Methods

The principal techniques used to date in reconstructions of terrestrial paleotemperatures for the lowland neotropics have been pollen in lake sediments and noble gases dissolved in groundwater. The collection and processing of pollen samples are broadly standardized around the techniques outlined by Faegri and Iversen (1989). The area of methodology that is worthy of review is in the quantification of paleotem-perature.

Lowland palynological temperature reconstructions are generally based on the movement of indicator taxa. Apparently, stenothermic pollen taxa, or ones that at least occupy identifiable habitat ranges, are selected as indicator species. The modern range of these taxa is then compared with past ranges, and from this an inference is made about past climate. Van der Hammen and Gonzalez (1960) were the first researchers to correlate the downslope movement of pollen taxa to changes in temperature. They achieved this translation through the application of moist air adiabatic lapse rates. For instance, if in the past a species was documented 1000 m downslope of its present position, and the local moist air adiabatic lapse rate was 5°C of cooling per 1000 m of ascent, the paleotemperature change was 5°C cooler than present values.

Most researchers have since adopted this method of calculating temperature change, but we should examine some of its assumptions.

The first assumption is that there has been no significant evolutionary change in the requirements of the pollen taxa. Given the short time interval (20-50 tree generations since the last glacial interval), it is unlikely that evolution is a major problem, especially when more than one pollen taxon is exhibiting the same trend.

A second assumption is that the moist air adiabatic lapse rate has not changed. Studies of moist air adia-batic lapse rates reveal that they are rigidly constrained by the physical properties of air and are unlikely to have wavered outside of a narrow range (Webster and Streten, 1978; Rind and Peteet, 1985). Moist air associated with cloud forests and the wet lowland forests has a lapse rate of ca. 5°C. At the other extreme, desert dry air can have lapse rates approaching 7°C (Webster and Streten, 1978). Thus, the greatest potential change would be 2°C /1000 m of ascent. For many years, changes in ice age lapse rates were used to explain the anomalously warm oceans compared with the cool Andes (e.g., Haffer, 1991). However, as will be demonstrated later, when there is evidence of cooling, the forests are mesic or humid, indicating the presence of moist air. In other words, the Alnus, Hedyosmum, Weinmannia, Podocarpus, and Drimys populations that spilled down the flank of the Andes were of species adapted to the moist conditions of the cloud forest in which they now live. With paleoecological evidence to show that elevations as low as 1000 m above sea level (asl) had saturated air, it is not possible to discount evidence of cooling on the basis of steepened lapse rates.

A further criticism of using pollen to describe annual average paleotemperature is that the range of plants (therefore the elevation at which they grow) is determined not by mean annual temperatures, but by absolute minima. The distribution of plants is determined by the coldest night they survive rather than mean temperature. One way to test whether observing minimum temperatures rather than mean temperatures would yield more information on species ranges is to plot both mean and minima data against elevation, and fit a regression line for each set of values. If the regression lines have a similar slope, then it is legitimate to use plant ranges to derive mean temperature values. Of course, the minima are likely to be more ecologically revealing in terms of determining the cause of the distribution, but that is a separate issue.

Detailed long-term climatic data on temperature maxima and minima are scarce for the neotropics, but a data set that provides a transect of daily minimum temperatures from Manaus, across lowland Ecuador, to the crest of the Andes is shown in Fig. 1. This data set is far from perfect, and some records were kept for only a few years. A relatively short run of data will not affect the mean temperature values, but may underestimate occasional bouts of extreme cooling. However, the lowland records did include an episode of friagem cooling, and it is unlikely that much lower temperatures would be experienced under modern conditions. As a first approximation, this data set clearly makes the point that tropical temperature minima are generally closely correlated to mean temperatures.

Climatic requirements of Araucaria—e.g., mean winter temperature, number of days of frost, and length of dry season—are used by Ledru (1991, 1992, 1993) and De Oliveira (1992) to infer paleoclimates associated with a Pleistocene range expansion of this genus. If it is assumed that Araucaria distributions are bound by these variables, a comparison of climatic data from the modern range with that of the Pleistocene range pro-

FIGURE 1 Modern mean annual and minimum temperatures for Ecuadorian weather stations (Centro Ecuadoriano de Investigación Geográfica [CEDIG] 1983) plotted against elevation. The regression line through the mean temperature data represents a 5°C /1000 m ascent, representing moist air adiabatic lapse rate. The line through the minima data represents a best-fit regression line. Note how the two lines for minimum and mean temperatures are virtually parallel.

FIGURE 1 Modern mean annual and minimum temperatures for Ecuadorian weather stations (Centro Ecuadoriano de Investigación Geográfica [CEDIG] 1983) plotted against elevation. The regression line through the mean temperature data represents a 5°C /1000 m ascent, representing moist air adiabatic lapse rate. The line through the minima data represents a best-fit regression line. Note how the two lines for minimum and mean temperatures are virtually parallel.

vides estimates of changes in temperature and precipitation. This technique is freed from assumptions about lapse rates and, therefore, provides a valuable alternate means to measure paleotemperature. The above-cited authors used the movement or expansion of Araucaria forest from southern into southeastern Brazil (20°-25°S) to infer past-climate change. Behling and Lichte (1997) adopted a similar technique as they documented the movement, or expansion, of subtropical grassland from southern into southeastern Brazil. They found Pleistocene assemblages rich in subtropical grassland species approximately 7° of latitude farther north than their present range. Basing their climatic inference on modern weather data for the two areas, they infer an ice age cooling of between 4° and 8°C.

Another way to assess temperature using whole community values rather than indicator species has formed the basis of conventional transfer functions (e.g., Imbrie and Kipp, 1971; CLIMAP Project Members, 1976; Bonnefille et al., 1990). A number of problems are inherent in this approach, such as the lack of modern analogs for past assemblages and an inherent tendency toward underestimating any change. Many pollen taxa within an assemblage provide no detailed climatic information and can be regarded as catholic. If a full range of analog sites existed, the diluting effect of many catholic species would not matter, but without a full array of analogs, the presence of catholic species inevitably moderates the signal of climate change. The solution is to exclude the catholic species from the analysis and use a selection of stenothermic species. This compromise between using single indicator species and whole communities can be used to estimate response surfaces for precipitation or temperature. This technique could provide a paleothermometer that is independent of lapse rates.

It has been suggested that biome boundaries could be used to model past-climate change. The strength of the biome approach is that it is independent of lapse rates, it does not rely on modern communities being exactly those of the past (though intermediate vegetation types between recognizable modern biome types are a problem), and it should reduce subjectivity in interpretation. However, this technique also has problems that are particularly severe in the tropical lowlands (Marchant et al., in press).

Biome models assume that there will always be a biome to replace the existing one, but in the case of the lowland tropics there is none. Applying such models to lowland tropical paleoecology brings into focus a philosophical problem inherent in the concept of the biome—that modern conditions are normal. But, they are not. Glacial age conditions were the norm of the last 2 million years, and modern times are oppressively hot.

At 0° latitude and at 50 m elevation in the middle of the vast Amazonian plain, there is nowhere to retreat when it gets warmer. Some of the most stenothermic species that flanked the Andes escaped upslope to cooler climates at the beginning of the Holocene and will stay there until normal conditions return. The majority of species stay where they are because there are no hotter adapted species to displace them. Thus, the lowland tropics are unique—they really cannot show a warmer than usual signal (remember glacial conditions are the norm), other than an upslope migration of a few species. Biome models may be more appropriate in other settings, but they will fail in the lowland tropics because the lowland tropic biome is an endpoint in the biome continuum.

A second problem with taking the results of biome models at face value is more mechanical. Because the models treat a biome as a uniform climatic mass, the only changes indicated are when one biome replaces another. In other words, two regions that occur within the same biome—say, Atlanta and New York, which both occur in a temperate forest biome—would be accorded the same climate. It is clear that there could be substantial climatic change and yet no change in biome. Where biomes do change, relatively massive changes in climate are inferred. Neighboring areas that experience similar climatic change, but are judged to be biome constant, are suggested to have had a constant climate. Not all climate effects are geographically gradual, but we suggest the biome is too coarse a descriptive unit to elucidate paleoclimatic change in the tropics.

The only possible biome change that could be registered in Amazonia would be a transition from forest to savanna. Clearly, it is unsatisfactory to reduce all possible climatic variants to a simple "either savanna or forest." Under this kind of biome construction, vast areas will show no climatic change, and within the constructs of their model they are precisely correct. During the last glacial period, savanna did not replace large areas of forest, nor did lowland forests give way to Paramo grasslands or even to montane forest. Given the observed vegetation changes documented in the Amazon basin, over the greatest portion of the area there were no changes in biome; but this does not mean that there was not a significant change in temperature or precipitation.

0 0

Post a comment