In the context of the AHP, regime shifts in northern and eastern Africa were not uniformly distributed or uni-causal. Maximum summer insolation values across the Sahara and Sahel peaked between 10,000 and 9000 years BP (Berger and Loutre, 1991), which is generally regarded as the zenith of the AHP (Gasse, 2000; Prentice and Jolly, 2000; Wanner et al., 2008). It is difficult to generalize the vegetation for such a large and topographically diverse area, and there are currently no full-scale vegetation maps of the entire region based on proxy data, but coarse approximations can be made (Figure 2). Hély et al. (2014) provide a latitudinal reconstruction of vegetation based on proxy data, but the paucity of sampling sites has inhibited time-transgressive longitudinal studies. Paleoclimatic models often conflict in the details of the spatial distribution, but the overall consensus of both models and proxy data indicate that the present-day Sahara was significantly wetter and included much higher plant diversity than is found in the region today (Watrin et al., 2009; Hély et al., 2014; deMenocal, 2015).

The floristic composition of the Sahara during the AHP does not appear to have an analog in today's Sahara. During the zenith of the AHP, plant communities distinctly characterized today as “Guineo-Congolian” (moist tropical woodlands), “Sudanian” (grassy woodlands), “Sahelian” (wooded grassland), and “Saharan” (xeric-adapted grasses) lived side-by-side between 12° and 20°N (Watrin et al., 2009; Hély et al., 2014). Riparian areas in now-desiccated sections of the Sahara were comprised of woodlands (e.g., Alchornea sp., Piliostigma sp., Celtis sp.), and grasses (Calligonum sp., Ephedra sp.) were distributed up to 24°N by 4200 years BP (Watrin et al., 2009; Hély et al., 2014). Based on the composition of plant taxa alone, the maximum zone of precipitation during the AHP likely fell between 15° and 20°N (Hély et al., 2014).

However, weakening of the summer monsoons associated with reduced solar insolation (Figure 3A) following the Holocene Climatic Optimum (HCO) induced new atmospheric dynamics across the entire region, beginning in the northern Sahara and gradually extending southward. After 8200 years BP, landscape changes trending toward xeric conditions are documented in the northeastern Sahara (Hoelzmann et al., 2001) with periodic recharges in some lake basins until 4500 years BP, after which the region achieves the desert-like conditions that predominate today (Gasse, 2000). An abrupt regional transition from grassy to shrub-dominated vegetation is recorded in two sediment cores from the mouth of the Nile River at 8700 and after 8000 years BP, but there is no concurrent rapid reduction in river runoff or precipitation suggesting that monsoon activity was only slowly decreasing at that time (Figures 3B,C; Hennekam et al., 2014; Blanchet et al., 2015). There is consistent agreement in the proxy record that, beginning in northern Africa by ca. 8200 years BP and spreading south into what is now the Sahel and eastern Africa by 4500 years BP, the landscape transitioned from generally pluvial to arid or semi-arid ecological conditions, even though the timing, pace and magnitude of the transition was variable (Shanahan et al., 2015).

Climate models consistently predict that changes in orbital precession weakened inland monsoon flow over the northern Sahara (>20°N) by 8200 years BP, and this effect spread southward (to 12°N) over the next 3500 years (Claussen et al., 1999, 2013; de Noblet-Ducoure et al., 2000; Foley et al., 2003; Renssen et al., 2006, 2012; Liu et al., 2007; Hély et al., 2009; Watrin et al., 2009; Lézine et al., 2011b; Harrison et al., 2015). Ecological boundary conditions in regions affected by the termination of the HCO have been attributed to their position relative to the Intertropical Convergence Zone (ITCZ)—once the rain belt migrated southwards, their predominant source of moisture was substantially reduced (Liu et al., 2007; Castañeda et al., 2016). As will be detailed below, the speed with which the transition from high- to low-biodiversity landscapes occurs is variable according to both longitude and latitude.

In simulations of the AHP termination, it has been proposed that ecosystems that hosted diverse plant communities reduced the sensitivity of the Saharan ecosystem to climatic effects (Claussen et al., 2013). In a study of vegetation response to AHP termination from Lake Mbalang in northern Cameroon, the impacts of steadily decreasing precipitation on the floristic composition of the site were gradual, which is attributed to the pre-event “stability of vegetation” that preceded climate change (Vincens et al., 2010). Similar landscape dynamics have been documented from pollen records recovered from Lake Tilla (Salzmann et al., 2002), Lake Yoa (Lézine, 2009), Lake Mega-Chad (Amaral et al., 2013), portions of the eastern Sahara (Neumann, 1989), and from an offshore sediment core near Senegal (Niedermeyer et al., 2010; Schefuß et al., 2015) where the vegetation was slow to respond to changing rainfall patterns until clear human impacts are documented in the proxy record.

On the other hand, there are many reconstructions of hydrological cycles and vegetation indexes to indicate abrupt termination of the AHP. There is a rapid decrease in the abundance of C4 vegetation running off into the Nile River between 8500 and 7800 years BP and 6500 to 6000 years BP, which is out of phase with the more slowly decreasing quantities of river discharge (Blanchet et al., 2015). A similar abrupt changeover from C3 to C4 vegetation is recorded at I-n-Atei in southern Algeria at 7400 years BP (Lécuyer et al., 2016). A spike in terrigenous dust from collected offshore from the western Sahara (Figure 3G; deMenocal et al., 2000; Adkins et al., 2006), abrupt reductions in the distribution of Guineo-Congolian plant taxa after 4500 years BP north of 20°N (Hély et al., 2014) and lake level reconstructions from Lake Mega-Chad (Armitage et al., 2015) and Ethiopia (Gillespie et al., 1983; Gasse and Van Campo, 1994; Gasse, 2000) infer abrupt hydrological regime shifts to dry conditions (see also Tierney and deMenocal, 2013). Curiously, potassium content of sediments from the now-dry Lake Chew Bahir, located in southern Ethiopia indicates a slow, but steady transition to a xeric landscape in phase with orbital precession (Foerster et al., 2012), which is in contrast to nearby hydrological proxies in northern Kenya that show rapid changes in the hydrological cycle that culminated in an 80 m regression in the level of Lake Turkana at 4500 years BP (Garcin et al., 2009, 2012; Junginger et al., 2014; Bloszies et al., 2015). Such discordance in proxy data, even at scales of hundreds of kilometers, is typical for the terminal AHP.

Precipitation and vegetation dynamics were often out of sync with orbital precession, reflecting localized landscape pressures and feedbacks. In a synthetic review of AHP dynamics based on a complex, non-linear δD_wax record from Lake Bosumtwi, Ghana, Shanahan et al. (2015) argue that the termination of the AHP was locally abrupt in the northern and eastern portions of Africa, but was buffered in the southern Atlantic regions by less variability in peak summer insolation values throughout the Holocene. This phenomenon is explained by two potential forcing mechanisms: either the summer monsoon belt (vis-à-vis the ITCZ) shifted south or there were synoptic-scale changes in Indian Ocean circulation patterns that reduced inland moisture flux over the continent (Shanahan et al., 2015). These mechanisms are not necessarily mutually exclusive. Atlantic Ocean sea surface temperatures also contributed to West African monsoon dynamics during the middle Holocene (Weldeab et al., 2005), but do not appear to have impacted local vegetation patterns as significantly as in northern Africa. Tierney et al. (2011) and Tierney and deMenocal (2013) attribute rapid termination of the AHP in northeastern Africa to constriction of the Congo Air Boundary (CAB) and ITCZ during the middle Holocene (see also Costa et al., 2014; Junginger et al., 2014; Bloszies et al., 2015; Castañeda et al., 2016). The CAB and ITCZ previously trapped tropical Atlantic moisture into a much wider area within the greater Sahara and Sahel region during the AHP, but with weakening summer insolation during the termination phase, the eastern and northern extents of these pressure convergence zones constricted between 5° and 7° relative to their maximum geographic extents (Figure 2; Gasse and Roberts, 2004).

Due to the spatially discordant nature of the evidence for the timing of the termination of the AHP, there is a need to consider alternative devegetation mechanisms beyond orbital parameters that could have amplified terrestrial-atmospheric feedbacks. The term “neolithization” is used to explain transformative landscape processes related to the transition of human societies from fully foraging to fully farming economies (Kuzmin and Orlova, 2000; Rispoli, 2007; Goring-Morris and Belfer-Cohen, 2011; Kim, 2014). Contrary to other places in the world, neolithization in northern Africa was not associated with the first use of ceramics and other forms of storage as so-called Aqualithic communities had developed pottery by 10,000 years ago (Stewart, 1989; McDonald, 2016). Instead, neolithization in the northern African context represented a shift in landscape tenure processes, which resulted in three primary changes to the human relationship with the environment: (A) transformation of the biotic environment from one that inherently produced resources for humans into one that must be cultivated to yield resources, (B) the net primary productivity (NPP) of the landscape diminished even as the anthrocentric productivity (ACP) initially increased, and (C) part and parcel of increasing the ACP resulted in a reduction in overall biodiversity and enhanced albedo. Increasing the ACP conforms to general theories of niche construction in which human communities manipulate the parameters of the ecosystem to improve their access to plant and animal resources (Smith, 2011). The tradeoff between NPP and ACP is an unintentional, yet repeated phenomenon in neolithization across the world. If NPP suffers to the detriment of the total ecology of a region, ACP may collapse, which ecologists typically call land denudation (Smil, 2000; Lal, 2012; Li et al., 2012).

Neolithization is used to argue in favor of the hypothesis of a “long Anthropocene” in which humans have been transformative agents on landscapes for many thousands of years, particularly following the introduction of agricultural techniques (Fuller et al., 2011; Ellis et al., 2013; Foley et al., 2013; He et al., 2014; Certini and Scalenghe, 2015; Lyons et al., 2016). In this view, humans have co-evolved as natural earth processes rather than acting as external forcing agents like orbital precession, and their influence on shaping the landscape can be traced deep into prehistory (Foley et al., 2013). As an apex species, humans have always played a significant role in the ecology of the landscapes they inhabit dating as far back as the Pleistocene (Smith, 2011; Ellis et al., 2013). In more recent times, large-scale land clearance resulted in significant deforestation and soil erosion during the early Neolithic periods of southwestern Asia (Yerkes et al., 2012), China (Zong et al., 2007), and northwestern Europe (Innes et al., 2013). A pollen study from the Tibetan Plateau documents the transformation from a Poaceae- to Kobresia-dominated landscape in intervals that correlate to changes in the monsoon system dating back to 6000 years ago (Schlütz and Lehmkuhl, 2009). Although such correlations can only provide a tentative basis for causation, apex species have been documented in numerous instances of profoundly altering regional vegetation and geomorphology (Terborgh and Estes, 2013).

This process is difficult to document in paleoecological records, so synchronic LTER conducted in historical settings provide the basis for understanding the processes involved with anthropogenic regime shifts. For example, in New Zealand climate change provided the backdrop for stress on rainforest ecosystems, but direct anthropogenic effects of deforestation, burning and the introduction of invasive species ultimately forced the system into a new regime following European settlement of the islands (Johnstone et al., 2016). Prior to the 1840s, the few European settlers in New Zealand were scattered around the coastal regions, but by the 1880s there were settlements throughout the interior of both North and South Islands, most of which were economically focused on sheep herding (Peden, 2011). Prior to European colonization, approximately two-thirds of North Island and one-fourth of South Island were covered in temperate rainforests (McGlone, 2009). By the end of the nineteenth century, forest cover in New Zealand had been reduced in half, wetlands were drained, and tussock grasslands were burned off to make way for sheep herding and cereal grain agriculture (Parsons and Nalau, 2016). The induced regime shifts within a 50-year period have proven to be irreversible even with aggressive efforts to restore the ecosystem to its pre-1840 condition (e.g., Whitehead et al., 2014; King et al., 2015). This phenomenon occurred with total accumulated population levels below 800,000 people on a landmass of 268,000 km² (Easton, 2011). Climate stress alone cannot sufficiently explain such a rapid transformation of the New Zealand landscape from forest to grasslands, so obvious contributing factors must be considered.

More analogous to the African context, the introduction of domesticated livestock by Euroamericans into semi-arid and arid regions of the Americas profoundly altered the ecosystem, inducing regime shifts in many regions. Grazing and browsing ungulates evolved in the Americas during the Cenozoic and were a critical component of the ecological matrix (Grayson, 2011; Woodburne, 2012). Prior to Euroamerican settlement, vast prairie grasslands spanned the interior upland regions of both North and South America. However, with the exception of Highland South America, there were no domesticated grazers present before the arrival of European settlers. Cattle (Bos taurus) introduced a new pressure to the landscape that spatially and temporally correlates to a regime shift from grassland to scrubland (Van Auken, 2000).

Two studies conducted in climatically analogous ecosystems to the terminal AHP are instructive for understanding the relationship between the introduction of domesticated livestock and ecological regime shifts. Research in southern Texas, USA attributes the replacement of indigenous savannas by mesquite (Prosopis glandulosa)-dominated shrubland with a combination of overgrazing, fire suppression, and climate change dating back to early Euroamerican colonization of the region (Archer, 1989). A study of tree rings and fire history matched with 125 years of historical records of settlement in south-central Oregon, USA finds positive correlations between the introduction of livestock and growth of the western juniper (Juniperus occidentalis) (Miller and Jeffrey, 1999). More generally, a synthetic quantitative review of multiple diachronic studies of rangelands conducted throughout North America finds consistent correlations between the introduction of domesticated livestock to xeric ecosystems, commensurate decline in floral biodiversity and increase in shrub growth (Jones, 2000; see also Morris and Rowe, 2014). Once the threshold is crossed, experiments demonstrate that it will not return to its pre-disturbance state without significant human intervention (e.g., Laycock, 1991; Curtin, 2002; Loeser et al., 2007).

The potential of humans to transform landscapes and prompt threshold crossing has been extensively documented in both prehistoric (Boivin et al., 2016) and historic (Jones, 2000) contexts. Multiple factors lead ecosystems to the verge of regime change, but anthropogenic effects documented in historical contexts have been demonstrated as profoundly capable of reducing the biomass of a landscape. Such effects accelerate rates of devegetation and soil loss and can spatially cascade as albedo and dust entrainment are enhanced and a terrestrial-atmospheric feedback loop is created (Pausata et al., 2016). That the Sahara and eastern Africa stood at the precipice of a large-scale landscape transformation at the end of the AHP is not in dispute. The question of whether human agents were sufficiently capable of accelerating the termination of the AHP in some regions relies on evaluating diachronic evidence of landscape change coeval with the introduction of new land tenure systems.

When viewed through the lens of anthropogenically-induced regime change, humans are potential amplifying agents in reducing NPP, and in areas where there are coeval archeological and palynological records, local landscape change often occurred at profound scales. In a large-scale analysis, sediment runoff in the Nile River progressively increases from the early termination phase of the AHP (ca. 8000 years BP), which is inferred to be the result of intensified use of land associated with farming and animal pastoralism (Ehrmann et al., 2016). More localized studies demonstrate the mechanisms by which large-scale changes could have been induced. In the western Mediterranean, the arrival of domesticated taxa at Ifri Oudadane by 7300 years BP was concurrent to a significant reduction in native arboreal pollen (AP) and grasses with significant increases in shrubs, specifically maquis species (Figure 4A; Morales et al., 2013). The growth of maquis vegetation has been associated with livestock grazing and burning during the Neolithic colonization of the Mediterranean region. Maquis growth promoted evergreen holm oaks (Quercus ilex), strawberry trees (Arbutus unedo), and basswood (Phillyrea latifolia) at the expense of deciduous oaks (Quercus sp.), manna trees (Fraxinus sp.), terebinths (P. terebinthus), wild service trees (Sorbus sp.), and elms (Ulmus sp.) (Geddes, 1983; Naveh, 1987). There is a progressive decline in AP concurrent with an increase in NAP and cereal cultivars at the site of Ifri n'Etsedda, Morocco between 7400 and 6800 years BP (Figure 4B; Linstädter et al., 2016). At the site of Tin-a-Hanakaten, Algeria, a similar transformation of flora are documented along with evidence of significant eolian activity at 7200 years BP, which is stratigraphically associated with the first occurrences of substantial quantities of cattle remains in the archeological deposits (Figure 4C; Aumassip, 1984).

Further south, a sediment core from Lake Chad provides evidence for a gradual increase in non-arboreal pollen (NAP) and shrub vegetation at the expense of AP and non-wetland grasses between 6700 and 5000 years BP (Amaral et al., 2013) with an abrupt switch to arid conditions after 5000 years BP (Figure 3F; Armitage et al., 2015). This time period has been posited, though not yet archeologically proven, to be when domesticated plants and animals arrived in the region (MacEachern, 2012). Nearby, a curiously abrupt reduction in Guinean and Sudanian plant taxa by 3300 years BP in favor of Sahelian shrublands is recorded in the Manga Plateau west of Lake Chad (Figure 3E; Salzmann and Waller, 1998; Salzmann et al., 2002). However, the hypothesis that human agency played a role in the transformation is questioned by the latter study's authors. The pollen record at Lake Yoa in the northern catchment of the Megalake Chad basin, shows the transition to a semidesert plant community on disturbed dunes after 4600 years BP and true desert communities are found after 2700 years BP (Figure 3D; Kröpelin et al., 2008; Francus et al., 2013). This corresponds to the time when archeology indicates that pastoralism replaced hunting and gathering as the primary subsistence economy (Van Neer, 2002).

In eastern Africa, the arrival of domesticated animals is difficult to precisely ascertain because the point of first entry likely occurs in the Lake Turkana region when lake levels were approaching their nadir (Figure 4D; Wright et al., 2015). However, by 4500 years BP, domesticated cattle and an accompanying Neolithic complex, including megalithic burial features, had taken root in the region (Grillo and Hildebrand, 2013) and had nominally spread southward into the equatorial regions at approximately the same time (Wright, 2007; Prendergast, 2010; Lane, 2013). Initial pastoral settlement of the Ethiopian Highlands also occurs at ~4500 years BP (Lesur et al., 2014). Coincidentally, there is a near-simultaneous monotonic regression in lake levels across the region recorded in central Ethiopian lakes (Gasse, 2000) and Lake Turkana (Bloszies et al., 2015), whereas lake levels and vegetation changes in the equatorial regions of Lake Victoria (Berke et al., 2012) and closer to the Indian Ocean coast (Tierney et al., 2011) were more attenuated. Although abrupt changes in hydrological systems were common in lakes on the Ethiopian Plateau drainage basin prior to 4500 years BP, the final regression is linked to the change in position of the CAB with vegetation-feedback factors considered as unlikely contributors (Tierney and deMenocal, 2013; Nutz and Schuster, 2016).

Changing climatic conditions progressively trending toward more arid conditions presented the backdrop for a radical transformation of the northern African landscape. However, the variable tempo and intensity of the termination of the AHP is spatially correlated to local transitions to shrubland environments and accelerated rates of soil erosion (Kutzbach et al., 1996; Braconnot et al., 1999; Foley et al., 2003), which, critically, also correspond to increasing human populations (Manning and Timpson, 2014) and the spread of the domesticated animal economy in the same locations at the same times (Supplementary Material). In tropical western Africa where pastoralism took root later and was a minor component of the subsistence economy relative to plant cultivation (Ozainne, 2014), the termination of the AHP was significantly later, progressed at a slower pace, occurred with reduced magnitude relative to the subtropical regions to the north and took place in a region with a robust ambient floral biodiversity (Shanahan et al., 2015). On the Ethiopian Plateau, pre-colonial endemic floral species biodiversity was likewise high in locations where pollen records exist (Lamb et al., 2004; Umer et al., 2007), which may have attenuated the impact of vegetation feedbacks on the hydrological cycle even though the speed of the AHP termination was high.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.