地 学 雑 誌 Joumal of Geography 103(7)867-885 1994
Snowline Altitude and Climate in the Peruvian
Andes (5-17•‹ S) at Present and during the
Latest Pleistocene Glacial Maximum
Andrew N. FOX * and Arthur L. BLOOM *
Abstract
The present snowline in the Peruvian Andes (5-17°S), rises from as low as 4.7•}0.1km on the
eastern (windward) to more than 5.3 •} 0.1km on the western (leeward) side of the central Andes.
The effect of temperature on snowline altitude is isolated from the effect of precipition by
subtracting the altitude of the mean annual 0•Ž isotherm from the altitude of the snowline.
This difference, defined as the normalized snowline altitude, increases with decreasing pre-
cipitation.
The lowest late Pleistocene snowline rose from east to west and ranged in altitude from 3.2 to
4.9 (•}0.1) km. Both the present and lowest late Pleistocene snowlines indicate that moisture at
both times was derived principally from tropical easterly winds. An east-west precipitation
gradient steeper than present is inferred for the eastern slopes of the centralAndes from the
steeper late Pleistocene snowline gradient. Mean annual temperatures were 10•}1.9•Ž cooler
that today at 3.52 km, as calculated from a late Pleistocene snowline as much as 1.4•}0.2 km
lower than today. Mean annual precipitation was 25 to 50% less than today along the eastern
side, and more than 75% less on the western side of the central Andes. These estimates of
lower temperature and decreased precipitation are more extreme than previous estimates. They
imply that the amount of glacial-age cooling elsewhere, such as in western North America,
may also have been underestimated by previous researchers because they did not adequately
consider the effect of reduced ice-age precipitation on snowline lowering.
America (Fig. 1). The single region chosen I. Introduction for discussion here is the portion of Peruvian
This paper is a review of portions of the Andes between latitude 5 •‹S, and about 17 •‹S,
Ph. D. dissertation of the first author (Fox, and westward of 69.5 •‹W longitude (Fig. 1).
1993), summarized and edited by the second The limits of the region were defined by the author, to illustrate some new techniques for availability of good quality topographic maps estimating and interpreting present and former at a scale of 1 : 100,000 with a contour interval snowlirie altitude in mountain regions. Fox of 50 m, compiled from stereographic aerial
(1993) analyzed snowline altitudes in three photographs. The methods illustrated here regions of the Andes Cordillera of South are applicable to any other mountainous region
* Department of Geological Sciences , Cornell University. 2122 Snee Hall, Ithaca, NY 14853, USA.
867 Fig. 1 Map of the central Andes showing the location of the three regions discussed in Fox (1993) Area A, the Peruvian Andes, is discussed here. The thin solid line is the 3 km topographic contour for geographic reference in subsequent figures. The thin dashed lines are international borders. with maps of at least comparable quality, Thematic Mapper images were extensively climatic records, and aerial photographs or used to compile the extent of modern snowfields other documentation of snowlines, both past and to estimate the extent of snow cover during and present. In the original study, Landsat the latest Pleistocene glacial maximum. The
868 age of the latest glacial maximum in the production and is shown by blue contour lines
Peruvian Andes is not known, but was probably with a contour interval of 50 m. about 18,000 to 20,000 years ago. Snow cover is unevenly distributed along
Many authors have discussed the probable the axis of the central Peruvian Andes and is effect of variable precipitation on estimates of notably concentrated on mountain peaks higher snowline lowering and cooling during the last than 5 km elevation from 8.5 •‹S to 13.5 •‹S glacial maximum. In particular, if the glacial latitude, and along the eastern and western climate was generally drier than at present, margins of the Peruvian Altiplano (Fig. 2). the snowline on mountains would not have Clapperton (1991) gave a good general review lowered as much as it would have if precipita- of the tectonic and volcanic controls on the tion had remained the same and only a colder spatial distribution of high topography and temperature had determined the amount of glaciation in the Andes. snowline lowering. Conversely, estimates of Polygons were digitized around the perimeter tempera ture decrease during the latest glacial of all snow-covered areas shown on the topo- maximum that are based on an assumption of graphic maps using the ESRI ARC/INFO ge- no change in precipitation will be systematically ographic information system (GIS). Total snow
too small. This creates special problems for cover in the Peruvian Andes, calculated by interpreing temperature change in the tropics summing the area of the snow cover polygons,
during the latest glacial maximum. Many anal- is approximately 3,700 km2.
yses of tropical mountains predict snowline Point elevations were recorded where the
lowering on the order of 1,000 m during the perimeter of each snow-covered region crosses
latest glacial maximum ; based on a moist adia- a topographic contour line to determine the batic lapse rate of 6 •Ž/1,000 m, tropical moun- local or orographic snowline altitude. Points
tain cooling of at least 6 •Ž is commonly infer- were digitized at horizontal intervals of approx-
red. If however, the climate was drier and imately 1 km to record local maxima and
the snowline was therefore lowered less, the minima details of the orographic snowline. A
cooling would have been even greater. The total of 5,730 points were digitized, with eleva-
implicat:.ons of this effect are discussed in the tions ranging from 4,050 m to 6,000 m.
final section of this paper. The position of the snowline determined
from the topographic maps is assumed to be II. Present Snowline close to its perennial position. Temporal dif-
Present perennial snow cover and the snow- ferences in the snowline altitude are generally line altitude in the Peruvian Andes was deter- small in tropical mountains because snowline mined from the snow cover shown on the is primarily controlled by the altitude of the topographic map series published at a scale of 0 •Ž isotherm, which has a smaller seasonal and 1 : 100,000 by the Instituto Geografico Militar, interannual range in the tropics than in moun- Lima, Peru. The maps were produced from tains at higher latitudes (Kuhn, 1981). stereographic aerial photographs that were The regional snowline altitude was determined
taken between 1955 and 1963. Snow cover was by digitally contouring the orographic snowline
photogrammetrically transferred from the aerial point observations (Fig. 2). Contours show the
photographs to the topographic maps during calculated mean altitude of the point measure-
869 Fig. 2 Extent of present snow and ice digitized from the Peruvian 1 : 100, 000 scale topographic map series (Fox, 1993, Appendix A) Contours show the altitude of the present regional snowline in meters. Polygons located east of the available topographic map coverage represent the additional extent of snow cover observed on 1 : 250, 000 scale Landsat MSS images.
ments within contiguous 20•~20 km regions. melting of snow on west-facing slopes (Has-
The mean, rather than the maximum or mini- tenrath, 1967), Contours were interpolated bet- mum altitude, was selected because it removes ween snow-covered areas using the altitude of the bias introduced by preferential melting snow-free peaks and Seltzer's (1987) glaciation or preservation of snow due solely to local threshold estimates as controls. The error in slope aspect. Individual snowline observations determining the altitude of the snowline is est- generally vary from the mean by only 100 m imated to be •} 100 m, based on the local vari- or less, and are slightly higher on east- ability of the orographic snowline altitude ob- northeast than on west-southwest slopes beca- served on the topographic maps. use clouds, more common in the afternoon than Snowline contours generally parallel topog- in the morning, block insulation and reduce raphic contours, but rise from 4,700 m in the
870 northern and eastern Peruvian Andes to more the previous reports estimated the snowline at than 5,300 m in the south and west (Fig. 2). elevations greater than 5,500 m. Landsat Mul-
Snowlin e gradients as steep as 11 m/km occur tispectral Scanner images acquired between perpendicular to the axis of the mountain belt October 1972 and May 1982 and Landsat TM on the east side of the Altiplano at 14 °S lati- images acquired between July 1984 and Septem-
tude where regional topographic and precipi- ber 1986 support the lower snowline altitudes
tation gradients are also steep, but an east- shown in Fig. 2.
west snowline gradient of 5 m/km is more III. Climatic Controls on the Present representative of the northern region. Snowline Altitude The snowline contours shown in Fig. 2 are
similar in trend and magnitude to previous Many previous studies have sought to identify
estimates of the present snowline altitude that and understand the factors controlling the
were calculated from far fewer orographic present snowline altitude so that paleoclimatic
snowline measurements. On the basis of inferences can be made from measurements of
of about 30 observations, Nogami (1976) showed lower snowlines in the past (Charlesworth,
the present snowline increasing from 4,800 m 1957 ; Ostrem, 1966, 1972 ; Bradley, 1975 ; Ch-
in the northern Peruvian Andes to 6,000 m in inn, 1975 ; Miller et al., 1975 ; Porter, 1975,
the southwest. Seltzer (1987) contoured single 1977 ; Wright, 1983 ; Seltzer, 1987 ; Broecker
values of the modern glaciation threshold that and Denton, 1989 ; Fox, 1991). Temperature
he determined for 57 of the 160 topographic and precipitation have been identified for a
maps utilized in this study and showed snowline long time as the primary climatic controls on
contours increasing from 4,700 m in the north snowline altitude. Topography influences the
to 5,200 m in the southwest, including a local snowline in alpine regions by its indirect con-
maximum of 5,200 m in the vicinity of the trol on temperature as a function of altitude and
Quelccaya Ice Cap (14 •‹S, 70 •‹W). Glaciolo- on precipitation as a function of the orographic
gical investigations by Thompson (1980) show- lifting and blocking of moist air masses.
ed that the annual snowline on the Quelccaya Snowlines are influenced by other climatic
Ice Cap is at 5,250 m, which compares well variables, including cloudiness (Porter, 1975 ;
with the value of 5,300 m shown in Fig. 2. A Wright, 1983 ; Jordan, 1991), shortwave
meridianal transect along the eastern margin irradiance and sublimation (Johnson, 1976 ;
of the Peruvian Andes by Satoh (1979) showed Paterson, 1981), the proportion of rain to snow
the present snowline increasing from a (Miller et al., 1975), and the effect of alloch- minimum altitude of 4,700 m in the north to thonous snow accumulation due to wind drifting 4,900 m in the south. Meridianal transects by (Charlesworth, 1957). These variables are Hastenrath (1967, 1971 b) showed comparable generally unquantified in the Peruvian Andes,
increases in the present snowline altitude from as in many other remote alpine regions of the
both north to south and east to west. world, but their combined effects may be
The largest differences between all of the only minor and are probably related to the
previous estimates of snowline altitude and the principal controls discussed below.
snowline contours shown in Fig. 2 are on the 1) Temperature
western side of the Peruvian Altiplano where Air temperature decreases non-linearly with
871 Fig. 3 Relationship between elevation and mean annual, summer, and winter temperature for climate stations in the Peruvian Andes (Fox, 1993, Appendix B) increasing altitude in the Peruvian Andes. Sto- close to 6.5 •Ž/km is often used to estimtea
ne and Carlson (1979) found that temperature the decrease in temperature with increasing lapse rates in the lower tropical troposphere altitude, such a linear fit to the data would are within 20 % of the non-linear, moist adia- overestimate temperatures both at sea level and batic lapse rate at altitudes above the planetary at the elevation of the 0 •Ž isotherm. Second- boundary layer (•`900 mb). Therefore, a moist order polynomial regression lines fit the temper-
adiabatic temperature lapse rate describes the ature observations better than linear regression
observed decrease in temperature with increa- lines. The resulting equations, when differen-
sing altitude in the tropical Andes better than tiated, closely approximate the non-linear,
an assumed dry adiabatic lapse rate (linear) of moist adiabatic temperature lapse rates which
6.5 •Ž/km. Vertical transfer of heat in the predominate at tropical latitudes (Rennick,
tropics is dominated by moist convection in 1977). They are : y= 24.60-2.04 x-0.54 x2 (R2=--
all seasons and there is little seasonal change 0.94) Summer (Dec-Feb), y= 24.29-2.16 x-0.65
in mean daily temperature. Since the moist x2 (R2 0 . 94) Winter (Jun-Aug), y = 24.51-1.93
adiabatic temperature lapse rate is dependent x-0.62 x2 (R2-= 0.94) Annual.
only upon land-surface temperature, it has very The annual and seasonal temperature lapse
little seasonal change in the tropics. rates in northern (5-12 •‹S) and southern Peru
Figure 3 shows the relationships between the (12-18 •‹S) were calculated separately and do
elevation and mean summer, winter, and annu- not differ significantly from the equations
al temperature at 167 climate stations in Peru shown above. Others have also noted that the
(Fox, 1993, Appendix B). Although an assum- moist adiabatic temperature lapse rate does not
ed or calculated linear temperature lapse rate vary significantly with latitude in Peru (Has-
872 tenrath, 1971 a; Johnson, 1976) These obser- altitude of the snowline in the Peruvian Andes vations all confirm interpretations of radiosonde (about 4,700 m) is coincident with the altitude data by Hastenrath (1967) and Satoh (1979), and of the ablation season (winter) freezing level show that there is no latitudinal variation in in the humid northern and eastern Peruvian the altitude of the annual 0 •Ž isotherm. How- Andes. The snowline rises westward from ever, the moist adiabatic temperature lapse this level toward drier regions. In fact, the rate does vary slightly throughout the year so modern snowline is located at or above the that the mean altitude of the 0 •Ž isotherm ablation season freezing level everywhere in
ranges between 4,675 m in whiter and 5,120 m the tropics (Hastenrath, 1985 ; Allison and Pe- in summer (Fig. 3). The altitude of the annu- terson, 1989 ; Young and Hastenrath, 1991).
al 0 •Ž isotherm is 4,920 m, and is similar to 2) Precipitation previous estimates of 4,950 m (Johnson, 1976) Moisture is transported to the Peruvian An-
and 5,078 m (Seltzer, 1987) that were based on des primarily by humid easterly winds from
fewer climate station records, and estimates of Atlantic Ocean and Amazon Basin sources. Asa
4,800 to 5,000 m (Hastenrath, 1967) and 4,783 m result, the eastern side of the cordillera receives
(Satoh, 1979) that were based on radiosonde more annual precipitation than the western
data. The winter temperature lapse rate is side (Fig. 4). More than 1,000 mm/yr of rain
steeper than the summer rate because the moist falls everywhere along the eastern margin of
adiabat is inversely related to land-surface the Andes below an elevation of about 3,000 m,
temperature (Stone and Carlson, 1979 ; Eagle- decreasing westward to less than 100 mm/yr
man, 1985). As a result, higher elevations ex- along the Pacific coast. Local precipitation
perience a larger annual range of temperature reaches maxima of 4,000 mm/yr and 6,400 mm/
(Fig. 3). The altitude of the annual 0 •Ž iso- yr at about 8 °S and 13 •‹S respectively. Pre-
therm interpolated from available climate sta- cipitation is produced and concentrated along
tion and radiosonde temperature data is believed the eastern side of the Andes as moist air
to be within 100 m of its true position. is orographically lifted over the mountains,
Despite an annual range of mean daily tem- with consequent adiabatic cooling and conden-
peratures of less than 5 •Ž everywhere in Peru, sation. Local slope aspect is also important,
the diurnal temperature range at high altitudes with windward slopes typically receiving more
can exceed 20 •Ž and still maintain an average precipitation than leeward slopes (Johnson,
daily temperature at or below freezing. Al- 1976). High regional topographic gradients
though melting may occur daily at the snow- are responsible for high regional precipitation
line, winter is the ablation season in the Peru- gradients. The east-to-west precipitation gra-
vian Andes because virtually no snow accumu- dient is steep along the eastern side of the
lates at this time, and the mass balance of Peruvian Andes and has a horizontal gradient
snow and ice is negative. of about -15 mm/km on average and a maxi-
Previous studies have identified the altitude mum gradient of -145 mm/km at 14 •‹S lati-
of the 0 °C isotherm as the primary control tude. In contrast, the horizontal precipitation
on the altitude of the snowline in the tropical gradient is only -3 mm/km on the central Pe-
Andes (Hastenrath, 1967 ; Satoh, 1979 ; Kuhn, ruvian Altiplano to the southwest of the maxi-
1981, 1989 ; Seltzer, 1987). The lowest mum precipitation gradient.
- 873 Fig. 4 Mean annual precipitation (mm) digitized from Hoffman (1975) Squares show the location of all known climate stations. Shaded region is the land surface above 3 km.
Because the annual 0 •Ž isotherm varies with snowline altitude include a measure of the latitude, a direct comparison of the relationship degree of continentality (Kuhn, 1981) and the between precipitation and snowline altitude vertical height range of the periglacial zone at different latitudes in the Andes is not possi- (Humlum, 1988). ble. The snowline must first be normalized to The normalized snowline altitude has a sys- the altitude of the annual 0 •Ž isotherm to tematic relationship with mean annual precipi- remove the temperature influence and isolate tation for 21 snow-covered regions in the trop- the influence of precipitation. The term "nor- ical Andes between 5 •‹ and 20 •‹S latitude malized snowline altitude" has been introduced (Fig. 5). The driest portion of the northern
(Fox, 1993, p. 9) to describe the elevation of Chilean and Bolivian Andes, as well as the the snowline above or below the annual 0 •Ž Peruvian Andes, are included in Fig. 5 to in- isotherm, and is primarily a function of pre- vestigate the influence of the present range of cipitation. Quantities similar to the normalized precipitation extremes in the tropical Andes on
874 Fig. 5 Relationship between the present mean annual precipitation and normalized snowline altitude for 21 perennially snow-covered locations in the Andes between 5° and 20•‹S
Numbered boxes on the right refer to locations on the left half of the figure. Bars on the
plot symbols show the range of snowline elevations and precipitation measured in each region. the present normalized snowline altitude, and best quantified with a logarithmic equation of to identify possible modern analogs for paleo- the form, N= b-a (log P), where the normalized climates inferred from the normalized snow- snowline altitude (N) decreases with increas- line altitude during the late Pleistocene glacial ing mean annual precipitation (P). Similar maximum. The normalized snowline altitude non-linear relationships between snow accumu- in the Peruvian Andes was calculated by sub- lation and glacier equilibrium line altitudes tracting the altitude of the annual 0 °C iso- (ELA's) have been reported by some workers therm (4,920 m) from the snowline altitude (Fig. (LiestƒÓl, 1967 ; Kasser, 1973, 1977) and linear
2) and in. Bolivia by subtracting the altitude approximations by others (Porter, 1977). The of the annual 0 •Ž isotherm (4,800 m: Satoh, logarithmic relationship between the present
1979) from the observations of the snowline normalized snowline altitude and present mean altitude compiled by Nogami (1976). The esti- annual precipitation provides the tool for using mated error in the normalized snowline altitude a past normalized snowline altitude to infer is •}200 m because it is calculated from the paleoprecipitation. altitudes of the snowline and the annual 0 °C IV. Lowest Late Pleistocene Snowline isotherm, both of which have errors of ± 100 m.
The mathermatical relationship between the The few available radiocarbon dates of gla-
present normalized snowline altitude and mean cial deposits place the last glacial maximum in
annual precipitation is non-linea (Fig. 5) and is the Peruvian Andes between 24 and 12 ka, with
875 an earlier maximum before 43 ka (Seltzer, locations and altitudes were digitized and com-
1990). In addition, there is indirect evidence piled from the topographic map series at a scale for a penultimate glaciation in Peru assumed of 1 : 100,000 (Fig. 6). Cirque floor altitudes to date between 130 and 170 ka (Clapperton, range in elevation from 3,200 m to 4,850 m,
1983 ; Wright, 1983). Minor late-glacial ad- with basins less confidently identified as cirques vances occurred between 12 and 10 ka, but there as low as 3,150 m between 6 •‹ and 6.5 •‹ south is little evidence for Holocene glacier activity latitude. These observations confirm an earlier except for the minor fluctuations associated report of cirque floor altitudes lower than 3,300 with the Neoglacial interval of the last 5,000 m in the northern Peruvian Andes (Seltzer, years. Despite the lack of detailed chronologic 1987). The lowest cirques, like the lowest control, the temperature and precipitation con- present snowline, are located in the northern ditions associated with the most extensive late and eastern Peruvian Andes.
Pleistocene glaciation, referred to here as the The altitude of the lowest late Pleistocene latest Pleistocene glacial maximum, can be snowline was reconstructed by contouring the inferred from the lowest cirques even though lowest cirque floor altitudes recorded in each their exact age is unknown. 20 X 20 km region (Fig. 6). The lowest cirques
Cirques identified in this study clearly range were used because they formed under glacial in age, commonly forming chains up a single climatic conditions most different from the pre- valley where the highest ones may have been sent interglacial climate. The snowline alti- occupied by ice as recently as the end of the tude was calculated at a spatial resolution of last century (Seltzer, 1987). The question of 20 km in order to take advantage of the large cirque age is not a critical factor in this study number of cirque floor observations, although because the primary objective is to estimate the snowline calculated at spatial resolutions the maximum change in climate recorded by lower than 20 km is not significantly different
the lowest cirques in each region, which are than that shown in Fig. 6. inferred to be contemporaneous. Although Late Pleistocene glacial maximum snowline the absolute age of the lowest cirques is un- contours generally parallel topographic con- known, Satoh (1979) made an argument for the tours and present snowline contours, but rise contemporaneity of the lowest cirques along a from east to west with a horizontal gradient single mountain range and even on neighbor- steeper than at present. The horizontal snow- ing ranges because they occur at a remarkably line gradient reached a local maximum of uniform altitude (•}100 m). The contour line 46 m/km a 10 •‹ S during the late Pleistocene marking the position of the abrupt change in glacial maximum (Fig. 6), but a horizontal stream gradient, generally also the contour line gradient of 20 m/km was more representative immediately below the tarn, defines the outer along the eastern side of the Peruvian Andes limit of the cirque and was used to determine where the present horizontal snowline gradient the altitude of the cirque floor. An error of •} is less than 11 m/km.
100 m is assigned to the regional snowline alti- The altitude of the lowest late Pleistocene tude based on the local variability of cirque snowline presented in Fig. 6 supports earlier floor altitudes. descriptions of the late Pleistocene snowline in
A total of 11,108 point observations of cirque Peru. Hastenrath (1967, 1971 b) showed the
876 Fig. 6 Location of 11, 108 cirque floor altitudes identified on the Peruvian 1 : 100, 000 scale topographic map series (Fox, 1993, Appendix D) Polygons located east of the available topographic map coverage show the limit of glacial landforms observed on 1 : 250, 000 scale Landsat MSS images. Contours show the inferred altitude (m) of the lowest late Pleistocene snowline.
Pleistocene snowline rising from an altitude of sects across the Peruvian cordillera. Satoh about 3,800 m in the north to 5,000 m in the (1979) showed the Pleistocene snowline increas- south along a snowline transect down the west- ing in altitude from a minimum of about 3,400 ern side of the Peruvian Andes. Hastenrath m in the northern Peruvian Andes to about (1971 b) also showed the late Pleistocene glacial 3,500 m in the south along the eastern side of maximum snowline rising from about 4,200 m the cordillera. His snowline is higher along to 4,500 m along two northeast-southwest tran- the western side of the Peruvian cordillera,
877 reaching a maximum altitude of 4,800 m in the southwest. Snowline contours maps by Seltzer (1987) and Nogami (1976) are also supported by the much greater number of points plotted in Fig. 6. Both show a rise in the Pleistocene snowline from about 3,200 m in the north to about 4,800 m in the southwest.
V. Snowline Difference and Paleoclimatic Implications
The altitudinal difference between the pres-
ent and a former snowline, sometimes referred
to as snowline depression, is a standard basis
for estimating the glacial-age climate in moun-
tainous regions (Flint, 1971). The lower alti-
tude of the late Pleistocene glacial maximum
snowline compared with the present snowline Fig. 7 Difference in altitude (m) between the clearly indicates that temperatures were lower present and lowest late Pleistocene
in the Peruvian Andes. The lack of a precipi- snowline Values in parentheses are not associated tation lapse rate in most mountainous regions, with contour lines. however, has dissuaded many workers from
making a quantitative estimate of paleoprecipi- cordillera and on most of the Peruvian Altip-
tation from snowline difference. As a result, lano. The error of •}200 m assigned to the
the entire observed snowline difference has too difference in snowline altitude is based on the
frequently been ascribed to paleotemperature combined estimated •} 100 m errors in deter-
without considering the effects of paleopre- mining the latitudes of both the present and
cipitation. The present study eliminates these lowest late Pleistocene snowlines.
traditional problems by avoiding the use of an As is the case everywhere in the tropics to-
assumed temperature lapse rate and by using day, the lowest altitude of the present snowline
the normalized snowline altitude instead of in the Peruvian Andes is close to the annual
the absolute snowline altitude. Use of the 0 •Ž isotherm, but never below the winter 0 •Ž
normalized snowline altitude isolates the isotherm (4,700 m), even where precipitation is
precipitation effect on the altitude of the high. Because the temperature lapse rate dur-
snowline from that of temperature. ing the late Pleistocene glacial maximum could
The difference in altitude between the pres- not have been significantly different from that
ent (Fig. 2) and the lowest late Pleistocene of today in the tropics (Rind and Peteet, 1985 ;
(Fig. 6) snowlines in the Peruvian Andes is Betts and Ridgway, 1992), it can be assumed variable (Fig. 7). A maximum snowline differ- that the snowline was not below the winter ence of 1,400•} 200 m is calculated for the east- 0 •Ž isotherm. Therefore, the winter 0 •Ž ern side of the Peruvian Andes, rapidly decreas- isotherm must have been lower than today ing to a minimum of 600 m in the western by an amount equal to or greater than the
878 Fig. 8 Normalized snowline altitude (m) during the late Pleistocene glacial maximum Values in parentheses are not associated with a contour line. maximum snowline difference of 1,400 m. be known or assumed, however, to calculate
An altitude of 3,520 m can be inferred for paleotemperature at altitudes above or below the annual 0 •Ž isotherm during the late Pleis- 3,520 m. An error of •} 1.9 •Ž is calculated tocene glacial maximum from an annual 0 •Ž for the estimate of paleotemperature from a isotherm 1,400 m lower than today (Fig. 3). total uncertainty of •}300 m in determining the
A mean annual temperature 10 •Ž cooler than altitude of the annual 0 •Ž isotherm during the today in the Peruvian Andes at an elevation late Pleistocene glacial maximum, including of 3,520 m can be calculated from the differ- uncertainties both in the snowline depression ence between the present (10 •Ž) and glacial- (•} 200 m) and in the present altitude of the age (0 •Ž) mean annual temperatures at that annual 0 •Ž isotherm (•} 100 m). altitude. This method of estimating paleotem- The normalized snowline altitude during the perature is preferred over the traditional meth- late Pleistocene glacial maximum in the Peru- od of multiplying snowline difference by an vian Andes can be used to make an estimate assumed temperature lapse rate because it does of paleoprecipitation. The present normalized not presume knowledge of the temperature snowline altitude is controlled only by precipi- lapse rate during the late Pleistocene glacial tation because the direct effect of temperature maximum. The temperature lapse rate must (annual 0 •Ž isotherm) on the snowline altitude
879 Fig. 9 Inferred mean annual precipitation (mm/ Fig. 10 Inferred difference (percent less) in yr) during the late Pleistocene glacial mean annual precipitation from present maximum in the Peruvian Andes during the late Pleistocene glacial maximum has been removed (Fig. 5). Assuming that the Values in parentheses are not associated relationship between the normalized snowline with a contour line. altitude and mean annual precipitation was the same during the late Pleistocene glacial maxi- cipitation (Fig. 5). Given the •} 200 m uncer- mum as it is now, an estimate of paleoprecipi- tainty in determining the normalized snowline tation can be made from the normalized snow- altitude, an error of •} 25 % is assigned to esti- line altitude. The normalized snowline altitude mates of paleoprecipitation. Mean annual pre- shown in Fig. 8 was calculated by subtracting cipitation ranged from more than 1,000•}250 the inferred maximum altitude of the glacial- mm along the eastern side of the northern and age annual 0 •Ž isotherm (3,520 m) from the central Peruvian Andes to less than 200•}50 snowline altitude shown in Fig. 6. The mm along the western side of the cordillera. normalized snowline altitude was lowest along The inferred high precipitation along the east- the eastern side of the Peruvian Andes and ern margin of the Peruvian Andes demonstrates rose to the west reaching a maximum of that the primary moisture source for late
1,200 m on the western margin of the Peruvian Pleistocene glaciers was from the east as it is
Altiplano. today. The east-to-west precipitation gradient,
An estimate of mean annual precipitation however, was steeper during the late Pleistocene during the late Pleistocene glacial maximum glacial maximum than at present, ranging from
(Fig. 9) can be calculated from the normalized about -15 mm/km to -50 mm/km compared snowline altitude (Fig. 8) and the inverted loga- with a modern representative horizontal rithmic relationship between the present nor- precipitation gradient of about -16 mm/km. malized snowline altitude and mean annual pre- The percent decrease in mean annual precip-
880 Table 1 Summary of climatic changes Inferred for the Peruvian Andes during late Pleistocene glaciations
ication from present amounts during the late literature reviews of pollen, lake level, and
Pleistocene glacial maximum is shown in Fig. glacial evidence by Markgraf (1989) and Seltzer
10. Precipitation may not have been significa- (1990) indicate that, in general, glacial cli-
ntly less than today along the eastern side of mates were colder and drier in the Peruvian
the northern and central Peruvian Andes dur- and neighboring tropical Andes during the
ing the late Pleistocene glacial maximum, in late Pleistocene than at present. Simulations
contrasi: to a much larger precipitation differ- of the climate during the last glacial maximum
ence to the south and west where it reached a by a number of general circulation models sup-
maximum reduction of about 80 % (or 20 % port this interpretation (Manabe and Hahn,
of the present mean annual precipitation) on 1977 ; Kutzbach and Guetter, 1986). However,
the Peruvian Altiplano. If the late Pleistocene the magnitudes of the temperature and precipi-
glacial maximum cooling was actually greater tation reductions presented here are much
than 10 •Ž at 3,520 m, then precipitation greater than earlier estimates (Table 1). Early
amounts (Fig. 9) would have been less and the estimates of termperature difference by Hasten-
percent difference in mean annual precipitation rath (1967, 1971 a) and Satoh (1979) from snow-
(Fig. 10) would have been greater. line difference were only qualitative. Wright
(1983) inferred a mean annual temperature 2 •Ž VI. Comparison with Previous lower than today during the late Pleistocene Paleoclimatic Estimates from an assumed modern temperature lapse
A number of independent estimates of ice- rate of 7 •Ž/km and a conservative estimate
age climates in the Peruvian Andes and else- of snowline difference (300 m) in the cordillera where during the late Pleistocene can be com- west of Laguna Junin at about 11 °S. Seltzer pared with the results of this study. Recent (1987) examined the present relationship be-
881 tween snowline altitude and climate to estimate at present (-100 to 400 m) supports an inter- a temperature reduction of 6.5 •‹ to 7.0 •Ž near pretation of more arid conditions (Figs. 5 and
Nevado Huaytapallana at about 12 •‹S in the 8). central Peruvian Andes. The maximum alti- The climate of the last glacial maximum tude of the annual 0 •Ž isotherm (3,520 m) in- has been simulated by several groups using gen- ferred in this study from the altitude of the eral circulation models (Table 1) and ice-age snowline during the late Pleistocene glacial boundary conditions reconstructed by CLIMAP maximum supports previous estimates in the project members (1976, 1981). Early models tropical Andes. A lowering of the annual 0 •Ž by Gates (1976) and Manabe and Hahn (1977) isotherm from its present position down to used CLIMAP project members (1976) bound- about 3,500 m in Colombia was inferred from ary conditions for July at 18 ka to estimate a
Pollen data by van der Hammen (1974) and by reduction of mean July temperature by about
La Fontaine (1988) in the tropical Andes using 5 •Ž and little or no change in mean July the NCAR-CCM simulation of the last glacial cloudiness and precipitation in the central Peru- maximum (Kutzbach and Guetter, 1986). vian Andes. The CLIMAP project members
Previous estimates of paleoprecipitation from (1981) reassessment of the sea surface tempera- snowline difference in the Peruvian Andes vary ture data for 18 ka resulted in warmer tempera.- significantly (Table 1). Hastenrath (1971 a) tures than in the 1976 reconstruction. Subse- and Wright (1983) reasoned that a larger snow- quent models by Manabe and Broccoli (1985), line difference along the east side of the Kutzbach and Guetter (1986), and Rind (1987),
Peruvian Andes than to the west would have using the warmer CLIMAP project members resulted from a southward shift of the equa- (1981) sea surface temperatures, estimated less torial zone of high precipitation associated with cooling (0 •‹ to 4 •Ž) than earlier model results, the ITCZ. The contemporaneous shift of the and seasonal decreases in precipitation of less mid-latitude westerly wind regime northward than 4 mm/day (<38 %) from present condi- during glacial intervals of the Pleistocene (Ha- tions in the Peruvian Andes (Table 1). The stenrath, 1971 a) would have effectively narro- decreases in temperature and precipitation sim- wed the subtropical arid zone from its present ulated by general circulation models for the width. However, Satoh (1979) and Seltzer last glacial maximum are, therefore, much less
(1987) argued from snowline observations that than the decreases inferred from snowline conditions in the Peruvian Andes were drier difference. than today during late Pleistocene glaciations Small changes in the climate of the last gla- and that snowline difference resulted primarily cial maximum simulated by general circulation from a decrease in temperature. General cir- models cannot be reconciled easily with the culation model simulations support their inter- much larger decreases in temperature and pre- pretation and do not show a southward shift in cipitation inferred in this study from snowline the position of the ITCZ during the last glacial difference at high altitudes. One way to recon- maximum (La Fontaine, 1988). The fact that cile a 10 •Ž cooling at high altitudes (> 3 km) normalized snowline altitudes were higher in with a 2° to 4 •Ž cooling at sea level is to the Peruvian Andes during the late Pleisto- invoke an unrealistic steepening of the ice-age cene glacial maximum (-100 to 1,200 m) than temperature lapse rate. Rind and Peteet (1985)
882 stated, however, that a large divergence of the Survey Professional Paper 1386-H. last glacial maximum temperature lapse rate Betts, A. K. and Ridgway, W. (1992) : Tropical boundary layer equilibrium in the last ice age. J. from the moist adiabatic value is unlikely be- Geophys. Res., 97-D 2, 2529-2534. cause moist convection is, and probably was, the Bradley, R. S. (1975) : Equilibrium-line altitudes, dominnnt process of vertical heat transport in mass balance, and July freezing-level heights in the Canadian High Arctic. J. Glaciol., 14, 267- the troposphere at low latitudes. However, 274. general circulation models do show a slightly Broecker, W. S. and Denton, G. H. (1989) : The steeper temperature lapse rate, greater cooling role of ocean-atmosphere reorganizations in glacial cycles. Geochim. Cosmochim. Acta, 53, 2465-2501. at higher altitudes than at lower altitudes, and Charlesworth, J. K. (1957) : The Quaternary Era, an increased annual temperature range at 18 ka with special reference to its glaciation. Edward Arnold, London, 2 v., 1700 p. (Rind and Peteet, 1985 La Fontaine, 1988). Chinn, T. J. (1975) : Late Quaternary snowlines The discrepancies between inferred cooling and cirque moraines within the Waimakariri wa- and the effects of ice-age precipitation decrease tershed. MSc Thesis, Univ. of Canterbury, Christ- church, New Zealand. are not restricted to the South American An- Clapperton, C. M. (1983) : The glaciation of the des. Porter et al. (1983, p. 78, 82) noted that Andes. Quat. Sci. Rev., 2, 83-155. late Wisconsin cooling in the Brooks Range Clapperton, C. M. (1991) : Influence of tectonics on the extent of Quaternary glaciation in the Andes. and the Alaska Range has been underestimated Bolibian IG-ISP, Publication Especial, 8, 89-108. because of decreased precipitation. Similar CLIMAP project members (1976) : The surface of conclusions were reached for the western cor- the Ice-Age Earth. Science, 191, 1131-1137. dillera of the United States (Porter, et al., 1983, CLIMAP project members (1981) : Seasonal recon- struction of the Earth's surface at the last glacial p. 103,, 107). General application of the con- maximum. Geological Society of America Map and cept of a normalized snowline altitude, as presen- Chart Series MC-36. ted in this paper, is likely to increase most Eagleman, J. R. (1985) : Meteorology : The atmo- sphere in action. 2 nd ed. Wadsworth Publishing previous estimates of mountain climatic cool- Company, 394p. ing during the latest glacial maximum. The Flint, R. F. (1971) : Glacial and Quaternary effect will be to increase the contradictions geology. Wiley and Sons, New York, 392 p. Fox, A. N. (1991) : A quantitative model for describ- between CLIMAP project members (1981) esti- ing snowlines in the central Andes mountains. Bo- mates of sea-surface cooling and most estimates livian IG-USP, Publicacion Especial, 8, 75-88. Fox, A. N. (1993) : Snowline altitude and climate of cooling in mountain regions. Especially in the central Andes (5-28 ° S) at present and dur- in the Andes, the cooling estimated in this ing the late Pleistocene glacial maximum. Ph. D. paper (Table 1) is much greater than that in- Dissertation. Cornell University, Ithaca, NY, 504 p. Gates, W. L. (1976) : Modeling the ice-age climate. ferred for the coast of Peru by the CLIMAP Science, 191, 1138-1144. study. It appears that climatologists must un- Hansen, B. C. S., Wright, H. E. and Bradbury, J. P. dertake a major review of the CLIMAP report, (1984) : Pollen studies in the Junin area, central Peruvian Andes. Geol. Soc. Am. Bull., 95, 1454- which has served as such an important refer- 1465. ence for almost two decades. Hastenrath, S. L. (1967) : Observations on the snow line in the Peruvian Andes. J. Glaciol., 6, 541- References 550. Hastenrath, S. L. (1971a) : On the Pleistocene Allison, I. and Peterson, J. A. (1989) : Glaciers of snow-line depression in the arid regions of the Irian Jaya, Indonesia. In Williams, R. S. Jr. and South American Andes. J. Glaciol., 10, 255-267. Ferrigno, J. G. eds. : Glaciers of Irian Jaya, Hastenrath, S. L. (1971 b) : On snow line depression Indonesia, and New Zealand. U. S. Geological and atmospheric circulation in the tropical Amer-
883 icas during the Pleistocene. South African Geogr. 18, 289-310. J., 53, 53-69. Miller, G. H., Bradley, R. S. and Andrews, J. T. Hastenrath, S. L. (1985) : A review of Pleistocene (1975) : The glaciation level and lowest equilibr- to Holocene glacier variation in the tropics. Z. Gl- ium line altitude in the High Canadian Arctic : estcherkunde and Glazialgeologie, 21, 183-194. Maps and climatic interpretation. Arctic and Al- Hoffmann, J. A. J. (1975) : Climatic atlas of South pine Res., 7, 155-168. America. World Meteorological Organization, Nogami, M. (1976) : Altitude of the modern snow- Geneva, Swizerland. line and Pleistocene snowline in the Andes. Tokyo Humlum, O. (1988) : Rock glacier appearance level Metropol. Univ. Geogr. Rept., 11, 71-86. and rock glacier initiation line altitude : A meth- Ostrem, G. (1966) : The height of the glaciation odological approach to the study of rock glaciers. limit in southern British Columbia and Alberta. Arctic and Alpine Res., 20, 160-178. Geografiska Ann., 48 A, 126-138. Johnson, A. M. (1976) : The climate of Peru, Bo- Ostrem, G. (1972) : Height of the glaciation level livia, and Ecuador (Ch. 4). In Schwerdtfeger, W . in northern British Columbia and south eastern ed.: Climates of Central and South America. Alaska. Geografiska Ann., 54 A, 76-84. World survey of climatology, 12, 147-218. Paterson, W . S. B. (1981) : The physics of glaciers. Jordan, E. (1991) : Die Glestcher der bolivianischen 2 nd ed. Pergamon Press, Oxford, 380 p. Anden. Erdwissenschaftliche Forschung, Bd. 23, Porter, S. C. (1975) : Equilibrium-line altitudes of Steiner, Stuttgart, 365 p. Late Quaternary glaciers in the Southern Alps, Kasser, P. (1973) : Fluctuations of glaciers 1965- New Zealand. Quat. Res., 5, 27-47. 1970. IAHS-UNESCO, Paris, 357 p. Porter, S. C. (1977) : Present and past glaciation Kasser, P. (1977) : Fluctuations of glaciers 1970- threshold in the Cascade Range, Washington, 1975. IAHS-UNESCO, Paris, 269 p. U. S. A.: Topographic and climatic controls, and Kuhn, M. (1981) : Climate and glaciers : Internation- paleoclimatic implications. J. Glaciol., 18, 101- al Association of Hydrological Sciences, pub- 116. lication, 131, 3-20. Porter, S. C., Pierce, K. L. and Hamilton, T. D. Kuhn, M. (1989) : The response of the equilibrium (1983) : Late Wisconsin mountain glaciation in the line altitude to climate fluctuations : Theory and western United States. In Wright, H. E., Jr. ed. : observations. In Oerlemans, J. ed.: Glacier fluc- Late Quaternary environments of the United tuations and climatic change. Kluwer Academic States, v. 1. In Porter, S. C. ed.: The Late Publishers, 407-417. Pleistocene, 71-111. Kutzbach, J. E. and Guetter, P. J. (1986) : The influ- Rennick, M. A. (1977) : The parameterization of tro- ence of changing orbital parameters and surface pospheric lapse rates in terms of surface temper- boundary conditions on climate simulations for the ature. J. Atmospheric Sci., 34, 854-862. past 18,000 years. J. Atmospheric Sci., 43, 1726- Rind, D. (1987) : Components of the ice age circu- 1759. lation. J. Geophys. Res., 92-D 4, 4241-4281. La Fontaine, C. V. (1988) : Comparison of the sim- Rind, D. (1988) : Dependence of warm and cold ulated climate and geologic observations from climate depiction on climate model resolution. J. equatorial land regions for the past 18,000 years. Clim., 1-10, 965-997. MSc thesis, University of Wisconsin-Madison, 62 p. Rind, D. and Peteet, D. (1985) : Terrestrial condi- Liestol, O. (1967) : Storbreen glacier in Jotunhei- tions at the last glacial maximum and CLIMAP men, Norway. Norsk Polarinstitutt Skrifter, 141, sea-surface temperature estimates : Are they con- 63 p. sistent ? Quat. Res., 24, 1-22. Manabe, S. and Broccoli, A. J. (1985) : The influ- Satoh, H. (1979) : On the snow-line altitude in the ence of continental ice sheets on the climate of an central and southern Andes of the modern age and ice age. J. Geophys. Res., 90-D 1, 2167-2190. diluvian epoch. In Horie, S. ed.: Paleolimnology Manabe, S. and Hahn, D. G. (1977) : Simulation of Lake Biwa and the Japanese Pleistocene, 7. of the tropical climate of an ice age. J. Geophys. 387-415. Res., 82-27, 3889-3911. Seltzer, G. O. (1987) : Glacial history and climatic Markgraf, V. (1989) : Palaeoclimates in Central and change in the central Peruvian Andes. M. S. The- South America since 18,000 BP based on pollen and sis, University of Minnesota, Minneapolis, 92 p. lake-level records. Quat. Sci. Rev., 8, 1-24. Seltzer, G. O. (1990) : Recent glacial history and Meierding, T. C. (1982) : Late Pleistocene glacial paleoclimate of the Peruvian-Bolivian Andes. equilibrium-line altitudes in the Colorado Front Quat. Sci. Rev., 9, 137-152. Range : A comparison of methods. Quat. Res., Stone, P. H. and Carlson, J. H. (1979) : Atmospher-
884 ic lapse rate regimes and their parameterization. Peruvian Andes. Geografiska Ann., 65 A, 35-43. J. Atmospheric Sci., 36, 415-423. Young, J. A. T. and Hastenrath, S. (1991) : Glaciers Thompson, L. G. (1980) : Glaciological investiga- of Africa, In Williams, Jr., R. S. and Ferrigno, tions of the tropical Quelccaya Ice Cap, Peru. J. G. eds. : Glaciers of the Middle East and Af- J. Glaciol., 25-91, 69-84. rica. U. S. Geological Survey Professional Paper Van der Hammen, T. (1974) : The Pleistocene chan- 1386-G. ges of vegetation and climate in tropical South (Accepted 18 November, 1994) America. J. Biogeogr., 1, 3-26. Wright, H. E., Jr. (1983) : Late-Pleistocene glacia- Cornell University Department of Geological tion and climate around the Junin Plain, central Sciences Contribution No. 881.
885