δ¹³C and δ¹⁸O in ratite eggshells can be used as a surrogate for dietary and climatic conditions [16]. The extant African ostrich is a mixed feeder adapted to subhumid to arid conditions. As an opportunist feeder, its diet consists of C3, C4 and CAM plants. Because they are non-obligate drinkers, modern African ostriches obtain their water requirements from leaf-water. Several studies revealed that eggshell δ¹⁸O is linked to bird body water which is in turn dependant on the influence of relative humidity on plants and average δ¹⁸O_{meteoric water} values [13,16]. In arid areas evaporation due to low relative humidity and high soil temperature promotes enrichment in ¹⁸O in environmental water. Later, soil water may be enriched in ¹⁸O in leaves during evapotranspiration. Plants present different photosynthetic pathways due to discrimination against ¹³C. Therefore they have distinct isotopic signatures: the mean δ¹³C of C3 plants is −27‰ (−35‰ to −21‰) and −12‰ (−15‰ to −10‰) for C4 plants whilst CAM plants have intermediate values from −10‰ to −22‰ [4,19]. C3 plants have larger isotopic ranges than C4 plants because the C3 photosynthetic pathways is more sensitive to the [CO₂]/[O₂] ratio in the atmosphere, light intensity, temperature, and humidity and moisture conditions [10,11]. Low radiation and low temperature during the growing season control the distribution of C3 plants; in contrast high radiation and warm conditions during this period control the distribution of C4 and CAM plants. In the Namib, because of the pattern of wind circulation, two different ecosystems prevail related to two rainfall seasons:

• • South of Lüderitz rain falls during the winter because of the westerlies wind system; the area is dominated by C3 and CAM plants;
• • North of Lüderitz rain falls during the summer because of a tropical rainfall system; C4 plants occur (Fig. 3).

In general, eggshell carbonate is enriched by 15–16‰ in ¹³C compared to the ingested food [13,43]. Mean δ¹³C values for modern ostrich eggshell from southern and central Namib regions are distinct: −6.10 ± 1.90‰ and −3.65 ± 2.40‰, respectively and these values are in agreement with the plant distribution in the two areas. For the oxygen, the mean δ¹⁸O is 9.55 ± 2.70‰ in the south and 14.20 ± 2.15‰ in the central Namib. These differences are related to the influence of enriched plant water by evapotranspiration on the bird body water and eggshell δ¹⁸O values [13,16]. This difference is also linked to different moisture conditions between the two regions. In the aeolianite outcrops, isotopic studies of the fossil ratite eggshells led to the establishment of an isotopic sequence through the last 20 Ma [33]. The main results were that the δ¹⁸O values, although variable, are always lower in the southern compared to the central Namib up to 4–6‰ [33] probably related to moisture conditions due to the proximity of the Atlantic Ocean in the south. δ¹⁸O trends during the Miocene differ for the two regions: in the southern Namib the mean values are around 6‰; in the central the mean values are between 10 to 16‰. But the trends track each other during the Late Miocene and Pliocene revealing that regional distinction and global climatic change have been recorded by the fossil eggshells. Throughout the Miocene δ¹³C values for ratite eggshells from both the central and southern Namib regions are indistinguishable showing that the flora comprised C3 (δ¹³C mean < −9‰) plants. The relative positive mean during the Lower Miocene may suggest a small amount of CAM in the dietary habits of these birds or a relative isotopic high value for the atmospheric CO₂ [33]. The mean δ¹³C values are significantly distinct for these regions as known today. During the Pliocene the mean δ¹³C values are –9.50‰ in the south and −8.20‰ in the central Namib. Proliferation of C4 plants is observed in the central Namib after ∼5 Ma. This expansion was probably related to energy budgets and rainfall seasonality shifts resulting from large-scale atmospheric and oceanic circulation reorganisation (with the development of cold upwelling cells along the coast) [33].

Eggshells also occur in the Middle Miocene of South Africa and comparisons have been made at a regional scale. The coastal aeolianites at the Prospect Hill site contain eggshells of Diamantornis wardi (10–12 Ma), which have been compared with the Namibian ones. The mean values for δ¹³C and δ¹⁸O of eggshells from Prospect Hill are more negative (about –13‰ for the carbon and of 1‰ for the oxygen) than those from southern and central Namib (Fig. 4). The δ¹³C isotopic range (between −13‰ to −9‰) is related to the isotopic variability of C3 plants to climatic conditions. The oxygen variation is related to evaporation conditions in the ecosystems. Isotopic data show that during the Upper Miocene the environment at Prospect Hill was clearly more humid (wet) and/or cooler than the Namib with vegetation dominated by C3 and CAM plants. Similar conditions occur today in the Prospect Hill area and similar isotopic values can be found in modern eggshells around this site [13].

Coeval with Arrisdift in Namibia, the Libyan site of Gebel Zelten (17.5 Ma) has yielded fossil bracket fungus [14] along with tropical forest trees, and a woodland fauna. In the Upper Miocene, the fossils from Sahabi [7] indicate the presence of wooded savannah in northern Libya some 7 Ma.

A suite of micromammals from a vast system of karstic deposits at Gebel Bellorat (11 Ma), includes several forest-adapted lineages such as tree porcupines, galagids and fruit bats [28]. Fossil ratite eggshells have also been recorded in North Africa, which allow us to realise some isotopic comparison at the intracontinental scale. In Egypt, isotopic studies of eggshell remains from the Late Miocene–Early Pliocene sediments at Wadi El-Natrun shows that the mean δ¹³C and δ¹⁸O values are of −9.4 ± 06‰ and 6.3 ± 1‰ respectively (n = 10). These values are quite similar to those obtained for fossil eggshells from the southern Namib [33] and indicate that northern Egypt was dominated by a C3 ecosystem indicating moist conditions with winter rainfall. The data seem to reveal a Mediterranean climate in Egypt at that time, but this needs to be confirmed by further research.

Fossil piscivorous crocodiles from 12–10 Ma deposits in central Tunisia indicate the former presence of large and deep rivers flowing northwards across the country, an indication borne out by the presence of extremely thick (up to 30 m) point bar deposits in the sedimentary successions of Central Tunisia [20].

In contrast to Northern African regions, the Congo Basin is today covered in tropical forest which grows on Miocene and Pleistocene aeolian sand deposits up to 250 m thick [12]. These deposits cover much of Congo, northern Angola and Gabon and even reach as far east as Western Uganda where Middle Miocene (13–12 Ma) sand deposits containing large gypsum crystals attest to semi-arid conditions in what is now a very humid part of the Albertine Rift Valley [27,37,39]. Between 10 and 8 million years ago, tectonic activity in the Western Rift led to the creation of a depression and the environment became more and more humid and tropical as indicated by the fossil plants and animals. The fossil fruits and woods were very diverse and similar to those which grow today in North Kasai in the Democratic Republic of Congo [8,39] (Fig. 5).

Studies of the fossil records of these different areas of Africa reveal a complex history of desertification, with different areas going arid at different times, and some areas going arid at the same time that neighbouring zones were becoming more humid. The geographic position of arid and humid areas appears to have changed throughout the Neogene.

In East Africa, uplift related to tectonic activity during rifting had an important impact on the environments during the Neogene [5,27]. In the Middle Miocene, due to global and local events, the forested environments began to dry out and the faunal composition changed in relation to altitude as shown by the differences in the environments of the Tugen Hills and Lothagam during the Upper Miocene [22,37]. The faunas and floras in the Lukeino Formation are more humid [17,18,24,36] than those of Lothagam [15]. In the Latest Miocene, the vegetation of the Congo Basin contrasted strongly with the more open vegetation in Ethiopia and Kenya which consisted of gallery forests, wooded savanna and savanna [2,6] (and Bamford, pers. comm.). In East Africa expansion of grassland began in the Upper Miocene around 7–8 Ma, but it remained a low component of the environments until the Late Pliocene [3,31]. Then, during the Plio-Pleistocene, desertification took place in low-lying areas, for example in the horn of Africa.

The earliest evidence for hyperarid conditions documented in the Cainozoic of Africa occurs in the Namib Desert, where thick aeolianites started accumulating ca 17–16 Ma in zones that had been well wooded and subhumid before (Fig. 8) [25]. The Namib has been arid to hyperarid ever since, with fluctuations in aridity indicated by changing faunas and the development of different kinds of palaeosols. The earliest evidence for aridity in the Sahara is much younger than that in the Namib, certainly younger than 10 Ma, and likely to be in the region of 7–6 Ma [23,28–30]. In East Africa, desertification progressed from the Late Pliocene (ca 3 Ma) into the Pleistocene and persists in parts of the region to this day.

Interpretation of Cainozoic African fossil localities in terms of the type of phytochore that existed at the time of deposition reconstructed on the basis of the faunas and floras that they have yielded (Fig. 6) (Table 1) shows an overall trend towards aridification in both southern and eastern Africa, with southwestern Africa becoming more arid earlier than East Africa did. This is revealed by the position of the younger localities in the more arid phytochores and the older localities in the more humid phytochores [22].

A major consequence of the early development of hyperarid environments in southern Africa, at the end of the Early Miocene, as far away from Eurasia as it is possible to go, is that plants and animals in the region had a considerable period of time to adapt to these conditions with minimal interference from plants and animals from Eurasia that may have been adapting to similar environments. The result was the evolution of a whole suite of autochthonous plant and animal lineages adapted to arid, climatically unstable environments. Lineages involved include crocodiles, ostriches, various rodent families (Pedetidae, Bathyergidae), macroscelidids (elephant shrews or senges), prohyracid hyracoids, tubulidentates, some carnivores and bovids among others. Plants were also involved in this evolutionary activity, but the details are less well-known due to a paucity of palaeontological studies, but grasses came to dominate ecosystems in Namibia much earlier than they did in East Africa, and in several mammal lineages this led to the evolution of a wide range of dental strategies (hypsodonty, ptychodonty, cementodonty, pliocodonty, polylophy) to extend the life of cheek teeth that were experiencing excessive wear due to the presence of opal phytoliths in the grass that they were consuming.

Adaptation of animals and plants to arid environments continued unhindered in Namibia for up to 10 million years before the first signs of aridity became evident in the Sahara and some 14 million years before East Africa started going dry. An interesting point is that when the environment in northern and eastern Africa began to open up due to increasing aridity, local plants and animals had no time to adapt before they were displaced by arid-adapted lineages spreading from South Africa on the one hand, and from arid zones in Eurasia on the other. This is demonstrated in Fig. 7 which shows the changes in the ratio between brachyodont to hypsodont mammalian species in Africa during the Neogene. Each fossil site has been allocated a colour (dark in brachyodonts, light in hypsodonts) on the basis of the presence or absence of hypsodont and/or brachyodont species. For example, in the Lower Miocene, hypsodont and brachyodont species are equal in number in southern Africa, but in East Africa, brachyodont species clearly dominate the fauna. In the lower Middle Miocene, the hypsodont species are still better represented in South than in East Africa. The East African fossil record thus shows an interesting history of faunal changes implicating lineages from southern Africa (pedetids, bathyergids, some suids, some bovids, aardvarks, ostriches, and crocodiles, for example) and from Eurasia (porcupines, leporids, hyaenas, giraffes, some suids, some bovids, equids, and camels, among others). Many East African animal lineages either went extinct (the crocodiles Euthecodon and Rimasuchus for instance) or retreated west as the forest cover shrank (osteolamine and cataphractine crocodiles, and tragulids for example). In contrast, fossil faunas from North Africa, such as those at Sahabi, Libya, tend to show much more evidence of faunal interchanges with Eurasia than do East African and South African faunas, although more fossil evidence is required from the Late Miocene of the latter regions to throw light on the details of the exchanges.

As a consequence of the complex history of desertification in Africa, it is essential to study the fossil record from all parts of the continent in order to obtain a holistic view of how desertification progressed, when it occurred, and what effects it had on the development of faunas and floras in the different regions of the continent. Only then can it be appreciated that the Namib Desert, even though it is tiny compared to the Sahara, had a considerable impact on the development of African plant and animal communities, simply because it has been arid for so much longer (since the Middle Miocene) and was so far removed from other arid areas in Europe and Asia. The animals and plants that live in the Sahara, in contrast, show a great deal of influence from Eurasia [1]; indeed much of the northern Sahara (Maghreb, Libya, Egypt) is populated by plants and animals (deer, bears, various land snails) with Palaearctic affinities, although it is clear from the fossil record that there were periods when these same areas had a substantial presence of animals that had spread northwards from equatorial and southern Africa (gelada baboons, aardvarks, wildebeest, phacochoeres).

When the ratio between brachyodont and hypsodont mammals is considered for the Neogene of Africa (Fig. 7), it becomes obvious that brachyodont animals dominated the faunas in the Early Miocene and hypsodont animals during the Pleistocene. However, the changes did not occur simultaneously throughout the continent. Hypsodont taxa started to dominate the faunas in Namibia long before they became dominant in eastern and northern Africa.