The habitat utilization and environmental resilience of Homo heidelbergensis in Europe

Abstract

The European populations of Homo heidelbergensis may have contributed to the genetic heritage of modern Eurasians. A better understanding of the possible effects of palaeoenvironmental alterations on the evolution of ancient humans can help to understand the origin of developed traits. For this purpose, the spatiotemporal alterations of physical factors were modelled in Europe for the period of 670–190 ka, covering the existence of Homo heidelbergensis in Europe. The factors included the following: paleoclimatic conditions, climatic suitability values of ancient humans, two prey species, and the European beech. Furthermore, the distribution and features of wood used for toolmaking were also investigated. Finally, changes in the relative mortality risk, the percentage of the body covered by clothing, and daily energy expenditure values in the coldest quarter of the year were modelled. The results suggest that H. heidelbergensis inhabited dominantly temperate regions in Europe where prey such as red deer were present. In the northern regions, European beech trees were abundant. When making wood tools, they preferred relatively light but not the strongest woods, which were readily available in the vicinity of the sites. Although hard and heavy woods were also occasionally used, at a European level, significant changes were observed in the relative mortality risk, the percentage of the body covered by clothing, and daily energy expenditure values during the period of 670–190 ka. However, substantial differences between archaeological sites in these values, indicating somewhat ecological variations, were not found during the studied period.

Introduction

Homo heidelbergensis occupies a central position in the cladogenesis of later Homo species, such as Homo sapiens, Neanderthals, and Denisovans (Di Vincenzo and Manzi 2023; Timmermann et al. 2022). The phylogenetic position of H. heidelbergensis has been interpreted in various ways over the last few decades. For instance, it was considered a stem species (Mounier et al. 2011) or a chronospecies (Stringer 2012a), potentially representing the direct ancestor of modern humans and Neanderthals (Buck and Stringer 2014; Meyer et al. 2016; Stringer 2012b). Alternatively, some researchers have suggested it may be a sister group of Neanderthals (Gómez-Robles 2019). Based on anatomical features, it has been proposed that the transition from H. heidelbergensis to Homo neanderthalensis in Europe can be exemplified by fossil specimens such as the parietal bone found in Casal de’ Pazzi, Rome (Di Vincenzo and Manzi 2023). If any of these hypotheses are accurate, including, e.g., the notion that Homo bodoensis was indeed a real ancient human taxon (Roksandic et al. 2022), H. heidelbergensis would have contributed to the genetic heritage of modern Eurasians.

By recognizing the central evolutionary role of H. heidelbergensis in the development of derived Homo species, it becomes crucial to explore how palaeoecological conditions and their changes influenced the physical and behavioural adaptations of this species. This is particularly important because the Middle Palaeolithic represented a period of rapid human intelligence growth and an expansion of habitat ranges, closely linked to climatic shifts (Petersen 2023). In light of H. heidelbergensis' ability to thrive in colder environments in Europe, it is very likely that this ancient human taxon acquired clothing and other anti-cold strategies and tools to counteract the risk of hypothermia in the cold northern climates of western Eurasia (Mondanaro et al. 2020). To delve into this topic, it is worthwhile to harken back to the time when the Homo genus first emerged in Africa. It is highly plausible that in the birthplace of Homo, alterations in precipitation played a pivotal role in shaping the range and survival of early Homo species (Trájer 2023). For instance, consider Ledi-Geraru in Afar, Ethiopia, where the oldest Homo remains were discovered (Villmoare et al. 2015). During the mid-Pliocene, the mean temperature of the coldest quarter at this archaeological site was ca. 24 °C, with an approximately 10 °C difference between the warmest and coldest quarters, and the annual precipitation was just 430 mm (PaleoClim data: Brown et al. 2018). Climatic simulations suggest that in their original East African homeland, Homo did not have to cope with cold conditions in winter times but rather had to adapt to high summer temperatures and seasonal aridity (Mondonaro et al. 2020; Trájer 2023), leading to water and food shortages. The findings from Dmanisi suggest that approximately 1.8 million years ago, Homo left Africa for the first time (Lordkipanidze et al. 2007, 2023). The initial Eurasian arrivals settled in the territory of present-day Georgia, which at the time had a subtropical climate (Messager et al. 2010, 2011). This region was cooler than the habitats previously inhabited by human populations in Africa.

Subsequently, the adaptation of early humans to the colder climate of Eurasia likely constituted a gradual and protracted process. In the case of human societies, due to the dual biological-technological dimension of human existence, both the availability of raw materials for tool-making purposes and the physiological adaptation level of human beings are crucial factors in the peopling of an area. From a technical perspective, it can be argued that the availability of chert and radiolite for flint-making (Wagner et al. 2011) or the presence of suitable woods like, e.g., spruce wood (Conard et al. 2020) for spear-making could also be of great importance for Middle Pleistocene humans in Europe. It should be mentioned that new evidence suggests that at 476 ka, the African contemporaries of the European H. heidelbergensis were already able to create complex structures from processed wooden materials (Barham et al. 2023).

However, factors such as the presence of different wood species were also strongly influenced by climatic conditions. The colonization of continental territories probably necessitated the acquisition of specific skills (Mondanaro et al. 2020), including cloth-making (Wales et al. 2012), fire control (Stahlschmidt et al. 2020; MacDonald et al. 2021), and food preservation techniques, such as the utilization of fermented or putrefied meats (Speth et al. 2017). The investigation of the major changes in the rate of climate niches of ancient Homo species suggests that H. heidelbergensis acquired these adaptive skills (Mondanaro et al. 2020). The earliest potential evidence for the consistent ability to create fire by hominins was discovered at Gesher Benot Ya ‘aqov, Israel, dating at 790 ka (Alperson-Afil et al. 2017), although the use of fire in Europe is attested only later, in several sites dated at 400–300 ka (Roebroeks and Villa 2011; Sanz et al. 2020; Pieruccini et al. 2022). At Vértesszőlős, Hungary, H. heidelbergensis demonstrated the innovative use of bone structures with radial geometry to regulate fire intensity as early as 315 ka (Dobosi 2006). The ability to exploit available raw materials and selectively utilize them might have also played a crucial role. It is documented that H. heidelbergensis began utilizing different pine woods for toolmaking at least 400 ka (Ennos and Chan 2016). However, the decision-making logic behind the selection of specific raw materials, such as different types of wood, remains relatively obscure.

As aforementioned, given the African origin of Homo, it is plausible that archaic species such as Homo antecessor and Homo erectus may have been susceptible to the continental (cool) environmental conditions and their fluctuations in northern Eurasian climates (Raia et al. 2020). In contrast, late Neanderthals appear to have been relatively well-adapted to the harsh environment of Upper Pleistocene Eurasia (Lacruz et al. 2019). Based on the evaluation of their body form and palaeoecology, it was assumed that Neanderthals were cold-adapted, well-suited to power locomotion, and predominantly lived in woodland habitats (Stewart et al. 2019). In light of these realizations, it is understandable that they extended their range to continental regions such as the southern Ural (Serikov and Chlachula 2014) and the Altai Mountains, where they interbred with Denisovans (Krivoshapkin et al. 2018), yet another advanced Homo species. It is worth noting that even during the last interglacial period, approximately 130 ka, the mean temperature during the coldest period in the Denisova Cave archaeological site was as low as some -17 °C (PaleoClim data: Brown et al. 2018). Phylogenetically speaking, H. heidelbergensis occupies an intermediate position between archaic and recent Homo species, suggesting that his environmental resilience may have fallen somewhere between these two groups. This is significant because some researchers have proposed that the rapid climatic and environmental changes of the Middle Pleistocene could have played a role in the extinction of ancient European human populations (Finlayson and Carrion 2007; Raia et al. 2020).

While H. heidelbergensis populations did inhabit relatively cool temperate regions, it is plausible that they did not specifically tolerate continental or cold climatic conditions (Wagner et al. 2011). Parfitt et al. (2010) similarly concluded that the overall distribution of Middle Pleistocene Homo populations primarily encompassed tropical forest savannah and temperate (Köppen Cs-Cw) rather than interior continental areas with extreme seasonal fluctuations and boreal regions. It is worth noting that, prior to the appearance of Neanderthals, Homo in Europe generally did not extend beyond the 53° latitude circle in Western Europe and the 45–47° latitude circle in Eastern Europe (Wiseman et al. 2020). While this does not necessarily imply that the species' northern distribution limit was determined by a specific thermal value, the European northernmost extension of Homo during the Middle Pleistocene approximately aligns with the current 9 °C annual average temperature isotherm line (Trouet and Van Oldenborgh 2013). It is important to note that the past course of mean temperature isothermal lines in Europe, in terms of the general pattern and not the absolute temperature values, may have been similar to the current conditions (Davies and Gollop 2003), but this cannot be definitively stated.

The level of physical and behavioural adaptation of H. heidelbergensis is difficult to address. Since H. heidelbergensis could plausibly be the ancestor of H. neanderthalensis and could also contribute to the genetic heritage of modern humans, either as the direct ancestor of H. sapiens (Mounier et al. 2009) or via genetic transfer from Neanderthals to modern humans (Weyrich et al. 2017), it is important to understand the ecological requirements of this ancient Homo species. The aim of the study is to characterize the Middle Palaeolithic paleoclimate of Europe in relation to the occurrence of H. heidelbergensis and various physical and biological factors that affected the habitability of the ancient European environment.

Materials and methods

The workflow of the study

In the entire study, the coldest quarter of the year received significant attention, as it can be assumed that it is strongly related to the length of the winter (harsh) season and cold-related mortality and can also be considered an indicator of the possible effects of glacial periods on human distribution (Davies and Gollop 2003; Gamble et al. 2004; Klein et al. 2021).

The detailed steps of the present study were as follows:

• Selected archaeological sites yielding evidence related to H. heidelbergensis were georeferenced using Quantum GIS 3.4.4 (Neteler and Mitasova 2002).

• In the process of site selection for analysis, a key criterion was the precision of age determination, assessed through variance or standard deviation. Notably, sites such as Clacton, Ehringsdorf, Mala Balanica, Petralona, and Steinheim were deliberately excluded due to the inherent uncertainty in age determination, surpassing 50–40 thousand years. Conversely, sites with cultural strata falling within the 670–190 thousand-year range, corresponding to the existence of early and late pre-Neanderthal populations in Europe, were included in the analysis. Notably, human remains were not discovered at all the selected sites, underscoring the diverse nature of the dataset.

• The paleoclimatic values of the archaeological sites were reconstructed via the age-related δ¹⁸O values (obtained from Billups and Schrag 2002) of the sites and two reference climatic models.

• The reconstruction of the paleoclimatic values was based on the δ¹⁸O‐inferred technique of Trájer (2022a), which required the use of the climatic data of an interglacial-glacial climatic simulation pair.

• Due to the uncertainties in age determination, the values of the determined δ18O-based modification factors were calculated using the LOESS (Locally Weighted Scatterplot Smoothing) technique, utilizing the average values of archaeological dates.

• The maxima and minima of the applied 19 climatic factors related to the ancient human fossils were selected as the distribution-limiting climatic extrema. Then, the potential climatic suitability patterns of H. heidelbergensis were modelled.

• The comparison of the reconstructed paleoclimatic values of the sites and the degree of similarity between these values and the present-time climatic categories were studied using K-Nearest Neighbors algorithm and by modelling climatically analog regions of Europe today. In these cases, the Köppen-Geiger climate classification of Beck et al. (2018) provided the basis for evaluations.

• As a biological indicator of oceanic and humid continental climate regions in Western Eurasia and Asia Minor (Fang and Lechowicz 2006; Zimmermann et al. 2015), the range of European beech (Fagus sylvatica) was modelled.

• The potential climatic suitability values of red deer (Cervus elaphus) a species in the temperate and humid continental regions of Europe that co-occurred with Middle Pleistocene archaic humans (Iannucci et al. 2021), were also modelled. Artiodactyls, particularly C. elaphus, were favored game for prehistoric hunters, as evidenced by discoveries in the Arago Cave (Rivals et al. 2004). Cervus elaphus stands out as the most common element among the three deer species in the large mammal assemblages found in Aroeira related to Middle Pleistocene archaeological strata (Croitor et al. 2019). Notably, remains of Cervus elaphus were excavated at various locations, including Atapuerca level TD8 (Van der Made et al. 2017), Casal de’ Pazzi (Briatico et al. 2023), and Castel di Guido (Saccà, 2012). The upper layers of the Middle Pleistocene fluvio-lacustrine sediments in the Ceprano Basin, housing Acheulean artifacts, reveal a diverse large mammal fauna, including the remains of C. elaphus (Ascenzi and Segre 2000). The distribution-limiting climatic extremes of Cervus elaphus were based on Trájer (2022b). It should be noted that it is challenging to find large mammal species currently living that could be modelled for the occurrences in the Middle Palaeolithic and be relevant from the perspective of H. heidelbergensis. This is partly due to the fact that most contemporary large mammal species have already become extinct in Europe. Others, like Capra ibex, did not exist throughout the entire examined period. Rangifer tarandus, being a northern species primarily inhabiting the tundra, is not considered in relation to the examined archaic human species. Furthermore, similarly to the European beech, these species can also be used as biotic indicators of temperate and humid continental climate regions in Europe.

• The potential range of European silver fir (Abies alba), European spruce (Picea abies), Scots pine (Pinus sylvestris), and European yew (Taxus baccata), all native European species whose wood was used for tool-making in Clacton (Allington-Jones 2015) and Schöningen (Richter and Krbetschek 2015), was modelled. It provides the opportunity to determine whether these woods were common or rare in the making of spears and other wood artifacts. Furthermore, the physical features of these woods were compared with each other and the wood of an additional 16 European tree species.

• The calculation of relative mortality risk, the percentage of the body surface covered by clothes, and the daily energy expenditure values were modelled based on the reconstructed paleotemperature values.

Spatial–temporal frames

A temporal frame with a period of 670–190 ka was selected. This period allows the modeling of former patterns over a 40-year period. To maintain a manageable number of model sub-periods, it is crucial to consider that 12 results will be produced for each case. The decision to limit the number of model sub-periods is driven by practical considerations, as the analysis involves generating results for specific time intervals within the overall period. Therefore, the model sub-periods were defined as follows: 229–190, 269–230, 309–270, 349–310, 389–350, 429–390, 469–430, 509–470, 549–510, 589–550, 629–590, and 670–630 ka.

For the sampling of modelled values, the following delimiting coordinates were used: -10 and 30° W and 35 and 62° E. This sampling grid covers the entire South, West, and Central Europe, as well as the southern and western regions of East Europe.

Palaeoclimatic reconstruction

Since high-temporal-resolution georeferenced paleoclimatic models are currently limited, the initial step involves reconstructing past climates based on age-site data. The available paleoclimate models cover only specific periods, leaving gaps in the representation of climatic conditions. However, the used timeframe aligns with some of the existing climatic model periods. This challenge was addressed through the following logical steps:

• Determining the lowest and highest temperature reference points based on the δ¹⁸O-content of benthic foraminifera corresponding to the Last Glacial Maximum (LGM) and the Last Interglacial Period (LIP). The source of the paleoclimatic model was the PaleoClim database (Brown et al. 2018).

• Due to the 1000-year resolution of the δ¹⁸O record from Timmermann et al. (2022) and the inherent variability in the determined age of the fossil archaic human as well as the Middle Pleistocene archaeological strata, LOESS (Locally Weighted Scatterplot Smoothing) regression was employed to calculate the δ¹⁸O-based modifier factor (refer to the explanation below).

• Calculating a modifier factor based on the age of the H. heidelbergensis fossil and the LGM using the ${\delta }^{18}$O-contents of the benthic foraminifera between the LGM and LIP periods.

The accurate reconstruction of previous global thermal conditions is often grounded in measurements of benthic foraminifera ¹⁸O/¹⁶O concentrations (referred to as δ¹⁸O) and Mg/Ca ratios (Liu et al. 2009; Billups and Schrag 2003). These measurements result from temperature-dependent carbon and oxygen isotopic disequilibria in deep-sea benthic foraminifera (Graham et al. 1981). This occurrence enables the recreation of atmospheric paleotemperature conditions or the estimation of the size of ancient ice sheets using δ¹⁸O values (Billups and Schrag 2003). The Paleoclim database was the source of the paleoclimatic information (Lisiecki and Raymo 2005; Brown et al. 2018).

The following formula served as the basis for the modifier factors:

$$m=1-\frac{d{18O}_{LGM}-d{18O}_{N}}{{d18O}_{LGM}-{d18O}_{current}}$$

where m is a modifier factor, ${\delta }^{18}$O_LGM denotes the ${\delta }^{18}$O concentration (‰) during the LGM, ${\delta }^{18}$O_N denotes the concentration of oxygen in the N^th period, and ${\delta }^{18}$O_current is the concentration of oxygen in the current period. The proportional bioclimatic factor of the location in the given time span was then computed using the following formula based on the bioclimatic factors of the reference LGM and current periods and the modifier factor.

$$r{bio}_{1-19,N}={bio}_{1-19,LGM}+m\times ({bio}_{1-19,current}-{bio}_{1-19,LGM})$$

where rbio_1-19,N is the reconstructed 1 to 19 bioclimatic factors of the N^th period, bio_1-19,LGM is one of the 1 to 19 bioclimatic factors of LGM, m is a modifier factor, and bio_{1-19,ncurrent} is one of the 1 to 19 bioclimatic factors of the current period.

As mentioned earlier, the δ¹⁸O data is available at approximately a 1000-year resolution for the Middle Pleistocene period. However, the age of the strata containing archaeological and archaic human remains is known with some uncertainty, often spanning tens of thousands of years. LOESS offers an optimal solution in this case as a non-parametric method that fits a smooth curve through data points, accommodating local variations (Niiler 2020). In essence, this method takes into account the δ ¹⁸O-based 'm' values in the local environment at a given time point. Instead of computing a simple mean value or conducting a straightforward data imputation through linear regression, LOESS introduces a parameter, denoted as λ (lambda), which governs the weight assigned to a value based on its distance from the time point. Another parameter denoted as ‘frac’ determines the width of the local neighborhoods used for regression. In the current study, the applied values for λ and 'frac' were set to 0.01 and 0.02, respectively. A 'frac' value of 0.02 indicates a narrow bandwidth, resulting in a fine-scale analysis of the data.

In summary, LOESS entails employing local polynomial fits weighted by a distance-dependent function. These local fits are subsequently combined to generate a smooth curve that captures the overall trend in the data. This approach facilitates a precise estimation of modifier values. LOESS analysis was performed in the Spyder (Scientific Python Environment) version 5.2.2 (Hanif 2024), utilizing a data table comprising δ¹⁸O values and the mean age of the investigated archaeological sites.

Table 1 shows the age-related mean δ¹⁸O values of the 12 Middle Pleistocene periods and the determined modifier factor values applied in the paleoclimatic reconstructions.

Table 1 The age-related mean δ¹⁸O values and the corresponding modifier factors related to the 12 model periods in 190–670 ka

Table 2 lists the archaeological sites inhabited by H. heidelbergensis to create climatic suitability models and two additional archaeological sites of wooden tools that were involved in the study. along with the age-related δ¹⁸O and modifier factor values applied in the paleoclimatic reconstructions.

Table 2 The H. heidelbergensis archaeological sites used for modelling with their corresponding δ¹⁸O and m values and similar values of two other archaeological sites where wooden tools were found

Occurrence sites

A total of 26 European Homo heidelbergensis-bearing archaeological sites, as well as certain sites where human remains were not found but in a chronological and spatial sense, can be included in this cohort, covering the period from MIS16 to MIS6 were involved in the study based on Di Vincenzo and Manzi (2023), Roksandic et al. (2018), and Marra et al. (2017). Alternatively, these sites can also be considered early and late pre-Neanderthals sites, which does not falsify the taxonomic classification of these findings and is in accordance with the moder viewpoints of the evolution of ancient Eurasians (Roksandic et al. 2018; Di Vincenzo and Manzi 2023). The selected H. heidelbergensis fossil sites are shown in Fig. 1.

Fig. 1

The H. heidelbergensis archaeological sites used in this study. a: European and b: Italian overview

Supplementary Table 1 presents these parameters by site, resulting in a total of 19 × 2 variables.

Comparison of the reconstructed climatic values of the sites

Spyder version 5.2.2 was used for plotting the cluster-heat map diagram based on the reconstructed 19 bioclimatic variables of the sites. Before the analysis, the bioclimatic values were normalized using a z-score normalization method to ensure comparability and mitigate the impact of differing measurement scales among variables. The heat map aspect of the visualization uses colours to represent the values of the variables, allowing for the identification of trends, variations, and clusters within the dataset. The hierarchical clustering organizes the sites based on similarities in their bioclimatic profiles, enabling the identification of groups that share similar climatic conditions.

Data and its processing related to tree species

The distribution data of European beech (Fagus sylvatica), European silver fir (Abies alba), European spruce (Pinus picea), Scots pine (Pinus sylvestris), and European yew (Taxus baccata) were obtained from the map created by Caudullo et al. (2017). To georeference the map, I employed the georeferencer function of Quantum GIS 3.4.1189, using the Polinom 3 technique as a fitting method to ensure high spatial accuracy. Subsequently, the software's sample function was applied to extract data from all the studied bioclimatic factor layers based on the presence points of the rasterized map. Furthermore, 2–2 percentiles were cut from the extremes to filter the sampled climatic data to remove potentially non-relevant values. It is worth noting that the resolution of the paleoclimate models used in this study is 2.5 arcminutes (approximately 5 km in mid-latitudes), which aligns with the accuracy of the site coordinates.

Supplementary tables 2–6 show the determined distribution-limiting climatic extrema of the studied tree species.

To compare the properties of four types of wood used by ancient humans in Clacton, UK, and Schöningen, Germany, during the Middle Pleistocene, the physical characteristics of 20 common European tree species and natural hybrids found in temperate and continental regions were examined. In Clacton, European yew (Taxus baccata) was used to craft the spear, while in Schöningen, spears were made from European spruce (Picea abies) and Scots pine (Pinus sylvestris) (Schoch et al. 2015). Additionally, European silver fir (Abies alba) branch bases are believed to have been used as hafting implements for stone artifacts (Bigga et al. 2015).

The physical properties considered included average dried weight, Janka hardness, modulus of rupture, elastic modulus, and crushing strength (Martins et al. 2022). Average dried weight represents the specific weight of dried wood, which can influence the usability and transportability of wood tools. Janka hardness is a measure of workability and durability for the wood. The modolus of rupture is employed for the prediction of material strength (Reeve 1965). Elastic modulus is associated with a material's resistance to bending (Cheng and Cheng 1998). Crushing strength signifies the highest compressive stress a solid material can withstand without breaking (Huang et al. 2020). To compare these physical variables, a principal component analysis (PCA) was conducted in Past326b statistics software (Hammer and Harper 2001). Before PCA, all values were normalized based on their respective variable types using the z-score normalization method. The data were sourced from the Wood Database (Meier 2015) (Table 3).

Table 3 The studied European tree species and the physical features of their woods

Software and statistics

Köppen-Geiger characterisation of the climatically analogous regions

The characterization of the climatically analogous regions was based on the same factors and equations as can be seen in the case of the past suitability value modelling of H. heidelbergensis. In this case, the range-limiting climatic values of H. heidelbergensis were displayed for the current climate, and then the Köppen-Geiger climatic class of those regions where the climatic suitability currently reaches the 89% value was presented. The Köppen-Geiger climatic class of these reference sites was classified based on the Köppen-Geiger map of Beck et al. (2018). The WorldClim database (Fick and Hijmans 2017) was the source of the current (1970–2000) bioclimatic values, and the mapped model of Beck et al. (2018) provided the present Köppen-Geiger climate classification maps at 1-km resolution.

In the next step, the K-Nearest-Neighbors (KNN) algorithm (Kramer 2013) was utilized to classify archaeological sites based on their feature values and visualize the results through a scatter plot. A supervised machine learning algorithm, KNN is commonly employed for classification tasks such as identifying the former Köppen climate types of archaeological sites based on reconstructed paleoclimatic data.

In this study, paleoclimatic data from archaeological sites served as features, with the Köppen climate types of 5912 European random sites designated as labels. Climatic layers were sampled with random points using the points sampling tool in Quantum GIS 3.4.4 (QGIS project 2020). The machine learning task was conducted in the Spyder version 5.2.2 environment. Nineteen climatic factors, encompassing all bioclimatic factors (Waltari et al. 2014) available in the WorldClim dataset, were considered.

The process involved the following steps:

• 1. Each archaeological site was represented as a data point in a multidimensional space, where each dimension corresponds to a specific paleoclimatic variable.

• 2. Before applying KNN, the paleoclimatic features were standardized to ensure that all variables contributed equally to the distance calculations. Standardization involved scaling the features to have a mean of 0 and a standard deviation of 1.

• 3. After that, the KNN model was created and trained on the standardized paleoclimatic data. During training, the algorithm memorized the relationship between the paleoclimatic features and their corresponding Köppen climate types.

• 4. Then, when presented with the reconstructed paleoclimatic data of a new archaeological site, the KNN algorithm identified the k-nearest neighbours of that site in the feature space. The "k" neighbours were the data points with the most similar paleoclimatic features.

• 5. The algorithm then looked at the Köppen climate types of these neighbours and assigned the most frequently occurring climate type (majority vote) as the predicted climate type for the new site.

• 6. Finally, the output of the KNN algorithm was the predicted Köppen climate type for each archaeological site. This prediction was based on the similarity of the site's paleoclimatic features to those of its nearest neighbours in the training dataset.

Model identifications

The climatic suitability pattern modelling of Homo heidelbergensis, five ligneous plants, and two large mammal species was performed. All modelling processes presented in this part of the study were conducted in Quantum GIS 3.4.4 (QGIS project 2020) and Grass GIS 7.4.1 software (Neteler and Mitasova 2002). The sampling tool and raster calculator functions were utilized. The Lambert Azimuthal Equal Area (EPSG:3035) was used as the projection system to present the model results.

Homo heidelbergensis and the five tree species

The upper and lower extremes of the 19 selected bioclimatic parameters were utilized to create numerical models for potential climatic suitability patterns. These values were produced as the minimum and maximum values of the bioclimatic variables and then used for suitability modelling purposes in the case of H. heidelbergensis and the five tree species. Georeferenced climatic suitability values were calculated for the time frame spanning from 670 to 190 thousand years ago (ka), with each period lasting 40 thousand years (e.g., 229–190 ka, 269–230 ka, and so on; see Table 1). The suitability values were generated by combining the individual maximum and minimum-determined potential distribution maps. The range of the maximum values for each specific bioclimatic factor is represented by a distribution function within the extremities.

The equation for the aggregated distribution area is as follows:

Deterministic unit-step functions for H. heidelbergensis and the four tree species can be expressed in the following way using bioclimatic factors:

$$1\left({bio}_{1-11}^T\right)=\left\{\begin{array}{c}0\;\;if\;\;{{bio}_{1-11}^T}_{upper\_limit}\leq{{bio}_{1-11}^T}_{min}\leq{{bio}_{1-11}^T}_{lower\_limit}\\1\;\;if\;\;{{bio}_{1-11}^T}_{upper\_limit}>{{bio}_{1-11}^T}_{min}>{{bio}_{1-11}^T}_{lower\_limit}\end{array}\right.$$

$$1\left({bio}_{12-19}^P\right)=\left\{\begin{array}{c}0\;\;if\;\;{{bio}_{12-19}^P}_{upper\_limit}\leq{{bio}_{12-19}^P}_{min}\leq{{bio}_{12-19}^P}_{lower\_limit}\\1\;\;if\;\;{{bio}_{12-19}^P}_{upper\_limit}>{{bio}_{12-19}^P}_{min}>{{bio}_{12-19}^P}_{lower\_limit}\end{array}\right.$$

where bio^T1-11 denotes the temperature-based bioclimatic variables' georeferenced climate model data. The upper and lower limiting factors (extrema) are indicated by the variables bio^T1-11_{upper_limit} and bio^T1-11_{lower_limit}, respectively; the georeferenced climate model data for the temperature-based bioclimatic variables is shown by bio^P12-19. The upper and lower limitation factors (extrema) are the bio^P12-19_{upper_limit} and bio^P12-19_{lower_limit}, as can be observed. The value of the limitation factor is independent of the climate model.

The equation of the aggregated distribution area is as follows:

$$A\left({bio}_{1-11}^{T};{bio}_{12-19}^{P}\right)={bio}_{1-11}^{T}-\sum_{i=1}^{11}0({bio}_{1-11}^{T})\bigcap {bio}_{12-19}^{P}-\sum_{i=1}^{8}0\left({bio}_{12-19}^{P}\right)$$

After taking into account the constraints based on temperature- and precipitation-based bioclimatic parameters, A(bio^T_1-11; bio^P_12-19) depicts the probable distribution region of H. heidelbergensis, which includes the remaining areas.

The distinct presence-absence maps were then aggregated, as in earlier instances. The suitability values were the proportion of satisfied climatic criteria in each grid. The distribution's upper bound was set at 95% appropriateness.

Red deer

According to Trájer's approach (Trájer 2022b), 6 × 2 factors were applied in the cases of Cervus elaphus. Red deer (Cervus elaphus) distribution limiting values were taken from the same source. The used distribution limiting factor values can be found in Supplementary Table 7.

In the case of the red deer, the equation of the aggregated distribution area is as follows:

$$1\left({bio}_{1,10,11}^T\right)=\left\{\begin{array}{c}0\;\;if\;\;{{bio}_{1,,10,,11}^T}_{upper\_limit}\leq{{bio}_{1,10,11}^T}_{min}\leq{{bio}_{1,10,11}^T}_{lower\_limit}\\1\;\;if\;\;{{bio}_{1,10,11}^T}_{upper\_limit}>{{bio}_{1,10,11}^T}_{min}>{{bio}_{1,10,11}^T}_{lower\_limit}\end{array}\right.$$

$$1\left({bio}_{,12,18,19}^P\right)=\left\{\begin{array}{c}0\;\;if\;\;{{bio}_{12,18,19}^P}_{upper\_limit}\leq{{bio}_{12,18,19}^P}_{min}\leq{{bio}_{12,18,19}^P}_{lower\_limit}\\1\;\;if\;\;{{bio}_{12,18,19}^P}_{upper\_limit}>{{bio}_{12,18,19}^P}_{min}>{{bio}_{12,18,19}^P}_{lower\_limit}\end{array}\right.$$

and then:

$$A\left({bio}_{1,10,11}^T;{bio}_{12,18,19}^P\right)={bio}_{1-11}^T-\sum\limits_{i=1}^30({bio}_{1,10,11}^T)\bigcap{bio}_{12,18,19}^P-\sum\limits_{i=1}^80\left({bio}_{12,18,19}^P\right)$$

After taking into account the limits brought on by temperature- and precipitation-based bioclimatic parameters, A(bio^T_1,10,11; bio^T_12,18,19) depicts the probable distribution region of Cervus elaphus, which includes the remaining areas.

Modelling of biomes

The approach of Trájer (2022a) served as the foundation for the biome modelling. The area-delimiting curves of the updated Whittaker biome plot were made by the author to rebuild the biome patterns (Stefan and Levin 2020; Deshmukh and Singh 2019). The equations of the upper or lower delimiting curves related to the biome areas were employed for modelling purposes. The chosen biomes were as follows: tundra and ice (Tu + Ic; including the Alpine) biomes, temperate seasonal forest (TeSeFo), boreal forest (BoFo), woodland or shrubland (Wo/Sh), and temperate grassland or cold desert (TeGr/CoDE). The temperate grassland/cold desert biome's upper delimiting curve can be calculated as follows:

$$-0.3498\times {Tm}^{2}+24.436\times Tm+159.92=Ps$$

The woodland/shrubland biome's upper delimiting curve has the following equation:

$$-0.0086\times {Tm}^{4}+0.2955\times {Tm}^{3}-2.1072\times {Tm}^{2}+41.12\times Tm+242.93=Ps$$

The temperate seasonal forest biome's upper delimiting curve has the following equation:

$$2235.23+(47.91482-2235.23)/(1+\left(\frac{Tm}{7.782309}\right)^{6.265971})=Ps\;and\;if\;Tm>-2.5^\circ C\;and\;Tm<22.5^\circ C$$

(6)

The following is the equation for the boreal forest biome's upper delimiting curve:

$$4.8254\times {Tm}^{3}-77.874\times {Tm}^{2}+519.17\times Tm+318.67=Ps$$

The tundra and ice biome's lower delimiting curve has the following equation:

$$4.8254\times {Tm}^{3}-77.874\times {Tm}^{2}+519.17\times Tm+318.67<Ps$$

where Tm is the annual mean temperature (°C), and Ps is the annual precipitation sum (mm).

Relationship between clothing and temperature

Based on the study of Wales (2012), the correlation between the percentage of the body surface covered by clothes in the coldest month and the mean temperature of the coldest month can be calculated according to the following equation:

$$C(\%)=\frac{0.5+{\text{arctan}}(1.24148-{T}_{mc}\times 0.19828)}{\pi }\times 100$$

where C(%) is the percentage of the body surface covered by clothes; T_mc is the mean temperature of the coldest month.

The RR₀₁ values were modelled for 670–190 ka. It means 479 time points, according to the δ¹⁸O data of Lisiecki and Raymo (2005). Then the modelled values were sampled according to the following delimiting coordinates: -10 and 30° W and 35 and 62° E.

The connection between the environment's temperature and the relative risk of mortality

Using data from 210 major Chinese cities. Chen et al. (2018) investigated the nature of the relationship between ambient temperature and relative mortality risk. The relationship between ambient temperature and relative mortality risk was established by the authors. They developed an equation to forecast the probable relative risk for the coldest quarter of the year based on their data and a quadratic method.

$${RR}_{CQ}=1.546526-0.0769347\times {T}_{CQ}^{1}+0.007552273\times {T}_{CQ}^{2}-0.0003916162\times {T}_{CQ}^{3}+0.000007121212\times {T}_{CQ}^{4}$$

where RR_CQ is the relative January risk of mortality and T_CQ is the mean ambient temperature of the coldest quarter.

The RR₀₁ values were modelled for 670–190 ka. It means 479 time points, according to the δ¹⁸O data of Lisiecki and Raymo (2005). Then the modelled values were sampled according to the following delimiting coordinates: -10 and 30° W and 35 and 62° E.

Daily energy expenditure

Ambient thermal conditions have a significant impact on work performance and productivity. affecting both mental and physical aspects (Li et al. 2016; Kjellstrom et al. 2009). According to Lan et al. (2010), air temperature also influences employees' productivity, workload, and willingness to work. It is well-known that the total daily energy expenditure is determined by the basal metabolic rate and the level of physical activity unique to each gender (Churchill 2009):

$$DEE=BMR\times PAL$$

where PAL is the degree of physical activity unique to a given gender, DEE is the daily energy expenditure, and BMR is the basal metabolic rate, both values are in cal d⁻¹.

In the case of intensive work, the PAL is at least 2.09 for men and 1.81 for women (WHO 1985). For the following calculations, these minimal values connected to the criterion of heavy physical work were used.

The possible body mass of males and females was based on the study of Will et al. (2017), who found that the average weight of Middle Pleistocene humans, including H. heidelbergensis was 68.75 kg. To model the weight of males and females, the Anthropometric Reference Data for Children and Adults of the United States for each race in 2015–2018 was used (Fryar et al. 2021). Based on this data, the weights in kilograms for adult males and females aged 20 years and over are 90.6 kg and 77.5 kg, respectively, which means that the weight ratios compared to the mean weights are 1.00779 for males and 0.92207 for females. Applying these ratios to the average weight of 68.75 kg, the weights of the males and females of H. heidelbergensis were 74.11 and 63.39 kg, respectively.

For males, the formula for calculating the gender-specific basal metabolic rate is as follows:

$${BMR}_{m}=\left(14.7\times 74.11\, {\text{kg}}\right)-\left(5.6\times {T}_{m}\right)+735$$

and in the case of females:

$${BMR}_{f}=\left(9.2\times 63.39\, {\text{kg}}\right)-\left(3.8\times {T}_{m}\right)+852$$

where BMR_m and BM_Rf are the basal metabolic rates for men and women under ambient thermal circumstances, respectively, in cal per day, and Tm is the mean temperature in degrees Celsius.

Therefore, about 2.5–3.0 kcal would be present in 1 kg of lean beef (Calories.info n.d). Finally, using an approximation of the following equation and calculating with the higher energy content value, the computed BMR_m values were converted to lean meat equivalent (lme) values:

$$lme\cong DEE\times 0.33 kg{cal}^{-1}$$

where lme is lean beef equivalent in kg, DEE is the is the daily energy expenditure in cal.

The RR₀₁ values were modelled for 670–190 ka. It means 479 time points, according to the δ¹⁸O data of Lisiecki and Raymo (2005). Then the modelled values were sampled according to the following delimiting coordinates: -10 and 30° W and 35 and 62° E.

Results

Paleoclimatic characterisation

Based on the results of the KNN classification, it is evident that during the studied period, most of the H. heidelbergensis archaeological sites featured temperate or humid continental climates. Specifically, the Visogliano site exhibited a humid subtropical (Köppen-Geiger Cfa) climate. The Casal de Pazzi, Clacton, Grotte Vaufrey (based on level X), Maastricht-Belvedere (based on unit IV), Montmaurin, Notarchirico di Venosa (based on the lower archaeological stratum), Pontnewydd Cave, Sedia del Diavolo, and Swanscombe sites fall into the oceanic (Köppen-Geiger Cfb) climate category. Castel di Guido and Isernia la Pineta had a dry and hot summer temperate (Mediterranean, Köppen-Geiger Csa) climate. Aroeira, Ceprano, Fontana Ranuccio, Pofi, and Ponte Mammolo had a dry and warm summer (Köppen-Geiger Csb) climate. Arago, Atapuerca, Boxgrove, Hoxne, Mauer, Schöningen, and Vértesszőlős had a cold winter-warm summer humid continental (Köppen-Geiger Dfb) climate. Bilzingsleben's climate was cold summer and cold winter continental (Köppen-Geiger Dfc). Cueva de Bolomor and Notarchirico di Venosa (based on the upper archaeological stratum) fall under the cold arid steppe (BSk) climate (Fig. 2a).

Fig. 2

The Köppen-Geiger climatic class results of the KNN of the reconstructed bioclimatic values of 27 Homo heidelbergensis-bearing archaeological sites based on the climatic values of 5911 random sites in Europe (a) and the bart chart of the results (b). Köppen-Geiger climatic classes in the climatically absolutely analogous regions considering the distribution-limiting climatic extrema of H. heidelbergensis in present-day Europe (c). 1: Arago, 2: Aroeira, 3: Atapuerca, 4: Bilzingsleben, 5: Boxgrove, 6: Casal de’ Pazzi, 7: Castel di Guido, 8: Ceprano, 9: Clacton, 10: Cueva de Bolomor level XIV, 11: Fontana Ranuccio, 12: Grotte Vaufrey level X, 13: Hoxne, 14: Isernia la Pineta, 15: Maastricht-Belvedere unit IV, 16: Mauer, 17: Montmaurin, 18: Notarchirico di Venosa lower lower stratum, 19: Notarchirico di Venosa upper stratum, 20: Pofi, 21: Ponte Mammolo, 22: Pontnewydd, 23: Schöningen, 24: Sedia del Diavolo, 25: Swanscombe, 26: Visogliano, 27: Vértesszőlős. Köppen-Geiger climatic classes are: BSh: steppe climate. Cfa: humid subtropical. Cfc: oceanic. Csa: hot summer Mediterranean climate. Dfa: hot summer humid continental climate without dry season. Dfb: warm summer humid continental climate without dry season. Dfc: boreal climate. Dsa: hot and dry summer humid continental climate. Dsb: warm and dry summer humid continental climate. ET: tundra climate

The results of the KNN classification show that the oceanic climate (Cfb) was the most frequent (n = 11) Köppen-Geiger climatic class among the studied sites. After that, the cold winter humid continental climate (Dfb) is the second most frequent (n = 8) climate type. With 4 sites, the dry and warm summer temperate (Csb) is the third most frequent climate type (Fig. 2b).

Considering the reconstructed distribution limits of 26 H. heidelbergensis-related sites in present-day Europe, the climatic range of this ancient human species extended from the Mediterranean (Köppen-Geiger Csa and Csb) to the southern boreal regions. Nevertheless, oceanic Köppen-Geiger Cfc and warm summer humid continental Köppen-Geiger Dfb climates dominate the climatically analogous regions today in Europe. Hot summer Mediterranean (Köppen-Geiger Csa, Csb), humid subtropical (Köppen-Geiger Cfa), and boreal (Dfc) climates played a less prominent role. The Köppen-Geiger BSk (steppe), Dfa (hot summer humid continental climate without a dry season), and Dsa (hot and summer humid continental) climates were only found in small areas. In conclusion, H. heidelbergensis primarily inhabited oceanic and humid temperate regions in Europe (Fig. 2c).

The cluster-heat map diagram shows the palaeoclimatic similarity of the studied sites. It indicates that the Cueva de Bolomor (level XIV), Aroeira, and Castel di Guido sites had a distinct paleoclimate compared to the others. However, at the same time, the climate of these sites was similar to each other. Then, the paleoclimates of Clacton, the upper stratum of Notarchirico di Venosa, and Swanscombe belong to the same cluster. However, at the same time, the climate of these sites was similar to each other. The remaining 23 site-time pairs belong to two major groups:

Atapuerca, Arago, Casal de’ Pazzi, Ceprano, Fontana Ranuccio, Isernia la Pineta, Notarchirico di Venosa (lower stratum), Pofi, Ponte Mammolo, and Sedia del Diavolo form the first group, and Bilzingsleben, Boxgrove, Clacton, Grotte Vaufrey level X, Hoxne, Mala Balanica, Maastricht-Belvedere (unit IV), Mauer, Montmaurin, Pontnewydd, Schöningen, Visiogliano, and Vértesszőlős belong to the second one. Based on the results of the PCA, the last two main clads containing most of the sites had milder, subtropical-like paleoclimate in the time of the existence of ancient humans, and the second one had cooler, oceanic (or humid continental)-like paleoclimatic conditions (Fig. 3).

Fig. 3

The cluster-heat map diagram shows the reconstructed palaeoclimatic similarity of the studied archaeological sites

The modelled suitability patterns by periods

Homo heidelbergensis

Based on the reconstructed climate of these sites, the distribution model indicates that ancient humans could have inhabited various regions of south, central, and western Europe between 670 and 190 ka. Nevertheless, the distribution patterns suggest significant past habitat contractions and expansions in response to climatic changes. Homo heidelbergensis likely did not occupy the northeastern regions of the European Plain. The northernmost extent of H. heidelbergensis-like humans in western Europe likely reached the Atlantic coasts of Lower Saxony, Germany, on the European mainland. Generally, Homo heidelbergensis did not extend beyond the northern forehills of the Sudetes and the Carpathian Ranges. However, some occurrences, like the archaeological site in Tunel Wielki, Southern Poland, indicate that this archaic human species had advanced adaptation strategies to the living conditions, which made it possible to inhabit the colder continental regions of Europe (Kot et al. 2022). These observations suggest that the potential distribution of H. heidelbergensis was constrained to the north by continental climatic conditions with harsh winters. Using 95% of climatic suitability limit as an approximate boundary for the species' distribution, it can be inferred that during warmer and wetter periods (229–190 ka, 349–310 ka, 429–390 ka, 509–470 ka, 589–550 ka, and 629–590 ka), when the potential habitat for the species was at its highest, nearly continuous habitable areas may have existed for ancient humans between the Bosporus and Western Europe. These areas could have served as migration routes between Asia Minor and the Atlantic coasts. In contrast, during the cold and dry periods of 469–430 ka and 670–630 ka, there may have been significant contractions in the distribution of H. heidelbergensis in Central Europe (Fig. 4).

Fig. 4

The modelled H. heidelbergensis suitability patterns (a: 229–190 ka. b: 269–230 ka. c: 309–270 ka. d: 349–310 ka. e: 389–350 ka. f: 429–390 ka. g: 469–430 ka. h: 509–470 ka. i: 549–510 ka. j: 589–550 ka. k: 629–590 ka. l: 670–630 ka). The Middle Pleistocene sea-level changes that caused coastline alterations were not modelled

European beech

The modelled climatic suitability ranges of F. sylvatica, an indicator species of European temperate seasonal forests in humid oceanic and continental climate regions, exhibit notable similarities to the previously presented model results concerning H. heidelbergensis. European beech was generally present in western Europe, including the territories of present-day France, the Benelux States, the Apennine Peninsula, the Northern Balkans, the Cantabrian Mountain Ranges, and the Pyrenees. However, the suitability patterns exhibit significant fluctuations across different time periods. During warmer periods, such as 230–190 ka or 429–390 ka, this tree species likely also extended into the southern part of present-day Great Britain. Central Europe and Eastern Europe. In contrast, during colder periods like 469–430 ka or 670–630 ka, the European beech was absent in a substantial portion of the Iberian Peninsula, Eastern Europe, Scandinavia, the northern part of Great Britain, and Ireland (Fig. 5).

Fig. 5

The modelled suitability patterns of European beech (Fagus sylvatica) (a: 229–190 ka. b: 269–230 ka. c: 309–270 ka. d: 349–310 ka. e: 389–350 ka. f: 429–390 ka. g: 469–430 ka. h: 509–470 ka. i: 549–510 ka. j: 589–550 ka. k: 629–590 ka. l: 670–630 ka). The Middle Pleistocene sea-level changes that caused coastline alterations were not modelled

Red deer

Like the range of the European beech, the distribution of the red deer showed notable fluctuations in the studied era. The Iberian and Apennine Peninsulas, as well as the coastal regions of the Balkan Peninsula, generally exhibit high climatic suitability values for C. elaphus, one of the most important common mammal prey species in most of the periods. The climatic suitability values of central Europe are relatively high in most cases. suggesting the highest potential range in the period 429–390 ka. In contrast, in 469–430 and 670–630 ka, red deer could lack in west and central Europe. The south-eastern part of present-day Great Britain could be suitable for C. elaphus only in the warmest periods (Fig. 6).

Fig. 6

The modelled red deer (Cervus elaphus) suitability patterns (a: 229–190 ka. b: 269–230 ka. c: 309–270 ka. d: 349–310 ka. e: 389–350 ka. f: 429–390 ka. g: 469–430 ka. h: 509–470 ka. i: 549–510 ka. j: 589–550 ka. k: 629–590 ka. l: 670–630 ka). The Middle Pleistocene sea-level changes that caused coastline alterations were not modelled

Results related to wood usage

In the period from 337 to 300 ka, A. alba exhibited the highest climatic suitability in central Europe, the Benelux States, the western part of present-day France, the Northern Balkan region, the Pyrenees, and along the Apennine Mountains. Schöningen showed relatively high suitability values. suggesting that A. alba may have been present in the narrower region of the site but likely wasn't the dominant forest-forming tree species. Except for the Pyrenees, A. alba was absent in the Iberian Peninsula, the British Isles, the Scandinavian Peninsula, and the southern Balkans (Fig. 7a). According to the model's high climatic suitability values, Picea abies was likely present from 337 to 300 ka in Scandinavia, Central Europe, the northern regions of Eastern Europe, and the higher mountainous regions of the peri-Mediterranean area. It may have been absent in the British Isles and significant portions of southern and western Europe. During the occupation of Schöningen by ancient people, this pine could have played a role as a forest-forming species (Fig. 7b). Pinus sylvestris may have been present in large regions of Europe, including the British Isles, western and central Europe, and the mid- to higher regions of southern Europe, except for the southern Iberian Peninsula (Fig. 7c). Around 400 ka, the climate of Europe, excluding its southernmost regions, would have been suitable for Taxus baccata, including the British Isles (Fig. 7d).

Fig. 7

The climatic suitability patterns of Abies alba (a), Picea abies (b), Pinus sylvestris (c) and Taxus baccata (d) in Europe in the time of the making spears or hafting implements found in Clacton (a-c) and Schöningen (d)

The modelled biomes at the time when the Clacton spear was used exhibit warmer climatic conditions compared to those observed in the time span of Schöningen. When the Clacton spear was crafted, most of Europe was covered by woodland, shrubland, and temperate seasonal forest biomes (Fig. 8a). In contrast, during the Schöningen spear-making period, temperate seasonal forests dominated the northern and central regions, while woodland/shrubland biomes prevailed in the southern and western regions of Europe (Fig. 8b). During the ancient human occupation of Clacton, the dominant biome was woodland/shrubland, whereas during the Schöningen occupation, the territory bordered temperate seasonal and boreal forests, as well as boreal forest-covered regions.

Fig. 8

The modelled biomes in Europe at the time of the making of spears found in Clacton (a) and Schöningen (b)

The PCA result indicates that based on the features of its wood, Ab. alba and Pic. abies are nested with relatively light and less resistant to physical effects woods like Po. tremula or Po. nigra. Pinus sylvestris has a harder and heavier wood than Ab. alba and Pic. abies, and it is like the wood of Al. glutinosa. In contrast, the wood of Ta. baccata has a relatively heavy and strong structure (Fig. 9).

Fig. 9

PCA of characteristic temperate climate-dwelling woods native to Europe. A: average dried weight, B: Janka hardness, C: modulus of rupture, D: elastic modulus, E: crushing strength. 1: Abies alba, 2: Acer platanoides, 3: Acer pseudoplatanus, 4: Alnus glutinosa, 5: Carpinus betulus, 6: Fagus sylvatica, 7: Fraxinus excelsior, 8: Larix decidua, 9: Malus sylvestris, 10: Picea abies, 11: Pinus sylvestris, 12: Populus tremula, 13: Populus nigra, 14: Prunus avium, 15: Pyrus communis, 16: Sorbus aucuparia, S, domestica, S, torminalis, 17: Salix alba, 18: Salix x fragilis, 19: Taxus baccata, 20: Ulmus procera

Temporal alterations of habitability factors

Body surface covered by clothes

In parallel with long-term temperature fluctuations, the modelled mean values of relative mortality risk in Europe during the coldest quarter exhibited significant variations from 670 to 190 thousand years ago (ka). During the periods of 471–418, 648–636 ka, and 554 ka, the average mortality risk values exceeded the 6.0 relative risk threshold multiple times. Similarly, during the periods of 266–247, 357–343, 225–222 ka, and 299 ka, the average relative mortality risk values for the coldest quarter both reached and exceeded the 5.0 relative risk mark. In contrast, during periods such as 418–400 and 325–324 ka, this value dipped below 1.5, indicating more favourable climatic conditions. Interestingly, advantageous climatic conditions also prevailed during the coldest quarter around 619, 596, 578, 495, 405, 331, 242, and 222–197 ka (Fig. 10a).

Fig. 10

The modelled mean values (blue points) of the relative mortality risk (a) and the percentage of body covered by clothes in the coldest quarter (b) in Europe (delimiting coordinates are -10 and 30° W and 35 and 62° E) Black lines are the 10-period moving average values in 670–190 ka

The percentage of the (presumed) body covered by clothing exhibited high, more than 82.5% values, as can be seen, e.g., in the periods of 672–623, 565–529, 481–416, 371–341, 303–296, 276–248, 235–222 ka. Very low average values in Europe occurred during the periods of 617, 577, 418–402, 325–320 and 197 ka. During these times, it was advisable for ancient humans to cover less than 70% or more of their bodies with clothing during European winters on average (Fig. 10b).

The mean daily energy expenditure in the coldest quarter in Europe exceeded the 3890 kcal/day in males and the 3090 kcal/day values in females in 668–628, 552–548, 471–416, 357–340, 270–263; 253–245; and 224–222 ka (Fig. 11a, b). In contrast, in 617, 418, 405, 401, 400, 325, 318, and 194 ka, the mean daily energy expenditure values averagely were less than 3795 kcal/day values in males and 2345 kcal/day values in females (Fig. 11a, b).

Fig. 11

The modelled mean values (blue points) of the daily energy expenditure values in males (a) and females (b) in Europe (delimiting coordinates are -10 and 30° W and 35 and 62° E) Black lines are the 10-period moving average values in 670–190 ka

In the case of the studied H. heidelbergensis fossil sites, the mean relative mortality risk in the coldest period was relatively low. 1.42 (SD: 0.23). The mean value of the percentage of body covered by clothes was 64.2% (SD: 15.4%). The daily energy expenditure was 3776 cal/day (SD: 42.4) in the case of males and 2576 cal/day (SD: 24.9 cal/day) in the case of females; those values are about equal to 1.13–0.77 kg of lean meat per day (SD: 0.01 kg). The reconstructed values do not exhibit notable differences between the sites (Table 4).

Table 4 The relative mortality risk in the coldest quarter (Rm_CQ), the percentage of the body covered by cloths (BCCQ), the daily energy expenditure values of males (mDEE_CQ) and females (fDEE_CQ), and lmbe values are the lean meat equivalents of the DEE values in the case of males and females

Considering the lower thermal distribution limit related to the coldest quarter, it can be observed that the percentage of the area within the studied European frame (-10° to 30° W and 35° to 62° E) exhibited dramatic changes over the study period. During the coldest episodes, such as 632 and 436 ka, the proportion of areas with the coldest quarter above -5 °C barely exceeded 40% within the study area. In contrast, during certain warm periods, like 405 and 325 ka, this ratio exceeded 95%. In some periods, the expansion of habitable regions, based on the temperature threshold of the coldest quarter, occurred rapidly. For instance, between 436 and 425 ka, following the -5 °C distribution limit associated with H. heidelbergensis related to the coldest quarter, the growth rate of habitable land per thousand years reached a value of 1.6% (17.6% over 11 kys). Conversely, in other periods, the decline in habitable regions could be swift, such as the period from 400 to 393 ka, when this rate was 3.88% (27.3% over 7 kys) of the total area per thousand years (see Fig. 12).

Fig. 12

The modelled percentage of the total area where the mean temperature of the coldest quarter reached or overwhelmed the related distribution limiting value (-6.2 °C) of H. heidelbergensis populations in Europe (delimiting coordinates are -10 and 30° W and 35 and 62° E) Black lines are the 10-period moving average values in 670–190 ka

Discussion

The results suggest that although the environmental range of H. heidelbergensis extended from the Mediterranean to milder boreal climatic conditions, the species preferred temperate climatic conditions. Generally, their habitats had a temperate (Köppen-Geiger Cfa, Cfb) character. Additionally, H. heidelbergensis also occurred in the forest steppe biome, but in Europe, it avoided the definitive steppe biome. It is also true for the typical boreal biome. The reconstructed paleoclimates, predominantly oceanic and humid subtropical-like, as well as climatically analogous ones, align with the archaeozoological record. For example, in the Vértesszőlős archaeological sites, the remains of Bison schoetensacki, an extinct woodland bison, and C. elaphus, the red deer (Jánossy 1990), which is typical of woodlands and temperate seasonal forests in Europe (Mania et al. 1992), indicate that H. heidelbergensis at this site occupied a forested biome. The palaeofauna of Mauer, the type locality of H. heidelbergensis, consists of a fully interglacial fauna with typically temperate seasonal forest and woodland-occupying mammal taxa like S. scrofa and Capreolus capreolus priscus (von Koenigswald 2007). Interestingly, Hippopotamus was also present at this site, indicating an important presence of water and humid conditions but not necessarily a warm climate, as this genus was widely distributed in the Early and Middle Pleistocene of Europe, excluding the northern and easternmost regions (Fidalgo et al. 2023). The initial dispersal of Hippopotamus in Europe was associated with humid and possibly warm climatic conditions in the late Gelasian (Iannucci et al. 2023). It is known that lakes fed by hot springs have also created the necessary living conditions for hippopotamus species, as found, for example, in Üröm Mt., Central Hungary (Jánossy 1986). The observation that Mid- and Late Pleistocene hippopotamus species tended to be relatively larger during warmer and somewhat more humid intervals but did not grow as large during colder and comparatively drier times (Mazza and Bertini 2012) indicates that these European taxa were not necessarily species with subtropical climate requirements. In other cases, the presence of megaherbivores of tropical-subtropical origin in northern regions, such as the presence of hippos in the British Early Pleistocene fossil record, can be explained by short unrecognized humid and warm climatic periods (Adams et al. 2022).

The H. heidelbergensis-bearing levels of Visogliano yielded remains of Dama clactoniana, Capreolus capreolus, Apodemus sylvaticus, and Apodemus flavicollis, among other species (López-García et al. 2021), which align with the proposed oceanic climate reconstruction of this study. Although similar current climatic analogues were not found in some cases, this is not surprising because present-day Europe is not an exact analogue of the past. In these cases, it can be assumed that the sites had relatively warm and humid or semi-humid paleoclimatic conditions. For example, in Aroeira, H. heidelbergensis co-occurred with Dama sp., Cervus sp., and equid species, indicating a former open-woodland landscape based on temperate conditions (Daura et al. 2017). However, these assumptions can be held with caution since many Middle Pleistocene large mammal species display wide ecological tolerance and plastic feeding behaviours (Rocca et al. 2023). These uncertainties highlight the importance of studies such as the current one, which aim to determine the former climatic types through paleoclimatic reconstructions.

It was found that the range of the European beech roughly coincided with the modelled range of H. heidelbergensis and C. elephus. The co-occurrence of these species indicates a typical biogeographical zone in present-day Europe characterized by the dominance of deciduous trees, which form temperate seasonal forests and woodlands. These forests are home to edible fungi species such as the summer truffle (Tuber aestivum) (Moser et al. 2017), golden chanterelle mushroom (Cantharellus cibarius) (Maciejczyk et al. 2015), black chanterelle (Craterellus cornucopioides) (Bulam et al. 2018), and coral tooth fungus (Hericium coralloides) (Boddy et al. 2011). Pine and acid-soil beech forests are also known to be rich in edible fungi, including species like the pine bolete (Boletus pinophilus) (Benedek and Pál-Fám 2006). Scots pine is commonly associated with edible fungi like saffron milk cap (Lactarius deliciosus) (Guerin-Laguette et al. 2003), dirty tricholoma (Tricholoma myomyces seu terreum) (Harrington and Mitchell 2002), and cauliflower fungus (Sparassis crispa) (Sharma et al. 2022). This observation suggests that the environmental exploitation by Middle Pleistocene ancient humans in western and central Europe may not have been limited solely to activities such as hunting mammals, collecting fruits, and using plants. Edible fungi could have also contributed to human nutrition. While we currently lack direct evidence for the consumption of fungi by H. heidelbergensis groups, similar hypotheses have been proposed regarding Neanderthals (Power et al. 2018; Diaz-Villaquiran 2019).

As shown, typical temperate-boreal region-dwelling pine species in Europe, such as the European spruce or the European silver fir, co-occurred with deciduous tree species in the humid continental regions. For example, in present-day Germany, we find coexistence between the European beech and the European spruce (Schönfelder 1999). In other words, the presence of these pine species in Middle Pleistocene archaeological sites in Europe does not necessarily indicate boreal conditions. The results of the PCA of the wood from typical European tree species suggest that H. heidelbergensis favoured specific wood types for toolmaking. These preferred species were gymnosperms, and, with the exception of the European yew, they had relatively light but not particularly hard woods. This might indicate the following considerations for ancient humans: 1) Softer wood species were preferred due to the ease of processing, and 2) it required less effort to transport lighter trees to the hunting site and, if possible, back home. Notably, in Clacton, pine wood was also used, despite it being relatively hard to work. This implies that ancient humans, when selecting wood for toolmaking, were likely influenced more by geographical proximity than a specific recognition of the value of pines and their wood. Even when lighter and softer angiosperm woods were available, the choice of pines may have been associated with their regional presence. Given that distribution models and modelled biomes indicate the occurrence of these pine species on the outskirts of Middle Pleistocene western and central Europe, it can be inferred that the use of these trees for toolmaking may have been widespread across a significant part of the continent.

As shown here, there were significant fluctuations in both the average relative mortality risk, the percentage of the body covered by clothing, and the daily energy expenditure values during the Middle Pleistocene in Europe. However, these values, when calculated for the 23 studied H. heidelbergensis sites, do not exhibit noteworthy differences. This paradoxical situation plausibly indicates that we possess human remains from those periods when and where the population of H. heidelbergensis was relatively high. This assumption aligns with the general principles of palaeontology, including a fundamental concept known as the Signor–Lipps effect (Signor et al. 1982). According to this concept, observing the exact times of species origination and extinction in the fossil record is challenging because, in the case of extremely rare vertebrates like hominins, we only have fossil data from periods when these species were abundant (Rosas et al. 2022). In summary, the evidence presented supports the ‘accretion model’, suggesting that H. heidelbergensis was susceptible to climatic fluctuations, especially rapid glacial cooling events. These climate disruptions could have induced bottleneck effects, fostering the gradual development of surviving populations (Dean et al. 1998). Raia et al. (2020) and Timmermann et al. (2022) have proposed a similar mechanism, attributing it to the evolution of various Homo species, including the emergence of Neanderthals from late H. heidelbergensis groups in western Eurasia. Based on the findings of the present study, it can be hypothesized that at least five major habitat constriction-expansion events occurred between 670 and 190 thousand years ago, corresponding to glacial-interglacial phases.

The decline of deciduous forests could have been a significant contributing factor to the decline of H. heidelbergensis populations. This is because glacial episodes not only led to a southward shift in biomes but also, due to the overall drier global climatic conditions, reduced the habitability of the northern hemisphere for deciduous forests. For instance, during the Last Glacial Maximum (LGM), deciduous forests almost entirely disappeared from large regions of Europe. Instead, graminoid vegetation and forb tundra, along with cold evergreen needle-leaf forests in humid and milder areas, covered West and Central Europe (Binney et al. 2017). Considering that the δ¹⁸O values, for example, 636 and 436 thousand years ago, were as high (exceeding 4.90) as during the LGM (Lisiecki and Raymo 2005), it can be hypothesized that these cold conditions resulted in a similar disappearance of deciduous forests in most parts of Europe, as occurred during the LGM.

The environmental changes mentioned above, coupled with the relatively northern habitat of H. heidelbergensis groups, may have left genetic traces on their genome. As was shown, H. heidelbergensis inhabited mainly humid subtropical and oceanic climate regions in Europe. It is an important finding since it is known that these regions receive a notable part of the annual rainfall in autumn and winter, which is generally associated with frontal rains, heavy cloud cover, and, in the oceanic regions, frequent fog in Europe. However, sunlight has several important positive impacts on human health, and the genetic regulation of skin colour varies according to the latitude in native populations, parallel to the annual mean solar radiation conditions (Diamond 2005). The MC1R gene (Melanocortin 1 receptor) is known to be responsible for skin pigmentation in humans. Primates with thick, pigmented fur do not require protection against harmful ultraviolet radiation, resulting in their pale skin colour (Bradley and Mundy 2008). In contrast, humans lacking body fur in open environments like savannas, under the strong equatorial sun, require extensive skin pigmentation.

Studies have shown that a specific gene variant associated with dark skin pigmentation in sub-Saharan human populations originated about 1.2 million years ago (Rogers et al. 2004), predating the appearance of H. heidelbergensis (Boyle and Wood 2020). This suggests that the ancestors of advanced Homo species, which lost most of their protective body hair (Weasel 2022), may have had reduced body fur, which did not provide sufficient protection against the cool winters of Pleistocene Europe. In addition, ultraviolet radiation dose follows a characteristic annual run in the middle latitudes of the Northern Hemisphere, reaching its minimum in winter, and there is a reverse correlation between mortality risk caused by mainly respiratory diseases. For example, in the US, about 25% higher death rates occur in the winter than in the summer (Grant et al. 2017). On the other hand, solar ultraviolet-B irradiance also reduces the risk of septicaemia (Grant 2009), which may be important in the case of a population living in nature, where individuals are constantly exposed to injuries during hunting and gathering. Since solar ultraviolet-B doses decrease to vanishingly negligibly low values at higher latitudes in winter (Grant and Boucher 2022), light skin could be an adaptive trait among ancient Europeans.

It is important to consider which Neanderthal-origin genetic traits may have originated in H. heidelbergensis. For example, the SLC16A11 ethnic-specific risk haplotype and its role in the development of diabetes mellitus are Neanderthal-origin introgressed variants found in modern humans. Given that northern populations of H. heidelbergensis lived in oceanic and humid continental regions where the risk of starvation could be high in winter, genetic variations leading to increased fat storage could have been advantageous for survival (Aguilar-Salinas and Tusie Luna 2023). As was shown, in the glacial periods, the average daily energy expenditure values increased in both sexes, which supports the hypotheses that glacial phases could lead to the increased adaptive value of such alleles that are associated with diabetes mellitus in modern populations. Furthermore, because the cluster-heat map diagram results indicate that most of the archaeological sites in the time of the existence of the ancient humans belonged to an oceanic-humid continental and the other one was to a humid subtropical cluster, it can be assumed that the annual food supply could not be the same in the two areas and that different food acquisition strategies were necessary for survival, especially in the winter half-year.

Neanderthal introgression, including the melanocyte-stimulating hormone receptor gene MC1R (Ding et al. 2014), may also be related to the Middle Palaeolithic habitat of H. heidelbergensis. Although this study does not directly address solar radiation, it is known that in oceanic and humid continental regions, cloud cover is high (Chiacchio and Vitolo 2012), particularly from October to March, when covering a large area of the body with clothing due to cold weather was necessary, especially in central and eastern Europe. This practice could have been disadvantageous for cholecalciferol (vitamin D) production, which is essential for maintaining healthy bone structure and immunity against respiratory infections (Meza-Meza et al. 2022; Saeed et al. 2022). However, it was not only the cloudiness of the cold season, or the shortness of the days related to the relatively high latitude that led to the reduction of the UV dose. As shown, H. heidelbergensis populations mainly inhabited temperate seasonal forests and woodland/shrubland territories, characterized by biotopes mainly formed by beech trees or mixed beech-conifer forests in colder regions. This is precisely characterized by the fact that the canopy closes at a high level, allowing only a very small amount of sunlight to reach the ground surface in summer. For example, in the Slovenian mixed silver fir-beech forest, solar radiation reaching the surface can be reduced by up to 95% (Rozenbergar et al. 2011). Since, according to the results, H. heidelbergensis populations lived in forested areas in Europe, this alone can explain the development of light pigmentation.

Moreover, it appears that modern humans inherited immunomodulatory genes from Neanderthals, which helped them cope with diseases prevalent in colder Eurasian regions. Interestingly, these genes may increase the risk of depression in humans (Nowak and Pawełczyk 2023). The long and wet winter half year in Western Europe within the modelled range of H. heidelbergensis, and the possibility of adaptation to the coldest phase of the year through hibernation (Bartsiokas and Arsuaga 2020), suggest that such genetic traits could have had a role in modulating behaviour. Winter depression, common in contemporary European populations, particularly among women and younger individuals (Dam et al. 1998), may have reduced the mental and physical activity of archaic people, potentially helping them avoid exposure to harsher winter conditions and reducing food demand due to decreased activity.

Conclusions

The findings suggest the following conclusions: 1) H. heidelbergensis primarily inhabited oceanic and humid continental regions in Europe, where specific plant and mammal fauna were prevalent; 2) In most parts of Europe, H. heidelbergensis was associated with deciduous forests and did not show a preference for boreal or steppe climate regions; 3) They displayed adaptability by making selective use of available wood materials; 4) H. heidelbergensis populations were susceptible to glacial episodes, resulting in increased winter mortality, food shortages, and the decline of deciduous forest ecosystems.