A spatiotemporally explicit paleoenvironmental framework for the Middle Stone Age of eastern Africa

Middle and Late Pleistocene climates of MSA occupations

We first examined MSA occupations (n = 84, Fig. 2) spanning the Middle to Late Pleistocene using simulated climate data (see Methods). We extracted mean annual temperature (bio01) and total annual precipitation (bio12) values from the climate model³³ within a 50 km radii, centred on the occupation’s mid-age date range rounded to the nearest 1000 year (kyr) time slice, to characterise environments across the wider logistical landscape (following Blinkhorn and Grove^10,11). The climatic conditions for each occupation can be found in Supplementary Table S1 and are illustrated in Fig. 1.

We found that average temperatures at eastern African MSA occupations varied between 9 °C and 25 °C, with 59 occupations falling within the 68% confidence interval of 14–23 °C. The warmest environments occupied were found in coastal regions, such as Abdur along the Red Sea coast of modern-day Eritrea (25 °C) and Panga Ya Saidi situated on the Kenyan coast (24 °C), as well as in the Lower Omo Valley of southwestern Ethiopia (24–23 °C). These hot environments were inhabited during MIS 5 and MIS 7. On the other hand, the coldest environments inhabited were at high altitude, at Fincha Habera in the Bale Mountains of southern Ethiopia (9–10 °C) and at Kenyan Rift Valley occupations of Marmonet Drift (10–14 °C) and Enkapune ya Muto (13 °C), most of which date to MIS 3. Average precipitation levels experienced by Middle to Late Pleistocene MSA populations in eastern Africa ranged between 396 and 1593 mm, with 59 occupations falling within the 68% confidence interval of 620-1150 mm, corresponding to the precipitation bracket of sub-humid landscapes. The wettest habitats were located on islands within and along the shore of Lake Victoria at Rusinga Nyamtia (1593 mm) and Karungu (1374-1499 mm) in MIS 3 and 5, as well as within the Ethiopian Rift Valley at Gademotta (1368 mm), the Ethiopian Highlands at Mochena Borango (1270-1297 mm), and the Kenyan Rift Valley at Marmonet Drift (1173-1368 mm) in MIS 3, 5 and 7. On the other hand, the driest occupations occurred at Laas Geel in Somaliland during MIS 3 (396 mm) as well as within the Lower Omo Valley (534-582 mm) during MIS 5 and 7.

Classifying biomes and ecotones at MSA occupations

We then used the modelled biome dataset (biome4output)³³ to classify the local ecology of each MSA occupation within a 50 km radius. We found that 38% of the occupations (n = 32) had access to only tropical xerophytic shrubland within their logistical landscape (see Fig. 3. for modern examples of this biome), and a further 42% with this biome among others within a 50 km radius (n = 35). Tropical xerophytic shrubland was persistently occupied throughout the Middle to Late Pleistocene (Fig. 1), and whilst it was the most prevalent biome type available, representing 61.9% of the biomes present during occupational phases across the region (Supplementary Fig. S2 and Table S2), eastern African MSA adaptive systems were likely specialised for engagement with tropical xerophytic shrubland, and its modulation may therefore have influenced patterns of Middle to Late Pleistocene human distribution. Nonetheless, the proportion of occupations with access to tropical xerophytic shrubland was significantly higher using a 2-sample proportion test than the proportion of the biome available across the region throughout MSA occupational phases (Z-value = 3.38, p-value = 0.0007; Supplementary Table S2), suggesting preferential occupation of tropical xerophytic shrubland and emphasising it as an important ecosystem for MSA populations.

In total, 57% of the occupations had a logistical landscape falling on the boundary between multiple biomes (n = 48; Supplementary Table S1). The majority of these ecotonal sites are situated between ‘open’ and ‘closed’ biome types, supporting the assertion of Basell⁹ that access to wooded ecologies was vital for MSA populations. Forest biomes made up relatively low proportions of the available environments available throughout the Middle to Late Pleistocene; however, importantly, we found the proportions of forest biomes occupied by MSA occupations to be significantly higher than would be expected based on the prevalence of these biomes, especially in MIS 3 and MIS 7 (see Supplementary Fig. S2 and Table S2), supporting the contention that MSA hominins preferred the rarer habitats that were near to woods and forests. The most common ecotone occupied during the eastern African MSA was that between tropical xerophytic shrubland and temperate conifer forest, which is seen as far north as Goda Buticha in southeastern Ethiopia, and as far south as Mumba in Tanzania. However, the region to the east of Lake Victoria shows the most intense occupation of this ecotone, the boundary of which fluctuates through time and space (Supplementary Table S1).

We found that MIS 7 saw the preferential occupation of closed ecotones between temperate conifer forest and warm mixed forest, as well as tropical xerophytic shrubland and associated ecotones which are generally occupied throughout the period. MIS 5 saw a slight increase in habitat diversity, though expansions primarily involved the tracking of tropical xerophytic shrubland environments (as shown by all occupations in MIS 5 having access to this biome within 50 km) with exposure to new ecotones occurring at the peripheries. This can be seen at occupations distributed widely across the region; for example, certain occupations at Omo would have involved engagement with deserts alongside tropical xerophytic shrubland, whereas some MSA populations at Panga Ya Saidi had access to tropical deciduous forest and tropical savannah environments within their logistical landscape. MIS 3 saw the greatest variety in the ecologies occupied, where expansions can be seen into new and previously uninhabited environments, such as steppe tundra and warm mixed forest, with a distinct emphasis on temperate conifer forest rather than tropical xerophytic shrubland. Importantly, a chi-square test revealed that the relative proportions of biomes in the region do not differ significantly between the Marine Isotope Stages (χ² = 9.07, p-value = 0.99), strongly suggesting that variation in the environments occupied through time reflects a shift in preference as opposed to fluctuation in the underlying ecology (see Supplementary Table S2).

Characterising MSA environments throughout the Middle to Late Pleistocene

We used cluster analyses to group the occupations based on their climatic values to assess patterns in habitat choice. To do this, we scaled and combined the temperature and precipitation data and employed an automated clustering algorithm (the average silhouette method) to ascertain the optimal number (k) of clusters in the data. The algorithm found ten clusters to represent the best division of the data (Fig. 4, Supplementary Fig. S1).

Most of the occupations (n = 45) fall within warm to temperate sub-humid clusters (2,4,5 and 7) with a broad temperature range of 13–19 °C and a precipitation range of 613-1297 mm. These clusters are dominated by tropical xerophytic shrubland and temperate conifer forest environments and their ecotones. We found that only two clusters (8,9) did not include occupations with access to tropical xerophytic shrubland, indicating that this biome was present across a large portion of the MSA climatic range, except at the coldest extreme. We found that the coldest cluster, cluster 9 (temperature range 9–10 °C), was the most ecotonal, with all occupations situated at high altitude where populations would have had access to steppe tundra, temperate conifer forest, temperate sclerophyll woodland and warm mixed forest, the complex topography allowing diverse biomes to appear closer together than is usually possible³⁴. Extremely humid occupations from around Lake Victoria (Karungu and Rusinga Nyamita) formed cluster 10 (1374-1593 mm precipitation). These occupations have moderate temperatures (16–18 °C) and occupy an ecotone between tropical xerophytic shrubland and temperate conifer forest. Panga Ya Saidi and Laas Geel form their own respective clusters (3 and 6) due to their distinctively hot temperatures; however, at Panga Ya Saidi, this is coupled with moist sub-humid conditions and a diverse tropical environment (24 °C, 996-1153 mm), whereas Laas Geel possesses the lowest annual precipitation of all the occupations (18 °C, 396 mm), making its hot-dry environment unique for the eastern African MSA. However, the occupation at Laas Geel falls within the tropical xerophytic shrubland biome, with access to some open conifer woodland within 50 km, suggesting that whilst occupying a climatic extreme, this distinct habitat represents an extension of the types of environments that eastern African MSA populations were already well-adapted to.

Phased habitability models

We used the precipitation and temperature data from the occupations as the parameters to produce phased ‘habitability’ models for the more abundantly populated interglacial phases of the MSA, demonstrating the extent of the landscape that experienced comparable climatic settings to occupations dated within that period. The climatic range produced by each phased subset was projected throughout every 1000-year time interval for that MIS, and then the percentage of ‘habitable’ cells (i.e., cells that remain within that climatic range) was calculated to identify areas that were persistently habitable, as well as the geographic range and temporal scope of impersistent habitable landscapes.

Figure 5 demonstrates the temperature, precipitation, and combined habitability models for each phase. MIS 9 shows the most limited habitable zone out of the interglacial phases, however the lower number of occupations available to construct the distribution likely has impacted the construction of the models. MIS 7 marks a period of expansion, with the region surrounding Lake Victoria and the Eastern Rift Valley Lakes and the Ethiopian Highlands showing the most persistent habitability across the region. For temperature, large areas of the Horn and modern-day Sudan show less persistent habitability (ca. 40–50% cells falling within the temperature range of 12–23 °C seen at MIS 7 occupations), with pockets of unsuitability along the coast of the Baab el Mandeb and the border between modern-day Ethiopian and Somalia. However, arid zones of the southern Sahara are completely uninhabitable in terms of precipitation (0% of cells fall within the precipitation range of 582-1368 mm at MIS 7 occupations), as is the tip of the Horn. Precipitation is thus the limiting factor when considering habitability for MIS 7, as the area deemed habitable in terms of precipitation is more geographically restricted than that derived from temperature. MIS 5 sees the largest increase in habitable area for temperature, with all cells showing temperature values within the MIS 5 occupation range of 13–25 °C for at least 60% of the period. Precipitation habitability, that we considered here to be ranging between 554-1385 mm, is however more fragmented, with pockets of uninhabitability forming around the northeast edge of Lake Victoria, in the region to the south of Lake Tana, and within modern-day Tanzania. Like MIS 7, this means that habitability is limited by precipitation in MIS 5. However, the habitability models for MIS 3 demonstrates the opposite pattern. Temperature habitability, defined as between 9–19 °C by the sites dating to MIS 3, shows the most restricted distribution of all the models, with habitable areas concentrated to the areas around Lake Victoria and the Ethiopian highlands, which are linked towards the southeast of Lake Turkana. Yet, MIS 3 shows the most persistent and widely distributed zone of habitability for precipitation, where much of eastern Africa, except towards the Sahara and the very tip of the Horn of Africa, remains persistently within the range of precipitation values experienced by MIS3 occupations (396-1593 mm). Overall, these models propose that interglacial MSA occupations, especially in MIS 5, may have been much more spatially diverse than presently known, however we note that these distributions are based purely on climatic data and ignore the potential effects of volcanic eruptions and subsequent ashfalls that have also been argued to have conditioned habitability in this region⁹.

Exploring the relationship between climate and Middle Stone Age occupations

We then examined the extent to which patterns of variability in stone tool assemblage composition and raw material use correlated with environmental conditions within a 50 km radius at the mid-age of occupation of each assemblage, as well as a suite of other variables recorded by Blinkhorn and Grove¹¹ (see Methods and Supplementary Methods S1 details). Figure 6 demonstrates the relationships between these variables and toolkit composition and raw material use, revealed using simple Mantel tests (Table 1 and Supplementary Table S4–S5). We found that MSA assemblage composition was correlated with differences in both mean annual temperature (adj. p = 0.001; Table 1) and total annual precipitation (adj. p = 0.003; Table 1), and raw material use also shows statistically significant relationships with both mean annual temperature (adj. p = 0.001; Table 1) and total annual precipitation (adj. p = 0.003; Table 1). With the use of Pleistocene climate models at high temporal resolutions, these results refine the findings of Blinkhorn and Grove¹¹, which relied on comparisons of the climatic extremes of the LGM and LIG.

We next employed multiple matrix regressions to resolve independent, significant correlations between differences in either stone tool assemblage composition or raw material use, and differences across the different variables. Table 2 demonstrates that four variables exhibit independent significant effects on toolkit composition (F = 68.0, multiple R² = 0.15, p = 0.001)—raw material use, roughness, site type and precipitation. In the multiple matrix regression for raw material (Table 2), five variables were found to have independent effects (F = 54.8, multiple R² = 0.12, p = 0.001) – toolkit composition, cost path, roughness and simple age, with precipitation being very close to significant. In each instance, these results complement earlier findings, but the use of highly resolved climate models illustrates the independent and significant relationship between MSA behaviour and precipitation that was not evident from previous analyses that were reliant on models for climatic extremes¹¹. This highlights the benefit of using highly chronologically resolved model data to more accurately characterise the climatic parameters of MSA occupations in eastern Africa and to reveal the relationship between environmental and behavioural variability.

Source: Ecology - nature.com