ELSEVIER 0 0 0 6 - 3 2 0 7 ( 9 5 ) 0 0 1 2 2 - 0

Biological Conservation 76 (1996) 219228 Copyright (C 1996 Elsevier Science Limited

Printed in Great Britain. All rights reserved 0006-3207/96/$15,00+0.00

A M A Z O N I A N CONSERVATION IN A C H A N G I N G WORLD

Mark B. Bush Department of Botany, Duke University, Durham. NC 27708, USA

(Received 1 March 1994; accepted 3 September 1995)

Abstract

Prioritization of areas for conservation in Amazonia is based on estimates of modern biodiversity and the distri- bution of endemic species. The refugial hypothesis pro- vided an important conceptual basis for understanding the effects of climatic" change on these reserve areas. The hypothesis predicted that ice-age aridity in Amazonia would have been the dominant force that resulted in mod- ern patterns of endemism. However, recent paleoecologi- cal data indicate that cooling, rather than drying, was the predominant climatic influence on the ice-age Amazon ,forests, and this leads' to a re-evaluation of forces structur- ing Amazonian diversity patterns. Modern Jorest clearance may result in a warmer and drier Amazon basin," conditions non, seen to be without past analog. In the light of these data, assumptions regarding the survival o f forest isolates in a drying landscape must be revised. Habitat functions in the sense of hydrogeomorphic processes and climate are recommended as conservation goals rather than explicitly attempting to save biodiversity. Reserve areas should be established in the expectation of future climatic change and be large enough to allow the ensuing migration of species. Copyright © 1996 Elsevier Science Limited.

Keywords: Amazonia, biodiversity, climatic change, conservation, Holocene, hydrology, Pleistocene, paleo- ecology, refugia.

INTRODUCTION

The goal of most conservationists is to maintain as much biodiversity as possible, even when substantial areas of forest are converted for economic use (Gins- burg, 1987; McNeely, 1992; Western, 1992). Unlike the more developed nations who have almost nothing left to protect, nations with Amazonian territories still have the choice of where to locate nature reserves, how extensive they should be and how many should be established. In Amazonia, the opportunity to establish wildlife reserves may not arise again, for once develop- ment has started in an area, wresting land from com- peting uses is virtually impossible. Recognizing the importance of planning reserves before the area is

Correspondence to: Mark B. Bush. Tel.: (919) 684 4003; Fax: (919) 684 5412; e-Mail: mbush@acpub.duke.edu

developed, the countries sharing Amazonia have each established conservation areas. This planning has taken account of phytogeographic variability, and biodiver- sity. However, for these plans to be successful in the long term, hydrological and climatic regimes must be maintained at both a local and regional scale. The true value of the reserves lies in their existence for perpetuity, and therefore, they should be planned to accommodate as much climatic and environmental change as possible.

Present estimates of Amazonian deforestation sug- gest that about 6% of the natural forest cover has been replaced with degraded forest or grassland (Skole & Tucker, 1993). However, fragmentation of once forested areas and the area indirectly affected by defor- estation through 'edge effects' (Lovejoy et al., 1986), doubles the area of Brazilian Amazonia disturbed by development to about 12% (Skole & Tucker~ 1993). Conservation planning must encompass the likely short-term protection of sites from fragmentation, and the edaphic and erosional consequences of deforesta- tion (Fearnside & Ferreira, 1984~ Myers, 1988). Plan- ners must also look to longer time-scales to estimate the robustness of proposed reserve structures given global climate change on a millennial scale. If we do not intend to save reserves in perpetuity there seems lit- tle point in starting to set aside any land for national parks. Therefore, the reserves must be designed to withstand the scale of climatic change, anthropogenic and natural, that can be predicted.

PREDICTIONS OF CLIMATIC CHANGE IF A M A Z O N I A IS DEFORESTED

The local effect of deforestation is to increase greatly surface soil temperatures, reduce moisture availability in the upper horizons, extend the length of dry seasons, and hence increase soil water deficits (Dickinson, 1991: Salati & Nobre, 1991; Saunders et al., 1991). An increasingly xeric upper soil environment may lead to a general decrease in evapotranspiration, with a conse- quent loss of local humidity, and reduced precipitation (Salati et al., 1979). Reduced humidity may also decrease cloud cover, which could lessen albedo and lead to heating, or decrease the insulation and lead to cooling (Lean & Warrilow, 1989; Salati & Nobre, 1991). Furthermore. changes to surface texture, a

219

220 M. B. Bush

feature with strong correlation with predicted climatic change in General Circulation Models (Pitman et al., 1993), are also likely if humid forests are replaced by more xeric forest or savanna.

If these patterns of local climate and vegetation change are repeated over a wide enough area, the cli- mate and hydrology of the whole Amazon basin may be altered. Salati et al. (1979) estimate that as much as 50% of regional rainfall in Amazonia is due to local evapotranspiration and re-precipitation. Computer sim- ulations for an Amazonia reduced to pasture suggest a 1-3°C warming, a 26% reduction in precipitation, and evapotranspiration reduced by 38% (Shukla et al., 1990). A GCM simulation of a 6-year period under conditions in which Amazonia is deforested suggests a more modest warming, about 0.69°C (Pitman et al., 1993), but similar reductions in precipitation and evapo- transpiration to the estimates of Shukla et al. (1990). The effect of global warming, as a result of atmo- spheric enrichment by greenhouse gases, is likely to be accelerated by widespread Amazonian deforestation (Eden, 1990) and may increase Amazonian tempera- tures by I ~ ° C within the next century (Goddard Insti- tute for Space Studies, pers. comm.). Such a change would exacerbate existing drought stress and lead to the expansion of xeric habitats at the expense of moist forests. The forests most likely to be affected are those that are ecotonal, forming the boundary between moist rainforest and cerrado; a boundary that is most often water- or fire-dependent (Eden, 1990).

HOW MUCH LAND MUST BE SAVED, AND IN WHAT CONFIGURATION?

Nature reserves and national parks often function as islands within a sea of development, and hence island biogeography plays a considerable role in determining optimum size and arrangement of reserves (Diamond, 1976; Sharer, 1990). Based on MacArthur and Wilson's (1967) theory of Island Biogeography, it has been sug- gested that if 50% of an area is maintained as natural vegetation, most species can persist (Odum & Odum, 1972). However, the area required to save different groups of organisms is likely to vary considerably. In Norwegian forests, maintaining avian biodiversity required about one-quarter of the land area needed to safeguard all plant species (S~etersdal et al., 1993). Insects are such an important component of tropical forest biodiversity (Wilson, 1992) that they should figure centrally in conservation planning. As insects are often chemotaxonomically restricted to a particular host foodplant, maintaining insect diversity may require areas similar to those needed to protect plant, rather than bird, diversity. Furthermore, the study of Norwegian forests, like many biogeographic studies, was based on species presence, not on sustainable pop- ulations, nor on the likelihood of survival given contin- ued global climate change (S~etersdal et aL, 1993). The

massive diversity and individual rarity of species within neotropical forests (Richards, 1952) would tend to increase the percentage of the total area that must be protected in order to include all species, and to protect them through time.

Forest fragmentation by development requires plan- ners to consider the arrangement of reserve lands, a dilemma summarized as SLOSS (Single Large Or Sev- eral Small; e.g. Diamond, 1976; Simberloff & Abele, 1976; Terborgh, 1976). The fear exists that if too few reserves are established, the spatial arrangement may preclude the inclusion of all habitat types, phytogeo- graphic units, or endemic populations. If small reserves are chosen they may include the full diversity of the biota but they may not be large enough to maintain minimum viable populations of all species (Diamond, 1976; Terborgh, 1976). Examples of the loss of species following the isolation of an area, termed 'relaxation' are many, e.g. the loss of jaguar and tapir and c. 25-30% of bird species from Barro Colorado Island, Panama, since 1913 (Willis, 1974). The sheer species diversity of Amazonian forests dictates that individual reserves must be large. Estimates of minimum areas to support populations have been as low as 100,000 ha needed to maintain avifaunal biodiversity (Willis, 1979), to more than 250,000 ha required to maintain full faunal and floral diversity (Terborgh, 1974; Myers, 1979; Higgs & Usher, 1980; Terborgh & Winter, 1980). Most recently, it has been suggested that a minimum size for individual reserves should be 1 - 2 million ha if full biodiversity and heterozygosity of large predators is to be maintained (J. Terborgh, pers. comm.). These estimates assume a constant environment; none includes a parameter for future climatic change.

It is politically unlikely that Amazonian forests can be saved in their entirety, so it falls upon the planners to identify areas for varying degrees of development. International pressure, and a growing domestic aware- ness of the resource that they hold, has spurred the Brazilian government into a series of initiatives designed to prioritize areas for conservation. In 1"965, legislation was established in Brazil to ensure that 50% of Amazonia was maintained with some kind of forest cover (Pandolfo, 1974; Goodland & Irwin, 1975). Since then land quotas have been suggested for zoning Brazilian Amazonia that provide c. 150 million ha (40% of Amazonia) as wildlife areas, 110 million ha (29%) for forestry and agro-forestry, 70 million ha (18%) for Indian lands, and 50 million ha (13%) for agricultural colonization (Dub& 1980). Just as 50% of Amazonian land is unlikely to be protected, so too is the 40% sug- gested by Dub6 (1980); a more conservative estimate is that 15-20% may eventually be designated as National Parks (Pandolfo, 1974: Myers, 1979, 1984). At present just 4% of Amazonia lies within National Parks (Eden, 1990).

Since the mid-1970s the Brazilian government has requested the aid of ecologists and biogeographers in

A m a z o n i a n conservat ion 221

helping to identify and prioritize further conservation areas. This action should be lauded as it signifies a commitment to a thoughtful and planned development of Amazonia.

CONSERVATION AND THE REFUGIAL HYPOTHESIS

The proposed conservation areas in Brazilian Amazo- nia have been raised on the perceived distribution of endemic species. A lesser, but still central, construct for defining these areas has been the concept of the exis- tence of Pleistocene refugia. The refugial hypothesis (Hailer, 1969) proposed a speciation model for Amazo- nia that sought to explain pockets of high species endemism among birds. The hypothesis suggested that during the Pleistocene ice-ages Amazonia became dry - - too dry to support tropical rainforest - - and that in its stead great grasslands developed. Only a few refugia of tropical rain forest remained as islands in a sea of savanna. These isolates were in the wettest locations, either along the western flank of the Amazon basin, or on elevated ground. Hilltops tend to be wetter than the adjacent lowlands because of precipitation due to oro- graphic rainfall, and thus the refugia frequently coin- cided with high ground. Implicit assumptions of the 'Refugial Hypothesis' were (1) tha t modern endemic patterns result from Quaternary speciation events; (2) that refugia contained unchanging forests of lowland taxa throughout the Quaternary; (3) that Amazonian temperatures at the last-glacial maximum were within 2°C of present; (4) that precipitation was reduced suffi- ciently to fragment the forests; and (5) that reduced gene flow resulted in allopatric speciation.

Refugialists argue that the isolated forests were arks for the plants and animals of the rainforest, were genet- ically disjunct from other forest areas, and were sites of allopatric speciation. With the return of moist inter- glacial conditions, the forests spread across Amazonia and ousted the grasslands, but the signature of the refuge was still evident in the high proportions of endemic species on the hilltops relative to the interven- ing areas. This hypothesis was so attractive that numer- ous biogeographers have analyzed their data sets in a similar manner and produced patterns of refugia. Data for nymphalid butterflies (Brown, 1982, 1987a), Anol is lizards (Vanzolini, 1970; Vanzolini & Williams, 1981), amphibians (Duellman, 1972, 1982), stingless bees (Camargo, 1978), scorpions (Lourenco & Florez, 1991), and some plant families (Prance, 1982, 1987) have been presented as refugial maps for the last ice-age. By amalgamating these data it was assumed that the greater the concordance of the patterns, the stronger the definition of the refugial boundary. Geomorpholog- ical evidence was also raised to support the refugial hypothesis. Buried stone-lines beneath the terra firme forests of Amazonia were suggested to have been formed during times of reduced vegetation and aridity

(Tricart, 1974; Journaux, 1975). Ab'Sfiber (1982) sug- gested that the stone-lines were formed during a single widespread period of aridity during the terminal Pleis- tocene, and that the presence of white sand soils and sand dunes within the Amazon basin has also been tied to dry events in the Quaternary. As these deposits were undated, their relationship to past climatic events was always uncertain. The best supporting geomorphic evidence for an arid Pleistocene was the deltaic deposition of sands rich in feldspar in sediments of Pleistocene age near the mouth of the Amazon (Damuth & Fairbridge, 1970). As feldspars degrade under humid conditions, it was suggested that their presence was further evidence of Pleistocene aridity (Damuth & Fairbridge, 1970).

Building on the prevailing view of Pleistocene arid- ity, Wetterberg et al. (1976) produced an influential report that prioritized conservation areas using phyto- geographic areas and, with equal weighting, the 'known' Pleistocene refugia. This report has provided the foundation for many later assessments of conserva- tion needs in Amazonia (Rylands, 1991). In 1979, the Brazilian government adopted the location of proposed Pleistocene forest refuges as being the mainstay of its conservation plan (Cgmara, 1983). These areas were designated if (1) the area was indicated to have been a Pleistocene refugium for two or more different groups of organisms; (2) a postulated refugium and unusual or distinct vegetational formations co-occurred there; and (3) they were of apparent ecological significance, not included in the former categories (Cfimara, 1983).

More recently 'Workshop 90' attempted to update the areas prioritized for conservation. Modern biodi- versity of a wide range of organisms governed the loca- tion of proposed conservation areas, although refugial patterns were also included in the analyses (Fig. 1).

A REAPPRAISAL OF THE REFUGIAL HYPOTHESIS

The refugial hypothesis was raised on modern biogeo- graphic distributions of endemic species - - patterns that have been challenged on methodological grounds (Nelson et al., 1990). The timescale of the speciation events that the refugial hypothesis sought to explain has been challenged (Endler, 1982; Heyer & Maxson, 1982) and alternative hypotheses raised through the application of cladistic biology (Cracraft & Prum, 1988; Patterson et al., 1992). The geomorphological data have been re-interpreted and alternative explana- tions advanced (Milliman & Barretto, 1975: Irion, 1984; Salo, 1987). Furthermore, the refugial argument has not been supported by radiocarbon-dated paleo- ecological evidence from Amazonia (Colinvaux, 1987; Salo, 1987).

Data quality for endemic concentrations Nelson et al. (1990) compared the number of plant col- lections made across Amazonia with the number of

222 M. B. Bush

Fig. 1. Sketch map of Amazonia showing [] the location of areas postulated with 60-80% confidence to have been Pleistocene rainforest refugia (Brown, 1987a); ~, areas representing c. 28% of Amazonia, given the top two ranks for conservation prioritiza- tion, as identified by the 'Workshop 90-Priority areas for Conservation in Amazonia', Manaus, 10-20 January 1990.

II, areas where postulated refugia and conservation zones overlap. With permission from Conservation International.

endemic species recorded for each locality. They found a correlation between collecting effort and reported diversity of some families, bearing out the maxim 'if you look you will find'. Sites along the Amazon River, e.g. near the major settlements of Manaus, Santar6m, and Bel6m, and around research bases, appeared from the biodiversity data to have unusually high concentra- tions of endemic species, but following the analysis of Nelson et al. (1990) these appear to be methodological artifacts. It should be emphasized that not all of the data for endemic concentrations can be dismissed in this way, and some local faunas and floras genuinely appear to have concentrations of endemic species, especially at the periphery of Amazonia (Nelson et al., 1990; Bush, 1994).

The reappraisal of geomorphic evidence The geomorphic evidence cited to support refugia has been contested and reinterpreted by Irion (1984) and Salo (1987). The stone-lines appear to be either the leached products of weathering that form pisoliths, or relictual from the formation of the Belterra Clays about 2-5 million years ago. The dunes appear to be pre-Quaternary in age (see Salo (1987) and Colinvaux (1987) for reviews).

The feldspars in the deltaic deposits of the Amazon have also been reinterpreted. The alternate explanation is that the feldspars are reworked materials, eroded during periods of intense fluvial downcutting in response to lowered ice-age sea levels (Milliman & Bar- retto, 1975; Irion, 1984). Thus the refugial argument may not be able to rely for support on available geo- morphic evidence.

Paleotemperature: empirical data Recent papers have described vegetation changes asso- ciated with ice-age cooling in the lowland neotropics (Liu & Colinvaux, 1985; Bush & Colinvaux, 1990; Bush e ta[ . , 1990, 1992; Piperno e ta[ . , 1990; De Oliveira, 1992; Ledru, 1993). Fossil pollen, phytoliths, wood and diatoms provide evidence of montane vegetational ele- ments invading the bottomlands. During times of extreme cooling at c. 33,000 years BP to 30,000 years BP, and again from 14,000 BP to 12,000 BP a vegeta- tional descent of about 1500-1800 m is recorded in Colombia, Ecuador, Brazil and Panama (van der Ham- men et al., 1981; Hooghiemstra, 1984; Bush et al., 1990; Bush, 1994). Paramo, subparamo and Andean forest elements shuffled downslope, bringing C3 grasses, Drimys, Podcarpus, and Alnus as low as 1000 m elevation in eastern Ecuador. Of particular interest is the descent of grasses using the C3 photosynthetic pathway. In tropical South America these grasses are presently restricted to paramo and subparamo ecosys- tems, suggesting that they moved about 1500 - 2000 m vertically downslope. The scale of this descent of vege- tation is replicated in records from the high Andes to the interior of lowland Brazil as far south as Minas Gerais (Hooghiemstra, 1984; Bush et al., 1990; Hooghiemstra & Sarmiento, 1991; De Oliveira, 1992; Ledru, 1993; Bush, 1994). Such a shift of vegetation corresponds to about a 7.5 - 9°C change in tempera- ture (assuming a moist air adiabatic lapse rate of 5°C per 1000 m of ascent - - Canadas, 1983; Bush et al., 1990; Bush, 1994). These extremely cold temperatures lasted for several millennia at a time, but for most of

A m a z o n i a n conservation 223

the Pleistocene the neotropics experienced a cooling of about 5°C relative to present (Colinvaux, 1987, 1993; Bush & Colinvaux, 1990; Bush, 1994).

These terrestrial paleoecological data appeared irre- concilable with sea-surface temperature estimates that invoked a cooling of just I°C at 18,000 aP (CLIMAP, 1976). However, new paleotemperature evidence from Caribbean corals indicates that the sea temperature close to land was cooled by about 5°C at 18,000 Bp (Guilderson et al., 1994). The coralline data, based on radiometric analysis of gtsO/~60, and ratios of stron- tium/calcium, are the strongest oceanic paleotempera- ture data yet obtained from the equatorial region; their concordance with the terrestrial record is striking.

Paleoprecipitation: empirical and modeled data Not all paleoecological records from lowland South America show a pure cooling signal. Records from the periphery of Amazonia reveal histories of lowered Pleistocene lake levels. Savanna elements are observed to have encroached into what are now the drier areas of Amazonia (Absy & van der Hammen, 1976; Absy et al., 1991). However, pronounced drying was brief, rela- tive to the overall length of the Pleistocene (van der Hammen et al., 1981; Absy et al., 1991), and does not appear to have been of sufficient strength to eradicate rainforest from many areas of Amazonia. Sites investi- gated in the western and northwestern Amazon reveal histories of continuous cool moist forest presence in the Pleistocene and warm moist forest throughout the Holocene (Liu & Colinvaux, 1985, 1988; Bush & Colinvaux, 1988; Bush et al., 1990). It appears that the ice-ages were times of cooling and modest drying in the Amazon basin.

Perhaps the best estimate for ice-age drying comes from the climate models of Gates (1976), Manabe and Hahn (t977), and Kutzbach and Guetter (1986) which suggest a 10-20% reduction in precipitation at 18,000 aP. In an ice-age world with atmospheric CO~ concen- trations reduced to 170 ppm (Barnola et al., 1987), plants are likely to transpire greater amounts of water as their stomates must remain open longer in order to accumulate sufficient CO2 for photosynthesis. Thus the reduced concentrations of CO, are likely to heighten drought stress. Bush (1994) suggests, therefore, that the larger of the modeled precipitation reductions be accepted as most closely approximating the effective drought stress experienced by Amazonian plants.

The precipitation map of Amazonia can be redrawn (Fig. 2) to accommodate this 20% reduction in rainfall, and to show the areas of overlapping endemism for birds, plants and butterflies (Brown, 1987). This model of a 20% reduction in precipitation accommodates the observed vegetation changes in every paleoecological datum from the neotropical region (Bush, 1994). Paleo- ecological records that show pronounced Pleistocene drying are seen to lie in regions supporting less than 2000 mm of rain, and therefore were ecotonal areas that may have supported forest, cerrado, or savanna, at various times during the ice-age. Areas receiving more than 2000 mm of rain are likely to maintain a forest (albeit with a novel assortment of lowland species). In these areas that were permanently wet the paleoecologi- cal data show the continuous presence of moist forest. In receiving less than 1500 mm the vegetation would have been a closed-canopy dry forest, cerrado oz" savanna, according to local edaphic conditions or

Fig. 2. Sketch map of Amazonia showing the location of paleoecological sites, and whether they indicate the expansion of non- forest elements, relative to a modeled 20% reduction in precipitation during the last glacial cycle (after Bush, 1994). U, area pos- tulated to receive >2000 mm precipitation with 20% drying; [~, area postulated to receive 2000-1500 mm precipitation with 20% drying; 1--], area postulated to receive <1500 mm precipitation with 20% drying; ~ , paleological datum indicating drying (and

consistent with cooling); ©, paleological datum indicating cool, moist conditions.

224 M. B. Bush

fire regimes. At precipitation levels close to 1500 mm per annum there is an increased likelihood of the lake drying out completely, either through evaporative loss or through the lowering of the water table. Lakes that lie in the region receiving <1500 mm rainfall per annum often show protracted dry periods in the late Plei- stocene, or only started to hold water with the rise of Holocene water tables.

If these precipitation data are compared with the postulated centers of endemism (Brown, 1987), it is evi- dent that the loci of endemic richness lie at the edge of the moist ice-age forests, not at their center (Fig. 3). This pattern is inconsistent with the refugial hypothe- sis, as endemism appears to reflect those areas most likely to be subject to environmental change, rather than those likely to have a constant environment. The 'refugia' lie in areas most likely to be invaded by mon- tane taxa surging down the flank of the Andes, or in ecotonal areas where small changes in precipitation result in savanna expansion during dry phases and the maintenance of rainforest during wetter ones.

The cooling hypothesis vs the refugial hypothesis The empirical data deny the possibility that Amazonian hilltops had temperatures within 2°C of present and that these areas could have supported an unchanged flora throughout the northern hemispheric glacial max- ima. Indeed, the hilltops should be seen as the areas most likely to change as they would have been too cold for many lowland rainforest taxa to persist. Amazonia was not warm and arid at the last glacial maximum as suggested by the refugialists, but cold and slightly drier than present.

The critical point to appreciate from these two differ-

ent accounts is not the actual speciation mechanism, but the implied environmental differences in the endemic-rich regions relative to the core of Amazonia and the ecological stability, or lack of it, of isolated patches of forest. In the refugial account of events, the forest isolates are purported to have survived as wet areas in an overall dry landscape. According to the new paleoclimatic evidence, the Pleistocene was not warm and arid, but cool, and the rainforest was not broken up into isolates. During the Pleistocene the modern centers of endemism were often relatively dry areas (or areas invaded by cold-adapted Andean plants) lying at the periphery of a vast moist forest block. The center and western portion of the Amazon basin were not savanna but moist forest moderating climatic change.

IMPLICATIONS OF A REVISED SPECIATION MODEL FOR CONSERVATION BIOLOGY

Prioritization of conservation areas has been based on modern biogeographic areas of high biodiversity (Rylands, 1991). For these areas to be valuable reserves, it is necessary to predict how they will be affected by development elsewhere in Amazonia. The 'refugial hypothesis' has played a vital part in formu- lating a conservation strategy for Amazonia. This hypothesis was valuable in introducing a temporal ele- ment to the debate over conservation, but it may have provided a false sense of ecosystem stability for por- tions of the Amazon basin. Worse, it may have sug- gested that areas of high biodiversity could, unchanged, withstand landscape alteration in the surrounding region. Because many of the areas of high biodiversity were thought to have been refugia, it was assumed that

Fig. 3. Sketch map showing the location of endemic centers relative to precipitation assuming a 20% effective drying during glacial times. Endemic centers lie either in the marginal forest areas or within the range of migration of the Andean flora (after Bush, 1994). i , overlap of areas of bird, butterfly and plant endemic richness; ~, overlap of two areas of bird, butterfly, or plant

endemic richness; I~, subtropical vegetation. Estimated precipitation: 17~, >2000 ram; R, 2000-1500 mm; I-7, <1500 mm.

Amazonian conservation 225

they survived in isolation for thousands of years during glacial conditions. As one of the predictions of environ- mental change wrought by deforestation of Amazonia is increasing aridity, the refugial hypothesis offers the hope that these prioritized conservation areas would be well located to survive drying, just as they did in the Pleistocene. However, the falsification of the climatic assumptions of the refugial hypothesis leads to the need to reassess conservation strategies based on that model.

The endemic-rich areas can be divided into two basic categories: those that have very high rainfall (>3000 ram), and those that, given a 20% drying, become eco- tonal between moist forest and cerrado. The wettest sites are primarily the ones along the Andean flank, and these will be sensitive to changes in temperature, particularly cooling, but may be relatively insensitive to changes in precipitation. Contrastingly, the ecotonal areas, far from being areas capable of surviving the buffets of increasing aridity, are most sensitive to cli- matic change. They were not separated by savanna during ice-age time, but lay at the margin of a massive forest block that still received more than 2000 mm of rain, with high evapotranspiration rates helping to maintain a moisture-laden environment. The forest would have moderated some of the effects of drying, and would have offered a wetter retreat for some species during times of exceptional drought. Thus the ecotonal areas, identified as areas of exceptionally high biodiversity, cannot stand alone. To predicate conser- vation strategies based on their existence as isolates is unwise.

Reliance on small islands of vegetation for conserva- tion will lead to the reduction of both faunal and floral diversity, especially as these areas come under the pres- sure of climatic change. Estimates of minimum sized areas to protect populations, e.g. the 250,000 ha reserve size (Terborgh, 1974; Myers, 1979; Higgs & Usher, 1980), are only appropriate so long as the regional cli- mate remains constant (Peters & Darling, 1985). Reserves of this size may be large enough to maintain breeding populations of most organisms under steady- state conditions, but they are too small to be self-regu- lating climatic units in an increasingly arid landscape (Peters & Darling, 1985; Peters, 1998).

Rather than choosing conservation areas for endemic richness alone, it would be better to preserve enough of the core Amazonian forest to mitigate local climatic change, and to concentrate on the areas that are likely to stay moist enough, given the scale of deforestation, to support tropical forest. This may mean taking a sec- tion of the ecotonal area and a portion of the adjacent forest, so that if the climate becomes drier, species can migrate into a wetter area.

Other climatic changes may drive species to migrate between elevations. At the edge of massifs, species will respond to local or global temperature changes by migrating up or down mountain sides, e.g. the Andes, the Guianan Shield or the Atlantic coastal mountains.

At present it is easiest to envisage a warming trend that would lead to an upslope migration of taxa, but a renewed stadial, such as the little ice-age of c. 1400 - 1800 AD, could at some point result in the descent of vegetation types. It is essential, therefore, that parks allow future migration of taxa in response to climatic change. Ideally protected lands would form belts rang- ing from the highest peak down into the lowlands. In Ecuador, such a park has already been established, e.g. the 270,000 ha Sangay National Park that extends from 5000 m elevation down to 800 m elevation (Anhalzer, 1988). Another example is the Brazilian Pico da Neblina reserve, close to the Venezuelan border, that includes an elevational range from 3000 m down to about 100 m above sea level.

DEVELOPMENT OF NON-RESERVE LANDS

Preserving Amazonian biodiversity is not simply a mat- ter of identifying conservation areas. These national parks will not function in isolation, and it is critical that the intervening areas are developed in such a way that the moisture cycles of the Amazonian climate are maintained. Not all of the core forests of Amazonia have to be set aside as untouched wilderness areas, but their development should protect them from reduction to pasture. No studies are available detailing the long- term moisture flux from sustainable silviculture or agroforestry, but climate models indicate that they would provide a surface that more closely resembles natural forest than would pasture (Shukla et al., 1990). Sustainable silviculture, mixed perennial polycultures or interplanting of crops with economically important trees may all be appropriate land uses that would pro- vide an income while minimizing local climatic change.

The rights of indigenous peoples have come to the forefront of the conservation debate, and present plan- ners with a difficulty. Their right to a national home- land is undeniable, and the designation of areas as protected 'Indian lands' may preserve large tracts of forest, but this relies on the indigenous people choosing to maintain forest on their land. As the control of these lands is passed to the indigenous population, no guar- antee exists as to their future use (Eden, 1990). Fur- thermore, it is possible that their goals, even their vision of what constitutes sustainability, do not match those of the conservationists (Redford & Maclean Stearman, 1993). The hope is that the traditional ways of hunting and harvesting would maintain the forest, but there is increasing evidence of those traditions being abandoned in favor of 'progress' (Redford & Maclean Stearman, 1993).

About 60% of present forest destruction in Amazo- nia can be attributed to slash-and-burn cultivation, a non-sustainable cycle of land use that leads to erosion, falling crop yields and further forest destruction (Jacobs, 1981). Clearly a complete conservation policy for Amazonia should include an educational element

226 M. B. Bush

that promotes the sustainable development of the land, and this includes the promotion of these ideals with the new generations of indigenous peoples. Furthermore, any concept of sustainability must include not only the harvesting of the land, but also the maintenance of the local climate. I f the interstices between conservation areas are degraded, the parks themselves will surely fail.

CONCLUSIONS

Preserving endemic species and maximal biodiversity must be an essential plank of Amazonian conservation strategies, and this means prioritizing areas with high species concentrations. However, given future climatic change, protecting these areas alone will not provide a successful conservation policy. A configuration of reserves without taking into account the likely effects of future climate change may lead to severe management problems in the future. The assumption that areas of high modern biodiversity are stable during times of cli- matic change, as suggested by the refugial hypothesis, is inconsistent with empirical paleoecological evidence. Far from being stable 'refugia', these areas appear to have been the ecotonal edge of the Pleistocene rain- forests. If Amazonian development results in decreased evapotranspiration and reduced precipitation, these areas will once again be ecotonal, but this time without the core of moist forest, and their species subject to enhanced likelihood of extinction. Conservation initia- tives directed to maintaining maximal biodiversity should allow for the migration of species in the face of future climate change. Such change may be manifested as a cooling or a warming in association with drying, and reserves should be established with as much eleva- tional and moisture range as possible. 'Retreat areas', where species can migrate during times of change, should be incorporated into the establishment of reserve areas. Reserves should be arranged so as to provide large continuous blocks of vegetation contain- ing a full range of natural habitats. It is unlikely that all of the sites presently identified as being of excep- tional biological interest can be incorporated into such large tracts of reserve land. Such pragmatic decisions will lead to the sacrifice of some areas of high biodiver- sity so that some very large reserves can be maintained. Even with this trade-off, less biodiversity would be lost than within undersized island reserves during times of climate change.

In the near future, the local climate of Amazonia is likely to become warmer and drier as a result of defor- estation. The result will be environments without past analog. The climatic divergence and hence the vegeta- tion change will become more extreme a s increasing areas of the core forest are converted to pasture. In order to maintain areas presently rich in endemic species, it is essential that the development of central Amazonia, an area generally considered to be poor in

these species, be done in such a way that present climatic conditions are maintained. Areas between the conservation areas, wherever possible, should be treated as buffer zones in which land management, if not actu- ally promoting conservation, is not prejudicial to its success (sensu Saunders et al., 1991). It is recom- mended, therefore, that management plans prioritize central Amazonia as a zone of regulated land use centered on sustainable silviculture or agroforestry, or (with the reservations noted above) for use as 'Indian lands'.

A CK N O W LED G EMEN TS

My thanks to Daniel Livingstone. Paulo De Oliveira, George De Busk, Adrienne Pilmanis, Renato Cintra and Peter Jipp for comments on the manuscript. Paul Colinvaux and his research team were integral to gath- ering the data for the paleoecological data presented and for many stimulating debates. The paleoecological work was funded by NSF grants ATM 8709685, BSR 9007019 and SES-881110.

REFERENCES

Ab'Sfiber, A. N. (1982) Paleoclimate and paleoecology of Brazilian Amazonia. In Biological diversification in the trop- ics, ed. G. T. Prance. Columbia University Press, New York, pp. 41-59.

Absy, M. L., Cleef, A, Fournier, M. et al., (1991). Mise en 6vidence de quatre phases d'ouverture de la for~t dense dans le sud-est de L'Amazonie au cours des 60,000 derni6res ann6es. Premiere comparaison avec d'autres regions tropicales. C. R. Acad. Sci. Paris, Series II, 312, 673-8.

Absy, M. L. & van der Hammen, T. (1976). Some palaeoeco- logical data from Rond6nia, southern part of the Amazon Basin. Act. Amazon., 6, 293-9.

Anhalzer, J. (1988). Sangay Parque nacional. Azuca, Quito. Barnola, J. M., Raynaud, D., Korotkevitch, Y. S. &

Lorius, C. (1987). Vostoc ice core provides 160,000 year record of atmospheric CO 2. Nature, Lond., 329, 408-14.

Brown, K. S. Jr (1982). Paleoecological and regional patterns of evolution in neotropical forest butterflies. In Biological diversification in the tropics, ed. G. T. Prance. Columbia University Press, New York, pp, 25-308.

Brown, K. S. Jr (1987). Biogeography and evolution of neotropical butterflies. In Biogeography and quaternary his- tory in tropical America, ed. T. C. Whitmore & G, T. Prance. Oxford Science Publications, Oxford, pp. 66-104.

Bush, M. B. (1994). Amazonian speciation: a necessarily complex model. J. Biogeogr., 21, 5-17.

Bush, M. B. & Colinvaux, P. A. (1988). A 7000-year pollen record from the Amazon lowlands, Ecuador. Vegetatio, 76, 141-54.

Bush, M. B. & Colinvaux, P. A. (1990). A long climatic and vegetation record from lowland Panama. J. Veget. Sci,, 1, 105-18.

Bush, M. B., Colinvaux, P. A., Wiemann, M. C., Piperno, D. R. & Liu, K.-b. (1990). Pleistocene temperature depression and vegetation change in Ecuadorian Amazonia. Quat. Res., 34, 330-45.

Bush, M. B., Piperno, D. R., Colinvaux, P. A., De Oliveira, P. E., Krissek, L. A., Miller, M. C. & Rowe, W. E. (1992). A 14,300 year paleoecological profile of a lowland tropical lake in Panama. EcoL Monogr., 62, 251-75.

Amazonian conservation 227

C/tmara, I de G. (1983), Tropical moist forest conservation in Brazil. In Tropical rain forest ecology and management, ed. S. L. Sutton, T. C. Whitmore & A. C. Chadwick. Black- well Scientific Publications, Oxford, pp. 413-21.

Camargo, J. M. F. (1978). O grupo Partamona (Partamona) testacea (Klug); suas especies, distribuicao e differenciacao geografica (Meliponinae, Apidae, Hymenoptera). Acta Amaz., 10 (suppl.), 1-175.

Canadas, L. C. (1983). El mapa biocliomatico y ecologico del Ecuador. Ministerio Agricultura Pronareg, Quito.

CLIMAP Project Members (1976). The surface of the ice-age Earth. Science, N. Y., 191, 1131-7.

Colinvaux, P. A. (1987). Amazon diversity in the light of the paleoecological record. Quat. Sci. Rev., 6, 93-114.

Colinvaux, P. A. (1993). Pleistocene biogeography and diver- sity in tropical forests of South America. In Biological rela- tionships between Africa and South America, ed. P. Goldblatt. Yale University Press, New Haven, CT, pp. 473-99.

Cracraft, J. & Prum, R. O. (1988). Patterns and processes of diversification: speciation and historical congruence in some neotropical birds. Evolution, 42, 603-20.

Damuth, J. E. & Fairbridge, R. W. (1970). Equatorial Atlantic deep-sea arkosic sands and ice-age aridity in tropi- cal South America. Bull. Geol. Soc. Amer., 81, 189-206.

De Oliveira, P. E. (1992). A palynological record of late Quaternary vegetational and climatic change in southeast- ern Brazil. PhD thesis, Ohio State University, Columbus, OH.

Diamond, J. M. (1976). Island biogeography and conserva- tion: strategy and limitations. Science, N. K, 193, 1027-9.

Dickinson, R. E. (1991). A commentary on: probable impact of deforestation on hydrological processes. Climatic Change, 19. 175.

Dub& Y. (1980). Em busca de uma politica florestal para a Amaz6nia brasiliera. Superintend~ncia do Desenvolvi- mento da Amaz6nia, Beldm.

Duellman, W. E. (1972). South America frogs of the Ityla rostrata group (Amphibia, Anura, Hylidae). Zool. Meded., 47, 177 92.

Duellman, W. E. (1982). Quaternary climatic-ecological fluc- tuations in the lowland tropics: frogs and forest. In Biologi- cal diversification in the tropics, ed. G. T. Prance. Columbia University Press, New York, pp. 389402.

Eden, M. J. (1990). Ecology and land management in Amazo- nia. Belhaven Press, London.

Endler, J. A. (1982). Pleistocene forest refuges: fact or fancy? In Biological diversification in the tropics, ed. G. T. Prance. Columbia University Press, New York, pp. 641-57.

Fearnside, P. M. & Ferreira, G. de L. (1984). Roads in Rond6nia: highway construction and the farce of unpro- tected forest reserves in Brazil's Amazon forest. Environ. Conserv., 11,358-60.

Gates, W. L. (1976). Modeling the ice-age climate. Science, N. K, 191, 1138-44.

Ginsburg, J. R. (1987). What is conservation biology? Trends' Ecol. & Evolut., 2, 2624.

Goodland, R. J. A. & Irwin, H. S. (1975). Amazon jungle: green hell to red desert? Elsevier Scientific Publishing, Amsterdam.

Guilderson, T. P., Fairbanks, R. G. & Rubenstone, J. L. (1994). Tropical temperature variations since 20,000 years ago: modulating interhemispheric climate change. Science, N. Y., 263, 663 5.

Haffer, J. (1969). Speciation in Amazonian forest birds. Sci- ence, N.Y., 165, 131 7.

Heyer, W. R. & Maxson, L. R. (1982). Distributions, rela- tionships, and zoogeography of lowland frogs. In Biologi- cal diversification in the tropics, ed. G. T. Prance. Columbia University Press, New York, pp. 375 88.

Higgs, A. J. & Usher, M. B. (1980). Should nature reserves be large or small? Nature, Lond., 285, 568-9.

Hooghiemstra, H. (1984). Vegetational and climatic history oJ the high plain of Bogota, Colombia: a continuous record of the last 35 million years. Gantner Verlag, Vaduz.

Hooghiemstra, H. & Sarmiento, G. (1991). Long continental pollen record from a tropical intermontane basin: late Pliocene and Pleistocene history from a 540-meter core. Episodes, 14, 107-15.

Irion, G. (1984). Sedimentation and sediments of Amazonian rivers and evolution of the Amazonian landscape since Pliocene times. The Amazon. Limnology and landseape of a might)' tropical river and its basin, ed. J. Sioli. Dr W. Junk, Dordrecht, pp. 2014.

Jacobs, M. (1981). An introduction to tropical Jorests. Springer Verlag, Hamburg.

Journaux, A. (1975) Recherches geomorphologiques en Ama- zonie Bresilienne. Bull. Centre Geomorph. Caen ( CNRS), 20, 1-67.

Kutzbach, J. E. & Guetter, P. J. (1986). The influence of changing orbital parameters and surface boundary condi- tions on climate simulation for the past 18,000 years. J. Atrnos. Sci., 43, 1726-59.

Lean, J. & Warrilow, D. A. (1989). Simulation of the regional climatic impact of Amazonian deforestation. Nature, Lond., 342, 411-13.

Ledru, M.-P. (1993). Late Quaternary environmental and climatic changes in Central Brazil. Quat. Res., 39, 90-8.

Liu, K.-b. & Colinvaux, P. A. (1985). Forest changes in the Amazon basin during the last glacial maximum. Nature, Lond., 318, 556-7.

Liu, K.-b. & Colinvaux, P. A. (1988). A 5200-year history of Amazon rain forest. J. Biogeogr., 15, 23148.

Lourenco, W. R. & Florez, E. (1991). Scorpions (Chelicerata) from Colombia: IlI. The scorpion thuna of Pacific region (Choco), with some biogeographic considerations. Amazo- niana, 1|, 119-34.

Lovejoy, T. E., Bierregaard, R. O. Jr, Rylands, A. B., Mal- colm, J. R., Quintela, C. E., Harper, L. H., Brown, K. S. Jr, Powell, A. H., Powell, G. V. N., Schubart, O, R. & Hays, M. (1986). Edge effects and other effects of isolation on Amazonian forest fragments. Conserv. Biol. ed. M. Soule. Sinauer Assoc. Inc. Sunderland, MA. pp. 257 85.

MacArthur, R. H. & Wilson, E. O. (1967). The theory o[" island biogeography. Princeton Press, Princeton, NJ.

McNeely, J. A. (1992). The biodiversity crisis: challenges for research and management. In Conservation of biodiversity for sustainable development, ed. O. T. Sandlund, K. Hindar & A. H. D. Brown. Scandinavian University Press, Oslo, pp. 15-26.

Manabe, S. & Hahn, D. G. (1977). Simulation of the tropical climate of an ice-age. J. Geophys. Res., 82, 3889-911.

Milliman, J. D & Barretto, H. T. (1975) Relict magnesium cal- cite oolite and subsidence of the Amazon shelf. Sedimentol- ogy, 22, 13745.

Myers, N. (1979). The sinking ark. A new look at the problem of disappearing species. Pergamon Press, Oxford.

Myers, N. (1984). The primary source. Tropical jorests and our future. W. W. Norton, New York.

Myers, N. (1988). Tropical forests and their species: going, going...? In Biodiversity, ed. E. O. Wilson & F. M. Peter. National Academy Press, Washington, DC, pp. 28 35.

Nelson, B. W., Ferreira, C. A. C., da Silva, M. F. & Kawasaki, M. L. (1990). Endemism centres, refugia and botanical collection density in Brazilian Amazonia. Nature, Lond., 345, 714-16.

Odum, E. P. & Odum, H. T. (1972). Natural areas as neces- sary components of man's total environment. Trans. NA Wildl. Nat. Resour. Conj., 37th, Mexico City, pp. 178 89.

228 M. B. Bush

Pandolfo, C. (1974). Estudios basieos para o estabelecimento de uma politica de desenvolvimento dos recursos florestais e de uso racional das terras da Amaz6nia, 2nd edn. Superin- tend~ncia do Desenvolvimento da Amaz6nia, Bel6m.

Patterson, B. D., Pacheco, V. & Ashley, M. V. (1992). On the origins of the western slope region of endemism: systematics of fig-eating bats, genus Artibeus. Mem. Mus. Hist. Nat., UNMSM, 21, 189-205.

Peters II, R. L. (1988). The effect of global climatic change on natural communities. In Biodiversity, ed. E. O. Wilson. National Academy Press, Washington, DC, pp. 450-61.

Peters, II, R. L. & Darling, D. J. S. (1985). The greenhouse effect and nature reserves: global warming would diminish biological diversity by causing extinctions among reserve species. BioSeience, 35, 707-17.

Piperno, D. R., Bush, M. B. & Colinvaux, P. A. (1990). Paleo- environments and human settlement in late-glacial Panama. Quat. Res., 33, 108-16.

Pitman, A. J., Durbidge, T. B. & Henderson-Sellers, A. (1993). Assessing climate model sensitivity to prescribed deforested landscapes. Int. J. Climatol., 13, 879-98.

Prance, G. T. (1982). Forest refuges evidence from woody angiosperms. In Biological diversification in the tropics, ed. G. T. Prance. Columbia University Press, New York, pp. 137-56.

Prance, G. T. (1987). Biogeography of neotropical plants. In Biogeography and Quaternary history in tropical America, ed. T. C. Whitmore & G. T. Prance. Oxford Science Publi- cations, Oxford, pp. 46-65.

Redford, K. H. & Maclean Stearman, A. (1993). Forest dwelling native amazonians and the conservation of biodi- versity: interests in common or in collision? Conserv. Biol., 7, 248-55.

Richards, P. W. (1952). The tropical rain forest. An ecological study. Cambridge University Press, Cambridge.

Rylands, A. B. (1991). The status of conservation areas in the Brazilian Amazon. World Wildlife Fund, Washington, DC.

S~etersdal, M., Line, J. M. & Birks, H. J. B. (1993). How to maximize biological diversity in nature reserve selection: vascular plants and breeding birds in deciduous wood- lands, western Norway. Biol. Conserv., 66, 131-8.

Salati, E. & Nobre, C. A. (1991). Possible climatic impacts of tropical deforestation. Climatic Change, 19, 177-96.

Salati, E., Dall'Olio, A., Matsui, E. & Gat, J. R. (1979). Recycling of water in the Amazon basin: an isotopic study. Water Resour. Res., 15, 1250-8.

Salo, J. (1987). Pleistocene forest refuges in the Amazon: evaluation of the biostratigraphical, lithostratigraphical and geomorphological data. Annal. Zool. Fenn., 24, 203-11.

Saunders, D. A., Hobbs, R. J. & Margules, C. R. (1991). Bio- logical consequences of ecosystem fragmentation: a review. Conserv. Biol., 5, 18-32.

Shafer, C. L. (1990). Nature reserves. Island theory and conservation practice. Smithsonian Institution Press, Wash- ington, DC.

Shukla, J., Nobre, C. & Sellers, P. (1990). Amazon deforesta- tion and climate change. Science, N. E, 247, 1322-6.

Simberloff, D. S. & Abele, L. G. (1976). Island biogeography theory and conservation practice. Seience, N. Y., 191, 285-6.

Skole, D. & Tucker, C. (1993). Tropical deforestation and habitat fragmentation in the Amazon: satellite data from 1978 to 1988. Science, N. Y., 260, 1905-10.

Terborgh, J. (1974). Preservation of natural diversity: the problem of extinction prone species. BioScience, 24, 715-22.

Terborgh, J. (1976). Island biogeography and conservation: strategy and limitations. Science, N. Y., 193, 1029-30.

Terborgh, J. & Winter B. (1980). Some causes of extinction. In Conservation biology." an evolutionary-ecological perspec- tive, ed. M. E. Soul6 & B. A Wilcox. Sinauer Associates, Sunderland, MA, pp. 119-33.

Tricart, J. (1974). Existence au Quaternaire de p6riodes s6ches en Amazonie et dans les r6gions voisines. Rev. Geo- morph. Dynam., 223 145-58.

van der Hammen, T., Barelds, J., De Jong, H. & De Veer, A. A. (1981). Glacial sequence and environmental history in the Sierra Nevada del Cocuy (Colombia). Palaeogeogr., Palaeoclim., Palaeoecol., 32, 247-340.

Vanzolini, P. E. (1970). Zoologia sistematica, geografia e a origem das especies. Inst. Geogr. S~o Paulo, Serie Teses e Monogr., 3, 1-56.

Vanzolini, P. E. & Williams, E. E. (1981). The vanishing refuge: a mechanism for ecogeographic speciation. Pap. Avulsos Zool. ($6o Paulo), 34, 251-5.

Western, D. (1992). The biodiversity crisis: a challenge for biology. Oikos, 63, 29-38.

Wetterberg, G. B., Padua, M. T. J., de Castro, C. S. & de Vascocellos, J. M. C.,(1976). Uma analise de prioridades em conservacao da natureza. Projeto de Desenvolvimento e pesquisa florestal ( PRODEPEF) PNUD/FAO/IBDF/BRA- 45, Serie Tecnica, 8.

Willis, E. O. (1974). Populations and local extinctions of birds on Barro Colorado Island, Panama. Ecol. Monogr., 44, 153-69.

Willis, E. O. (1979). The composition of avian communities in remanescent woodlots in southern Brazil. Pap. Avulsos Zool. (Sdo Paulo), 33, 1-25.

Wilson, E. O. (1992). The diversity of life. Norton, New York.