Figure 5(a). Positions of continents and their vegetational zones, (a) about 60 million years ago, (b) about 40 million years ago, (c) about 20 million years ago, (d) present vegetational zones. Vegetational zones in (a) (b) and (c) have been compiled with simplifications and modifications from Wolfe (1985), and  (d) from The Times Atlas. Base maps are from Irving (1982). The old polar deciduous forest has been labeled P to distinguish it from later, botanically different, mixed northern hardwood forests.

Figure 6. Some ways of forming mountains. The diagrams are much simplified and are intended only to illustrate general principles, (a) Collision of continents; cross-section through India and Tibet showing one possible way of explaining the doubling of the crust and the elevation of Tibet, (b) Subduction of ocean plate and creation of mountainous volcanic arcs or islands; the example shown is from the high-island archipelago, (c) rifting apart of a continent; cross-section of an early stage of formation of the North Atlantic.

Figure 7. Possible distribution of uplands about 55 million years ago. Arrows denote convergent (C) and divergent (D) margins corresponding to Figures 6(b) and (c).

Figure 8. (a) Earth's mean temperature at sea level, compiled from Baron (1987). (b) Variation at mid latitudes, of mean sea temperature, complied from Savin (1977) much simplified.

Global Climatic Change
Of course, it is not sufficient to say that because there is a suitably hilly landscape, such as that depicted in Figure 7, that rhododendrons would actually have grown there. A suitable climate is needed also, and the Earth's climate is known to have changed drastically since rhododendrons first appeared.

Today there is a year-round ice cap at both poles, the difference in mean surface temperature between pole and equator is over 40°C (Fig. 8(a)), and at latitudes above about 55° the mean surface temperature is below freezing. Earth is said to be in a "glacial period". Earth has been in a glacial period for the past 2 to 3 million years, during which there have been large oscillations roughly every 50,000 years. At times (called ice ages) glaciers have extended down into intermediate latitude (Fig. 9). At other times (called interglacials) they have diminished to about their present extent with glaciers at sea level only near the poles. For the past 10,000 years glaciers have been generally in recession. However, from 100 to 40 million years ago it was much warmer in high latitudes and the pole-to-equator temperature difference was only 10 or 15°C, three times less than at present. There appears to have been no ice at sea level, and everywhere the mean surface temperature exceeded freezing (0°C). These are the characteristics of a "non-glacial period" and define Earth's normal climatic regime. Today, climatically speaking, we live in exceptional times.

Figure 8(b) is a much simplified graph showing the variation with time of mean surface temperature in middle latitudes. During the early history of ericaceous plants, the Earth had an equable non-glacial climate (Fig. 8(b)). This was followed by a long slow decline to about 3 million years ago when the present glacial period commenced, although there is evidence of some glacial ice in Antarctica as long ago as 15 million years.

Figure 9. Maximum of ice sheets during glacial period of past few million years.

Vegetation Zones
For any configuration of global climate, the principle factor controlling climatic patterns is latitude. During the early history of rhododendrons 60 to 40 million years ago, North America and Eurasia were close to or north of the positions they occupy today; they were squarely lodged between 30 and 60°N latitude, with large areas north of 60°N (Figs. 5(a) and (b)). The climate, however, was warmer and more equable, and forest grew at much higher latitudes, because the Earth was then in its non-glacial state (Fig. 8). As a consequence, there appeared to have been only about four major vegetational zones, compared to about three times that number today. These are respectively from north to south, "tropical" rain forest ("tropical" placed in inverted commas because such forest extended well outside the tropics), rain forest, para-tropical rain forest, broad-leaved evergreen forest, and polar deciduous forest (labeled P in Figs. 5(a) and (b)). These ancient forests can only be compared to modern forests in a general way because many modern forest trees had not yet evolved. For example, the dominant coniferous trees were not the spruces, pines and firs of today, but instead were, for the most part, members of the bald cypress family (Taxodiaceae), which includes such trees as the dawn redwoods, the redwoods and the bald cypress, whose range is nowadays fragmented and constricted.

The polar deciduous forest, which included deciduous hardwoods and conifers such as the dawn redwood, occurred in northern North American and northern Siberia well inside the Arctic Circle. Around and just south of the Arctic Circle there was a belt of broad-leaved evergreen forest including palms, followed by a belt of warm parat-ropical rain forest. Both belts probably extended into lower latitudes in upland areas, and both probably provided habitats suitable for rhododendrons. "Tropical" rain forests were the dominant vegetation, which, apparently, covered most low-lying regions in latitudes lower than about 45°. Within it, most habitats suitable for rhododendrons are likely to have occurred only in upland regions, but these could have been widespread as already noted (Fig. 7).

It seems probable therefore on geographic, climatic and ecological grounds that rhododendrons, during their early history, could have extended more or less continuously from North America to Greenland and Europe, and eastward into China and northeastern Asia. It is probable that they moved freely between North America and Europe across the narrow gulfs of the nascent North Atlantic, rather then across Beringia, a route that was then at very high latitude. The early population of rhododendrons, although probably widespread, may not have been very diverse (that is, had comparatively few species).

By 40 million years ago, there were about five global vegetational zones. "Tropical" rain forest occupied a smaller area than before (Fig. 5(b)). Polar deciduous forests persisted, and there was a notable development of a northern evergreen coniferous forest composed typically of members of the family Taxodiaceae. The belts of warm para-tropical rain forest and broad-leaved evergreen forest moved southwards and expanded. Except for the region of extreme relief and the high-island archipelago which had not yet taken shape, the ranges of these two zones embrace the present-day distribution of rhododendrons, including that of the cold-climate R. lapponicum . It appears that, in and between the places where rhododendrons are found nowadays, there existed habitats suitable for them. With the exception just noted, rhododendrons are still more or less where they always have been.

By 20 million years ago "tropical" rain forests were further restricted, and general cooling in middle and high latitudes had begun. These changes culminated in the glaciations of the the past few million years. Continental climates and grasslands began to develop, and there was a further increase in the number of vegetational zones from about six at 20 million years ago, to as many as eight today, twice the number present when rhododendrons began (compare Figs. 5(a) and (d)). In the north, extensive coniferous forest dominated by the pine family (Pinaceae), came into existence. Its present-day descendant is the boreal forest of northern North America and northern Eurasia. Mixed hardwood and coniferous forest for the first time occupied large areas of Eurasia and North America, but the great broad-leaved evergreen forests of the early Tertiary were much diminished, becoming restricted as they are today to southeastern China and the southeastern United States. These mixed hardwood and broad-leaved evergreen forests provided the habitats best suited to rhododendrons.

Between 20 million years ago and today, and especially in the last few million years, the climate in middle and high latitudes becomes much cooler than when rhododendrons first made their appearance. Large areas of desert, semi-desert and grassland developed in central and southwestern North America, across Africa and into Asia (Fig. 5(d)). This expansion of arid habitats eliminated vast tracts of rhododendron habitat. During the past 2 to 3 million years, vegetational zones fluctuated greatly in response to glacial advances and retreats, which disrupted the climate of lands bordering glaciated areas. There was a broad circumpolar belt of tundra and ice, and "tropical" rain forests shrank to a remnant of their former extent. Today there are broad belts of desert and semi-deserts in tropical and mid-latitudes which are bordered by broad expanses of tropical and temperate grasslands, something quite unknown 50 million or more years ago when rhododendrons first appeared (Fig. 5(a)). The region of broad-leaved evergreens and mixed hardwood and coniferous forest is confined to eastern and southeastern North America, Europe and Asia.

Figure 10. Great shear faults which define the river valleys in the region of extreme relief (RER). Compiled from Molnar and Taponnier (1975).

Extreme Relief and the Recent Speciation of Lepidote and Elepidote Rhododendrons
One of the most spectacular events in Earth's history is the long drawn-out collision of India with Asia which began about 40 million years ago and continues today. This collision set in motion the train of events which lead to the creation of the region of extreme relief and, we argue, to conditions highly favorable to the development of a rich rhododendron flora.

Prior to 120 million years ago, India formed part of the giant southern supercontinent, Gondwana, which comprised South America, Africa, Arabia, Antarctica and Australia, as well as India. Gondwana fragmented in piecemeal fashion. India and Australia were the last large fragments to break away. In its northward travel, India left behind an ever-widening rift which became the modern Indian Ocean. The ocean that once existed north of India slowly closed (Fig. 5). The Earth's crust beneath India is over two billion years old, and hence is cooler, denser and stronger than the much younger part of Asia with which it collided. As a consequence, India did not break and loose its identity, but relentlessly pushed northward, deforming much of southeastern Asia as it went. Immediately north of the zone of contact, the Earth's crust was doubled in thickness and was lifted up forming the Tibetan Plateau. South of the plateau, debris piled up forming the Himalayas (Fig. 6(a)). Together the Tibetan Plateau and the Himalayas form a vast region of high elevation unique on Earth (Fig. 3(a)). It is the only region where two continents of large mass are colliding with one another today.

The question of the timing of uplift of the Tibetan Plateau and the exact manner in which it happened is a matter of much current debate. The most recent estimates are that rapid uplift began about 20 million years ago, and that the plateau reached approximately its present elevation about 8 million years ago (Harrison et al. 1992). The exact timing of uplift is important for understanding the details of the general vegetational history, and ultimately for understanding rhododendron evolution, but for our present purposes it is sufficient to know that there has been a region of high altitude there for at least 8 million years.

The assembly of a landmass as large as Asia and the uplift of the Himalayas and Tibet has affected climate profoundly. Very low atmospheric pressure (centered over Pakistan and stretching northeast to Mongolia) develops in summer, drawing wet tropical air from across the equator, and causing heavy rainfall along the southeastern border of Tibet and the Himalayas (Fig. 3(d)). As uplift continued, river systems grew, their valleys etched along large crustal faults that had been developed as India pushed its way into Asia (Fig. 10). It is this close fan-shaped network of faults along the southeast border of the Tibetan Plateau that accounts for the geometry and bunching together of the deep valleys. The fault-controlled alignment of valleys provides ready access to rain-bearing prevailing winds (Fig. 3) ensuring that the slopes of the region of extreme relief are well watered. Through time, a landscape evolved which became home to a wide variety of habitats - sub-tropical valley bottoms, temperate slopes, and alpine peaks all very close together.

Speciation occurs more readily in small reproductively isolated populations of plants and animals than in large contiguous populations; favorable genetic modifications spread much more rapidly through small compact populations than through large diffuse ones distributed over a wide area; in a small population, genetic drift is more likely to cause changes in a population's characteristics. Such small reproductively isolated populations that undergo rapid diversification are referred to as "founder populations" (Mayr 1982).

What were the founder populations of the present-day, highly speciated rhododendrons of the region of high relief, where were they located, and where did they come from? As already explained, it is possible that during the period from about 60 to 40 million years ago, rhododendrons ranged more or less continuously across North America and Eurasia because suitable climates and topographies occurred there over a much wider area than at present (Figs. 5(a) and (b), and 7). Although rhododendrons may have been widespread and abundant, the uniformity of conditions may not have lead to the evolution of many species. As Asia was enlarged by the addition of India, and as the Tibet-Himalaya region was uplifted, new habitats ideal for rhododendrons came into being. It is possible that representatives of both lepidote (subgenus Rhododendron) and elepidote (sub-genus Hymenanthes) rhododendrons, but not azaleas, expanded into these newly established favorable habitats and became founder populations. Exactly where they came from cannot be determined in the absence of fossil records, but the mixed hardwood, coniferous and broad-leaved evergreen forests of eastern Asia about 20 million years ago are their most likely source (Fig. 5(c)).

Figure 11. Hypothesis of speciation during glacial cycling in region of extreme relief, (a) During cool period following an interglacial interval, species A, B and C are distributed across lower valley slopes and the central lower divides, (b) During a glacial maximum, the species disperse down slope becoming separated by the central divide, and B species in one valley and C in the other change to B' and C' through genetic drift, (c) Warming allows population to ascend, and five species are now everywhere present, (d) Warming continues so that this mixed population disperses upward and is split into three; genetic drift produces three new species, C" from C' on the left-hand slope, A' from A on the central ridge, and B" from B' on the right-hand slope, (e) Cooling causes populations to descend and remix. During the cycle, the number of species has been increased from three to eight. The whole cycle is assumed to take between 50,000 and 100,000 years.

It is of vital significance for rhododendron evolution that the rapid uplift of the Tibetan Plateau was occurring at a time when global climate was cooling. As Earth entered a glacial period, ice advances and retreats occurred repeatedly every 50 to 100 thousand years. In the region of extreme relief, as an ice age commenced and harsh conditions at high altitudes expanded, species ranges would be displaced downward to lower elevations (Fig. 11). Populations in valley bottoms, previously separated by unfavorable condition, could mix with populations from adjacent slopes and valleys, and species could cross after long intervals of separation to generate new hybrids, the potential progenitors of new species. Populations at higher elevations would become separated temporarily from one another by intervening glaciers or unfavorable alpine habitat. These now-separated populations would reproduce in isolation and, over time, diverge from each other by genetic drift following their own evolutionary path. During interglacial intervals, populations at higher elevation could migrate upward over ridges and divides, and formerly isolated populations would intermingle. The large vertical range of valleys and ranges meant that species distributions could be displaced rapidly upward and downward without annihilation, despite major climate shifts, because the actual distances travelled would be short. Consequently, fragmentation and coalescing of rhododendron habitats could have occurred repeatedly as a consequence of glaciation and deglaciation, so producing the present-day diverse flora.

In essence, mountains and deep valleys each act as islands of habitat which repeatedly join and separate as climatic zones and plant populations ascend and descend. These joining-separating cycles occur at about the right interval for them to serve as a forcing process for speciation. A period of 50 to 100 thousand years is generally considered sufficient for a small, physically isolated population to undergo changes great enough to ensure reproductive isolation. These separations and re-joinings would provide repeated opportunities for species to mix and separate repeatedly. The end result would be a very large number of closely related species many of which would retain the ability to hybridize. We suggest that it was precisely this combination of very high mountains situated in tropical latitudes, and cycles of glaciation and deglaciation, that provided the dynamic conditions for the extraordinary speciation of lepidote and elepidote rhododendrons in the region of extreme relief. According to this view, most of the speciation has happened in the last 3 million years during the current glacial period.

Lowland areas or mountains of modest size provide fewer opportunities for populations to separate and then rejoin (the hallmark of the above speciation mechanism). They provide no refuge in times of severe climate. In valleys of less than a certain critical depth, rhododendron populations would be eliminated or greatly reduced as climate deteriorated; as temperature fell and valleys become occupied by unfavorable habitats even filled with ice, there would be no possibility of downward migration; as rainfall diminished during warmer interglacial periods, there would be no nearby higher wetter slopes on which to find refuge.

Critical valley depth depends mainly on latitude. Western North America or the Alps, for example, have about one half the relief and are more northerly than the region of extreme relief. Much of northern North America was covered by ice during major glaciations, whereas to the south relatively dry climate prevailed. There were few suitable habitats nearby, and during ice ages the local rhododendron population must have suffered severely. In the region of extreme relief the correct balance appears to have been struck, so that although climatic changes were great, the tropical latitudes and the great depth of the valleys ensured that safe havens were nearby. In such a physical environment, latitudinally well-placed, climate change acts not as an agent of destruction, but as an incentive to diversification. Nowadays, because of its unique elevation, the region of extreme relief is isolated botanically from the rest of the Asian land mass of which it is a part. Moreover, the region itself is a vast intricate honeycomb of valleys whose degree of isolation changes as ice ages wax and wane and as continuing uplift and erosion modify their configurations. Figure 12 depicts the process of evolution proposed above.

Figure 12. Proposed paleogeographic evolution of rhododendrons. Except for the evergreen azaleas (Tsutsusi) which may always have been fairly restricted, rhododendrons were, we suggest, spread widely across the northern continents during the climatically mild interval 60 to 40 million years ago. Although widespread and abundant, they were not necessarily represented by many species (not strongly diversified). As climate in middle and high latitudes began to deteriorate about 20 million years ago, and as the North Atlantic opened, suitable habitats were restricted and rhododendrons became confined to isolated pockets. At about this time, we suggest that small founder populations (circled in the diagrams) of the subgenera Rhododendron (lepidote) and Hymenanthes (elepidote) colonized what is now the region of extreme relief as Tibet began to be uplifted, and the deep valleys on its southeastern border were eroded. These founder populations proceeded to speciate, especially during the past few million years. The vireyas, an offshoot from the region of extreme relief, were a late development which colonized the recently formed high-island archipelago.

Azaleas
Azaleas differ from other rhododendrons in that there are only about one tenth the number of species, and they are, so far as we have been able to determine, absent from the region of extreme relief.

Deciduous azaleas occur mainly in southeastern USA and in southeastern Asia and Japan, with scattered representatives in southern Europe, Asia Minor, southwestern Asia, and western North America. This pattern indicates that, as has happened with the subgenera Hymenanthes and Rhododendron, they were more generally distributed over the northern hemisphere, and that their range has since been reduced and fragmented by climatic deterioration and by the widening of the North Atlantic, which increasingly presented a barrier to dispersal.

Evergreen azaleas are found only in southeastern China and Japan, and it is possible that they have always remained confined to warm temperate broad-leaved evergreen forests. They may never have ventured far beyond their original ecological niches, and their range has, in consequence, shrunk along with that of broad-leaved evergreen forests (Fig. 5).

The azaleas are not plants of the region of extreme relief although they live close by in terrain of about one half or one third the altitude and relief. Presumably, because of their genetic make-up and historical circumstances, they were unable to take advantage of the dynamic environment of the region of extreme relief. Although well positioned to act as founder populations, they failed to do so. In this respect, the paleogeo-graphical evolution of azaleas appears therefore to have been very different from that of other rhododendrons (Fig. 12).

High-island Archipelago and the Speciation of Vireyas
The Indo-Malaysian region has an abundance of species from section Vireya of the subgenus Rhododendron. Other representatives of the genus are rare (the evergreen azalea R. subsessile grows in the Philippine Islands). Vireyas are also found, but in much fewer numbers, in the region of extreme relief, in the mountain ranges of Indochina and Malaya, and in adjacent areas of southeastern China and Taiwan (Fig. 1(d)). In contrast to many other sections, such as the section Ponticum (of the sub-genus Hymenanthes) which has representatives in North America, Europe, and southeastern Asia and a markedly discontinuous distribution, the vireyas are concentrated in one essentially continuous region. This suggests that they have originated comparatively recently, more recently, for example, than the section Ponticum. The terrains in which the majority of vireyas are found today are of recent origin indicating that the abundance and diversity of vireyas is also a recent phenomenon. Many of the mountain ranges of the high-island archipelago, have been created by volcanic activity as a result of recent subduction and melting of the Indian and Pacific Plates (Figs. 3(b) and 6(b)). In their current manifestations, they are probably of recent origin, no more than a few million years old. A case in point is the highlands of New Guinea, which exceed 4000 metres in relief and which began to be uplifted about 4 million years ago (Hill and Gleadow 1989). Because the region is near the equator, vegetational zones vary from tropical forest at sea level to alpine grassland with glaciers on the peaks, and they may have fallen as much as 1000 metres during ice ages (Hope 1986). Hence, a mechanism for speciation similar to that proposed for the region of extreme relief may also be operative here. We suggest that the founder populations of the section Vireya originated in the region of extreme relief within the past few million years. From there they have spread swiftly down the Malay Peninsula into the high-island archipelago where they speciated rapidly (Fig. 12(b)). One species has even reached Australia, where it occurs in the tropical rain forest of the Atherton Tableland.

Summary
The essence of our thesis is that, following a period of mild climate when rhododendrons were widely and more-or-less continuously distributed across North America and Eurasia, their range became much reduced as a result of global climatic deterioration. This deterioration began about 25 million years ago, was clearly marked by about 15 million years ago, and became extreme with the onset of the current glacial period about 3 million years ago. Small founder populations of the sub-genera Hymenanthes and Rhododendron (but not azaleas) situated marginal to their main range were able to enter and retain a foothold in the newly developing region of extreme relief on the southeastern fringe of the Tibet-Himalayan region. There they proliferated, and it was from there that the vireyas spread into the mountains of the high-island archipelago. By taking advantage of special newly developing conditions, these small, originally peripheral populations have become now the most numerous and diverse. The place of origin of rhododendrons is not known, but is was not the region of extreme relief where they are nowadays most diverse and abundant. Many of the arguments we have presented are very speculative, necessarily so because of the poor representation of rhododendrons in the fossil record. Nevertheless these arguments provide reasonable explanations of their present remarkable distribution and of their differing abilities to cross and recross.

Acknowledgements
We are very grateful to Darlene Upton and Lara Pearson for preparing the manuscript, to Gina Lesperance and Linda Bedard for preparing the figures, all of the Pacific Geoscience Centre; to Dr. Glen Rouse of the University of British Columbia for directing us to papers on the fossil record of rhododendrons and vegetation history; to Mr. Norman Todd and Dr. Alec McCarter of Victoria, to Dr. Geraldine Allen of the University of Victoria to Mr. Frank Doleshy of Edmonds, Washington, and to Dr. John Fyles of The Geological Survey of Canada for constructive criticisms of the manuscript.

References
1.  Barron, E.J., Paleoceanography, 2, 729-739, 1987.

2.  Chamberlain, D.F., Notes from the Royal Botanic Gardens Edinburgh, 39, 209-486, 1982.

3.  Cullen, J., Notes from the Royal Botanic Gardens Edinburgh, 39, 1-207, 1980.

4.  Harrison, T.M., Copeland, P., Kidd, W.S.F. and Yin, A., Science, 255, 1663-1670, 1992.

5.  Hill, K.C. and Gleadow, A.O.W., Australian. J. Earth Sci., 36, 515-539, 1989.

6.  Hope, G., In Barlow, B.A. (editor) Flora and Fauna of Alpine Australia, Ages and Origins, CSIRO, Melbourne, 131-145, 1986.

7.  Hollick, C.A. and Smith, P.S., Tertiary floras of Alaska, U.S. Govt. Publ. Off., Washington, D.C., 1-185, 1936.

8.  Hortus Third, Macmillan, 1976.

9.  Irving, E., Geophysical Surveys, 5,299-332, 1982.

10.  Kron, K.A. and Judd, W.S., Systematic Botany, 15, 57-68, 1990.

11.  Luteyn, J.L. (editor), Contributions towards the classification of Rhododendrons, New York Botanical Gardens, 1980.

12.  Mayr, E., The Growth of Biological Thought. Harvard University Press, Cambridge, Mass.

13.  Molnar, P. and Taponnier, P., Cenozoic Tectonics of Asia: effects of a continental collision, Science, 189, 419-426, 1975.

14.  Savin, S.M., Ann. Rev. Earth Planet. Sci., 5, 319-355, 1977.

15.  Sleumer, H., Flora Malesiana, series 1, Spermatophyta, Ericacaene, 6,469-914, 1972.

16.  Wolfe, J.A., Amer. Geophys. Union Geophys. Monogr., 32, 357-375, 1985.

Ted Irving, a retired geophysicist, is an emeritus scientist with the Geological Survey of Canada, Richard Hebda, a botanist with the Royal British Columbia Museum. Both are adjunct professors at the School of Earth and Ocean Sciences at the University of Victoria.