III. Aridity or Cooling at the Ice Age Equator?

Proxy data from ocean cores have been used to argue that glacial cooling was minimal at lows latitudes (CLIMAP 1976). Instead, many investigators have found evidence for reduced precipitation, yielding the concept of tropical ice-age aridity. This concept has guided recent biogeographical thinking, particularly the forest refuge hypothesis for the ice- age Amazon and the hypothesis of savanna corridors across the Isthmus of Panama. Ice-age aridity was described for Africa by Livingstone (1975), using the evidence of old water levels in ancient lakes, and elaborated by Street and Grove (1979). East African pollen diagrams are also consistent with reductions in ice-age precipitation that required large changes in the distribution and abundance of tropical forest species. But the pollen data do not require the existence of wet refugia for African forest as much as they suggest local persistence of forest species in low populations, perhaps widely dispersed in small favorable sites or along watercourses. Ice-age aridity in tropical Africa is plausible because climate models of the ice-age earth predict that precipitation in their monsoonal climates should have been reduced by 10 to 20 percent during parts of glacial cycles. In the Neotropics, however, monsoons are important only in the Caribbean and northern South America, the bulk of Amazonian rains coming from the South Atlantic. It is prudent, therefore, to base conclusions about Neotropical ice-age aridity only on data from America itself. A. Galapagos and Pacific coast Aridity. The first equatorial site do yield evidence of aridity at the American equator was on the Galapagos Islands, where El Junco lake dried completely sometime before 40,000 B.P, briefly ponded about 25,000 B.P.., and did not become a permanent lake again until 10,000 B.P. The cause of this increased Galapagos aridity is still not completely known. Colinvaux(1972) argued that the Intertropical Convergence Zone8(ITCZ) must have failed to reach the islands at any part of the year, and Newell(1973) suggested that this could have happened if the ITCZ was displaced far to the south by increased Hadley cell circulation. Houvenaghel(1974) and Simpson(1975) invoked increased oceanic circulation that extended the present system of cold surface water and bottomheavy, stable air masses. Whatever the true cause, Galapagos aridity is clearly related to regional events in the eastern Pacific Ocean, rather than to a global condition of tropical aridity. If regional, the effects might well be found on the western seaboard of equatorial South America, but not necessarily farther away. Data from the sediments of a lake, Yaguarcocha, close to the equatorial in the Inter-Andean Plateau of Ecuador are consistent with this expectation of late glacial aridity (Colinvaux, Olson, and Liu 1988). Radiocarbon inversions give some doubt about the age of lowest Yaguarcocha sediments, but the dating is consistent with a late-glacial age. Both pollen of Chenopodiaceae in the bottom sediments and diatom data strongly suggest a reduced lake in a semi-arid landscape. B. Pluvial Eastern Flank of the Andes One other lake recorded from the tropical Andes below the paramo spans glacial times, the classic section from the Sabana de Bogota, at 5 N latitude and 2550-m elevation on the eastern (Amazonian) side of the cordillera. The lake is dry now (the result of stream downcuting), but it held water during the last glacial period. Pollen from Isoetes and other aquatic plants record a

history of lake level rising and falling in synchrony with glaciations, but it is a pluvial story rather than a history of ice-age aridity, with the west periods roughly coinciding with glaciations and the dry periods with interglacials like the present. This record suggests that it is unreasonable to extrapolate aridity cycles of the eastern Pacific across the Andes Amazonian side. C. Ice-Age Aridity in the Monsoonal Caribbean. For Caribbean weather systems to the north of the South American landmass, patterns of reduced precipitation reminiscent of those in Africa are suggested by several records, including Lake Valencia in Venezuela, lakes of the Peten in Guatemala, and marsh deposits in the Guianas. These data are consistent with climate models that predict 10 to 20 percent reduction in monsoonal rains in the Northern Hemisphere. They suggest that dry seasons were prolonged over the northern part of South America as well as over the Caribbean. D. Absence of Evidence for Amazonian Aridity No unequivocal arid land features radiocarbon date to glacial times have been identified from the Amazon basin, although surficial erosional features of uncertain age can be detected on side-scan radar images(Tricart 1977). Reported buried stone lines are not relies of desert ablation but concretions formed in deeply weathered tropical soils. The one provocative datum comes from the work of Damuth and Fairbridge(1970), who identified significant quantities of feldspar minerals in ice-age sediments from marine cores off the mouth of the Amazon. Feldspars do not survive humid weathering, and Damuth and Fairbridge suggest that the ice-age feldspars were delivered to the Amazon system by rivers draining the Guiana Highlands to the north, when these were made even more arid than now by the reduction in monsoonal rains. To the extent that this explanation is correct, no arid demands are made on the Amazon lowlands. But a more cogent explanation for the feldspars is provided by Irion (1984) who suggests that feldspars were extracted from unweathered basements of deep channels cut by rivers when ice- age eustacy lowered sea levels in excess of 100m. Radiocarbon-dated pollen data to test the aridity hypothesis is the central amazon lowlands are still lacking. A detailed section radiocarbon dated to glacial times has now been reported from the Brazilian amazon, but is it on a 700-m plateau at carajas covered with savanna-like vegetation in a strongly seasonal climate (Absy et al. 1989:figure 1). Mineral and pollen data both suggest episodes of decreased precipitation in the last glacial period. This does not imply widespread Amazonian aridity because the carajas site is near the northern edge of the basin under the influence of northeast monsoons, the reduction of which also reduced precipitation in the Caribbean. Persistent tree pollen throughout the record, however, suggests tropical forest persisting in the region, presumably, as now, in bottom lands below the plateau. Thus those who would believe. On biogeographic grounds. That the amazon basin was semi-arid in gracial times should realize that have not a single paleoecological datum that requires, or even suggest, aridity in what aye now the forested bottom lands at times of northem glaciation (Salo 1987; Clinvaux 1987). Conversely, paleoecological data increasingly suggest that cooling was the primary climatic forcing of the ice-age Amazon.

E. The Andean Record of Cooling By far the best paleoecoligical record of the Neotropics comes from the Sabana de Bogota cores. Not only have these sediments accumulated continuously throughout the Pleistocene but the proximity of the lake to tree line lest relative glacial state be inferred in the pollen diagrams. Van der Hammmen and Gonzales (1960) and Hooghiemstra (1984, 1989) calibrated the percentage of total arboreal pollen with altitude from surface samples, using the results to infer the elevation of tree line at all stages in a glacial cycle. That this procedure gave realistic results is confirmed by the data of individual Adean pollen taxa like Wienmannia, Hedyosmum, and Podocarpus; or by the Quercus pollen, the percentage of whith falls in glacial maxima as the temperature sensitive oaks are restriced to lower elevations. The overwhelming implication of this pollen history is that temperatures were depressed whit each glacial cycle and raised to about those of the present day whit each interglacial. Cool and relatively wet in glacial times, not warm and dry, is the message of the Sabana de Bogota cores. Cool and moist glacial intervals for the Andes are fully in accord whit the evidence for past firn lines and moraines that show glacial descent of about 1500m, the same maximium descents shown by the tree line below them at Sabana de Bogota (Clapperton 1987; Hastenraht and Kutzbach 1985). Thus nearly all the well-estabilished, radiocarbon-dated, proxy measures of the continental climate of the American equator in glacial times have long suggest cooling, and possibly moisture . Records suggesting moderate aridity are generally confined to the periphery of the basin, particulary on the Caribbean coast. Near the equator itself, only the Galapagos and Yaguarcocha (figure 16.1) recods suggest aridity, and the parsimonious explanation of these is that the aridity was a local phenomenon drive by circulation patterns or air-mass movements in the eastern Pacific Ocean. F. Cooling in the Ice-Age Amazon While seeking ancient lakes in the Amazon rainforest of Ecuador in 1984, we found old lacustrine sediments rich in wood samples that were exposed by a road cut at Mera (figure 16.1). These yielded radiocarbon ages between 26,000 B.P. and 33,000 B.P. Both pollen and macrofossils suggested
Andean taxa growing locally, requiring that temperature sensitive trees had descended at least 700 m into what is now the upper part of the lowland tropical rainforest at 1100 m (Liu and Colinvaux 1985). Ecuadorians familiar with the forest told us of seeing other deposits with buried wood. In 1988 we undertook to search the rainforested foothills of the Ecuadorian Andes for additional sections. M.B. Bush and P. De Oliveira discovered a second site in a stream-cut near San Juan Bosco, about 100Km from Mera and at 950-m elevation. Radiocarbon dates show that the section spanned part of the same time interval, 31,000 b.p to 26,000b.p. The main features of the Mera and San Juan Bosco pollen diagrams can be understood from surface pollen spectra arranged in order of altitude. The Andean spectra from above 1000-m elevation are clearly separated from the Amazonian spectra below 1000m.The ice-age spectra from the fossil sections, however, include elements of both lowland and upland floras, are unlike any modern surface spectra, but are closely comparable to each other. In both diagrams the high Alnus percentage is strinking. Hooghiemstra and Grabandt regard such high alder percentages as indicating alnetum meadows, which they find not to occur below 2500m in Colombia. In surface samples from Ecuador we do not find high alder percentages below 3000m.

Wood samples fro both sections, and phytoliths from Mera, record the local presence of the Andean taxa Drimys, Magnoliaceae, Wienmannia, and Podocarpus,as well as taxa appropriate to modern rainforest elevations like palms. The intrusion of wood and pollen of Andean taxa cannot have been by stream transport from high elevation because of the simultaneously high pollen concentrations of both Andean and rainforest taxa and because stream transport could not have moved pollen, phytoliths, and wood in unison to redeposit them together. The data thus require descent of Andean taxa, from their interglacial position of higher than 2500m down to 1000-m elevations or lower, a descent consistent with the 1500-m descent of mountain glaciers and Colombian tree lines above them. Apparently the cooling long established for the high Andes was expressed at low elevations too. The cooling can be measured by applying the logic of lapse rates. We use a rate of 5C per thousand meters, which is the lapse rate shown by modern Ecuardorian meteorological records for all elevations from the Amazon lowlands to the high Andes, yielding a temperature depression of 7.5C. This is a minimal estimate that allows highland taxa to descend only 1500m to conform to the known tree line and glacial depressions. The cooling cannot be simply a local effect caused by catabatic winds from ice at high elevation. The areal extent of glaciers on the highest Andes was still too small to make this hypothesis plausible, and we have shown comparable temperature depression in lowland Panama, where the highlands are yet too low to have held glacial ice( see Glacial Cooling and the Exclusion of the Darien Refugium). Paleoecological data from the Andes, therefore, now demonstrate ice-age cooling in excess of 7 at all elevations, from snowline to the Amazon bottom lands. G. Reconciliation of Proxy Data from Land and Sea These data for significant cooling in equatorial America conflict directly with the CLIMAP (1976) reconstruction showing the temperature of the equatorial sea surface as only modestly depressed. Our conclusions, however, are consistent with those of pollen analysts working in New Guinea and Madagascar. Before the recent results from Ecuadorian Amazonia were published, the conflict appeared to be merely between estimates of sea surface temperature and depressions of tree lines and snow lines at high elevations. Attempts were made to reconcile these two data sets by speculating that lapse rates were altitude-sensitive in glacial times, so that warm lowlands could persist under relatively colder highlands, a concept involving some difficult assumptions. The new data showing temperature depression in the Amazon lowlands as well as in the high Andes appear to make these explanations untenable.

Proxy data for sea surface temperatures used by CLIMAP are assemblies of foraminiferans thought to be stenothermic. Temperatures proxies from the land are forest trees, together with such geological benchmarks as old morraines and inferred firn lines. To the extend that the conflict pits Foraminifera and forest trees as alternative proxies for past temperature, the results could be reconciled by postulating that foraminiferal assemblies record mean annual temperature, whereas trees are sensitive to the coldest intervals in runs of years. This explanation requires that the amplitude of ice-age temperatures varied widely even as the mean remained close to the Holocene mean. In modern times, lowest recorded temperature at all elevations in the Andes tracks mean annual temperature quite closely. Furthermore, the explanation does nothing to address the issue of firn line and glacier depressions comparable to depressions of montane vegetation below them, since it requires that limits to mountain glaciers or firn lines should be set in a similar way by the coldest years or days. This seems unlikely. An appealing alternative explanation is given by Livingstone in Chapter 15. This relies on the relative rarity of interglacial conditions like the present making unlikely the evolution of Foraminifera adapted to warmer water. As a result, modern Foraminifera of tropical surface

oceans are postulated to be the same species the lived in colder oceans in glacial times, but which are now at the limits of their range, occupying ocean areas from which they would have been denied if more thermophilic species were available. IV. Amazon Forest in Ice-Age Times. Cooling the Amazon by seven or more degrees celcius certainly had profound effects on the biota. Rainforest trees in bottom lands could not descent a further 1500m, for this would put them below the surface of the sea. It follows that existing taxa of Amazonian rainforest are all adapted to climate regimes subject to air temperatures significantly colder than those of modern times, except for those few taxa that may have evolved in the last 10000 years. Community composition in Amazonian forests should have been different in ice-age times, as is illustrated by the Mera and San Juan Bosco records. These show that communities replacing modern rainforests between the 900-m and 1100-m contours have no modern analogs. They are not merely high Andean forests displaced downslope but rather are assemblies built from selected species populations, many of which are now separated into lowland and highland floras. What appears to have happened is that glacial cooling allowed a number of Andean taxa to be able to compete at lower elevations, letting them arrive as immigrants into the upper reaches of the lowland forests. The relative abundance of lowland species then changed in response to competition from these immigrants, as well as directly to the cooling itself. The predictable homes of the more thermophilic rainforest species throughout the colder intervals of glacial cycles are in the Amazonian bottom lands 100-300m above modern sea level. Significant arcas of Brazilian Amazonia lie close to 100-m elevation, thus providing a broad habitat some 800m below our San Juan Bosco site. It is in this broad area that we should except to find the densest ice-age population of many Amazonian trees and their associated biota. A task of the next few years is to discover long pollen sections from lowland Amazonia in Brazil that can be used to test this hypothesis. Meanwhile we have pollen data only from our two sites in the Ecuadorian rainforest. The Napo region of Ecuador probably holds the most species rich forest in the world. Our records from Mera and San Juan Bosco show that this diverse rainforest was replaced in glacial times with moist forest of a kind for which there is no modern analog but in which both modern Andean elements and lowland taxa were prominent. Accordingly the Napo region was not a refugium in which modern rainforest persisted throughout a glacial cycle. Rather the Napo region of high species endemism was a place of invasion and counterinvasion, as the more temperature-sensitive plants extended ranges up and down slope with fluctuating temperatures of glacial cycles. The elevated Napo region, like all the other proposed refugia was thus a place of intense species interactions, apparently leading to the understandable outcome of being a center of speciation and endemism. The Ecuador evidence for cooling further suggests that the more thermophilic of rainforest organisms had their ranges compressed to low elevations. A accordingly sll the lowerlying parts of the amazon basin that are forested now were probably forested throughout the the glacial cycle, except for marginal regions with strongly seasonal climates. The forests of the central amazon were probably not markedly fragmented. The amazon basin as a whole was not arid, though savanna regions at the periphery were probably more extensive than now.

V. The Ice- Age Isthums of Panama Lake cores from La Yeguada and El Valle now provide a transglacial section for the Panamanian lowlands. They are in region of strongly seasonal climate where most of the 3-4m of annual rain falls between January and June. Despite prolonged dry seasons at both sites, the local vegetation before human disturbance was tropical forest. The combined sections suggest that the overriding environmental events of glacial times were a lowering of mean annual temperature and the consequent compression of tropical forest into the lowlands, although some evidence of reduced precipitation is present, probably manifest as increased duration of dry seasons. A strong inference from the records is that the uplands of Darien were not a forest refugium and that moist tropical forest was present across the isthmus at all stages of a glacial cycle. A .A Transglacial Pollen History from Lowland Panama. Laguna de la Yeguada lies at 650-m elevation of the Pacific side of the isthmus divide, which rises above the lake to 1500m. Pollen and phytolith analyses are available for the complete sequence since the lake first formed with the emplacement of its dam at the start of the late-glacial period about. The pollen and phytolith data show that the local late-glacial forests were rich in what are now mantane elements, including Quercus, Magnolia, Ilex, Gunnera, Symplocus, Carex, Ranunculus and Thalictrum. Many of these taxa are now restricted to altitudes above 1500m elevation, requiring downward displacements from present limits of at least 800 m in the lateglacial. Assuming the standard moist air lapse rate of 6C per 1000m, this yields a minimal temperature depression of 5C. As with the Ecuadorian forest records, the montane taxa in the pollen diagram do not represent a total replacement of lowland vegetation by montane communities but rather an intermingling of montane and lowland elements into communities for which there are no modern analogs. El Valle, site of an even older Panamanian environmental history, is a city built on the dried bed of a caldera lake, at 500-m elevation in a forested region of the Pacific slope of Panama. A 55-m sediment core of the underlying deposits spans from about 8000B.P at the top, through radiocarbon 30000B.P by the tenth meter, to basal samples thought to represent the last interglacial of isotope substage 5e. The pollen diagram confirms the conclusion from the La Yeguada data that the late-glacial vegetation of the Panamanian lowlands included montane elements like oaks that were growing 800m or more below their modern limits. At El Valle, however, this record is extended into full glacial times. Temperature depressions ranging between 4 and 6 C are suggested for the whole glacial period. B. Glacial Cooling and the Exclusion of the Darien Refugium. Temperature depression inferred from the La Yeguada and El Valle records strongly suggests that the cooling recorded at the Ecuadorian lowland sites was experienced throughout the entire Neotropical region. Certainly local explanations based on the descent of glacial ice cannot be applied to Panama, where no mountains were high enough to have been glaciated. The only discrepancy between Panamanian and Ecuadorian cooling is that temperature depression was slightly less in Panama, probably due to maritime effects from the nearby oceans. In Ice-age Panama as in the Amazon the more thermophilic plants of the tropical forest were restricted to bottom lands close to modern sea level. Higher elevations like the Darien were not refugia for tropical rainforests as has been argued but instead held moist forest communities

resulting from cooler temperatures in which those tropical rainforest taxa with broadest temperature tolerances were associated with species now restricted to higher elevations. The pronounced cooling was always insufficient, however, to allow modern montane taxa to be established in lowlands separating the elevation regions of Chiriqui and Darien. Taxa like Quercus that penetrated the isthmus, eventually reaching the Colombia Andes, yet had to disperse across gaps of lowlands tropical forest. That tropical lowlands remained a severe barrier to dispersal explains why 11 oak species are known from Chiriqui but only 1 from Darien. That the gaps were crossed at all apparently was made possible because they were narrowed by the cool moist conditions of glacial times. The data allow a strong inference that low-lying parts of the isthmus like the modern canal zone separating the elevated regions of Chiriqui and Darien together with some coastal regions were home to tropical forest through glacial and late-glacial times. Such is the variability o climate from one locale of Panama to another however that this lowland forest was doubtless a patchwork of forest communities depending on position relative to rain shadows though tropical rainforest should have been expected in the lowlands spanning the isthmus along the line of the modern canal. C. Falsifying the Savanna Causeway Hypothesis A consequence of having the Panamanian lowlands covered with tropical forest throughout entire glacial cycles is the exclusion postulated savanna causeways for migrating fauna and humans. This exclusion is consistent with the paleontological evidence used by Webb to show that no savanna causeway can have existed through the isthmus for the last million years. Localized semi-arid habitats were probably available however on coastal strips of land exposed by eustacy but now submerged. Wide shelves of shallow sea now about the southern shores of Panama and would have been exposed as dry land in glacial times. Because intermittent stretches of the modern south Panama coast are so strongly seasonal as to be semi-arid it is mot unreasonable to expect comparable aridity in the glacial period thus providing extensive savanna habitats along the south coast. Where costal aridity reflects rain shadows to northeast monsoons the ice-age dry areas of the coast should have been even more extensive that those of modern times from the diminution of monsoon rain. Evidence of lowered water levels in the lake at El Valle concurs with the Lake Valencia and Peten records in suggesting this. In modern times semi-arid patches do not line the entire coast there being rainforest to the seas edge at the eastern end. Our pollen data suggest that this forest barrier was always present and may indeed have been even more extensive in glacial times. It follows that all animals and plants using the Isthmus of Panama as a glacial cycle necessarily penetrated stretches of lowland tropical foorest.