The Potential Negative Impacts of Global Climate Change on Tropical Montane Cloud Forests

Earth-Science Reviews 55Ž. 2001 73–106 www.elsevier.comrlocaterearscirev

The potential negative impacts of global climate change on tropical montane cloud forests

Pru Foster ) Institute for Study of the Earth’s Interior, Misasa, Tottori-ken 682-0193, Japan Received 17 October 2000; accepted 12 March 2001

Abstract

Nearly every aspect of the cloud forest is affected by regular cloud immersion, from the hydrological cycle to the species of plants and animals within the forest. Since the altitude band of cloud formation on tropical mountains is limited, the tropical montane cloud forest occurs in fragmented strips and has been likened to island archipelagoes. This isolation and uniqueness promotes explosive speciation, exceptionally high endemism, and a great sensitivity to climate. Global climate change threatens all ecosystems through temperature and rainfall changes, with a typical estimate for altitude shifts in the climatic optimum for mountain ecotones of hundreds of meters by the time of CO2 doubling. This alone suggests complete replacement of many of the narrow altitude range cloud forests by lower altitude ecosystems, as well as the expulsion of peak residing cloud forests into extinction. However, the cloud forest will also be affected by other climate changes, in particular changes in cloud formation. A number of global climate models suggest a reduction in low level cloudiness with the coming climate changes, and one site in particular, Monteverde, Costa Rica, appears to already be experiencing a reduction in cloud immersion. The coming climate changes appear very likely to upset the current dynamic equilibrium of the cloud forest. Results will include biodiversity loss, altitude shifts in species’ ranges and subsequent community reshuffling, and possibly forest death. Difficulties for cloud forest species to survive in climate-induced migrations include no remaining location with a suitable climate, no pristine location to colonize, migration rates or establishment rates that cannot keep up with climate change rates and new species interactions. We review previous cloud forest species redistributions in the paleo-record in light of the coming changes. The characteristic epiphytes of the cloud forest play an important role in the light, hydrological and nutrient cycles of the cloud forest and are especially sensitive to atmospheric climate change, especially humidity, as the epiphytes can occupy incredibly small eco-niches from the canopy to crooks to trunks. Even slight shifts in climate can cause wilting or death to the epiphyte community. Similarly, recent cloud forest animal redistributions, notably frog and lizard disappearances, may be driven by climate changes. Death of animals or epiphytes may have cascading effects on the cloud forest web of life. Aside from changes in temperature, precipitation, and cloudiness, other climate changes may include increasing dry seasons, droughts, hurricanes and intense rain storms, all of which might increase damage to the cloud forest. Because cloud forest species occupy such small areas and tight ecological niches, they are not likely to colonize damaged regions. Fire, drought and plant invasionsŽ. especially non-native plants are likely to increase the effects of any climate change damage in

) Tel.: q81-858-43-3827; fax: q81-858-43-3450. E-mail address: [email protected]Ž. P. Foster .

0012-8252r01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved. PII: S0012-8252Ž. 01 00056-3 74 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 the cloud forest. As has frequently been suggested in the literature, all of the above factors combine to make the cloud forest a likely site for observing climate change effects in the near future. q 2001 Elsevier Science B.V. All rights reserved.

Keywords: cloud forest; tropical montane cloud forest; climate change impacts; relative humidity; paleovegetation; epiphytes; cloud formation changes

1. Introduction cycle. At first glance, this may seem to lead to increasing clouds, but detailed studies have indicated As one ascends in altitude up a tropical mountain, a tendency for a decrease in low-level cloudiness the trees typically decrease in stature and the leaves that envelopes cloud forestsŽ. see Section 3.3 . Long- become smaller, harder and thickerŽ Whitmore, term observationsŽ. Pounds et al., 1999 and prelimi- 1989. . If the mountain intersects the cloud cap, these nary modellingŽ. Still et al., 1999; Foster, 2001 at effects are magnified. Twisted stunted trees are en- the Monteverde Cloud Forest in Costa Rica suggest countered and the leaves become even more akin to that the height of the cloudbank is already rising, desert xeromorphic leaves. The dwarfed trees in the resulting in less cloud immersion, and thus driving cloud cap are laden with a heavy mass of epiphytes local extinctions through enhanced dryness. whose roots suspend in air and a new group of Exactly how the cloud immersion impacts the animal species distinct from below the cloud cap forest is unknown, but certain effects are obvious— is encountered. The frequent mists add to the high relative humidityŽ. RH and reduced sunshine atmosphere and give this special ecosystem its are two of the most important ones. The enhanced most commonly known name—the cloud forest RH contributes to the large epiphyte mass, as the Ž.Stadtmuller,¨ 1987 . aerial plants capture water directly from the clouds As the global climate changes, the hydrological or via enhanced condensation. In addition, a high RH cycle will undoubtedly change. Increasing sea sur- explains the presence of so many kinds of unique face temperatures will increase evaporation from the amphibian species. However, the explanation of the sea and pump more water into the tropospheric water stunted and twisted nature of the cloud forest trees

Nomenclature:

Notation CF cloud forest ENSO El Nino˜ Southern Oscillation ET evapotranspiration GCM global circulation model HP horizontal precipitation LAI leaf area index LGM last glacial maximum LMRF lower montane rain forest LRF lowland rain forest NPP net primary productivity RH relative humidity SST sea surface temperature TMCF tropical montane cloud forest UMRF upper montane rain forest WI warmth index P. FosterrEarth-Science ReÕiews 55() 2001 73–106 75 remains a puzzle. Many suggestions have been put Threats to the cloud forest from global climate forth but it remains unclear if any one of them change include changes in air quality, hurricane fre- explains all cloud forest attributesŽ. Table 1 . quency and duration, ultraviolet radiation, rainfall, This lack of understanding of the underlying and temperatureŽ. Hamilton et al., 1995b . We will mechanisms of cloud forest morphology make it not discuss direct CO2 or air quality impacts in this difficult to predict what the impacts of climate change reviewŽ. see Korner,¨ 1994 , but will address the other will be. A further difficulty lies in distinguishing effects. This review draws extensively from previous between local climate changes and global ones. reviews, including: a detailed overview of the cloud Changes in the global climate will cause changes in forestŽ. Stadtmuller,¨ 1987 , the first Tropical Montane local circulation patterns. Furthermore, local climate Cloud Forest symposium proceedingswŽ Hamilton et is likely to be affected by local events such as al., 1995a, especially its articles on hydrology deforestation or dam building. While local changes Ž.Bruijnzeel and Proctor, 1995 , and plant distribu- may be much stronger than global ones at a given tionsŽ.. Merlin and Juvik, 1995x , the 1998 special site, generalizing about local changes is difficult. In issue of Climatic ChangeŽ. Markham, 1998 on cli- order to mitigate against potentially hazardous mate change impacts on tropical forestswŽ especially changes at a given site, we must understand the BenzingŽ. 1998 on epiphytes and Donnelly and different components contributing to the change. This CrumpŽ. 1998 on amphibians .x , and an article on review aims to elucidate the role of global changes atmospheric change impacts on mountain ecosystems on the tropical montane cloud forest. Ž.Bensiton et al., 1997 .

Table 1 2. The cloud forest Listing of some of the principle attributes of the cloud forest Climatic characteristics 2.1. Characteristics of the cloud forest Frequent cloud presence Usually high relative humidity Low irradiance 2.1.1. Nomenclature A detailed account of the nomenclature used for Vegetation Characteristics cloud forests is given in Stadtmuller¨ Ž. 1987 , see also Abundance of epiphytes Hamilton et al.Ž. 1995b . Common names include: Stunted trees dwarf forest, elfin woodlands, mossy forest and mon- Small, thick and hard leaves High endemism tane rainforest or thicket in English; nuboselÕa, bosque() montano nubuloso, and selÕa nublada in Low ProductiÕity Spanish; foretˆ´´ nepheliphile and foret ˆ de nuage in Low NPP French; and Nebelwald and Wolkenwald in German. Low LAIŽ. although locally it can be very high For our purposes, the cloud forest will simply refer Slow Nutrient Uptake to forests whose characteristics are tied to the fre- Sap flow depressed quent presence of clouds and mist. There are various Low transpiration definitions that distinguish between fog and clouds. Low photosynthesis ratesŽ. though capacity not reduced One definition defines fog as any cloud which touches the ground and thus one might argue that these Soil and Litter Different High organic content in soil forests should be called fog forests. However, other High concentration of polyphenols in litter definitions that rely on droplet size or formation Soils wet mechanisms favor the cloud forest appellation. We will use the common names cloud forestŽ. CF and PositiÕe Water Balance tropical montane cloud forestŽ. TMCF . Additional moisture input from cloud stripping Stream flowrincident rainfall very high Several ways to subdivide cloud forests have been Low evapotranspiration and evaporation presented in the literature. Stadtmuller¨ Ž. 1987 stressed the difference between CFs that occur in regions of 76 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 abundant rainfall and those that occur on arid moun- CF are the Lomas Formations on the western coast tainsidesŽ Sugden and Robins, 1982a,b, 1983; Cave- of Peru and Chile, where fog water contributions lier and Goldstein, 1989. . The first can be difficult to allow the development of ferns, gymnosperms and distinguish from surrounding vegetation, whereas the flowering plants, in an otherwise desert, almost second can be very distinct from its neighbors. plantless, environmentŽ. Follman, 1963; Dillon, 1997 . Hamilton et al.Ž. 1995b distinguished band and patch The Lomas may well respond to climatic change in a type forests based on their extent and concluded that manner similar to the cloud forests. they typically thrive in continental versus islands climates, respectively. For the purposes of climate 2.1.2. Ecotones change impacts, a classification scheme that includes On moist tropical mountains, ecosystems grade all isolated cloud forests in one grouping and large from lowland rainforestŽ. LRF to lower montane band type forests in another will serve to highlight rainforestŽ. LMRF to upper montane rainforest the greater sensitivity of the isolated cloud forests. Ž.ŽUMRF Grubb et al., 1963 . . These zones are shown The isolated CF, regardless of its cause of isolation in Fig. 1. Grubb and WhitmoreŽ. 1966 concluded —such as the Elfin or dwarf cloud forests, will be that this zonation results from a gradation in cloud more sensitive to climate change by virtue of their frequency, from negligibleŽ. LRF to frequent Ž LMRF . small fragmented size, which leads to difficulties in to persistent cloudinessŽ. UMRF . Cloud forests occur relocation and repopulation. Similar to the isolated in LMRF or UMRF locations where cloud immer-

Fig. 1. Ecotone zonation on tropical mountains. Typical gradation of mountain ecosystem types is shown. From the lowland rainforest Ž.LRF through the lower montane rainforest Ž LMRF . up to the upper montane rainforest Ž UMRF . , tree stature decreases. Leaf size also decreases. In the presence of frequent cloudiness, the cloud forestŽ. CF can occur in either the LMRF or UMRF. Within the cloud forest, tree height and leaf size are even smaller than in standard forests at the same altitude. Accompanying the decrease in size in the CF, trees become twisted and gnarled and leaves become thick and hard. On exposed ridges and peaks these effects can be even more magnified and the crown of trees often assumes an umbrella-like shape. This subclass of cloud forest is sometimes referred to as an elfin forest. P. FosterrEarth-Science ReÕiews 55() 2001 73–106 77 sion frequency is enhanced relative to its surround- epiphytes. The following discussion will be most ings. relevant to those CFs which are most abundant in Following the ground contact of the clouds, the epiphytes, such as the elvin or dwarf CFs. An aston- CF typically occurs in narrow altitude belts, and on ishing percentage of the epiphytes in CFs are often ridges or mountain peaks, with species distributions endemic, especially within the isolated CFsŽ Young that are akin to island archipelagosŽ Merlin and and Leon,´ 1991; Gentry, 1992; Hietz and Hietz- Juvik, 1995; Vazquez-Garcia, 1995. . The altitude of Seifert, 1995; Benzing, 1998. . Thus, it is not surpris- cloud formation, and hence the cloud forest, shows ing that epiphytes can have important roles and even considerable variation as it depends on the moisture dominate many of the cycles in the cloud forest, content of the air, convective or advective cloud includingŽ. as summarized by Benzing, 1998 the formation processes, the velocity and direction of the following points: wind, the distance to the sea and macro and micro Ž.a The light cycle: the leaf area index Ž LAI . topographical effects. Nonetheless, a typical altitude rarely exceeds 20 in most humid tropical forest but for cloud forest on large, inland mountains is be- can be as large as 150 on branches densely covered tween 2000 and 3500 m. On coastal and isolated with epiphytes; productivity of epiphytes can exceed mountains, the cloud forest is found at lower eleva- other flora. tions, roughly 1000 m in Hawaii and as low as 500 Ž.b The water cycle Ž capture, storage and slow m in Micronesia and FijiŽ. Hamilton et al., 1995b . release. : water capture by epiphytes via horizontal This compression of ecotones, or the Massenerhe- precipitation, HP or cloud stripping, is typically 5– bung effect, occurs for other montane ecotones as 20% of annual rainfall, although values exceeding well. 100% have been recorded. Epiphytes also store water ŽŽ.up to 50,000 litersrha Sugden, 1981 ; 3000 litersr 2.1.3. Vegetation characteristics haŽ.. Richardson et al., 2000 and provide water to With increasing altitude, plant species change canopy animals. from lowland species to montane ones. Accompany- Ž.c The nutrient cycle Ž capture, storage and slow ing the species change, tree stature decreases with release. : up to half the total input of NHqy and NO increasing altitude. For example, in monsoon Asia, 43 and other important ions and nutrients to the forest the drop off in tree height can be quite step-like, can come from water stripped from passing clouds dropping by a factor of ;2 from LRF to LMRF and by epiphytes. Up to half the total nutrient pool of the again from LMRF to UMRF. Leaf sizes also de- canopy is sometimes found in the epiphytes and their crease with altitude, ranging from mesophyllous in associated humus. Richardson et al.Ž. 2000 also noted the LRF, to notophyllous in the LMRF and to micro- that of three forests studied, only in the dwarf forest phyllous in UMRFŽ. Ohsawa, 1995b . Epiphyte load did bromeliads achieve sufficient abundance to be- tends to increase with altitude. In the cloud forest, come significant in the litter nutrient cycle and to the these high altitude characteristics are enhanced. Fur- total net primary productivity. thermore, cloud forest trees are twisted, gnarled and These factors alone argue for the importance of often assume an umbrella-like crown. Trees are even the health of epiphytes, not only for the life cycle of more stunted on exposed ridges and peaks where the canopy, but also in some cases for the ecosystem many CFs thriveŽ Werner and Balasubramaniam, as a whole. Furthermore, epiphytes provide homes, 1992; Werner, 1998. . Leaves are thicker, harder and nesting materials and food for a vast array of animal smaller than in surrounding vegetation. Epiphyte species, including invertebrateswŽ which make a con- loads, diversity and endemism explode. Tree species siderable contribution to the food chainŽ Richardson from the Weinmanna family and tree ferns from the et al., 2000..x , frogs and birds, and even some pri- Cyatheaceae familyŽ. Merlin and Juvik, 1995 are matesŽ. Benzing, 1998 . common in the cloud forest worldwide. Epiphytes can play a critical role in the health of 2.1.4. Epiphytes the cloud forest, making their vulnerability to climate Epiphytes are a defining characteristic of cloud change an important issue. Evidence for their sensi- forests, where 1r4 of all plant species may be tive dependence on climate spans many scales. Along 78 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 mountain transects, specific epiphyte populations even specific mountain peaksŽ Raynor, 1995; Merlin appear and disappear following saturation deficit and Juvik, 1995. . Gentry estimates that endemism transectsŽ. Benzing, 1998 . The distribution of epi- can occur in cloud forests as small as 5–10 km2 phyte families also follows the saturation deficit. The Ž.Gentry, 1986 . In South American cloud forests, he vascular epiphytes, including orchids, ferns and estimates local endemism at 10–24%, suggesting bromeliads, have some capacity to store water, that perhaps an entirely different evolutionary mode whereas the nonvascular epiphytes do not. The non- is operating in these areasŽ. Gentry, 1992 . If we are vascular epiphytes, including mosses, lichens and to conserve a broad array of plant species, it seems liverworts, lack the simple epidermis of their vascu- obvious that the cloud forest is a good place to start. lar cousins such that they require almost constant In contrast to most cloud forests, one series of high relative humidity. Therefore, their biomass dis- isolated cloud forests, which is surrounded by arid tribution peaks at the altitude of minimum saturation vegetation in South America, shows uncharacteristi- deficit. This altitude is somewhat higher than the cally low endemism that might suggest a recent peak in vascular epiphytes’ biomass distribution, as originŽ. Sugden and Robins, 1983 . higher altitudes have lower temperatures and hence Animals also show high endemic rates within the tend to have higher relative humidityŽ Benzing, cloud forest. Thirty percent of Peru’s endemic birds, 1998. . mammals and anurans are found primarily in the Aside from altitude zonation, epiphyte species are cloud forestŽ. Leo, 1995 . A large-scale, global sur- localized within the cloud forest: some at the edge vey of restricted range birds found that over 20% of wŽ.cold mesic epiphytes and some at the center restricted range birds use the cloud forest extensively ŽŽ..warmth loving epiphytes Vazquez-Garcia, 1995x . and many of these birds are endangeredŽ see refer- Both altitude and intra-forest zonation are found ences in Long, 1995. . among terrestrial species as well, but the epiphytes show a further recognition of microhabitat distinc- tion: different epiphytes occupy different locations 2.1.6. Soil on a tree, from crown, to stem base to knotholes Montane slopes host a variety of soil types deriv- Ž.Hietz and Hietz-Seifert, 1995; Benzing, 1998 . Our ing from the abundance of parent rock types that are understanding of the photosynthetic process used by highly weathered from frequent rainfall and land- most epiphytes, which appear to be very tempera- slides. Steep slopes and weathering often results in ture-sensitive, also supports the observational evi- very shallow soils and little horizon development. dence for climate dependenceŽ. Benzing, 1998 . Furthermore, montane soils often, but not always, have low nitrogen valuesŽ. Tanner et al., 1990 and 2.1.5. Endemism low decomposition ratesŽ Hafkenscheid, 2000 and The intense microhabitat specialization of epi- references therein. . These factors may lead to the phytes explains in part their exceptionally high en- characteristically low primary productivity of mon- demism rate in the cloud forest. While tree species tane vegetationŽ. Grubb, 1977 , although high alu- diversity generally decreases with altitude, epiphytic minum and organic carbon soil contents or low biomass increases dramatically within the cloud for- levels of sunlight might also be partly responsible est, often accompanied by increases in speciation and ŽGrubb, 1977; Kitayama 1995; Bruijnzeel and extremely high rates of endemismŽ Leon´ and Young, Veneklaas, 1998; Werner, 1998, Section 2.2. . Cloud 1996. . Indeed, the cloud forests along the base of the forest soils are wet and frequently saturated, though Northern Andes and adjacent Central America host rarely waterlogged. Bruijnzeel and ProctorŽ. 1995 the distributional center of epiphytic and herbaceous concluded that of all the observations to date, no families, in contrast with the trees’ center which is cloud forest soil ever showed a water deficit, even within AmazoniaŽ. Gentry, 1992 . The fragmented during a severe drought. However, at some point the nature of the CF ecosystem also plays an important soil water must run out and there has been at least role leading to high levels of endemism, with some one observation of CF canopy dieback in response to plants endemic to specific islands or countries or a droughtŽ. Werner, 1995 . P. FosterrEarth-Science ReÕiews 55() 2001 73–106 79

The capacity of the soil and the epiphytes to cover can be drastic and substantial changes in vege- absorb and store a great deal of water provides the tation occur over a relatively small shift in altitude important services of erosion and flood control as across the boundary of frequent cloud occurrence well as dry weather stream flow. Žfor examples, see Sugden and Robins, 1982a,b; Werner, 1986, 1998; Merlin and Juvik, 1995; 2.1.7. The role of temperature Vazquez-Garcia, 1995; Young, 1998. . The displace- Current models of ecotone movement rely mainly ment of standard montane species andror morphol- on average temperature and precipitationŽ Holdridge, ogy by CF ones can be quite sharp, suggesting that 1947; Prentice et al., 1992; Neilson, 1995. . Box’s cloud cover can indeed control floristic assemblages, Ž.1995 model is more complex but still relies on especially of canopy speciesŽ Grubb and Whitmore, correlations between the prevailing climate and eco- 1966; Smith et al., 1995; Kitayama, 1995. . Grubb type. Plant responses to climate change may also and WhitmoreŽ. 1966 suggest that the altitude of depend on altered wind regimes, anthropogenic fac- cloud formation determines the distribution of not tors or even past forest statesŽ. Graumlich, 1994 . For only the cloud forest but all tropical montane forests. these reasons, models of ecotone movement must be Whether or not this is true, we know that the charac- based on mechanistic understanding of the climate teristic stunting and twisting of trees and small hard vegetation relationship rather than correlations. This leaves of the cloud forest is linked to frequent cloud is particularly obvious when we consider that future cover. How the clouds lead to these characteristics is climate states may have no analog in the present unknown but many suggestions have been put forth, world. see Section 2.2. Nonetheless, observations show that ecotones are strongly correlated with temperature. This is particu- 2.1.9. Massenerhebung effect larly true in eastern Asia, which has abundant rain- The Massenerhebung effect is the often noted fall throughout the year and in the absence of man’s telescoping, or compression of vegetation zones, on influence would be entirely forestedŽ Ohsawa, 1990, small and coastal islands. It is especially strong for 1995a; Fang and Yoda, 1990a,b, 1991. . The altitude the cloud forest biome, whose lower boundary is limit of the forests in humid monsoon Asia appear to typically 1500–2500 mŽ. Stadtmuller,¨ 1987 , but correlate quite strongly with a warmth index value of drops to as low as 500 m on isolated coastal moun- 158C monthŽ Ohsawa, 1990, 1993; WIssum of all tainsŽ. Merlin and Juvik, 1995 . Lower cloud forest monthly temperatures exceeding 58C. . This result altitudes on island mountains are to be expected, has been substantiated on the species level by Fang somewhat, as the water vapor content of air at the and YodaŽ. 1990a,b , who found that the total aver- foot of the island mountain will be higher than that aged value of WI for six forest species at their over the large continental mountain range. Wetter air altitude limit, at 12 different sites, was 15.28C month condenses at higher temperatures, and hence lower Ž.ss2.4 . Finally, Kitayama Ž. 1995 concludes that altitudes, than dry air, given equal lapse rates where the upper limit of the cloud forest on Mount Kina- GsydTrd z is the lapse rate, T the temperature balu in Sabah may be thermally controlled, with a and z the altitude. Further reducing the altitude of cut-off of WIs858C month. In the absence of water cloud formation on small isolated mountains is a stress, there appears to be evidence that forest eco- steeper lapse rate relative to large mountains. This tones are strongly temperature-controlledŽ Ohsawa, compresses the temperature profile on small moun- 1995a. . tains relative to large ones and increases the relative humidity at a given altitude, implying lower cloud 2.1.8. The role of clouds formation on the smaller mountainŽ Hastenrath, In humid Southeast Asia, the frequent occurrence 1991. . Given steeper lapse rates on shorter moun- of fog has less impact on the vegetation and it can be tains, the forest limitrtemperature correlations dis- difficult to pick out the cloud forest from the sur- cussed above will produce the telescoping effect. rounding vegetationŽ. Ohsawa, 1995b . However, However, there is no evidence that the lapse rates on elsewhere the hydrologic impact of frequent cloud small mountains are steep enough to reproduce the 80 P. FosterrEarth-Science ReÕiews 55() 2001 73–106

500–1000 m altitude differences. Furthermore, the growth and small thick leavesŽ see references in shifting of vegetation zones is not always linear. For Flenley, 1992c. . Under exposure to high UV-B, instance, in New Guinea, the UMRF appears to have plants protect themselves by producing phenolic been absent during the late Pleistocene, which would compounds, flavonoid and alkaloids, that absorb not be expected by a temperature gradient alone UV-B radiation. Fresh leaf litter on the floor of cloud Ž.Grubb and Whitmore, 1966, see also Section 7.2 . forests shows high concentrations of polyphenols The WI-forest limit argument might possibly avoid Ž.Bruijnzeel and Proctor, 1995; Hafkenscheid, 2000 . these failings as it is computed with a non-linear sum In turn, polyphenols may interfere with photosynthe- of temperatures exceeding a certain limit, but it also sis, cell division in fine roots, transpiration and ion fails to explain the distribution of all forest species in uptake. While the photosynthetic capacity of the CF the moist forests of ChinaŽ. Fang and Yoda, 1990a,b . does not appear to be systematically lower than other Flenley also argues that the lapse rate change–Mas- forests, the net productivity of the CF is reduced. senerhebung effect connection is not satisfactory, as Two measures of productivity, above ground NPP it does not explain the CF tree stunting. As discussed Ž.Ž.net primary productivity and LAI leaf area index , below, Flenley ties the stunting of cloud forest trees are markedly lower in the CF relative to other forests to excessive UV-B radiation. UV-B intensity in- at comparable altitudesŽ Bruijnzeel and Veneklaas, creases with altitude and may be enhanced by reflec- 1998. , although locally the LAI can be quite high tion off cloud surfaces. He argues that it is possible Ž.Benzing, 1998 . Transpiration rates also appear to that on island mountains, UV-B radiation is further be reduced relative to lowland forests and nutrient enhanced by reflection off the sea surface, thus uptake rates also operate at below their maximum moving the CF location down slopeŽ. Flenley, 1995 . capacityŽ Bruijnzeel and Proctor, 1995; Hafken- For the cloud forest at any rate, the Massenerhebung scheid, 2000. . Other impacts of high polyphenol effect is certainly tied to lower altitude cloud forma- concentrations in the soil are slow decomposition tion on shorter mountains. Ž.Kuiters, 1990 and the reduction of the impact of both aluminum toxicityŽ. Hue et al., 1986 and low 2.2. The cause of CF morphology nitrogen concentrationsŽ. Northup et al., 1995 in the Whatever the cause of stunting, one thing is clear: soil, all of which have been reported for a number of cloud forests are enveloped in clouds. Grubb and cloud forest soilsŽ. Hafkenscheid, 2000 The correla- Whitmore’sŽ. 1966 suggestion that the frequency of tion of the presence of polyphenols, low productivity cloud contact is the most important factor in limiting and uptake and stunted growth does suggest that the stature of montane forest is continually rein- polyphenols, perhaps driven by UV-B radiation, may forced by descriptive reportsŽ Bruijnzeel and play a significant role in stunting. But why the Veneklaas, 1998. . Somehow, the frequent cloud correlation with cloudiness? Furthermore, UV-B ra- cover is closely linked to the occurrence of short and diation only becomes sufficiently strong above twisted trees and xeromorphic leaves. Since wind 2500–3000 m to cause harmŽ. Caldwell et al., 1980 , commonly accompanies cloudiness, it might be re- well above the altitude of many cloud forests. sponsible for some of the odd morphologyŽ Merlin Several environmental characteristics of cloudy and Juvik, 1995; Werner, 1995. . Other factors that days may explain the enhancement of the UV-B accompany frequent cloud contact are: an extra hy- effect in cloud forests. Low levels of visible light drological input, a distinct chemical input, reductions and abundant water tend to make plants more sensi- in irradiance of 10–50% and persistent leaf wetness tive to enhanced UV-B radiationŽ. Ziska et al., 1992 . leading to a reduction of photosynthesisŽ Bruijnzeel In addition, the common daily regime of cloud for- and Veneklaas, 1998. . Perhaps one of these effects mation below the altitude of the cloud forest in the leads to the odd morphology. morning and enveloping clouds in the afternoon may increase the insolation of UV-B radiation in the 2.2.1. Polyphenols and UV-B radiation morning through reflection and prevent visible light In control experiments, plants that underwent en- driven repairs in the afternoon, making the CF habi- hanced UV-B radiation showed CF-like stunted tat particularly unacceptable to non-adapted plants P. FosterrEarth-Science ReÕiews 55() 2001 73–106 81

Ž.Flenley, 1992c . Thus, the UV-B hypothesis is con- tions of the unique morphologyŽ. Whitmore, 1989 . sistent with the Massenerhebung effect and can also Since the one aspect that binds all cloud forests neatly explain the disappearance of the CF at some together is the frequent contact with clouds, in the sites during the last glacial maximumŽ see Section rest of the paper we will concentrate on changes in 7.2. . These, in combination with the explanation of the cloudiness regime. CF tree stunting and the high concentration of phe- nolyps in the leaf litter, make the UV-B theory well worth further research. 3. Possible climate changes with relevance to the cloud forest 2.2.2. Other stunting hypotheses With high humidity and low insolation of the 3.1. High altitude sites cloud forest, one expects to find moisture-loving leaves rather than leaves that resemble drought-re- Projecting climate change on tropical mountain sistant onesŽ. Flenley, 1995 . This oddity led to the sites is fraught with difficulties. Current global cli- suggestion that periodic water shortages caused the mate models, GCMs, do not model current day xeromorphic characteristics and stunted tree growth, mountain climates well because they do not resolve but soil water shortages have never been observed in the topography, which is particularly important for a cloud forestŽ. Bruijnzeel and Proctor, 1995 . On the capturing precipitation and cloudiness. Further, not other hand, CF soils are usually wet and this may enough detail is known about mountain climates to lead to the observed low transpiration of plants by adequately test current mountain climate simulations. impeding root respiration, which in turn might cause This lack of basic understanding stems from a paucity the unusual morphologies to arise. Reduced nutrient of climate data, little theoretical attention to moun- uptake is generally seen in mountain sitesŽ Korner,¨ tain climatology and the complexity of mountain 1994. and is further reduced within the cloud forest, weatherŽ. Barry, 1994 . This ignorance is particularly which correlates well with the general trend for dangerous in light of the fact that higher altitude stunting on mountains and its exaggerated mani- sites may be affected by climate change even more festation in cloud forestsŽ. Hafkenscheid, 2000 . strongly than lowland sites. This arises because as BruijnzeelŽ. 1995 concludes that the answer lies the climate warms, more water will be pumped into somewhere within the soil and is perhaps linked to the atmosphere and the lapse rate is expected to exposure to harmful UV-B radiation as discussed decrease—approaching the saturated lapse rate. If above. this occurs, high altitude sites will be less cold Above the soil, another explanation for the twist- relative to the surface Ž.GsydTrd z is decreasing ing and stunting of trees is wind, which might ex- than they were in the cooler climate, i.e. high alti- plain the exaggerated stunting found on exposed tude sites will warm more than lower altitude sites. ridges and peaksŽ Merlin and Juvik, 1995; Werner, Tropical montane temperature recordsŽ Bensiton et 1995. . Another hypothesis is that low leaf tempera- al., 1997.Ž , the retreat of tropical glaciers Schubert, tures may cause stunting since temperature decreases 1992; Hastenrath and Kruss, 1992. , ice core tempera- with increasing altitude and is further reduced under ture recordsŽ. Thompson et al., 1993 , and freezing cloud cover. However, Bruijnzeel et al.Ž. 1993 heightsŽ. Bensiton et al., 1997 strongly suggest that showed that temperature depression was not the cause tropical high elevation sites may already be showing of stunting on Gunung Silam. Neither temperature this enhanced warmingŽ. Bensiton et al., 1997 . How- depression nor wind obviously explain the disappear- ever, recent results show that although the tropical ance of the CF in New Guinea during the Late lapse rate decreased prior to 1979, since then it has PleistoceneŽ. Flenley, 1992b, Section 7.2 , although a shown a slight increase. Furthermore, while glaciers separate cause or a complex chain of effects might continue to retreat, tropical freezing levels have low- be responsible for this effect. Given the many char- ered in the 1980s and 1990sŽ. Gaffen et al., 2000 . acteristics of the cloud forest, different factors may Many of these changes appear to be linked to sea be responsible for different aspects and manifesta- surface temperature changes, although the recent in- 82 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 crease in lapse rates does not appear to beŽ see Sabah—Kitayama, 1995, Hawaii—Chu, 1989. . Just Section 3.3. . how much drought, low soil moisture and increased water demands from increasing temperatures a forest 3.2. Increases in water stress and intense eÕents can withstand is very site-dependent, but certainly some cloud forests will not be able to survive the The moisture available to an ecosystem affects coming climate change. Decreases in cloud immer- both the kind and the amount of the vegetation that sion, discussed below, will also lead to increased will growŽ. Woodward, 1987 . The soil serves as an stress through decreases in water input and increases important water reservoir for terrestrial plants for in water loss. non-rainy days, dry seasons and droughts. Climate change will likely change the balance of current 3.3. Changes in cloudiness water tables by increasing loss through evaporation, via higher temperatures, and by changing rainfall 3.3.1. Cloud formation and rainfall inputs. Few studies have focused on climate change Clouds form when water vapor condenses to form impacts on the hydrological cycle in the tropics; but liquid water droplets. Theoretically, this can occur one study, the Hulme and Viner analysis of the UK via chance collisions, but this requires highly super- Met Office’s GCM, shows increasing water stress in saturated airŽ. of the order 300–400% that rarely many tropical regions. They find decreasing soil occurs in nature. Instead, clouds usually form via the moisture throughout the tropicsŽ although they cau- collection of water droplets onto a foreign substance, tion that this result is crude. . Furthermore, with the or an aerosol particle. For a given particle size, exception of east and northeast Africa through the condensation then occurs when the RH exceeds a Middle East and into monsoon India, the length of much lower critical value. Typically, this occurs the dry season in the tropics increased. The same when air rises and cools adiabatically, thus lowering regions also showed a decrease in the relative humid- its saturation vapor pressure and increasing its RH. ity field when averaged over four other GCMsŽ see For instance, over the warm tropical oceans, convec- Section 3.3. . These predictions suggest increasing tion acts to draw air upwards and thereby forms water stress, which if it occurs will cause damage to cumulus clouds. These types of clouds often provide trees and epiphytes, implying more fire and drought. a lot of the annual rainfall in cloud forests, but Fire and drought tend to degrade tropical forests to another type of cloud, orographic clouds, often pro- grasses, bushes and fire-tolerant trees, such as Pinus vides the daily regime of cloud immersion. Ž.Goldammer and Price, 1998 . Orographic clouds are formed when wind is forced Hulme and Viner’s analysis also points to more upward by topographic relief, and in the case of the intense rainfall events, as seen in other GCMs as tropical montane cloud forests, it is often the trade wellŽ Fowler and Hennessey, 1995, although Bensi- winds running into a mountain. The bottom of the ton et al., 1996 concludes that it is still uncertain clouds typically forms at the level of the lifting whether extreme events will increase. . Typhoons and condensation level. The top of the cloud is often hurricanes may also increase in frequency although limited by the trade wind inversion, which causes an results for this are even more uncertain. If extreme increase in temperature with altitude. If this occurs, events do increase in intensity, wind damage to trees the thickness of the cloud, and hence the size of rain and soil erosion will increaseŽ Emanuel, 1987; Hulme droplets that can form, are limited. If the droplets are and Viner, 1998. . Soil erosion leads to a multiple of prevented from becoming large enough to overcome woes—including drought, nutrient leaching and the frictional resistance of air, no rain will fall. complete slope failure, all of which can effectively Ž.Hastenrath, 1991; Houze, 1993 . This is essentially reduce available soil moisture with devastating con- the case for some cloud forests along the Caribbean sequences for mountain forests. El Nino˜ events may coast of South America, where annual rainfall is very increase in frequencyŽ. Timmermann et al., 1999 and low and lowland vegetation is of the arid type. Yet, many authors have linked drought at particular sites cloud forests thrive above about 800 m. Every day, to El Nino˜ conditionsŽ Sri Lanka—Werner, 1998, high altitude small clouds form above the mountain P. FosterrEarth-Science ReÕiews 55() 2001 73–106 83 peaks and in late afternoon, when temperatures cool, latter period implies greater convective instability the cloud base descends, covering the forest mainly and hence more cloudiness. However, lapse rates at night. In cooler higher sites, clouds cover the are only an indication of potential cloudiness— forests mainly in the afternoonŽ Cavelier and Gold- sufficient vertical velocity is also needed to initiate stein, 1989. . convection. A better indicator of cloudiness is hu- For the sake of the following sections, note that midity, as discussed previously. climate models are incapable of resolving the cloud GrahamŽ. 1995 was able to reproduce observed microphysics discussed above. Instead, cloud forma- trends in specific humidity using a GCM driven with tion has to be parameterized. In a study correlating SST changes alone. The specific humidity profile has observed potential cloudiness indicators to low level changed in the following way: over the Western stratiform cloudsŽ as in the trade wind-induced oro- Pacific, at and above 1000 mb, the specific humidity graphic clouds discussed above.Ž. , Slingo 1987 found increased from the early 1970s through the 1980s that an RH)80% and an updraft correlated well ŽFlohn and Kapala, 1989; Gaffen et al., 1991; Gut- with cloud occurrence. Grubb and WhitmoreŽ. 1966 zler, 1992. . The largest increase was observed at the found that 80% was a useful cutoff as well because 1000 mb layer with smaller increases both upwards the RH climbed to 80% when rain or cloudiness to the 700 and 300 mb layers and downwards to the interrupted clear spells in an LMRF in Ecuador. surface. In terms of relatiÕe humidity, the observed Many GCMs use schemes related to these observa- surface RH actually decreased slightly at all four tions, adding conditions that are more detailed but stations used in Gutlzer’sŽ. 1992 study. This hints at essentially relying on the relative humidity in a grid decreased surface cloud formation as cloud forma- box to determine if a cloud is present. The modeling tion correlates with RHŽ Wallace and Hobbs, 1977; of cloud formation and rainfall remains one of the Houze, 1993.Ž , especially low-level clouds Slingo, essential tasks of the climate community. 1987. . Observations of CFs on tropical Pacific Is- lands might provide a link to these changes in the 3.3.2. Sea surface temperatures and cloudiness hydrological cycle and thus provide an early warning In the late 1970s, tropical sea surface tempera- system of climate changes over the ocean. turesŽ. SSTs underwent a step-like increase of sev- Results of GCMs driven by SST increases also eral tenths of a degreeŽ Guilderson and Schrag, suggest decreased low-level cloudiness upon SST 1998. . This shift has been linked to a number of warning. For example, a GCM driven with a uniform recent changes in the hydrological cycle of the trop- SST increase showed a decrease in the globally ics including increased El Nino˜ frequency and inten- averaged low-level cloudinessŽ Schneider et al., sityŽ. Guilderson and Schrag, 1998 , as well as the 1978. . Other simulations that looked at cloudiness in 100-m upwards shift of tropical glaciers and freezing climates warmer than today also found decreases in heightsŽ. Diaz and Graham, 1996 . This SST shift low-level cloudinessŽ Rind, 1986, Mitchell and In- also correlates with changes in the trends of tropical gram, 1992.Ž . Surface observations Pounds et al., lapse rates, which decreased from 1960–1979 and 1999.Ž and modeling efforts Still et al., 1999; Foster, increased from 1979–1997. The decrease has been 2001.Ž at the Monteverde Cloud Forest see the next reproduced in an ensemble of simulations with a section. also suggest decreasing low level cloudiness GCM forced with observed SST values. However, at that site. In summary, GCM results suggest that an results of the second period, from 1979–1997, also increase in SST may result in decreased low-level predict a lapse rate decrease rather than the observed cloudiness, at least in certain locations, and concur- increaseŽ. Gaffen, 2000 , suggesting that SSTs are not rent observations of increased SSTs but decreased the only factor in driving the value of the tropical RH at the surface in the western Pacific and cloudi- lapse rate. Indeed, Santer et al.Ž. 2000 concluded that ness at Monteverde support the causal link. anthropogenic and volcanic effects may be partially responsible for the lapse rate shift. The lapse rate 3.3.3. Cloudiness at the MonteÕerde Cloud Forest observations suggest less cloudiness before 1979, Pounds et al.Ž. 1997, 1999 have linked changes in and more after 1979, as the greater lapse rate in the climate, cloud formation and animal abundance pat- 84 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 terns at the Monteverde Cloud Forest in Costa Rica. teverde has been rising. The subsequent decrease in Bird, lizard and anuran species abundances all corre- cloud immersion would increase the number of dry late with the number of dry days in the dry season, days but not decrease the overhead cloud cover, which is linked to cloud immersion frequency. In- which would cause the DTR to increase rather than cluded in the changes are bird species now coloniz- decrease as observed. Still et al.Ž. 1999 found consis- ing the cloud forest, which were previously restricted tently similar results in a study of expected shifts of to lower altitudes, highland endemic lizard declines ecotones for four cloud forest sites, one of which and disappearances, and in 1987 an exceptionally was MonteverdeŽ. see Section 4.2 . Incidentally, the large disappearance of 25 of 50 frog and toad species. increase in dry days is consistent with the upward Given the increase in the number of dry days, and a shift of ecotones discussed above, if the number of decrease in the diurnal temperature rangeŽ. DTR , dry days decreases with altitude. Since rainfall typi- Pounds et al. suggested that the cloudbank at Mon- cally increases with altitude, just such a negative

Fig. 2.Ž. a Changes in the surface relative humidity field from Ž 1960–1990 . to Ž 2060–2090 . as modeled in four GCMs. The results of four = GCM simulations for today and 2 CO2 concentrations are compared and shown for the four seasons. The relative humidity change is given as a fraction from 0.20 to y0.20Ž. or 20% to y20% . Most of the tropical land area shows a decrease in RH of 5–10%. This could imply decreases in low-level cloudiness and consequent damage to the cloud forest. See the text for details of the models and implications. Ž.b The standard deviation of the relative humidity averages shown in panel a. The value of the standard deviation of the relative humidity averages often is as large as the average itself, suggesting the average results are not entirely robust. However, a large portion of the deviation comes from one GCM and it does appear that there is a trend towards lower RH over tropical land areas. See the text for details. P. FosterrEarth-Science ReÕiews 55() 2001 73–106 85

Fig. 2 Ž.continued . correlation between dry days and altitude is expected 1987. Meanwhile, the Monteverde RH decreased, and, indeed, has been observed at the Luquillo Ex- both in the ECMWF data and in the simulation. This perimental Forest in Puerto Rico, which harbors a suggests that if the tropical tropospheric hydrological cloud forest at its peakŽ. Garcia-Martino et al., 1996 . cycle continues to increase, there may be further FosterŽ. 2001 has used the regional area model decreases in low-level cloudiness and consequently RAMS to test the hypothesis of a rising cloudbank at more damage to the cloud forestŽ see previous sec- Monteverde in more detail. Modeling 2 years that tion. . Since the enhancements in the tropical tro- bracketed the anuran species crash, 1979 and 1987, pospheric hydrological cycle may be driven by an- she was able to reproduce the increase in dry days thropogenic effectsŽ. Graham, 1995 , the drying at and indeed found it was due to a rise in the cloud- Monteverde may also be driven by human-induced bank. The driving force for modeling the decrease in climate change rather than natural fluctuations. moisture was the large-scale ECMWFŽ European Centre for Medium-Range Weather Forecasts. RH 3.3.4. The relatiÕe humidity changes of four GCMs = field, which among other atmospheric variables, was in a 2 CO2 world used as a boundary condition to drive the limited I looked at the expected response to climate area simulation. This large-scale RH field showed a change of the relative humidity field, as a proxy for very slight increase over the Pacific from 1979 to cloudiness, in four different global climate models. 86 P. FosterrEarth-Science ReÕiews 55() 2001 73–106

While these models are relatively coarse in resolu- RH increase. Except for the regions noted, the re- tion and do not show altitudinal zonation or moun- mainder of the cells over land tropics in all four tain climates, I expect local cloudiness responses to models primarily shows decreases in RH. These follow the value of the large-scale relative humidity results highlight the fact that the RH response to field. global warming is complex and even given similar I compared the evolution of the RH field at the assumptions, different predictions are possible. How- surface in the following four GCMs during the ever, there appears to be a trend towards decreasing = course of 2 CO2 transient IS92 simulations: RH in these models over most tropical land surfaces. Echam3rLSG-DKGG, CCSRrNIES CGCM-NIGG, Since surface RH is likely to be correlated to low HADCM3-HC3GG and CCCma’s CGCM-CCGG level cloudiness, this may indicate a trend towards models, available online from the IPCC at http:rr less ground touching clouds and hence may indicate ipcc-ddc.cru.uea.ac.ukripcc– ddcr. First, I calcu- a severe threat to cloud forest species. lated the seasonally averaged RH field for each of the four models: December–FebruaryŽ. DJF , March–MayŽ. MAM , June–August Ž. JJA and 4. Climatically suitable zone September–NovemberŽ. SON . Then, I averaged the seasonal averages into period averages from 1960– 4.1. Introduction 1990 and from 2060–2090, except for the Echam3 model which was only averaged from 2060–2085 In what follows, we will discuss the shifting of because subsequent years were not available. Next, I climatically suitable regions, so-called ecotones. differenced the period averages to find predicted Typically ecotones are defined by temperature and = changes in the RH field from today to a 2 CO2 precipitation regimes and do not consider cloud im- world. Using a nearest neighbor assignment, I inter- mersion that is so essential to the cloud forest. polated each field onto a 128=64 grid so that I Nevertheless, the general results of biogeography could average all four models and present the results models provide some insight into expected changes in Fig. 2a. To assess the robustness of these results, I on tropical mountains. Ecosystems as a whole will present the standard deviation of the averages of the not shift to new locations under different climates, four models in Fig. 2b. The pattern of the standard because ecosystems are composed of many species, deviation roughly follows the peaks in the difference which will each respond to climate change in its own plot. way. This is exemplified in the fossil record, which In Fig. 2a, the light blue–green color represents contains past climate states and vegetation distribu- little to no change in RH and this covers most of the tions, which have no analog in today’s world. This ocean surfaces. Blue to dark blue represents a de- may also be the case for future changesŽ Graumlich, crease in surface RH and appears over most land 1994. . Biogeography models attempt to overcome surfaces with an average value of about y5%, though this limitation by grouping plants by functional type some decreases are as large as 20%. Yellow Ž.q5% and linking these groups to prevailing climate condi- and red areas Ž.q15% are relatively rare appearing tions. However, these models are still fundamentally over central Africa and the Middle East in the tropi- co-relational rather than process-based and may not cal regions. This is dominated by the CCSRrNIES be appropriate for climate change scenarios. We also model, which has the largest RH increase of all four do not know if plants can migrate rapidly enough to models, with increases in Central Africa, the Middle keep up with the changes in climate that may occur. East and northern South America. Since both the Other barriers for successful migration are natural CCC and HADCM3 models show decreases in RH and man-made obstacles, such as rivers, valleys, over the entire globe, the areas of increase in the cities, roads and farms in the migration path. These CCSRrNIES model appear as regions of high stan- obstacles can both block movement and occupy land dard deviation in Fig. 2b. The largest increase in the of a potential new colony. This is especially true of averaged RH field appears over the Middle East, mountain habitats that are naturally fragmented and where the Echam3 model adds to the CCSRrNIES’ small. However, there is some evidence that at least P. FosterrEarth-Science ReÕiews 55() 2001 73–106 87 some cloud forest species have successfully leapt Ž.and therefore fewer ecotones in a warmer world are from mountain peak to mountain peakŽ Werner and paleoreconstructions of vegetation distributions dur- Balasubramaniam, 1992; Werner, 1998. . In the fol- ing colder eras that show narrower ecotone bands lowing section, we discuss where an ecosystem might Ž.Halpin, 1994, adapted from Flenley, 1979 , al- possibly thrive under the possible climate changes. though Colinvaux et al.’sŽ. 1997 pollen fossil study suggests that Andean vegetation zones were not 4.2. Biogeography models compressed during the last glacial period. In a study similar to Halpin’s Costa Rican transect study, Biogeography models, such as BIOMEŽ Prentice ScatenaŽ. 1998 used temperature and rainfall profiles et al., 1992.Ž , MAPSS Neilson, 1995 .Ž. , Box’s 1995 on the Luquillo mountains of Puerto Rico to estimate model and Holdridge’sŽ. 1947 life zones do a decent that either a 2.58C temperature rise or a 11% reduc- job of predicting the location of current day ecosys- tion in rainfall would be sufficient for the Tabonuco tems based on temperature and rainfall profiles, as forest to replace the next highest forest type, the well as sunshine and soil characteristics. Assuming Colorado forest. Either of the increases, which are = that the correlation relationships between the cli- well within the boundaries expected for a 2 CO2 matic variables and the dominant ecosystems hold, scenario, would result in no part of the mountain the models can be used to make predictions about having the climate conditions that the peak residing the change in location of ecosystems under changing dwarf cloud forest currently experiences. This cloud climates. For a uniform temperature rise, the prime forest, too, may be pushed out of existence. In any viable location for ecosystems on mountains shifts event, as the ecotone of the cloud forest moves upward to maintain the same temperature regime. upward, its viable land area will decrease as there is This has been summarized as Hopkins bioclimatic less land at higher altitudes on any given mountain. law, which states that for a 38C increase, species Aside from issues of migration, competition and ranges will shift upward 500 mŽ Peters and Darling, secondary growth, cloud forest diversity is expected 1985. . However, temperatures are not expected to to decrease simply because biodiversity scales with rise uniformly, either horizontally or verticallyŽ see areaŽ. see Fig. 4a . Section 3.3. , and other climate changes and vegeta- Instead of assuming simplistic uniform increases tion responses complicate the expected outcome. of temperature and precipitation, GCMs can be used For instance, rainfall patterns are sure to change. to predict the climatic changes. Still et al.Ž. 1999 One approach to modeling the effect of changing looked at several atmospheric variables thought to rainfall and temperature on ecosystems distributions correlate with cloud forest location in three GCM is to impose a uniform temperature rise and a uni- simulations of the GENESIS modelŽ Thompson and form percentage change in rainfall to the current Pollard, 1997. , corresponding to the present, the last = climate, and then apply biogeography models to the glacial maximum and a 2 CO2 atmosphere. They shifted climate. A transect study of the La Amistad found that the warmth indexŽ. see Section 2.1 and Biosphere Reserve, Costa Rica, found that for a 3.68 the absolute humidity surfaces both move upslope by temperature increase and a 10% precipitation in- 200–300 m at four cloud forest locations in a 2= creaseŽ as found in two GCM simulations for the CO2 world. The similar altitude shifts of the AH and Costa Rica areaŽ.. Smith et al., 1995 , there was a WI surfaces, representing water and temperature non-uniform reorganization of the current ecotones, regimes, suggest that the water balance of the CF with a general trend upslopeŽ. Halpin, 1994 . Three may well be reproduced at this shifted altitude. This of the ecotones along the transect were entirely lost, argument is strengthened by the 200–300 m down including the climate associated with the dwarf cloud slope shift of the WI and AH surfaces in the LGM forest, currently on the peak in this regionŽ Smith et simulation, which is consistent with pollen records of al., 1995; in agreement with Still et al.’sŽ. 1999 cloud forest species during the LGMŽ see Section conclusions. . The lower elevation ecotones expanded 7.1. . and moved upslope in the model with the alternate This predicted CF altitude shift of 200–300 m in = climate. Lending confidence to the trend for wider a2 CO2 world is somewhat smaller than other 88 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 estimates, probably because it is based on a very ment of rainfall may reduce the CF distribution by coarse GCM, which does not resolve the topography reducing the relative importance of horizontal pre- well. Other estimates for the shifting of tropical cipitationŽ. Juvik and Nullet, 1995 . It is likely that montane ecotones are: Costa Rica—500 mŽ Halpin, the impact of large-scale changes in the hydrological 1994.Ž. , Puerto Rico—400 m Scatena, 1998 , moun- cycle will be very site-dependent. tains everywhere—500 m for a 38C riseŽ Hopkin’s bioclimatic law, Peters and Darling, 1985. . These shifts are very significant relative to the current 5. Implications of climate change for the cloud altitude range of cloud forest species. Tropical eco- forest tones typically have very narrow elevation ranges because there are strong climate gradients along Exactly how direct contact with clouds changes tropical mountains, due to a lack of seasonal blurring the cloud forest ecosystem is unknown. What we do of temperature averages. HalpinŽ. 1994 surmises that know is that the postulate of Grubb and Whitmore the trees in Costa Rica live in altitude ranges of Ž.1966 , that long persistent cloudiness is the primary about 500 m. Estimates of cloud forest vertical ex- factor determining the stature of cloud forests, is tent are often around 500 m as wellŽ for example: borne out by many descriptive studies. Since we do Thailand, Werner and Santisuk, 1993; the Upper not know the underlying relationships, it is difficult Guayllabamba River Basin, Ecuador, Sarmiento, to predict what a change in cloudiness will mean for 1995; El Rey National Park, Argentina, Brown, 1995, the cloud forest. However, in the following we make Serranıa´ de Maeuira, Colombia, Sugden and Robins, some educated guesses. 1982a,b. , although larger ranges do exist. A 500-m shift in the viable climate range of an ecotone, which 5.1. Deforestation, flooding and dry weather flow spans 500 m, suggests complete replacement of the current range. Even smaller upward shifts of 200–300 Conservationists have argued that the cloud forest m may well push peak residing cloud forests out of plays an important role in reducing flooding and existence, with perhaps even less needed to push providing water during dry seasons for the forest many Central American cloud forests out of exis- itself as well as for downhill ecosystems and cities. tence. Cloud forest deforestation, via chainsaws, fire or climate change, may therefore result in increased 4.3. Latitudinal changes flooding and reduced stream flow in dry weather ŽZadroga, 1981; Stadtmuller,¨ 1987; Werner, 1998; Latitudinally, there may be less change in cloud Becker, 1999. . Since these consequences are so forest distributions. The high latitude limit of tree grave, we will now discuss the possible impact of type seems strongly tied to minimum annual temper- deforestation on the water budget in detail. atures whereas the equator-ward limit is apparently determined by competition, with no apparent high 5.1.1. The cloud forest’s water budget temperature limit. So the tropical evergreen broadleaf Hydrologically, the frequent contact of clouds on tree, which is the most competitive physiology given the vegetation has multiple implications. The vegeta- ample moisture and temperature, is expected to ex- tion gains an additional source of water by stripping pand its limits somewhat pole-ward. If the tropical water out of the passing clouds, either through direct cloud forest does expand pole ward, it may displace contact or through condensation, in a process known the Cryptomerica-dominated temperate cloud forest as horizontal precipitationŽ hereafter HP, sometimes Ž.Ohsawa, 1995b . The hypothesis that tropical eco- referred to as cloud stripping. . Bruijnzeel and Proc- tones will expand into current temperature ranges is torŽ. 1995 reviewed the hydrological cycle of cloud based mainly on temperature changes, not cloudiness forests and found that HP inputs varied from 70- changes that are of paramount importance to the CF. mmryearŽ. in Venezuela up to over 1160 mmryear Speculating about latitudinal changes in the cloudi- in Honduras, but that typical values were around ness regime is risky, we note only that an enhance- 5–20% of rainfall. This can be especially significant P. FosterrEarth-Science ReÕiews 55() 2001 73–106 89 during the dry season and droughts, or in cloud pasture sites for equal incident rainfall amounts. On forests located in the rain shadow of mountains the other hand, the through fall of water in the cloud Ž.Juvik, 1995 or in arid regions Ž Sugden and Robins, forest is enhanced relative to the standard montane 1983. . Subsequent estimates of HP have confirmed forestŽ. Bruijnzeel, 1999 , apparently through the ad- this rangeŽ Schellekens et al., 1998; Hafkenscheid et ditional input of horizontal precipitation and the al., 1998; Holder, 1998. , with an exceptionally high reduced evaporative demands. So given equal value of HPŽ 1365 mm of HP in a 100-day period, amounts of incident rainfall, the bucket is perhaps 406% of the rainfall. measured at Alakahi, Hawaii filled up most in the cloud forest, followed by pas- Ž.Juvik and Nullet, 1995 . ture land and perhaps least in the standard forest. In addition to increasing the water input, the This relation is strengthened by the rate at which the frequent presence of clouds reduces evapotranspira- bucket is emptied. Evapotranspiration of the standard tionŽ. ET , further reducing potential water stress. forest is greater than that of the CF. These conjec- BruijnzeelŽ. 1999 gives ranges of ET for various tural relationships between the water fluxes are shown montane forests fromŽ. a 275, Ž. b 500–700, Ž. c graphically in Fig. 3. 830–980 toŽ. d 1155–1380 mmryear for Ž. a dwarf This model concurs with observations that stream cloud forests,Ž. b TMCF in UMRF with several flow increased following deforestation of some hours of clouds per day,Ž. c TMCF in LMRF that ‘standard’ tropical forests in controlled experiments, only get modest amounts of cloud water, and finally dominated mainly by an increase in dry weather flow Ž.d cloud-free LMRF at similar or higher elevations. Žthe pasture’s bucket is fuller than the standard for- Bruijnzeel cautions against using these estimates est’s and thus provides more dry weather flow. . The blindly, as they are based on indirect evidence. Fur- CF’s full bucket implies the largest dry weather thermore, a recent detailed measurement of ET in a flow, which is corroborated by observations that the lowland tropical rain forest below 600 m gave values stream flow to incident rainfall ratio of the CF is of ET in excess of 2000 mmryearŽ Schellekens among the highest of all tropical forestsŽ Bruijnzeel, et al., 2000. —exceeding all of the rates given by 1990. . Further arguments that CFs are a good source BruijnzeelŽ. 1999 . Nevertheless, the strength of the of dry weather flow and flood prevention, are the decreasing trend of evaporation with increasing fact that CF soils are nearly always saturated cloudiness, as well as theoretical considerations, Ž.Bruijnzeel and Proctor, 1995 , the huge water stor- strongly suggests that evaporation is indeed reduced age capacity of the canopyŽ. Sugden, 1981 , and in the cloud forest. mounting observational evidence that cloud forest destruction leads to reduced dry weather flow 5.1.2. Bucket model Ž.Bruijnzeel, 1999; Becker, 1999 . The bucket model The impact of deforestation on the water balance is overly simplistic, of course. The consequences of is very complicated, but in essence, it depends on the forest loss will really depend on the relative amounts water holding capacity of the system. Water is pri- of water table input and output, as well as the soil marily held in the soil, so although the canopy can and canopy characteristics of each forest. also store a lot of water in the CF, we will ignore it for the time being. If the soil is considered as an 5.1.3. Soil compaction open bucket with infinite depth, it can absorb any In the event of a complete loss of forest, a further amount of rainfall, preventing flooding and slowly complication may be soil compaction: either via releasing the stored water, providing dry weather trucks, cattle or incident raindrops. In essence, the flow to systems down slope. Various factors compli- top of the bucket will be at least partly closed. In this cate this simple model. First, the amount of water case, flooding may be more common and more reaching ground level, the top surface of our bucket, intense. If flooding does occur, it will lead to further is different under the montane rain forests, pasture soil erosion and more flooding. A much-reduced dry land and cloud forests. The canopy of montane and weather flow is expected in the event of soil com- cloud forests intercepts and subsequently evaporates paction, as much less water will penetrate the soil to rainfall, so that less water reaches the soil than at fill up the bucket. Since some form of soil degrada- 90 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 hroughfall and dry nder the surface and again he cloud forest, the montane s figure demonstrates the water forest and grasslands. Theweather size flow are of greatest the init arrows the is cloud for greatest forest under a and theservices given least cloud that forest in water the and the flux least cloud montane is under forest. forest the an Similarly, probably montane the estimate provides: forest. amount increased of Relative of values dry the soil are weather moisture size speculative, flow is but of and represented based the by flood on flux the the prevention. shaded current relative area best to u estimates. the Thi other ecosystems. For example, t Fig. 3. The speculative relative water budgets of the cloud forest, montane forest and grassland. In this figure, three ecosystems are represented: t P. FosterrEarth-Science ReÕiews 55() 2001 73–106 91 tion is likely to accompany any form of deforesta- event of severe droughtŽ Werner and Balasubrama- tion, this outcome may be the most likely. Control niam, 1992. . If the drying trends outlined above are experiments that showed enhanced dry weather flow truly underwayŽ. Section 3.3 , the effects may first be following tropical rainforest deforestation may have seen in the epiphyte communities of cloud forests. been unrealistically gentle on the soilŽ Bonnell, BenzingŽ. 1998 stresses that epiphytes are important 1998. . components of several cycles within the CF. In some The results discussed in this section on the water instances, therefore, death of the epiphytes may di- budget are based primarily on intentional deforesta- rectly harm the underlying vegetation and alter re- tion rather than forest loss due to climate change. source allocation. However, there have been cases Therefore, caution must be exercised in extending where hurricanes have destroyed the epiphyte com- these conclusions to the future. For instance, if cli- munity and the tree community continued to thrive mate change does cause deforestation, it may not Ž.Scatena, personal communication . necessarily be harmful to the soil. While there is a chance that newly exposed soil will be compacted by 5.3. Impacts on animals incident raindrops, and hence effectively cut off from infiltration, it is more likely that grasses will grow Animals will be affected both directly and indi- and provide the soil with some protection, as oc- rectly by climate change. For example, consider the curred in the Andean paramoŽ. Sarmiento, 1997a,b . impact of the loss of epiphytes due to climate change. Almost all CF terrestrial invertebrate taxa and many 5.2. Epiphytes and climate change lower vertebrates use the moisture or nutrients that epiphytes store and produce, including seeds, fruit, The biomass of a forest, or individual plant for nectar and pollenŽ. Benzing, 1998 . Epiphytes also that matter, is strongly influenced by water availabil- provide nesting materials and water services to ity—no water, no plantsŽ. Woodward, 1987 . Trees canopy animals. Some animals, especially birds and can survive dry seasons and droughts by accessing insects, have co-evolved symbiotic relationships with underground water supplies that build up over the one specific plant species. Epiphyte reduction and course of previous wet periods. However, in the disappearance may lead to the death of many ani- event of a major climate change, with permanently mals and probably to some extinctions. Like the increased dry seasons, this water source may eventu- epiphytes, animals also play a key role in forest ally be used up and some forests will probably die, dynamics. For instance, many birds and mammals to be replaced by more drought-resistant ecosystems. distribute seeds, which highlights the complex web Epiphytes, however, respond much sooner to changes of interactions of life within a forest—death of one in the moisture cycle as they do not have access to component can have far-reaching consequences. Two ground water and rely entirely on atmospheric mois- high profile cloud forest mammals are the mountain ture. Furthermore, their extreme microhabitat colo- gorilla of VirungaŽ of which less than 500 survive in nizationŽ. see Section 2.1 argues for sensitivity to the wild. and the spectacled bear of South America. even small changes in moisture availability. Because While they are not likely to be directly affected by epiphytes occupy so many microhabitats in a given climate change, the collapse of their forest could canopy, there may be some ability for them to certainly cause their disappearance. reorganize within the canopy to take advantage of One group of cloud forest animals that may be shifted climate regimes. This may be limited, how- directly affected by climate change is the class Am- ever, because epiphytes are sensitive not only to phibia, including frogs, toads, salamanders and cae- climate, but also to other factors in their microenvi- cilians. Amphibian skins and eggs are permeable, ronment such as the light regime, substrate depth and making them sensitive to changes in temperature and so forth. Observations of epiphyte reactions to longer moisture as well as pollutants. Worldwide, amphib- than usual dry spells include the expected wilting of ian populations have been declining and while some epiphytesŽ Grubb and Whitmore, 1966; Lowry et al., decline is expected, rates of decline are greater than 1973. as well as whole scale canopy die back in the population model predictions. Many of these de- 92 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 clines are linked to habitat destruction, but some days at Monteverde and hence to the Monteverde have occurred in seemingly undisturbed places. There extinctions. Long-term animal and climate studies at are many postulates as to why these declines are other sites may soon reveal similar patterns. occurringŽ. Alford and Richards, 1999 . These in- clude the direct effect of desiccation on the moisture-sensitive animals, and the indirect effects of 6. Recovery from disturbances in cloud forests less energy for reproductive activities, suppression of the immune system, parasite growth and the loss of 6.1. Canopy openings and inÕasions food source, i.e. insects, which in turn are climate- sensitive. The possible reduction of water input to CF plants at the edge of the CF in Sri Lanka the CFŽ. Section 3.2 may cause CF amphibians the appear to be relatively frost-tolerant while those at following problems: difficulty in finding suitable the center are not well adapted to frostŽ Werner and egg-laying spots, crowding at ponds that do form, Balasubramaniam, 1992. . Similarly, CF plant species the drying up of ponds before pond-frog tadpoles appear to be intolerant of fire. So whenever there is a hatch, the drying up of eggs of leaf-laying frog eggs, canopy opening, and frost or fire is a threat, CF and so forth and so onŽ. Donnelly and Crump, 1998 . species along the new opening are likely to die, At this point one can only conclude that specialist which can result in a spread of the openingŽ Werner, species, be they dietary or habitat specialists, are 1995, 1998. . This arises partly because cloud forest likely to be more affected, just as we concluded for trees are relatively slow growing and thus will not the microhabitat colonization of epiphytes. And, since compete well with fast growing pioneer species in so many amphibian specialist species are found the openings. Non-native pioneer species can be within the cloud forest, once again, the cloud forest especially threatening as the resident cloud forest inhabitants face a heightened risk. Local extinctions species will not have evolved any defenses against will have impacts on the food chain, both up and them. down, as well as impacts on competitive balances Canopy openings can arise from a number of with other species. events—hurricanes, typhoons, logging and even Cloud forest frogs have not escaped the world- drought. While there is evidence that CF trees can wide decline in frog abundances. We discussed a withstand some amount of droughtŽ Bruijnzeel and striking case earlierŽ. Section 3.3 in which 25 of 50 Proctor, 1995. , drought-induced canopy die backs species of frogs and toads disappeared in 1987 from have been observed in the Sri Lankan CF. Subse- the Monteverde Cloud Forest. Some scientists claim quently, fire and frost, which were previously sup- that this is normal population dynamics, but it seems pressed in the CF interior, prevented tree growth in unlikely given the failure of 20 of those 25 lost the newly opened canopy and promoted the expan- species to reappear in the intervening timeŽ Pounds sion of grass areas. Now, Rhododendron arboreum et al., 1999. . Furthermore, anuran and other verte- is the only tree species found above 1500 m in the brate species abundances within the Monteverde grasslands that replaced the CF of Sri LankaŽ Werner, cloud forest are correlated to the number of dry days. 1995. . Similarly, in Tahiti, the montane forest has Bird species that were previously confined to alti- been invaded by the fast growing South American tudes below the CF are now regularly breeding in melastome tree following two destructive hurricanes what was strictly CF habitat. Two highland endemic Ž.Ž.Merlin and Juvik, 1995 . In a short-term 3 year lizard species have disappeared from the CF re- but detailed study of Andean mountain forest succes- cently, while one drier climate adapted lizard has sion in pastures, SarmientoŽ. 1997a,b showed that shown no distributional changes. The changes in all introduced non-native grasses limited the dispersal three classes, birds, lizards and anurans, are signifi- success of montane tree species into the pasture, via cantly correlated with the number of dry days in the seedling competition. In the past, anthropogenic dry season. Pounds et al.Ž. 1999 and Foster Ž. 2001 burning led to the spread of Andean grasslands, argued that a complete chain of reasoning connects paramo, but with native species with which the local anthropogenic climate change to the increase in dry shrubs and trees evolved and can eventually success- P. FosterrEarth-Science ReÕiews 55() 2001 73–106 93 fully compete. Sarmiento concludes that these non- species that can halt regeneration altogetherŽ Myers, native tussocks must be removed before the natural 1991; Werner, 1995; Merlin and Juvik, 1995; progression from grassland to shrubs to forest can Sarmiento, 1997b. . In one CF in Hawaii, the re- occur, otherwise the succession may follow the path moval of feral pigs has resulted in an unusually rapid of forest to pasture. In Hawaii, feral pigs damage the recovery. This has probably arisen because other CF vegetation and non-native species are invading common threats, especially fire, are lacking and that Ž.Ž.Medeiros et al., 1995 . Olander et al. 1998 found most of the alien plants there are disturbance special- that non-endemic species found at disturbed sites did ists. In this case, once the disturbance was removed, not spread far into the undisturbed cloud forest at the the native CFs were able to recover. In the event of Luquillo Experimental Forest. This suggests that the climate change-induced openings, recovery will CF may have some resiliency against invasions if it probably depend on the presence of further distur- remains undisturbed, but in areas with disturbed soil, bances including fire, frost, invaders and soil health. it looses the competitive battle against non-native invaders. This seems reasonable in light of the slow 6.3. Secondary forest— poor in biodiÕersity? growth rate of CF speciesŽ. see the following section and the lack of co-evolution which would prevent In cases where the cloud forest has recently re- the development of defense mechanisms against covered from destructive events, both the biodiver- non-native species. sity and the biomass tend to be very low. For instance in the Solomon Islands, which frequently 6.2. RecoÕery rates experiences typhoons and is also under pressure from an increasing human population, most of the Byer and WeaverŽ. 1977 concluded that the TMCF cloud forest is in some stage of regeneration. The has the slowest recovery rate of all tropical forests. flora on this island is markedly poor relative to other Olander et al.Ž. 1998 quantified recovery rates in the tropical Pacific islands, with numerous species poorly cloud forest of the Luquillo Experimental Forest, represented or completely absentŽ Merlin and Juvik, Puerto Rico. They found that cloud forest recovery 1995. . On the other hand, tree species richness might in areas where the root mat was removed is much be expected to recover relatively rapidly with respect slower than in other forest types at Luquillo and to other tropical forests. This is because there is less extrapolated their observations to estimate a 200- to tree diversity and hence fewer successional stages 300-year recovery time. In cases where the soil was within the CF. The case for the epiphytes also has not so disturbed, recovery rates were much faster. competing effects. Epiphytes are characteristically Ž.Weaver, 1990; Olander et al., 1998 . Compacted quick and successful colonizers and hence might be soil can also lead to cascading damage via enhanced expected to recover rapidly. However, it seems that flooding and more soil erosion. ScatenaŽ. 1995 points they will be limited in the secondary forest, which out that simple foot trails in Luquillo that have been lacks the old trees that harbor many microclimates nearly abandoned for 25 years were still recogniz- and carry the bulk of epiphytic mass in old growth able. Except for the possible increase of more high cloud forestsŽ. Hietz and Hietz-Seifert, 1995 . intensity rainfall events, climate change may not In summary, the combination of increased water have such drastic consequences for the soil. Conse- stress and increased strong wind events is likely to quently, under the pressure of changing climate the result in more frequent CF canopy die backs. Subse- CF may be able to reestablish themselves faster than quently, nonnative grass invasions, fire, frost, animal the 200–300-year recovery rate estimated for the grazing and so forth may slow or prevent cloud Luquillo Forest. However, it seems plausible that forest recovery and may even spread the canopy other stresses may arise, which in conjunction with gaps. In the event of recovery, it may take centuries climate change will counterbalance this shorter re- to rebuild the biomass of the trees and hence the covery rate. richness of the epiphyte community to the level of Another important factor in the recovery and relo- the old growth forests. But clearly, cloud forests cation of CF species is competition with nonnative have survived typhoons and extreme events in the 94 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 past. Perhaps the cloud forest can recover on Flenley, 1979.Ž. . According to Flenley 1997 , similar timescales longer than we have presently observed, phenomena have been reported for the sub-alpine maybe with input from nearby non-damaged cloud shrubbery in the Colombian AndesŽ an ecosystem forests. It is currently unclear if cloud forests can similar to the CF with stunted growth and small survive the combined stresses of increasing isolation thick leaves. and for the UMRF at Pedro Palo. and fragmentation, increased temperatures, drought, Alpine grasslands appear to have expanded to fill the rainfall and wind, as well as direct anthropogenic gap. Temperature decreases acting alone would shift pressures, not to mention the catastrophic conse- the vegetation zonation downwards, but would not quences of the loss of cloud contact with forest pinch out a given ecotone. An alternative explanation conjectured in Section 3 above! It is difficult to is the UV-B theory and it is discussed below. know if the above cases of slow recovery or failure The New Guinea cloud forests currently range to recover are our first glimpses of the large-scale from about 2800 to 3700 mŽ. Flenley, 1992a . Above breakdown of CFs or the result of peculiar combina- the tree line, a small scrub zone sometimes exists for tions of stresses. roughly 200 m in altitude, eventually grading into alpine grassland. Flenley argues that the upper limit of this mossy forest is apparently determined by low 7. Past climate change impacts on the cloud forest temperatures; low temperatures kill woody plant types including CF trees. Below the lower limit of the cloud forest, at about 2800 m, LMRF plants 7.1. Forest retreat and expansion thrive. Flenley argues that within the cloud forest, above 2800 m, UV-B radiation is so high that unless In the past, tropical forest have expanded and plants have some protective mechanism they cannot retreated in response to climate changes. The pollen survive. One such protective measure is the produc- fossil record of the tropics for the last 18,000 years tion of flavinoids in the leaves, which absorb harm- Ž. was concisely summarized by Flenley 1998 and ful UV-B radiation, and in turn lead to stuntingŽ see shows forest advances and retreats as well as altitude Section 2.2. . LMRF plants cannot survive in the high shifts in forest boundaries. By comparing records UV-B radiation at high altitudes, and the twisted from sites at various altitudes, Flenley estimated that stunted CF plants are at a competitive disadvantage the forest limit in the tropical regions of Latin Amer- at lower altitudes, thus a UV-B cutoff marks the ica, Southeast Asia, the west Pacific and in Africa LMRFrCF boundary at 2800 m. FlenleyŽ. 1992b,c descended approximately 1000 m during the Last conjectures that under climate change, this altitude Ž. Glacial Maximum LGM . Similarly, all three re- would probably not change much as it is dependent Ž gions show evidence for montane species including on the intensity of UV-B radiation, although cloudi- . cloud forest ones descending to lowland sites during ness may act to magnify this effect and could thus Ž the LGM descents of 1000 m and more are docu- play a role in adjusting the altitude of this boundary. mented in Latin America, Colinvaux et al., 1996a,b; During the late Pleistocene, when temperatures 1200 m in Indonesia, Van der Kaars and Dam, 1997; dropped, the forest limit dropped to about 2000 m, . 600 m in Africa, Maley, 1991 . below the UV-B cutoff. The zone where the special UV-B protecting characteristics of stunted vegetation 7.2. Disappearance of CF species in the pollen fossil have a competitive advantage, above 2800 m, was record then above the forest limit, and the CF species disappear. If this is indeed the main mechanism The altitude shifts for the lower and upper limit of driving CF morphology and its lower limit, then a given ecosystem are not expected to be the same, future climate change may drive a potentially large meaning ecotone altitude ranges will changeŽ see expansion of CF areas as the lower UV-B limit is Section 4.2. . In the Late Pleistocene, species of the expected to remain at the same altitude, but the tree mossy forest of New Guinea almost completely dis- line will probably rise. However, we must remember appeared from the pollen fossil recordŽ Walker and that the UV-B mechanism is probably a secondary P. FosterrEarth-Science ReÕiews 55() 2001 73–106 95 consideration to CF altitude distribution; cloud cover dance maximum in this cloud forest species, and apparently dominates CF distributions. However, the a concurrent widespread synchronous increase of UV-B mechanism could well limit CF distribution Podocarpus at several other mountains in equatorial and explains many aspects of the CF, whereas other Africa, suggest an increase in wet cloud forest habi- mechanisms seem to explain only one. tat. Seemingly in contradiction is the Lake Bosumtwi An alternative explanation for the loss of stunted core in western Africa. It shows a dramatic lake level vegetation zones during the LGM is the low CO2 drop of about 130 m at the start of the Podocarpus concentration documented for this eraŽ Street-Perrott, maximum, 3700r3800 years ago. Furthermore, there

1994. . Low CO2 levels are known to favor grasses is evidence that a nearby forest retreated a few Ž.C4 plants over trees Ž. C3 , perhaps explaining the hundred years later, from 3000–2000 years BP, al- expansion of the grasslands. This CO2 theory does though there are indications that the forest retreat not explain the Massenerhebung effect or stunting, may have started earlierŽ. Maley, 1991 . Apparently but it is not necessarily the same mechanism that some cloud forest species’ ranges are expanding causes stunting and controls the forest boundary. The during a time when at least one lake level dropped

CO2 or UV-B theory can also help explain apparent and other forests may have been shrinking. This temperature depression discrepancies during the picture is consistent with a decrease in overall rain- LGM. Pollen records from tropical mountains record fall, but enhanced cloudiness at mountain sites where a vegetation descent, which, if based solely on tem- cloud forests reside. This suggests an increase in perature, imply a decrease of some 6–108 during the stratiform clouds, which do not necessarily rainŽ see LGM. This is larger than estimates of SST cooling Section 3.3. but still provide the cloud forest charac- rates, even the larger rates found for corals of 58C teristics of enhanced horizontal precipitation and re- Ž.Ž.Guilderson and Schrag, 1994 . Flenley 1997 sug- duced evaporationŽ Maley, 1997; Maley and Brenac, gests that either the UV-B theory or the CO2 theory 1998. . could have been responsible for part of the observed descent rather than solely temperature responses. 7.3.2. Cloudiness changes and SSTs MaleyŽ. 1997 points out that the increased strati- form cloud formation from 3700r3800 to about 7.3. Tropical Africa, cloud types and Podocarpus 3000 years ago is also consistent with SST depres- sions in the Gulf of GuineaŽ Morley and Dworetzky, 7.3.1. The pollen record 1993. . As SSTs drop, uplift is replaced by subsi- Maley et al. have studied changes in forest distri- dence and stratiform clouds replace cumulous clouds. butions in tropical west Africa and found evidence of Žsee Zheng et al., 1999 for a recent study of correla- changes in the distribution of species found only in tions between tropical Atlantic SSTs and west African the cloud forest today, Olea capensis and Podocar- rainfall. . The cold SSTs are related to an enhanced pus latifolius Ž.Maley and Brenac, 1998 . These two upwelling in the ocean driven by high trade wind montane taxa sometimes show different reactions to stress, which in turn are linked to anti-cyclonic a given change in climate, O. capensis favors slightly activity. This is a close analogy to the Pacific’s El drier climates than P. latifolius ŽMaley and Brenac, Nino˜ phenomenon. In fact, the dry stratiform phase 1998. . This cautions us to be very careful in making in western Africa appears to be in phase with wet statements about the retreat or advance of entire conditions in South America and a wet Sahara— biomes based on just one species. The pollen fossil characteristics of current El Nino˜ modes. The period record of Lake Barombi Mbo shows a maximum in before 3700r3800 years ago has opposite moisture Podocarpus 3800–3400 years ago, when O. capen- trends and corresponds to a La Nina-like˜ phase. sis abundance is quite low. The shores of Barombi Perhaps the most striking lesson to be learned Mbo itself may not have hosted the Podocarpus, from Maley’s work in the context of this review is because Podocarpus pollen is able to move over that climate change dramatically alters cloud forest long distances, so the pollen may have floated in species extents. Maley also notes that the 3700r3800 from nearby Mount Kupe. However, the fossil abun- years ago climate shift coincides with the end of the 96 P. FosterrEarth-Science ReÕiews 55() 2001 73–106

Tenerean NeolithicŽ. Roset, 1987 civilization in the ley, 1996. . This plane flooding seems reasonable in Sahara and the Harapan–Indus civilization in India light of the documented descents of montane species Ž.Bryson and Swain, 1981 , which is another sobering and forest extensions throughout the tropics during lesson about the potential impacts of climate change. the LGM and evidence of apparently relic montane Further back in time, the Lake Bosumtwi pollen patches at low altitudes where there is persistent core shows the characteristic montane taxa descent cloudiness and fogŽ. Maley and Brenac, 1998 . during the LGM. From around 28,000 to 9000 years On the other hand, some authors interpret the BP, montane species dominated the core and lowland presence of the same species residing on different rain forest species practically disappear. One of the mountains as evidence that the seeds somehow jump typical montane elements found in the core for the from site to siteŽ Vazquez-Garcia, 1995; Werner, LGM period, Olea hochstetteri, today resides on 1998. . Common island species must have evolved in mountains 700 km away from the lake at an altitude this manner, so this theory is sometimes referred to of about of 1200 m asl. The lake is only at 100 m asl as the archipelago theory of speciation. Sugden and but nearby hills, with altitudes of 400–550 m asl, RobinsŽ. 1982a,b, 1983 favor such an interpretation may have hosted O. hochstetteri during the LGM, for the isolated Serranıa´ de Macuira cloud forest implying a minimum descent of 600 mŽ Maley, whose non-endemic, widespread, early successional 1991. . species suggest recent favorable cloudiness changes and the aerial arrival of CF seeds. Merlin and Juvik 7.4. Plant migrations Ž.1995 provide a useful survey of species at isolated island cloud forests in the Pacific and conclude that Whether or not plants will be able to survive the widespread occurrence of Weinmannia and the future shifts in their climatically optimal range is a tree fern Cyathea, common if not dominant CF complicated question. Barriers to successful recolo- species at these sites, suggests that they are dispersed nization include: by wind. Paleo-records of woody species in Europe Ž.1 the existence of a suitable site, both Ž. a climat- and North America during the Holocene show rapid ically, andŽ. b undisturbed by mankind; horizontal dispersal rates, 50–2000 mryear. This Ž.2 physical barriers such as rivers, mountains, suggests that there was either long distance seed oceans, roads, and cities in the migration path; dispersal, or that there may have been pre-existing Ž.3 migration rates versus climate change rates; ‘infection sites’ that spread under climate change Ž.4 competition at the new site with perhaps Ž.Melillo et al., 1996 . In summary, whatever the previously unencountered species; mechanism of montane plant migrations—flooding Ž.5 the slow maturation rate of the CF. of the plains during cold spells, mountain hopping or advance infection sites—there is abundant evidence that plants have migrated across long distances and 7.4.1. Methods of migration sometimes quite rapidly. We know that forest species have successfully shifted their latitudinal and altitudinal ranges in the 7.4.2. Migration rates past. We also know that even in the event of nearly However, future climate changes are expected to complete disappearance of species in the pollen require migration rates 10 times faster than that record at a given site, the species can reestablish which is typically observed in the pollen fossil record itself, as happened with cloud forest species in New Ž.Melillo et al., 1996 . Estimates for the migration Guinea during the LGMŽ. see above . Evidence for rates of montane species are much slower than those long distance distribution of montane taxa comes of lowland species. For example, the expansion from similar montane taxa inhabiting widespread growth rate of one dominant alpine species ŽCarex African forests. This has led some authors to con- curÕula.Žhas been estimated at 0.5 mmryear Grab- clude that montane taxa and animals descended to herr et al., 1978. —far too slow to track even the the plains during the last LGM and migrated to lower estimate of climate change shift rates of 2 different mountainsŽ Maley and Elenga, 1993; Ma- mryearŽ. see the following paragraph . P. FosterrEarth-Science ReÕiews 55() 2001 73–106 97

Apart from arriving at a new site, species must nature of cloud forests, CF may have already over- complete their life cycle before the climate envelope come this kind of barrier through long distance seed has moved on. Estimates for the establishment of dispersal. However, the loss of areas to colonize new montane forest indicate it is a slow process. represents a serious threat to the cloud forest, which Korner¨ Ž. 1994 concludes that paleo-ecological evi- may just be able to move upslope in time. For the dence for montane tree line migration indicates that many cloud forests that reside on mountain peaks, new mature montane tree lines take at least 200 these details are irrelevant—these cloud forests are years to establish. Walker and ChenŽ. 1987 found likely to be pushed out of existence. that doubling times for the areal expansion of vari- ous rainforest tax in the early Holocene, a time whose warming is thought to be similar to predicted changes in the next 100 years, were on order of 100 8. Summary years for many species. Podocarpus, a characteristic CF species, was found to have a doubling time of Mountains host an incredible array of climatic 200–400 years. This timescale is similar to Korner’s¨ variability, soil types and site exposure, resulting in 200-year estimate for montane tree line establish- small isolated eco-niches. This is even more pro- ment and Olander et al.’sŽ. 1998 estimate for recov- nounced in the cloud forest, which, of course, relies ery of the cloud forest. So let us assume it will take on regular immersion in low-lying clouds. Within the at least 200 years for the cloud forest to move—any- cloud forest itself are also microhabitat zonations, where. Given current cloud forest altitude bands of which the epiphytes and fauna have evolved to ex- about 500 m, and an altitude shift rate of 200 mr100 ploitŽ. Benzing, 1998 . The fragmentary nature of the years, a lower limitŽ. see Section 4.2 , we estimate cloud forest, and especially of its epiphyte and anu- that in 200 years the entire climate optimum belt of ran communities, promotes both explosive speciation cloud forest will be relocated. This is roughly the and exceptionally high rates of endemismŽ Gentry, same as the speculative relocation timescale dis- 1992; Leon´ and Young, 1996. , making cloud forests cussed above. Perhaps cloud forest species will be valuable as biological hotspots, but also highly sus- able to spread just fast enough to keep up with ceptibility to climate changeŽ not to mention direct climate change. anthropogenic disturbances—Young, 1994. . In a These relocation and reestablishment timescales summary statement of the 1993 TMCF conference, should not be confused with the timescales for other MarkhamŽ. 1998 concluded that of all tropical forests, possible climate changes. For instance, the impact of the TMCF may be especially vulnerable to climate increasing CO2 on patterns in the global temperature change and we can only concur. One hopeful aspect field may already be being detectedŽ Santer et al., of this highly differentiated climate is that the frag- 2000. . Other predictions of the impact of the in- mentation itself might provide an excellent opportu- crease of CO2 on particular aspects of the climate, nity for nearby relocations of climate sensitive species such as the predictions for the relative humidity field in the event of climate changeŽ Young, 1992; Peters in Fig. 2, are typically given for the time when the and Darling, 1985. . However, many other problems

CO2 content of the atmosphere has doubled relative in relocation are expected to arise, which may be to pre-industrialized levels. This is expected to occur especially difficult for the slow growing cloud forest sometime in the following century. to overcome. Aside from the migration race against climate In response to climate changes, the cloud forest is change, another barrier to migration will exist in the not expected to shift en masse to a new location future, which did not hamper past migrations. Fewer Ž.Colinvaux et al., 1997, 2000 . Plants have a vast and fewer natural areas are now available to colonize array of behaviors, from low temperature limits, to as humans continue to transform wild lands into reproduction based on day length, and consequently farms, roads and towns. It is often speculated that individual species will respond differently to climate such human structures may cause a physical barrier change. Some plants will move to new locations, to orderly migration routes, but given the island-like some will stay where they are and some will dieŽ see 98 P. FosterrEarth-Science ReÕiews 55() 2001 73–106

Fig. 4.Ž. a Possible altitude changes in the suitable climatic region. The current location of the ecotone boundaries is represented by solid lines and a projected ecotone shifting of about 500 m is represent by dashed lines. This amount of altitude shift is arrived at in several ways, but relies mostly on temperature increases. Lower altitude ecotones may expand more than upper onesŽ. see Section 4.2 , and pinch out some ecotones. This effect is represented in the figure as in increase in the LRF and UMRF’s altitude range. The current cloud forest location is represented by the dark clouds. If the UV-B hypothesis explains the lower limit of the cloud forest, rather than temperature constraints, it is possible that the extent of possible cloud forest habitat may expand in the future. Such a case is represented as the dashed semi-circle, but it seems most likely that the cloud forest will relocate where cloud contact is most frequentŽ. see Section 7.2 as represented by the light grey clouds in the figure. Cloud forests residing at the top of mountains may be pushed into extinction.Ž. b Relocation of cloud forest species. Speculative impacts on different species within the cloud forest are shown. Three different futures are represented: relocation to a new site Ž™ ., staying in the current location Ž.` and death Ž.= . Of the species that move, some may move upslope, some to other mountain sites and some off the tops of mountains into extinction. Higher altitude species within the cloud forest are more likely to still experience frequent cloud coverŽ. if cloud bases move up in altitude . Therefore, the figure shows a tendency for lower altitude species dying and higher ones surviving. However, all high altitude species may not survive as along with the reshuffling of species mixes, new competitive struggles will be encountered. Furthermore, although new climatically suitable regions may exist, there may be other problems involved in relocation, including migration barriers and rates and no undisturbed location to colonize. P. FosterrEarth-Science ReÕiews 55() 2001 73–106 99

Fig. 4b. . A period when there was no suitable region Ž.Bazzaz, 1998 . And of course, there are character- for CF species appears to have occurred during the istics of the cloud forest which are little understood, LGM in New Guinea when cloud forest species such as the tie between the stunting of trees, the disappeared from the pollen recordŽ Walker and xeromorphic leaves and the frequent cloud cover Flenley, 1979. and such a state with no analog Ž.Section 3.3 . Until this riddle is solved, it is difficult climate for CF species may occur again in the future. to predict the implications that climate change will In the case of epiphytes, not only must the microcli- have on the cloud forest. mate be reproduced but also the substrate, nutrient In the case of wholesale loss of old growth forest, input and anchoring conditions. In the case of cloud much will be lost. Loss or reduction of the cloud forests currently residing on peaks and ridges, it forest habitat will mean a loss of endemic species seems very plausible that they will be pushed into and loss of genetic diversity, with consequent losses oblivion. to crop cross breeding, local people’s livelihoods, If an analog climate does exist somewhere in an potential medicines and much more. As plants die, undisturbed site on the mountain side, cloud forest the animals that rely on them will also die. Some species’ seeds appear to be capable of traversing animals will themselves be directly impacted by long distances to reach a given site. There is evi- climate changesŽ. see Section 5.3 . Even if secondary dence for long distance dispersal of cloud forest forest can establish itself, it is markedly poorer in seeds in both paleo-recorded shifts of distributions biomass and biodiversityŽ. Section 6.3 . For instance, and the archipelago-like family grouping of cloud old growth trees harbor many more microclimates forestsŽ Merlin and Juvik, 1995; Vazquez-Garcia, for epiphytes, which in turn provide food, water and 1995. . But even if the seeds get this far, the battle is homes for animals. The CF epiphyte community is still not over. It may take the mature cloud forest very dependent on the high humidity of the CF and 200–300 years to become established, longer than thus is likely to be adversely affected by increases in other tropical forests. It is unclear if the suitable water stress from increasing temperatures, decreasing climate envelope will have moved on by this time cloudiness and other climate changesŽ. see Fig. 5 . Ž.see Section 7.4 . Part of the establishment of a BenzingŽ. 1998 argues that epiphytes play a large mature forest is a competitive battle with other plants. role in the light, water and nutrient cycles of the If non-native species are present, the battle can be forest. This suggests that their death may have cas- much harder to win, as the cloud forest species will cading effects on the CF. Aside from biological loss, not have evolved defenses against these intruders. destruction of the cloud forest may increase flooding This appears to have occurred in the tropical mon- and decrease dry weather flow to downstream tane forests of the Andes where non-native tussocks ecosystems and human settlements that rely on the of grass appear to be arresting the regeneration the cloud forest to provide these water servicesŽ Section cloud forestŽ. Sarmiento, 1997a and similarly in Sri 5.1. . This is particularly alarming as high intensity Lankan cloud forestsŽ. Werner, 1998 . rainfall events and droughts may both increase under Aside from the shifting of the climatically suitable global warmingŽ. Section 3.2 . Many of these impacts region, other climate change threats may increase. are presented in the flowchart of Fig. 5. For instance, there may be more typhoons, more It is often stated in the literature that the intense intense rainstorms and enhanced wind damage in the sensitivity of the cloud forest to climate can be used future. This is particularly devastating for the cloud as an early warning system. In addition to the tight forest as openings in the canopy of the cloud forest coupling of the life in the cloud forest to climate, the appear to be very slow to regenerate, and in a few changes in climate at cloud forest locations may cases did not recover at all. Fire, frost and soil change faster than in other regions. As the globe damage all act to spread the gap, and prevent regen- warms, we expect more moisture to be pumped into eration. And once again, the invasion of non-ende- the atmosphere. This is likely to have two effects mic grass species even seems to arrest regeneration that will directly affect the cloud forest. First, the Ž.Section 6 . Other climate threats include changes in increased atmospheric moisture may shift the value seasonal timing to which plants are very sensitive of lapse rate, the change in temperature with altitude, 100 P. FosterrEarth-Science ReÕiews 55() 2001 73–106 n outcomes of t comprehensive but it indicates if the effect is ted by solid lines the same q nnected to potential impacts Ž. if a decrease is expected. Potential results of a given climate change are listed below the climate change and are connected by solid lines of the y Ž. increasing dry seasons and extreme events: both will lead to increased soil erosion and wind damage. See the text for discussion. same color as the text of the original climatic change. Similarly, subsequentdoes results list of impacts some are of listed the below likely the driving outcomes forcing from and climate again change connec on the bottom line. The lines leading off of the figure at the top left and right represent commo color as the source climaticby change. the Impacts multicolored that thick could line. be Events the result that of may all cause five impacts climate printed changes, above such them as in water the stress chart or are loss of connected the by cloud dashed forest, lines are with co arrows. This chart is no expected to increase and a Fig. 5. Flow chart of potential impacts of climate change. At the top of the figure, potential climate changes are printed in five different colors, a P. FosterrEarth-Science ReÕiews 55() 2001 73–106 101 towards the saturated lapse rate such that high alti- areas that physically connect distinct TMCFs to al- tude sites experience a greater amount of warming low for migration and repopulation. than nearby lowland sitesŽ. see Section 3.1 . Second, The UNEP World Conservation Monitoring the cloudiness regime will certainly change. Several CentreŽ http:rrwww.unepwcmc.orgrforestr GCMs suggest that relative humidity in the tropics is cloudforestrenglishrhomepage.htm. is coordinating likely to decrease by several percent by the time that a much-needed conservation effort for the TMCF. the CO2 content of the atmosphere has doubled. The mutual support of their program and local peo- Other GCMs concur and predict that low-level ple will be essential in the program’s success, see the cloudiness will decrease with warming. A series of fascinating case study of the protection of a cloud studies at the Monteverde Cloud Forest concludes forest in EcuadorŽ. Becker, 1999 . Education will be that a recent changes in frog, bird and lizard abun- required to teach conservation measures—both how dances are linked to a rising cloud baseŽ Section and why, including how to sustainably farm or har- 3.3. . Perhaps the warning bell is already ringing. vest the forest. The excellent color brochure by While climate change is certainly a big threat to Bruijnzeel and HamiltonŽ. 2000 outlining the impor- the cloud forest, no discussion of the cloud forest tance of, and threats to, the cloud forest will cer- would be complete without mentioning the current tainly be a useful educational tool. Those of us who high-pitched frenzy of rainforest cutting. It has been live in developed countries must do our part to estimated that TMCFs comprise roughly 1r4 of all ensure that our personal and government’s money tropical montane forestsŽ. Persson, 1974 with an support sustainable forestry. More involvement from annual deforestation rate for tropical hills and moun- scientists in battles to help indigenous people save tains of 1.1%Ž. Doumenge et al., 1995 . This defor- their sustainable ways of life and protect the world’s estation rate exceeds that of other tropical forests forests is desperately needed. Ž.0.8% primarily because montane forests have been left relatively untouched until recently. The TMCFs high altitude and twisted trees, often accompanied by Acknowledgements steep slopes, make logging and agriculture difficult. For this reason, they often remain quite pristine and Thanks to Stephen Schneider for all kinds of help provide a habitat not only for their many endemic including discussions, advice and support. Likewise, species, but serve as a refuge for other montane thanks to Akimasa Sumi for support and discussions. species whose habitats have been destroyed. This Colleagues Chris Still and Alan Pounds were invalu- makes the 1.1%ryear deforestation rate even more able critics and great resources. Fred Scatena, Fausto disturbing. Sarmiento and Philip Bubb provided thoughtful and Deforestation pressures on the cloud forest are greatly appreciated comments. All remaining errors similar to those on other tropical forests, including are, of course, my own. A. Kushiro, M. Kono, and logging, mining, road building and clearance for the staff of ISEI and CCSR are greatly appreciated farming and grazing. Furthermore, all forests are for their support. Thanks also to all the researchers threatened by climate change, but the TMCFs appear who sent their reprints. Finally, thanks to Dakota and to be exceptionally vulnerable to climate change. To Michael. most effectively save these biologically rich regions will require understanding the roles of human-in- duced local and global climate change in order to References best establish mitigation measures. Yet, we already know enough to state that to save TMCFs requires Alford, R.A., Richards, S.J., 1999. Global amphibian declines: a establishing them as protected forests, as soon as problem in applied ecology. Annu. Rev. Ecol. Syst. 30, 133– possible, with ample buffer zones around their 165. Barry, R.G., 1994. Past and potential future changes in mountain boundaries to allow for potential altitudinal and lati- environments. In: Beniston, M.Ž. Ed. , Mountain Environments tudinal migration of TMCF species. These protected in Changing Climates. Routledge, London, pp. 3–33. areas should include mountain ranges and lowland Bazzaz, F.A., 1998. Tropical forest in a future climate: changes in 102 P. FosterrEarth-Science ReÕiews 55() 2001 73–106

biological diversity and impoact on the global carbon cycle. Chu, P.-S., 1989. Hawaiian drought and the southern oscillation. Clim. Change 39, 317–336. Int. J. Clim. 9, 619–631. Becker, C.D., 1999. Protecting a Garua´ Forest in Ecuador: the role Colinvaux, P.A., De Oliveira, P., Moreno, J.E., Miller, M.C., of institutions and ecosystem valuation. Ambio 28, 156–161. Bush, M.B., 1996a. A long pollen record from lowland Ama- Bensiton, M., Fox, D.G., Adhikary, S., Andressen, R., Guisan, A., zonia: forest and cooling in glacial times. Science 274, 85–88. Holten, J., Innes, J., Maitima, J., Price, M., Tessier, L., 1996. Colinvaux, P.A., Liu, K.-B., De Oliveira, P., Bush, M.B., Miller, Impacts of climate change on Mountain Regions. Climatic M.C., Kannan, M.S., 1996b. Temperature depression in the Change 1995: Impacts, Adaptations and Mitigation of Climate lowland tropics in glacial times. Clim. Change 32, 19–33. Change: Scientific–Technical Analyses. Cambridge Univ. Colinvaux, P.A., Bush, M.B., Kannan, M.S., Miller, M.C., 1997. Press, pp. 191–213. Glacial and postglacial pollen records from the Ecuadorian Bensiton, M., Diaz, H.F., Bradley, R.S., 1997. Climatic change at Andes and Amazon. Quat. Res. 48, 69–78. high elevation sites: an overview. Clim. Change 36, 233–251. Colinvaux, P.A., De Oliveira, P., Bush, M.B., 2000. Amazonian Benzing, D.H., 1998. Vulnerabilities of tropical forests to climate and neotropical plant communities on glacial time-scales: the change: the significance of resident epiphytes. Clim. Change failure of the aridity and refuge hypotheses. Quat. Sci. Rev. 39, 519–5440. 19, 141–169. Bonnell, M., 1998. Possible impacts of climate variability and Diaz, H.F., Graham, N.E., 1996. Recent changes in tropical change on tropical forest hydrology. Clim. Change 39Ž. 2–3 , freezing heights and the role of sea surface temperature. 215–272. Nature 383, 152–155. Box, E.O., 1995. Factors determining distributions of tree species Dillon, M.O., 1997. Lomas formations—Peru. In: Davis, S.D., and plant function types. Vegetatio 121, 101–116. Heywood, V.H., Herrera-McBryde, O., Villa-Lobos, J., Brown, A.D., 1995. Ecology and conservation of the Argentine Hamiloton, A.C.Ž. Eds. , Centres of Plant Diversity, A Guide montane forest. In: Hamilton, L.S., Juvik, J.O., Scatena, F.N. and Strategy for their Conservation. WWF, Information Press, Ž.Eds. , Tropical Montane Cloud Forests: Proceedings of an Oxford, UK, pp. 519–526. International Symposium. Springer-Verlag, New York, pp. Donnelly, M.A., Crump, M.L., 1998. Potential effects of climate 107–115. change on two neotropical amphibian assemblages. Clim. Bruijinzeel, L.A., 1990. Hydrology of Moist Tropical Forests and Change 39, 541–561. Effects of Conversion: A State of Knowledge Review, Paris. Doumenge, C.D.G., Perez, M.R., Blockhus, J., 1995. TMCFs: Bruijnzeel, L.A., 1999. Hydrology of Tropical Montane Cloud conservation status and management issues. In: Hamilton, Forests: A Re-Evaluation. Second International Colloquium on L.S., Juvik, J.O., Scatena, F.N.Ž. Eds. , Tropical Montane Hydrology and Water Management in the Humid Tropics, Cloud Forests: Proceedings of an International Symposium. Panama City. Springer-Verlag, New York, pp. 24–37. Bruijnzeel, L.A., Hamilton, L.S., 2000. Decision time for cloud Emanuel, K.A., 1987. The dependence of hurricane intensity on forests. UNESCO International Hydrological Programme: Hu- climate. Nature 326, 483–485. mid Tropics Programme Series No. 13. Fang, J.-Y., Yoda, K., 1990a. Climate and vegetation on China: Bruijnzeel, L.A., Proctor, J., 1995. Hydrology and biogeochem- III. Water balance and distribution of vegetation. Ecol. Res. 5, istry of tropical montane cloud forests: what do we really 9–23. know? In: Hamilton, L.S., Juvik, J.O., Scatena, F.N.Ž. Eds. , Fang, J.-Y., Yoda, K., 1990b. Climate and vegetation on China: Tropical Montane Cloud Forests: Proceedings of an Interna- IV. Distribution of tree species along the thermal gradient. tional Symposium. Springer-Verlag, New York, pp. 38–78. Ecol. Res. 5, 291–302. Bruijnzeel, L.A., Veneklaas, E.J., 1998. Climatic conditions and Fang, J.-Y., Yoda, K., 1991. Climate and vegetation on China: V. tropical montane forest productivity: the fog has not lifted yet. Effect of climatic factors on the upper limit of distribution of Ecology 79Ž. 1 , 3–9. evergreen broadleaf forest. Ecol. Res. 6, 113–125. Bruijnzeel, L.A., Waterloo, M.J., Proctor, J., Kuiters, A.T., Kot- Flenley, J.R., 1979. The Equatorial Rain Forest. Butterworths, terink, B., 1993. Hydrological observations in montane rain London. forests on Gunung Silam, Sabah, Malaysia, with special refer- Flenley, J.R., 1992a. Palynological evidence relating to distur- ence to the Massenerhebung effect. J. Ecol. 81, 145–167. bance and other ecological phenomena of rain forests. In: Bryson, R.A., Swain, A.M., 1981. Holocene variations of mon- Goldammer, J.G.Ž. Ed. , Tropical Forests in Transition. soon rainfall in Rajasthan. Quat. Res. 16, 135–145. Birkhauser, Verlag, Basel, Switzerland, pp. 17–24. Byer, M.D., Weaver, P.L., 1977. Early secondary succession in an Flenley, J.R., 1992b. Two problems of the late Quaternary in the elfin woodland in the Luquillo mountains of Puerto Rico. Sunda–Sahul region and a possible solution. Proceedings of Biotropica 9, 35–47. the International Symposium on Global Change. IGBP, Shin- Caldwell, M.M., Robberecht, R., Billings, W.D., 1980. A steep juku, Tokyo. latitude gradient of UV-B radiation in the arctic-alpine life Flenley, J.R., 1992c. Ultraviolet-B insolation and the altitudinal zone. Ecology 61, 600–611. forest limit. In: Furley, P.A., Proctor, J., Ratter, J.A.Ž. Eds. , Cavelier, J., Goldstein, G., 1989. Mist and fog interception in elfin Nature and Dynamics of Forest Savanna Boundaries. Chap- cloud forests in Columbia and Venezula. J. Trop. Ecol. 5, man & Hall, London, pp. 273–282. 309–322. Flenley, J.R., 1995. Cloud forest, the Massenerhebung effect, and P. FosterrEarth-Science ReÕiews 55() 2001 73–106 103

ultraviolet insolation. In: Hamilton, L.S., Juvik, J.O., Scatena, Guilderson, T.P., Schrag, D.P., 1998. Abrupt shift in subsurface F.N.Ž. Eds. , Tropical Montane Cloud Forests: Proceedings of temperatures in the tropical Pacific associated with changes in an International Symposium. Springer-Verlag, New York, pp. El Nino.˜ Science 281, 240–243. 150–155. Gutzler, D.S., 1992. Climatic variability of temperature and hu- Flenley, J.R., 1997. The Quaternary in the tropics: an introduction. midity over the tropical western Pacific. Geophys. Res. Lett. Journal of Quaternary Science 12Ž. 5 , 345–346. 19Ž. 15 , 1595–1598. Flenley, J.R., 1998. Tropical forests under the climates of the last Hafkenscheid, R.L.L.J., 2000. Hydrology and biogeochemistry of 30,000 years. Clim. Change 39, 177–197. tropical montane rain forests of contrasting stature in the Blue Flohn, H., Kapala, A., 1989. Changes of tropical sea–air interac- Mountains, Jamaica, PhD Thesis, Vrije Universiteit, Print tion processes over a 30-year period. Nature 338, 244–246. Partners Ipskamp, Enschede, The Netherlands. Follman, G., 1963. Nordchilenische Nebeloasen. UMSCHAU, Hafkenscheid, R.L.L.J., Bruijnzeel, L.A.S., de Jeu, R.A.M., 1998. Heft 4, 101–104. Estimates of fog interception by montane rain forest in the Foster, P.N., 2001. Regional scale modeling of recent changes in Blue Mountains of Jamaica. Conference on Fog and Fog the hydrological cycle at the monteverde cloud forest. Mon. Collection, Vancouver, IDRC, Ottawa. Weather Rev., submitted for publication. Halpin, P.N., 1994. Latitudinal variation in montane ecosystem Fowler, A.M., Hennessy, K.J., 1995. Potential impacts of global response to potential climatic change. In: Beniston, M.Ž. Ed. , warming on the frequency and magnitude of heavy precipita- Mountain Ecosystems in Changing Climates. Routledge Pub- tion. Nat. Hazards 11, 283–303. lishing, London, pp. 180–203. Gaffen, D.J., Barnett, T.P., Elliott, W.P., 1991. Space and time Hamilton, L.S., Juvik, J.O., Scatena, F.N., 1995a. Tropical Mon- scales of global tropospheric moisture. J. Clim. 4, 989–1008. tane Cloud Forests: Proceedings of an International Sympo- Gaffen, D.J., Santer, B.D., Boyle, S., Christy, J.R., Graham, N.E., sium. Springer-Verlag, New York. Ross, R.J., 2000. Multidecadal changes in the vertical temper- Hamilton, L.S., Juvik, J.O., Scatena, F.N., 1995b. The Puerto ature structure of the tropical troposphere. Nature 287, 1242– Rico tropical cloud forest symosium: introduction and work- 1245. shop synthesis. In: Hamilton, L.S., Juvik, J.O., Scatena, F.N. Garcia-Martino, A.R., Warner, G.S., Scatena, F.N., Rainfall, D.L., Ž.Eds. , Tropical Montane Cloud Forests: Proceedings of an 1996. Runoff and elevation relationships in the Luquillo International Symposium. Springer-Verlag, New York, pp. Mountains of Puerto Rico. Caribb. J. Sci. 32Ž. 4 , 413–424. 1–23. Gentry, A.H., 1986. Endemism in tropical versus temperate plant Hastenrath, S., 1991. Climate Dynamics of the Tropics. Kluwer communities. In: Soule, M.E.Ž. Ed. , Conservation Biology: Academic Publishers, Dordrecht. The Science of Scarcity and Diversity. Sinauer Associates, Hastenrath, S., Kruss, P., 1992. The dramatic retreat of Mount Sunderland, MA, pp. 1532–1581. Kenya’s glaciers 1963–87: greenhouse forcing. Ann. Glaciol. Gentry, A.H., 1992. Tropical forest biodiversity: distributional 16, 127–133. patterns and their conservational significance. Oikos 63, 19–28. Hietz, P., Hietz-Seifert, U., 1995. Structure and ecology of epi- Goldammer, J.G., Price, C., 1998. Potential impacts of climate phyte communities of a cloud forest in Central Veracruz, change on fire regimes in the tropics based on magicc and a Mexico. J. Vegetation Sci. 6, 719–728. GISS GCM-derived lightning model. Clim. Change 39, 273– Holder, C.D., 1998. Fog precipitation in the Sierra de las Minas 296. Biosphere Reserve, Guatemala. Conference on Fog and Fog Grabherr, G., Mahr, E., Reisigl, H., 1978. Nettoprimarproducktion¨ Collection, Vancouver, Canada, International Development und Reproducktion in einem krummseggenrasen,Ž Caricetum Research Centre. curvulae. der Otztaler¨ Alpen Tirol. Oecol. Plant 13, 227–251. Holdridge, L.R., 1947. Determination of world plant formations Graham, N.E., 1995. Simulation of recent global temperature from simple climatic data. Science 105, 267–268. trends. Science 267, 666–671. Houze, R.A.J., 1993. Cloud Dynamics. Academic Press, San Graumlich, L.J., 1994. Long-term vegetation change in mountain Diego. environments. In: Beniston, M.Ž. Ed. , Mountain Environments Hue, N.V., Craddock, G.R., Adama, F., 1986. Effects of organic in Changing Climates. Routledge, London, pp. 167–179. acids on aluminum toxicity of subsoils. Soil Sci. Soc. Am. J. Grubb, P.J., 1977. Control of forest growth and distribution on 50, 28–34. wet tropical moutains: with special reference to mineral nutri- Hulme, M., Viner, D., 1998. A climate change scenario for the tion. Annu. Rev. Ecol. Syst. 8, 83–107. tropics. Clim. Change 39, 145–176. Grubb, P.J., Whitmore, T.C., 1966. A comparison on montane and Juvik, J.O., Nullet, D., 1995. Relationships between rainfall, lowland rain forest in Ecuador: II. The climate and its effects cloud–water Interception and canopy throughfall in a Hawai- on the distribution and physiognomy of the forests. J. Ecol. 54, ian Montane forest. In: Hamilton, L.S., Juvik, J.O., Scatena, 303–333. F.N.Ž. Eds. , Tropical Montane Cloud Forests: Proceedings of Grubb, P.J., Lloyd, J.R., Pennington, T.D., Whitmore, T.C., 1963. an International Symposium. Springer-Verlag, New York, pp. A comparison on montane and lowland rain forest in Ecuador: 165–182. I. The forest structure, physiognomy and floristics. J. Ecol. 51, Kitayama, K., 1995. Biophysical conditions of the montane cloud 567–601. forests of Mount Kinabalu, Sabah, Malaysia. In: Hamilton, 104 P. FosterrEarth-Science ReÕiews 55() 2001 73–106

L.S., Juvik, J.O., Scatena, F.N.Ž. Eds. , Tropical Montane tropical pacific: some aspects of their floristics, biogeography, Cloud Forests: Proceedings of an International Symposium. ecology and conservation. In: Hamilton, L.S., Juvik, J.O., Springer-Verlag, New York, pp. 183–197. Scatena, F.N.Ž. Eds. , Tropical Montane Cloud Forests: Pro- Korner,¨ C., 1994. Impact of atmospheric changes on high moun- ceedings of an International Symposium. Springer-Verlag, New tain vegetation. In: Beniston, M.Ž. Ed. , Mountain Environ- York, pp. 149–162. ments in Changing Climates. Routledge, London, pp. 167–179. Mitchell, J.F.B., Ingram, W.J., 1992. Carbon dioxide and climate: Kuiters, A.T., 1990. Role of phenolic substances from decompos- mechanisms of changes in clouds. J. Clim. 5, 5–21. ing forest litter in plant–soil interactions. Acta Bot. Neerl. 39, Morley, J.J., Dworetzky, B.A., 1993. Holocene temperature pat- 329–348. terns in the South Atlantic, southern and Pacific Oceans. Leo, M., 1995. The importance of tropical montane cloud forest Global climates since the Last Glacial Maximum, H.E. e.a. for preserving vertebrate endemism in Peru: the Rio Abiseco Wright. Univ. Minnesota Press, pp. 125–135. National Park as a case study. In: Hamilton, L.S., Juvik, J.O., Myers, N., 1991. Tropical forests: present status and future out- Scatena, F.N.Ž. Eds. , Tropical Montane Cloud Forests: Pro- look. Clim. Change 19, 3–32. ceedings of an International Symposium. Springer-Verlag, New Neilson, R.P., 1995. A model for predicting continental-scale York, pp. 198–205. vegetation distribution and water balance. Ecol. Appl. 5Ž. 2 , Leon,´ B., Young, K.R., 1996. Distribution of pteridophyte diver- 362–385. sity and endemism in Peru. In: Camus, J.M., Gibby, M., Northup, R., Dahlgren, R.A., Yu, Z., 1995. Intraspecific variation Johns, R.J.Ž. Eds. , Pteridology in Perspective. Royal Botanic of conifer phenolic concentration on a maritime terrace soil Gardens, Kew, pp. 77–91. acidity gradient: a new interpretation. Plant Soil 171, 255–262. Long, A.J., 1995. Restricted-range and threatened bird species in Ohsawa, M., 1990. An interpretation of latitudinal patterns of tropical montane cloud forests: what do we really know? In: forest limits in south and east Asian mountains. Vegetatio 97, Hamilton, L.S., Juvik, J.O., Scatena, F.N.Ž. Eds. , Tropical 1–10. Montane Cloud Forests: Proceedings of an International Sym- Ohsawa, M., 1993. Forest patter along hierarchical habitat gradi- posium. Springer-Verlag, New York, pp. 79–106. ents in east Asia. In: Kawanabe, H., Ohgushi, T., Higashi, M. Lowry, J.B., Lee, D., Stone, B.C., 1973. Effect of drought on Ž.Eds. , Symbiosphere: Ecological Complexity for Promotting Mount Kinabalu. Ecol. Syst. 6, 178–179. Biodiversity. Biology InternationalŽ. IUBS . Special Issue, vol. Maley, J., 1991. The African rain forest vegetation and paleoenvi- 29, pp. 30–36. ronments during late Quaternary. Clim. Change 19, 79–98. Ohsawa, M., 1995a. Latitudinal comparison of altitudinal changes Maley, J., 1996. The African rain forest—main characteristics of in forest structure, leaf-type and species richness in humid changes in vegetation and climate from the Upper Cretaceous monsoon asia. Vegetatio 123, 3–10. to the Quaternary. Proc. R. Soc. Edinburgh 104B, 31–73. Ohsawa, M., 1995b. The montane cloud forest and its gradational Maley, J., 1997. Middle to Late Holocene Changes in Tropical changes in southeast Asia. In: Hamilton, L.S., Juvik, J.O., Africa and Other Continents: Paleomonsoon and Sea Surface Scatena, F.N.Ž. Eds. , Tropical Montane Cloud Forests: Pro- Temperature Variations. Third Millennium BC Climate Change ceedings of an International Symposium. Springer-Verlag, New and Old World Collapse. York, pp. 254–265. Maley, J., Brenac, P., 1998. Vegetation dynamics, palaeoenviron- Olander, L.P., Scatena, F.N., Silver, W.L., 1998. Impacts of ments and climatic changes in the forests of western Cameroon disturbance initiated by road construction in a subtropical during the last 28,000 years B.P. Rev. Paleobot. Palynol. 99, cloud forest in the Luquillo Experimental Forest, Puerto Rico. 157–187. For. Ecol. Manag. 109, 33–49. Maley, J., Elenga, H., 1993. The role of clouds in the evolution of Persson, R., 1974. World Forest Resources. Royal College of tropical African mountain paleoenvironments. Veille Clima- Forestry, Stockholm. tique Satellitaire, Lannion 46, 51–63. Peters, R.L., Darling, J.D.S., 1985. The greenhouse effect and Markham, A., 1998. Potential impacts of climate change on nature reserves: global warming would diminish biological tropical forest ecosystemsŽ. guest editorial . Clim. Change 39 diversity by causing extinctions among reserve species. Bio- Ž.2–3 , 141–143. science 35, 707–717. Medeiros, A.C., Loope, L.L., Hobdy, R.W., 1995. Conservation Pounds, J.A., Fogden, M.P.L., Savage, J.M., Gorman, G.C., 1997. of cloud forests in Maui CountyŽ. Maui, Moloka’i and Lana’i , Tests of null models for amphibian declines on a tropical Hawaiian Islands. In: Hamilton, L.S., Juvik, J.O., Scatena, mountain. Conserv. Biol. 11Ž. 6 , 1307–1322. F.N.Ž. Eds. , Tropical Montane Cloud Forests: Proceedings of Pounds, A., Fogden, M.P.L., Campbell, J.H., 1999. Biological an International Symposium. Springer-Verlag, New York, pp. response to climate change on a tropical mountain. Nature 223–233. 398, 611–615. Melillo, J.M., Prentice, I.C., Rarquhar, G.D., Schulze, E.-D., Sala, Prentice, I.C., Cramer, W., Harrison, S., Leemans, R., Monserud, O.E., 1996. Terrestrial biotic responses to environmental R.A., Solomon, A.M., 1992. A global biome model based on change and feedbacks to climate. In: Houghton, J.T.Ž. Ed. , plant physiology and dominance, soil properties and climate. J. Climate Change 1995: The Science of Climate Change. Cam- Biogeogr. 19, 117–134. bridge Univ. Press, Cambridge, pp. 449–481. Raynor, B., 1995. Montane cloud forests in Micronesia: status and Merlin, M.D., Juvik, J.O. et al., 1995. Montane cloud forest in the future management. In: Hamilton, L.S., Juvik, J.O., Scatena, P. FosterrEarth-Science ReÕiews 55() 2001 73–106 105

F.N.Ž. Eds. , Tropical Montane Cloud Forests: Proceedings of Changes: International Impact of Climate Change. Cambridge an International Symposium. Springer-Verlag, New York, pp. Univ. Press, pp. 146–179. 274–283. Stadtmuller,¨ T., 1987. Cloud Forests in the Humid Tropics: A Richardson, B.A., Richardson, M.J., Scatena, F.N., McDowell, Bibliographic Review. The United Nations University, Tokyo. W.H., 2000. Effects of nutrient availability and other eleva- Still, C., Foster, P.N., Schneider, S.H., 1999. Simulating the tional changes on bromeliad populations and their invertebrate effects of climate change on troical montane cloud forests. communities in a humid tropical forest in Puerto Rico. J. Trop. Nature 398, 608–610. Ecol. 16, 167–188. Street-Perrott, A., 1994. Palaeo-perspectives: changes in terrestrial Rind, D., 1986. The dynamics of warm and cold climates. J. ecosystems. Ambio 23, 37–43. Atmos. Sci. 43Ž. 1 , 3–24. Sugden, A.M., 1981. Aspects of ecology of vascular epiphytes in Roset, J.P., 1987. Neolithisation,´´ Neolithique et post-Neolithique ´ two Colombian cloud forests: II. Habitat preferences of au Niger nordoriental. Bull. Assoc. Fr. Et. 203–214. Bromeliaceae in the Serranıa´ de Macuira. Selbyana 5, 264– Santer, B.D., Wigley, T.M.L., Gaffen, D.J., Bengtsson, L., Doutri- 273. axu, C., Boyle, J.S., Esch, M., Hmilo, J.J., Jones, P.D., Meehl, Sugden, A.M., Robins, R.J., 1982a. The ecological, geographic G.A., Roeckner, E., Taylor, K.E., Wehner, M.F., 2000. Inter- and taxonomic relationships of an isolated Columbian cloud preting differential temperature trends at the surface and in the forest, with some implications for island biogeography. J. lower troposphere. Science 287, 1227–1232. Arnold Arbor. 63, 31–61. Sarmiento, F.O., 1995. Human impacts on the cloud forests of Sugden, A.M., Robins, R.J., 1982b. The vegetation of the Serranıa´ Upper Guayllabamba river basin, Ecuador, and suggested de Macuira, Guajira, Colombia: a contrast of arid lowlands management responses. Ecol. Studies 110, 284–295. and an isolated cloud forest. J. Arnold Arbor. 63Ž. 1 , 1–30. Sarmiento, F.O., 1997a. Arrested succession in pastures hinders Sugden, A.M., Robins, R.J., 1983. Determinants of species com- regeneration of Tropoandean forests and shreds mountain position in some isolated neotropical cloud forests. In: Sutton, landscapes. Environ. Conserv. 24Ž. 1 , 14–23. S.L., Whitmore, T.C., Chadwick, A.C.Ž. Eds. , Tropical Rain Sarmiento, F.O., 1997b. Landscape regeneration by seeds and Forest: Ecology and Management. Blackwell Scientific, Ox- successional pathways to restore fragile tropandean slopelands. ford, pp. 43–56. Mountain Res. Dev. 17Ž. 3 , 239–252. Tanner, E.V.J., Kapos, V., Freskos, S., Healey, J.R., Theobalkd, Scatena, F.N., 1995. The management of Luquillo elfin cloud A.M., 1990. Nitrogen and phosphorus fertilization of Jamaican forest ecosystems: irreversible decisions in a nonsubstitutable montane forest trees. J. Trop. Ecol. 6, 231–238. ecosystem. In: Hamilton, L.S., Juvik, J.O., Scatena, F.N. Thompson, L.G., Pollard, D., 1997. Greenland and Antarctic mass

Ž.Eds. , Tropical Montane Cloud Forests: Proceedings of an balances for present and doubled CO2 from the GENESIS International Symposium. Springer-Verlag, New York, pp. version-2 global climate model. J. Clim. 10, 871–900. 191–198. Thompson, L.G., Mosley-Thompson, E., Davis, M.E., Lin, N., Scatena, F.N., 1998. An assessment of climate change in the Yao, T., Dyurgerov, M., Dai, J., 1993. Recent warming: ice Luquillo Mountains of Puerto Rico. Third International Sym- core evidence from tropical ice cores with emphasis on central posium on Water Resources. Asia. Global Planet. Change 7, 145–156. Schellekens, J., Bruijnzeel, L.A., Wickel, A.J., Scatena, F.N., Timmermann, A., Oberhuber, J., Bacher, J., Esch, M., Latif, M., Silver, W.L., 1998. Interception of horizontal precipitation by Roeckner, E., 1999. Increased El Nino˜ frequency in a climate elfin cloud forest in the Luquillo Mountains, Eastern Puerto model forced by future greenhouse warming. Nature 398, Rico. Conference on Fog and Fog Collection, Vancouver, 694–696. IDRC, Ottawa, pp. 29–32. Van der Kaars, S., Dam, R., 1997. Vegetation and climate change Schellekens, J., Bruijnzeel, A.J., Scatena, F.N., Bink, N.J., Holw- in West-Java, Indonesia during the last 135,000 years. Quat. erda, F., 2000. Evaportation from a troipcal rain forest, Luquilla Int. 37, 67–71. Experimental Forest, eastern Puero Rico. Water Resour. Res. Vazquez-Garcia, J.A., 1995. Cloud forest archipelagos: preserva- 36, 2183–2196. tion of fragmented montane ecosystems in tropical America. Schneider, S.H., Washington, W.M., Chervin, R.M., 1978. In: Hamilton, L.S., Juvik, J.O., Scatena, F.N.Ž. Eds. , Tropical Cloudiness as a climatic feedback mechanism: effects on Montane Cloud Forests: Proceedings of an International Sym- cloud amounts of prescribed global and regional surface tem- posium. Springer-Verlag, New York, pp. 315–332. perature changes in the NCAR-GCM. J. Atmos. Sci. 35, Walker, D., Chen, Y., 1987. Palynological light on tropical rain 2207–2221. forest dynamics. Quat. Sci. Rev. 6, 77–92. Schubert, C., 1992. The glaciers of the Sierra Nevada de Merida´ Walker, D., Flenley, J.R., 1979. Late Quaternary vegetational Ž.Venezuela : a photographic comparison of recent deglacia- history of the Enga District of upland Papua New Guinea. tion. Erdkunde 46, 58–64. Philos. Trans. R. Soc. B 286, 265–344. Slingo, J.M., 1987. The development and verification of a cloud Wallace, J.M., Hobbs, P.V., 1977. Atmospheric Science: An prediction scheme for the ECMWF model. Q. J. R. Meteorol. Introductory Survey. Academic Press, New York. Soc. 113, 899–927. Weaver, P.L., 1990. Succession in the elfin woodland of the Smith, T.M., Halpin, P.N., Shugart, H.H., Secrett, C.M., 1995. Luquillo Mountains of Puerto Rico. Biotropica 22Ž. 1 , 83–89. Global forests. In: Smith, K.M.S.J.B.Ž. Ed. , If Climate Werner, W.L., 1986. A comparison between two tropical montane 106 P. FosterrEarth-Science ReÕiews 55() 2001 73–106

ecosystems in Asia: PidurutalagalaŽ. CeylonrSri Lanka and Young, K.R., Leon,´ B., 1991. Diversity, ecology and distribution Pangrango-GedeŽ. Java . Mountain Res. Dev. 6 Ž. 4 , 335–344. of high-elevation pteridophytes within Rio Abiseo National Werner, W.L., 1995. Biogeography and ecology of the upper Park, north-central Peru. Fern Gazette 16, 25–39. montane rain forest of Sri LankaŽ. Ceylon . In: Hamilton, L.S., Zadroga, F., 1981. The hydrological importance of a montane Juvik, J.O., Scatena, F.N.Ž. Eds. , Tropical Montane Cloud cloud forest area of Costa Rica. In: Russell, R.L.E.W.Ž. Ed. , Forests: Proceedings of an International Symposium. Tropical Agricultural Hydrology. Wiley, New York. Springer-Verlag, New York, pp. 343–352. Zheng, X., Eltahir, E.A.B., Emanuel, K.A., 1999. A mechanism Werner, W.L., 1998. The mountain forests of Sri Lanka—a world relating tropical Atlantic spring sea surface temperature and of its own. In: Domroes, M.H.R.Ž. Ed. , Sri Lanka: Past and west African rainfall. Q. J. R. Meteorol. Soc. 125, 1129–1163. Present; Archaeology, Geography, Economics. Bohler Verlag, Ziska, L.H., Teramura, A.H., Sullivan, J.H., 1992. Physiological Weikersheim, pp. 118–130. sensitivity of plants along an elevational gradient to UV-B Werner, W.L., Balasubramaniam, S., 1992. Structure and dynam- radiation. Am. J. Bot. 79, 863–871. ics of the upper montane rain forests of Sri Lanka. Tropical Forests. In: Goldammer, J.G.Ž. Ed. , Transition. Birkhauser Pru Foster got her PhD on the collapse Verlag, BaselrSwitzerland, pp. 165–172. of gaseous spheres in outer space with Werner, W.L., Santisuk, T., 1993. Conservation and restoration of Roger Chevalier from the University of montane forest communities in Thailand. In: Lieth, H.M.L. Virginia in 1993. She did a postdoctoral Ž.Ed. , Restoration of Tropical Forest Ecosystems. Kluwer fellowship studying triggered star for- Academic, Netherlands, pp. 193–202. mation and the injection of short-lived Whitmore, T.C., 1989. Tropical forest nutrients, where do we radioactive nuclides into the forming stand? A tour de horizon. In: Proctor, J.Ž. Ed. , Mineral Nutri- solar system with Alan Boss at the ents in Tropical Forest and Savanna Ecosystems. Blackwell Carnegie Institution of Washington. She Scientific, Oxford, pp. 1–13. then switched to climate change re- Woodward, F.I., 1987. Climate and Plant Distribution. Cambridge search under the guidance of Steven Univ. Press, Cambridge. Schneider and did a postdoc at the Uni- Young, K.R., 1992. Biogeography of the montane forest zone of versity of Tokyo’s Center for Climate System’s Research with the eastern slopes of Peru. Mem. Mus. Hist. Nat., U.N.M.S.M. Akimasa Sumi studying recent changes in cloud formation in the 21, 119–149. tropics. She is currently a research scientist at the Institute for Young, K.R., 1998. Composition and structure of a timberline Study of the Earth’s Interior in Misasa, Japan. forest in north-central Peru. Forest Biodiversity. In: Comiskey, F.D.J.A.Ž. Ed. , North, Central and South America and the Caribbean. Unesco, Paris, pp. 593–623.