are provided in the Supporting Information Appendix (Table 1).
Next, we scored each study in the corpus by giving relative weights to the environmental categories based on the results of the studies.Using the process detailed in Box 1 we converted the descriptions of the palaeoenvironments into a comparable quantitative format.Our data consists of 121 reconstructions representing 74 individual sites or site phases described in the corpus of 83 research papers.Following the processing and scoring of the environmental reconstructions according to six categories, we also collapsed the descriptions of vegetation into one dimension by calculating the "tree affinity" (Box 2) for each site.This variable gave us an estimate of the relative importance of trees in the environment at each site.By calculating this for our sites we were able to better understand potential habitat variation along the open-closed vegetation gradient.Additionally, we used correspondence analysis to reduce the dimensionality of the data and to illustrate the distribution of reconstructions in the environmental space.It is important to keep in mind that our data can have several reconstructions per locality.This potential pseudoreplication is considered in the formal statistical analysis, details of which are given in the Supporting Information Appendix.Since our research corpus is composed of a methodologically diverse set of studies, it is important to acknowledge that different proxies do not represent vegetation components equally well.For example, pollen productivity and dispersal vary across plant taxa which can bias fossil pollen-based diversity estimates. 106Furthermore, in herbivore dental isotopes it has been shown that the relationship between δ 13 C values and woody cover estimates is complex and largely decoupled, with only Proboscideans showing a strong link. 107I G U R E 1 A map with points indicating the geographic distribution of sites for which environmental reconstructions were extracted from the corpus.Many of the points represent several individual locations, for example in Nihewan, Northern China, what appears as one point is six sites within our data (Supporting Information Appendix and Table 1).The coloring of points corresponds with regions used in the analysis: Africa (Red), Asia (Purple), Europe (Blue), Levant, and Caucasus (Green).LITERATURE?

| Farewell to Savannahstan?
The analysis of published research indicates that grassland was a major component of Homo habitats in the early Pleistocene (Figure 2), with the category accounting for 38% (376/968) of all vegetation scores (Figure 2c).Together with shrubland and mixed category, grasslands form the most important component of human habitats in all regions, which appears to give support for the Savannahstan hypothesis.However, the overall pattern has more nuance.Figure 2a shows The results also indicate regional differences.The mixed vegetation category, which includes savannah and wooded grassland, shows the highest average scores in Africa, whereas forested habitats gain more prominence in Eurasia (Figure 2b,e).These differences are evident also in the distribution of tree affinity scores (Figure 2d between Africa and Eurasia is also statistically significant (Supporting Information Appendix).This supports the idea that the first human range expansion out of Africa was accompanied by expansion of niche size.Humans did not remain within their African niche but encountered and inhabited novel environments when they dispersed into Eurasia.
Interestingly, these key results are supported by Zeller et al. 108 who recently correlated Pleistocene human occurrences with climate model-driven vegetation model output.Zeller et al. 108 argue that the habitats of the early Pleistocene Homo in Africa were predominantly located in savannah and grassland environments, whereas in Eurasia, Homo appeared in a more diverse suite of habitats, with almost 60% of occurrences associated with temperate or tropical forests.They also argue that Homo in this period exploited and possibly even actively sought heterogeneous habitat conditions. 108e fact that the published proxy-based habitat reconstructions analyzed in this paper and the model-based reconstructions of Zeller et al. 108 suggest a similar pattern is a strong indication that we are capturing a real signal of the past human habitat variability.This signal directly contravenes the strict form of the Savannahstan hypothesis: The current evidence shows that the early Pleistocene Homo was not an obligate dweller of savannah grasslands, but managed in a heterogeneous set of environments that included fully forested habitats.
Given this ecological flexibility, it is interesting that it has been argued that the extinction of H. erectus in Indonesia was due to the expansion of forests during the previous interglacial.H. erectus, together with other presumably savannah and woodland specialists, would have found changes associated with forest expansion extremely challenging. 7,18,19However, considering our results, it seems less likely that the expansion of forest alone would have caused such devastating challenges to a species that was able to cope with a wide array of different environments already in the early Pleistocene.
0][111][112] As early Pleistocene Homo was able to sustain itself in the range of habitats from the temperate zone to the tropics, populations living under different environmental conditions could have started to develop adaptations to the specific conditions across the Old World.
Together with geographical distances between populations, this could have led to the morphological diversity with ≥4 Homo species found in the middle Pleistocene. 110,113If Homo in the early Pleistocene had been a savannah-grassland specialist remaining in its narrower African niche, it would not have been able to occupy the diverse set of environments.This in turn, could have impeded the development of the observed diversity.

| If
Homo falls in a forest, will anyone find the fossil?
While the current research places early Pleistocene hominins in Africa mainly into savannah-grassland environments and suggests a niche expansion when humans dispersed into Eurasia, these conclusions depend on how complete our understanding of African habitat variability is.There are multiple potential sources of bias and incompleteness in the African record of hominin habitats.First, we must consider the potential taphonomic and sampling biases which may affect the sample of fossil sites available for research.Under the poor preservational conditions of forested ecosystems in central and western Africa, fossilization is less likely to occur than in open environments. 114Thus, forests are less likely to be reconstructed from fossil-based proxies.Second, research attention has remained largely within the African Rift system and South Africa (Figure 1).While in many cases this is for valid reasons, including easy access to fossiliferous sediments and political instability in other potential target regions, one cannot help but be reminded of the Bernard Wood quote: "Paleoanthropologists are like the drunk looking for their lost keys under the lamppost" 115 and wonder if we are missing out on incredible hominin finds by remaining under the lamppost of wonderful sediment archives of eastern and southern Africa.These sources of sampling bias reduce the likelihood of sampling more forested environments, and thus including these within the range of Homo habitats. 114,116stly, even though the available African habitat reconstructions show a stronger tendency towards more open environments than Eurasian reconstructions, there is still considerable variation in the available African data, which covers all our vegetation categories, including forest.As Figure 2e shows, the environmental space covered by African localities is almost as big as European, although the large size depends on Ain Hanech (rec.n. 44) and Lower Busidima in Gona (rec.n. 35) that have exceptionally large forest components.
The African data also shows regional variation; early Pleistocene human habitats in the Karonga basin in Malawi rift (rec.n. 11-12) are described to be less open than east African habitats, with woody cover up to 70% in the basin localities. 36,105However, even the classic sites in the eastern African Rift such as Olduvai Gorge, show significant habitat heterogeneity and, at least local importance of habitats with dense woody cover. 37,43,45,117e crucial aspect is temporal variability, which becomes visible when sedimentary records with higher temporal resolution are used.
By using lipid biomarkers and the isotopic record preserved in lake sediments at Olduvai, Magill et al. 54 were able to reveal cyclic vegetation changes related to precession-driven changes in the moisture availability.Their results show how early Pleistocene hominin habitats in Olduvai repeatedly changed from open grassland to forest and back within the precessional cycle of 20,000 years.This suggests that early Pleistocene hominins in Africa were not unfamiliar with more closed environments and that coping with them was already part of the behavioral repertoire of early Pleistocene Homo when it started to expand its range out of Africa.

| How many trees make a forest?
An additional source of uncertainty related to reviewing a diverse research corpus arises from the qualitative language that is often used to describe paleoenvironments.Qualitative language is inherently subjective and as such extracting comparable research conclusions from qualitative terms can be difficult.Specifically, to interpret the corpus analyzed here, one needs to understand the complexities of the qualitative terms used to describe palaeoenvironments.For example, one reconstruction of Olduvai Bed I (rec.n. 24) indicates that the site was wooded to densely wooded, 45 which may indicate a woodland which increases in density grading to the forest (our interpretation from additional context).However, it may also indicate a woodland varying in density; or a savannah with woody cover that grades into woodland.In particular, "woodland" is a term which is used differentially depending on many factors both linguistic and cultural. 1180][121] The term "forest" is often treated as synonymous with woodland, 118 when functionally they should be considered as separate. 122,123With potentially variable usage of the terms, it is likely that some environments we categorized as "woodland" are in fact forests and vice versa.This may have masked the presence of closed environments in some regions while overemphasizing their presence in others.
Similarly, the savannah itself, a key concept in answering this research question, is a heterogeneous, ambiguously defined environment. 1246][127] Savannahs clearly constitute a heterogeneous environment containing grassland and woodland components in variable mixtures. 128,129A key aspect in this mixture is a ground layer dominated by grasses, 124 which allows for variation along a broad spectrum of overlying tree coverage.With different definitions of savannah placing varying emphasis on these different aspects, the biome is left to "range from grasslands with scattered trees to densely tree-covered woodlands." 127In this definition, we see that environments defined by some as savannah may be ascribed to the forest category by another.
If we were to categorize using this broad definition, then Homo would almost exclusively be found in savannahs.While variable meanings and use of environmental terms have likely influenced our interpretations of researchers' environment descriptions, similarities between our results and the model-based habitat reconstructions of Zeller et al. 108 nevertheless suggest that our interpretations of qualitative descriptions have not significantly biased our results.

| FUTURE DIRECTIONS
We can argue that a fairly robust pattern emerges from the analyzed research corpus, where early Pleistocene Homo occupied a diverse set of habitats and did not remain within their African niche, when expanding into Eurasia.Yet, as highlighted above, uncertainty and open questions remain.Faith et al. 130 have recently suggested several approaches to better deal with the uncertainty and facilitate more robust conclusions in the study of ecological drivers of hominin evolution.Here, we highlight three potential research lines that could increase our understanding specifically about habitat variability of the early Pleistocene hominins and the origins of plasticity in the genus Homo.

| Testing Homo heterogenus
Different theories about human habitat variability tend to assume that humans were actively selecting a certain fraction of the environment to avoid or seek particular characteristics.Knowledge of the strength of the preferences is important to our understanding of the origins and degree of plasticity in the Homo lineage.If Homo in the early Pleistocene was actively selecting heterogeneous habitats, this implies a generalist adaptive strategy which could represent the first link in the chain of evolution that leads to our own species generalist-specialist strategy.However, simply analyzing the distribution of humans across different habitat types does not necessarily give an indication of preferences; finding humans more often in savannah-like, or heterogeneous habitats may only reflect the prevalence of these habitats in the available environment.To make inferences about habitat preferences requires information about the distribution of available environments.This would then make it possible to analyze if the distribution of human occurrences significantly deviates from the expectation based on the distribution of available environments.Such information about the available environments cannot really be extracted from the proxy data.Thus, Zeller et al. 108 have proposed an approach based on climate and vegetation models and suggested that the model-based data indeed indicate a preference for habitat heterogeneity.Since climate and vegetation models vary in their outputs, it would be important that this finding is tested using different model simulations as they become available.

| About time
Regardless of whether habitat heterogeneity was actively sought by ancient humans or not, proxy records indicate that heterogeneity or mosaicism were important characteristics of hominin habitats, especially in Africa. 131However, when evaluating heterogeneity in palaeoenvironments it is important to consider the potential impact of temporal resolution on habitat reconstructions.Faith et al. 130 argue that time-averaging can lead to the reconstruction of environmental conditions that were never present.More specifically, time-averaging can increase the number of vegetation components in the reconstructions and consequently lead to overestimation of heterogeneity.As Magill et al. 54 have shown, precession-driven moisture variability typical to low-latitude environments can result in relatively rapid vegetation changes from forest to grassland, which, in a lower resolution, would show up as a signal of habitat heterogeneity.While the lake sediments used by Magill et al. 54 have sufficient resolution to detect temporal variation at this scale, this is not true for many palaeoenvironmental proxies.Hopley and Maslin 2010 132 have shown that terrestrial fauna undergoes precession-related time averaging if they accumulate over a period of time greater than onethird (c.7000 years) of the precessional cycle.They argue that this degree of time-averaging would be well within the expected range of terrestrial vertebrate bone accumulations.This means that terrestrial vertebrate assemblages may contain woodland, and grasslandadapted species in the same faunal unit when these species never coexisted in the area; resulting in the environment being inaccurately reconstructed as a woodland grassland mix.
For example, in our research corpus, one reconstruction of Olduvai Bed 1 contains wetland, grassland, shrubland, woodland, and forest components (rec.n. 25; Figure 2a). 46It is possible that this reflects the true mosaic nature of the environment that still characterizes many parts of Africa, but it is also possible that the high number of different components results from a time-averaged proxy assemblage.Perhaps the assemblage accumulated over a period when the locality changed from humid densely wooded conditions to a more arid grassland environment.While some sites in the research corpus do provide proxy data, which distinguish arid and humid phases of vegetation compositions, not all proxy records can provide such a high temporal resolution.Thus, it is possible that heterogeneity is overestimated in the data set.To estimate the influence of time-averaging on habitat reconstructions and assess the prevalence of the problem, it would be important to analyze the strength of the relationship between the age range of a proxy assemblage and the likelihood of obtaining reconstruction suggesting mosaic or heterogeneous habitat composition.

| Seeking quantification
Due to the focus of individual studies on specific locations or a limited set of sites, the priority is using the most appropriate methods that apply to that particular context, where relative statements are often sufficient to resolve a particular question.To make use of such local data for global comparative analyses, relative and often qualitative statements need to be converted to a common currency of some sort.For example, in our analyses, we have used vegetation scores and tree affinity.However, these methods are still based on our interpretation of environmental terms' meanings and dilute the context of authors' qualitative statements down into a quantitative value.As we highlighted, misinterpretations of meanings and qualitative statements can lead to biases in this type of comparative research.To overcome this and better facilitate future meta-analyses and understanding of hominin niche changes, the research community needs to continue the shift towards more quantitative representations of habitat characteristics using common currencies such as; vegetation cover, 87,92 net primary productivity, 133 or climate parameters. 25other solution could be the compilation of a large data set covering the geographic range of early Pleistocene Homo, which would then allow the application of a common measurement scale across the data.By controlling the measurement scale, it is also possible to reduce the uncertainty caused by the biases of different proxy measures (such as those discussed briefly in Section 2).For example, comparing the mean ordinated hypsodonty of large planteating mammals at hominin sites would give a representation of how similar or different the precipitation regimes controlling vegetation communities were across sites. 133,136,137Despite our call for a more quantitative approach, it is evident that the research of hominin habitats will always need multiple approaches, and that multiproxy, interdisciplinary analyses will provide the most complete picture of the palaeoenvironments which early Homo species occupied.The growing availability of open access databases (e.g., Neotoma, 134 the NOW database 135 ) and repositories can serve as an excellent resource in facilitating multiproxy quantitative approaches.

| SUMMARY
In this study, we have shown that it is possible to explicitly make inferences and extract quantitative data from published literature by applying a consistent treatment across the assembled corpus.This has allowed us to assess the current state of research in hominin habitat variability during the early Pleistocene.While we were able to synthesize meaningful conclusions by taking the literature as a whole, there remain some wider issues that need to be overcome to further advance our understanding of early Pleistocene human habitat variability.These include, for example, the influence of time-averaging in palaeodata and limitations of qualitative environmental descriptions for comparative analyses.
Nevertheless, it is clear that the research we have reviewed here does not place early Pleistocene Homo into a single biome.We can conclude that while grasslands and savannahs were prominent features of hominin environments in the early Pleistocene, Homo was able to exploit a wide variety of environments from open grasslands to forests.This indicates that the niche of early Homo was not bound to savannah environments.The current research also suggests that when Homo dispersed out of Africa for the first time, it expanded, not only its geographical range but also its niche.It follows that early Pleistocene Homo may have been more plastic in its ecology than the strict form of Savannahstan hypothesis proposes.The occurrence of early Pleistocene Homo in variable environments may be considered as sowing the seeds of plasticity which has come to characterize H. sapiens and enabled the species' ecological dominance.
that 6 out of 121 reconstructions indicate fully forested habitats and eight reconstructions show forests having the largest share among different vegetation components.Some of these cases are based on records that show an alternation of open and closed vegetation phases often interpreted as representing an alternation of arid and humid or glacial and interglacial conditions.For example, "wet" phases in Tres Pins (reconstruction number 85) and Nihewan Majuangou (rec.n. 71) and "interglacial" phase in Bovila Ordis (rec.n. 87) are reconstructed as fully or almost fully forested, 73,85 while "dry" or "glacial" phases in these sites show much more open F I G U R E 2 Early Pleistocene Homo habitat composition.(a) Proportional scores of different vegetation categories in individual reconstructions.The number in front of the bar refers to the reconstruction number in the research corpus (Supporting Information Appendix).(b) Average scores for each vegetation category in different regions.(c) Total scores are given for each vegetation category.(d) Distribution of tree affinity scores in different regions.(e) Distributions of environmental reconstructions along the first two dimensions of the correspondence analysis.The figure also shows convex hulls representing environmental envelopes covered by data from different regions.Points 35 (Gona Lower Busidima) and 44 (Ain Hanech formation B) are individually labeled to facilitate referencing in the main text.vegetation.This highlights the potential importance of temporal variability in human habitats.Thus, instead of unequivocally supporting the Savannahstan hypothesis, our analysis shows that the published reconstructions place early Pleistocene Homo into a wide spectrum of environments, ranging from open grasslands to full forests, and in many cases environments with varied or heterogeneous vegetation composition, which is consistent with the persistent idea that mosaic environments were important for early humans.As such, reconstructions of early Pleistocene Homo environments by the research community clearly indicate that the genus was not restricted to only one broad habitat type.
), which tracks the relative importance of trees in the environment and the width of the human niche along an open-closed vegetation gradient.While median tree affinity scores in all regions tend towards the open end of the spectrum, maximum values and, consequently, the distribution of tree affinity values is wider in Asia and Europe than in Africa.The difference in the interquartile range (0.05-0.95)