Snowline Depression in the Tropics During the Last Glaciation Stephen C

Quaternary Science Reviews 20 (2001) 1067}1091

Snowline depression in the tropics during the Last Glaciation Stephen C. Porter* Department of Geological Sciences and Quaternary Research Center, University of Washington, Seattle, WA 98195-1360, USA

Abstract

Five primary methods have been used to reconstruct Pleistocene snowlines or equilibrium-line altitudes (ELAs) in the tropics (23.53N}23.53S) during the last glaciation, but each has inherent errors that limit the accuracy of the results. Additional potential errors in determining ELA depression involve estimates of modern snowline altitude, dating resolution, topographic reconstruction of former glaciers, orographic e!ects, the presence of rockfall debris on glaciers, and calculation of regional ELA gradients. Eustatic sea-level lowering during the last glaciation is an additional factor in#uencing estimates of ELA depression (!ELA). In cases where modern snowline lies above a mountain summit, only a minimum value for !ELA can be obtained. At 12 tropical sites in Africa, the Americas (to 103S latitude), and Paci"c islands, estimates of average !ELA range from 440 to 1400 m, but most fall in the range of 800}1000 m (mean $1""900$135 m). In a regional study of ELA depression in the southern tropical Andes (8}223S), an average !ELA of ca. 920$250 m has been reported. Based on the assumption that glacier mass balance was controlled solely by ablation-season temperature, and assuming a full-glacial temperature lapse rate of !63C/km, depression of mean annual temperature in glaciated alpine areas was ca. 5.4$0.83C. If adjusted for a sea-level fall of !120 m at the glacial maximum, this value is reduced to 4.7$0.83C. The "gure is based on the (unlikely) assumption that accumulation on alpine glaciers has been invariant; nevertheless, it is similar to values of temperature depresson (5}6.43C) for the last glaciation obtained from various terrestrial sites, but contrasts with tropical sea-surface temperature estimates that are only 1}33C cooler than present. ! 2001 Elsevier Science Ltd. All rights reserved.

1. Introduction experiment, which used the CLIMAP SSTs as boundary conditions, and low-latitude terrestrial paleoclimate Recurring questions regarding the magnitude of tropi- proxy data. Their analysis employed pollen evidence and cal climate change during the last glacial age have estimates of snowline depression from four tropical sites emerged since publication of the CLIMAP Project Mem- (Hawaii, the Colombian Andes, equatorial Africa, and bers (1976, 1981) reconstruction of ice-age sea-surface New Guinea), and led them to conclude that the temperatures (SSTs). The CLIMAP reconstruction, CLIMAP reconstruction underestimated the amount of which focused on the last glacial maximum (LGM) re- tropical temperature depression, which likely amounted vealed large areas in the tropics to have had SSTs as to 5}63C. warm as, or even slightly warmer than, those of Renewed interest in this topic has been generated the present. The CLIMAP project considered the by evidence and modeling that point to colder tropical LGM to date to 18,000 C yr BP [21,648 (21,484) temperatures than those implied by the CLIMAP 21,313 cal yr BP; equivalent calibrated ages ($1") have reconstruction (e.g., Guilderson et al., 1994; Stute et al., been obtained using CALIB 3.03 (Stuiver and Reimer, 1995; Thompson et al., 1995; Bush and Philander, 1998; 1993) for ages (18,000 yr, and using Stuiver et al. (1998) Farerra et al., 1999). Basic to much of the discus- for ages '18,000 yr]. Rind and Peteet (1985) sub- sion about colder glacial-age tropics has been the sequently noted con#icts between ice-age paleotempera- question of snowline depression (e.g., Broecker, 1995; tures generated by a general circulation model Hostetler and Mix, 1999; Lee and Slowey, 1999), yet most of the snowline data used in the arguments has not been rigorously evaluated. Since Rind and Peteet (1985) questioned the CLIMAP conclusions more than a dec- * Tel.: #1-206-543-1904; fax: #1-206-543-3836. ade ago, additional information has emerged that now E-mail address: [email protected] (S.C. Porter). permits a more thorough assessment. This paper focuses

0277-3791/01/$- see front matter ! 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 3 7 9 1 ( 0 0 ) 0 0 1 7 8 - 5 1068 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 on the set of paleoclimatic data bearing on snowline sometimes been used as a convenient proxy for former depression in the tropics during the last glaciation. In ELAs (e.g., PeH weH and Reger, 1972; Nogami, 1972, 1976; assessing these data, potential sources of error are also Fox and Bloom, 1994). While this approach is reasonable considered in arriving at site-speci"c and globally aver- in situations where Pleistocene glaciers terminated at aged values. cirque thresholds, in such cases the cirque glaciers disap- peared when snowlines rose above cirque levels at the 1.1. Glacier equilibrium-line altitudes end of the Pleistocene, meaning that site-speci"c ELA depression cannot be calculated directly. Furthermore, in The snowline, de"ned as the lower limit of perennial many glaciated tropical mountain ranges and on large snow on the landscape, is equivalent to the "rn limit volcanoes, glaciers expanded beyond cirques to form on temperate alpine glaciers, which is the lower limit valley glaciers, and in these circumstances ELAs lay of snow at the end of the ablation season. On such below (often well below) the altitudes of cirque #oors. In glaciers, the "rn limit approximates the equilibrium line, such cases, snowline reconstructions based on cirque- the locus of points along which the annual mass balance #oor altitudes may substantially underestimate actual is zero. In most recent paleosnowline studies, the equilib- snowline depression. rium line is regarded as synonymous with the snowline, and its altitude, following Meier and Post (1962), is designated the equilibrium-line altitude (ELA). The dif- 2.2. Upvalley limits of lateral moraines ference between the modern ELA (ELA ) and that of For a glacier in a balanced (steady-state) condition, some earlier time (e.g., the last glaciation, ELA") is a measure of equilibrium-line (i.e., snowline) depression the upvalley limit of its contemporary lateral moraines lies at the equilibrium line, below which ice-#ow paths (!ELA). The mass balance of a glacier, and #uctuations of the are diverging and ascending. If lateral moraines of a glacier's equilibrium line, are controlled by a number former glacier are well preserved, then the altitude of climate-related processes. For most low-latitude of their upvalley limits may closely approximate the temperate glaciers, the most important controls are accu- former ELA (Fig. 1b). Whereas this method has been mulation-season precipitation and ablation-season tem- used with success in some areas (e.g., Andrews, 1975; perature. Together these parameters encompass a range Mahaney, 1990), in many alpine regions lateral moraines of possible conditions controlling the ELA. Therefore, are absent or poorly preserved and at best provide only a unique value for past precipitation or temperature lower limiting estimates for contemporaneous ELAs. Meierding (1982) considered the highest lateral moraine cannot be derived from the !ELA alone (Porter, 1977; Seltzer, 1994). In most published paleosnowline studies, altitude to be the least reliable of several methods for no di!erence in precipitation is assumed (in most cases determining Pleistocene ELAs in the Front Range of probably erroneously) between the present and the Colorado. LGM, and a change in temperature is obtained by assuming a "xed atmospheric lapse rate. In cases where independent evidence for one parameter (i.e., LGM pre- 2.3. Glaciation threshold cipitation or temperature) is available from another cli- The glaciation threshold (GT) for a speci"ed area (nor- mate proxy, then !ELA can provide an estimate of the other parameter. mally a 7.5 topographic quadrangle or its equivalent: e.g., ca 60 km at 453 latitude) is the mean altitude between the lowest mountain with a glacier on it and the highest without (Fig. 1c). Although this method is 2. Methods not applicable to isolated peaks, such as volcanoes, it has proved useful for assessing regional snowline In studies of glaciated low-latitude mountains, trends across mountain ranges (e.g., "strem, 1966; "ve common methods have been used to reconstruct Porter, 1975, 1977; Rodbell, 1992). Studies have shown former ELAs. Because the methods di!er in their that the GT essentially parallels the regional ELA trend, approach, the results they produce are not strictly but commonly lies 100}200 m higher (Meierding, 1982; comparable. S. C. Porter, unpublished data). A limiting problem when using the GT to determine Pleistocene snowline 2.1. Cirque-yoor altitude depression is the need to identify and map former glacierized and nonglacierized peaks for a speci"c time (e.g., When a glacier just "lls a cirque, its steady-state ELA the LGM) throughout a rather broad region. Such exten- typically lies not far above the average altitude of the sive "eldwork normally is impractical, and so subjective cirque #oor (Fig. 1a). Therefore, cirque-#oor altitude has assessments of the extent and age of past glaciation S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1069

Fig. 1. Common methods used to derive past equilibrium-line altitudes in the tropics. See text for details. Cirque-yoor method: The ELA of a cirque glacier is inferred to lie above, but not far above, the cirque #oor (CF). However, if a glacier expands beyond the cirque threshold, the ELA will be lower than the cirque #oor. Lateral-moraine method: The upglacier limit of a lateral moraine approximates the ELA of the glacier that constructed the moraine. Glaciation-threshold method: The average altitude between the highest nonglacierized summit (Sn) and lowest glacierized summit (Sg) de"nes the glaciation threshold (GT) in a restricted area. Altitude-ratio method: In the median-altitude variant of this method, the ELA lies midway in altitude between the head of the glacier (A#) and the terminus (A). In the terminus-head altitude ratio (THAR) approach, the THAR equals the ratio of the altitude di!erence between the terminus and the ELA divided by the total altitude range of the glacier. The ELA can be estimated by adding the altitude of the terminus to the product of the total altitude range and an assumed THAR. Accumulation-area ratio method: In using this method, an accumulation-area ratio (AAR) is used, based on the ratio of the accumulation area (Sc) to the total area of the glacier (where Sa is the ablation area). Empirical studies suggest that a steady-state (SS, when the mass balance"0) AAR of 0.65$0.5 is appropriate for most temperate, relatively debris-free glaciers. The surface topography of the former glacier is reconstructed based on glacial-geologic data. From the glacier's area}altitude distribution (here depicted as a cumulative curve) and an assumed AAR, an ELA value is obtained. are usually based on analysis of topographic maps or method has proved useful in assessing snowline depress- aerial/satellite imagery. Despite the inherent uncertain- ion in some areas (e.g., the Cascade Range: Porter et al., ties and subjectivity involved (Meierding, 1982), this 1983, Fig. 4-15). 1070 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

2.4. Altitude ratios line). The area between each pair of successive contours is then measured and used to generate a cumulative curve Use of the median altitude of a former glacier as that graphically displays the glacier's area/altitude distri- a proxy for past snowline altitude is based on the bution. Assuming a steady-state AAR of 0.65, the ELA empirical observation that the "rn limit on temperate can be determined from the graphical plot. Error limits glaciers at the end of the ablation season often lies about are derived by assuming a range of AAR values halfway between the head of a glacier and its terminus (e.g.,$0.05 or$0.10). (Fig. 1d). HoK fer (1879) used a variation of this approach by computing the arithmetic mean of the altitude of 2.6. Comparison of methods a glacier's terminus and the average altitude of the mountain crest at the glacier's head. The median altitude Meierding (1982) assessed the relative reliability of method, in theory, is easy to apply if adequate altitude various paleo-ELA methods based on data from the data are available (i.e., topographic maps with a resolu- Front Range of Colorado. Using "rst-order trend-surface tion of ca 30 m or less, "eld data based on altimetry analyses, he found that the cirque-#oor, median-altitude, measurements, or digital-elevation data), and if a former and lateral-moraine methods had the greatest root mean alpine glacier had a normally distributed area vs. altitude square error (RSME"97}148 m), whereas the THAR curve. Nevertheless, whereas determination of the lower ("0.40) and AAR ("0.65) methods produced the most limit of a glacier based on end moraines or outwash consistent results (RSME"ca 80 m). A similar study in heads may be relatively straightforward, assigning an Norway by Torsnes et al. (1993) also concluded that the upper altitudinal limit to the former glacier in the cirque AAR method produced the most reliable results. The region is generally subjective and arbitrary. High, steep only similar comparative study in low latitudes was made cirque headwalls can lead to a potential range of esti- by Osmaston (1989a) in his study of glaciated equatorial mates di!ering by tens to hundreds of meters. African mountains. He concluded that a modi"ed version The median altitude method assumes that the ratio of of the altitude-ratio method gave the best results. Over- a glacier's range in altitude above the equilibrium line to all, the general lack of reliable topographic information the total altitudinal range of the glacier is 0.5. A variation and detailed "eld mapping for many tropical glaciated of this method has also been used in which the ratio areas, as well as limited radiometric age control, means [termed the toe-to-head (i.e., terminus-to-head) altitude that errors inherent in most of these methods may be ratio, or THAR] is some lower value (Fig. 1c). For magni"ed at low-latitude sites. example, Meierding (1982) reported that THARs of 0.35}0.40 generated the most accurate results in the Col- 2.7. Additional potential sources of error orado Front Range. The resulting ELAs were ca 100}150 m lower than those derived using the median In addition to di!ering results obtained from the sev- altitude (THAR"0.5). eral ELA methods outlined above, as well as the errors peculiar to each method, several other sources of error 2.5. Accumulation-area ratio enter into the calculation of LGM snowline depression.

The accumulation-area ratio (AAR) of a glacier is the 2.7.1. Altitude of the modern snowline ratio of the glacier's accumulation area to the sum of its On many tropical mountains, glaciers are absent or the accumulation and ablation areas (Fig. 1e). Empirical modern snowline altitude is known only approximately. studies of modern glaciers have shown that under Furthermore, in a time of generally warming climate, the steady-state conditions the AAR typically falls between transient nature of the snowline means that values ob- 0.5 and 0.8 (i.e., 0.65$0.15) (Meier and Post, 1962), tained from direct observations a decade or more ago, or meaning that the accumulation area occupies approxim- from topographic maps based on them, may underesti- ately two-thirds of the glacier's total area. In calculating mate the present snowline altitude. Where direct obser- past ELAs using the AAR method, a steady-state condi- vational data are unavailable, modern ELAs obtained tion is assumed and the glacier's extent and topography from recent glacier maps or aerial photography and are determined using glacial-geologic data such as lateral employing the median altitude or AAR methods likely moraines, erratics, and trimlines (Porter, 1981). An initial o!er the best estimates. Nevertheless, errors of tens of (estimated) ELA is selected using the altitude ratio meters or more may result. method. Contours of the glacier surface are then drawn, consistent with principles of glacier #ow (contours of 2.7.2. Age of the LGM glacial limit a glacier in a balanced state typically are concave up- In few cases has the limit of the last glaciation been glacier in the accumulation area and convex in the abla- radiometrically dated in tropical mountains, and in no tion area, with the degree of concavity or convexity instance has it been closely bracketed by dates. There- increasing with increasing distance from the equilibrium fore, synchrony of moraines likely built during the LGM S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1071 commonly is inferred, based mainly on relative-age cri- (1994) suggested that the steady-state AAR on such gla- teria and the pattern of moraine sequences, and on infer- ciers might be reduced from ca 0.65 to as little as 0.10. red regional or global correlations. Estimates of paleo-ELAs for debris-covered glaciers Recent studies in several (nontropical) mountain based on the AAR or THAR methods therefore may ranges report that glaciers advanced repeatedly during result in anomalous ELA values and produce erroneous the last glaciation (marine isotope stage 2), and that the regional ELA gradients. resulting moraines often are nested or closely spaced. These ice advances typically date between ca 27,000 2.7.6. ELA gradient and 16,000 C yr BP [31,300 and 18,972 (18,876) An error can be introduced in calculating !ELA if the 18,784 cal yr BP] (e.g., Phillips et al., 1990; Gosse et al., regional snowline gradient is not considered. If there is 1995; Lowell et al., 1995; Swanson and Porter, 1999). no present or past ELA gradient and both the modern Although comparable moraine sequences may exist in and LGM ELAs are determined for glaciers on the oppo- the tropics (e.g., New Guinea: Blake and LoK %er (1971); site #anks of a mountain range, the !ELA on each #ank Andes: Clapperton (1987) and Thouret et al. (1996)), none will be the same (Fig. 2, !ELA). However, consider the has yet been closely dated. The actual age of an `LGMa common case where the modern and LGM ELAs are moraine may lie anywhere within this ca 12,500 yr range. determined for opposite #anks of a mountain (e.g., Por- Nevertheless, the juxtaposition of such moraines implies ter, 1979) or mountain range (e.g., Porter et al., 1983), comparable snowline depressions ()50 m di!erence) across which there is a marked precipitation gradient. If during these successive advances. regional trend surfaces of present and past ELAS are determined, then !ELA may be less than if no gradient 2.7.3. Paleoglacier reconstructions exists (assuming uniform lowering on both #anks) (Fig. 2, A potential source of error in the AAR method is !ELA), or the !ELA on one #ank may di!er from that associated with the topographic reconstruction of a for- on the opposite #ank if the modern and paleo-ELA mer glacier, which is necessarily subjective. Below the gradients converge or diverge (Fig. 2, !ELA ). Where equilibrium line, terminal and lateral moraines and trim- possible, therefore, trend surfaces of present and former lines provide altitudinal constraints along a glacier's for- ELAs should be calculated and their di!erence deter- mer margin, whereas above the former ELA little control mined in order to obtain the most reliable estimates of usually exists. Errors in circumscribing the accumulation ELA depression. area tend to be minimal because of steep valley walls upglacier from the equilibrium line; thus, an erroneous 2.7.7. Adjustment for lowered sea level altitude estimate for the glacier margin in this zone only In most ELA reconstructions, sea-level lowering at the minimally a!ects the lateral extent of the accumulation LGM is not considered. However, the eustatic fall of sea area. level had the e!ect of raising the altitude of mountain Errors related to topographic reconstruction are mini- summits by the amount of the sea-level drawdown. As- mized in the case of small glaciers with normally distrib- suming that sea level fell ca 120 m (e.g., Fairbanks, 1989; uted area}altitude curves. Large, complex glaciers, and Bard et al., 1990; Rohling et al., 1998), this amount should those having a trend perpendicular to the regional ELA be subtracted from the calculated !ELA to obtain an gradient, may generate unreliable results. adjusted ELA depression with respect to the changing world sea-level datum (Broecker, 1997). For example, if 2.7.4. Orographic ewects the present snowline (ELA ) on a glacier lies at 3900 m Small glaciers con"ned to deep cirques on leeward (Fig. 3) and the reconstructed ELA" at 3000 m, then the #anks of mountains, or shaded by steep mountain walls, apparent !ELA"900 m. However, during full-glacial may persist at altitudes well below those of glaciers on time, the ELA" lay at an altitude of 3120 m, rather than fully exposed slopes. In general, ELAs based on geomet- 3000 m. Therefore, the di!erence between the present rically simple glaciers in exposed sites are likely to pro- ELA (3900 m) and full-glacial ELA", (3120 m) is 780 m vide the most regionally consistent ELA values. In ("!ELA adjusted for sea-level fall, designated !ELA addition, orographic e!ects (e.g., unequal exposure to in Fig. 3). The sea-level factor becomes relevant if !ELA sun, unequal accumulation) may lead to a range of tens of is used in conjunction with an atmospheric lapse rate to meters in the altitude of the "rn limit on a given glacier. estimate temperature depression during full-glacial time, for it will reduce the estimate by ca 10}15% (see below). 2.7.5. Anomalies resulting from a cover of rockfall debris An extensive cover of rockfall debris can insulate an alpine glacier and greatly reduce ablation (Clark et al., 3. Tropical mountain glaciers (23.53N}23.53S) 1994). Such glaciers tend to be relatively insensitive to a warming climate and typically advance to lower alti- Data on full-glacial snowlines are available for 18 tudes than do nearby debris-free glaciers. Clark et al. mountain areas in tropical Africa, Central and South 1072 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

Fig. 2. Where no modern (ELA ) or full-glacial (ELA") snowline gradients exist, !ELA is the same on opposite sides of a mountain or mountain range (!ELA). Where ELA gradients exist, the !ELA may be less (!ELA) or greater (!ELA) than in the no-gradient case, or may di!er on opposite sides of a mountain or mountain range (!ELA, !ELA).

Fig. 3. When adjusted for fall of sea level from its modern level (SL ) to its full-glacial level (SL"), !ELA is reduced by an amount equivalent to the sea-level fall (!ELA). A fall in sea level of 120 m had the e!ect of raising the altitude of the summit and the reconstructed !ELA" by this amount.

America, and several glaciated Paci"c islands (Fig. 4). snowline, but the highest summits developed glaciers Some permit only minimum estimates of snowline during the last glaciation. The Simen Mountains (13314 depression because the highest summit lies below N) culminate in Ras Dejen (4543 m), the highest moun- the modern snowline, but for more than half, the full- tain in Ethiopia (Fig. 5a). Hurni (1989) mapped moraines, glacial snowline depression can be calculated. In addi- cirques, and periglacial features that presumably date to tion, regional data on snowline depression have been the last glaciation (no radiometric dates are available). Of generated for the tropical Peruvian, Bolivian, and 20 former glaciers, those that formed in NW- to NE- Chilean Andes. facing cirques terminated as low as 3760 m; those occupy- ing S-facing catchments reached only as low as 4400 m. 3.1. Africa Former ELAs were estimated on the basis of the median altitude of the glaciers (using end moraines and the top of 3.1.1. Ethiopian highlands cirque headwalls), as well as the altitude of the upper end The highlands of Ethiopia, which reach altitudes of of lateral moraines, giving an average value of ca 4250 m. more than 4000 m, are too low to intersect the modern Minimum !ELA was therefore 290 m (Table 1). S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1073

Fig. 4. Map showing location of glaciated areas in the tropics where ELA" and !ELA estimates have been obtained.

Glacial landforms (cirques, U-shaped cross-valley pro- eastward on Kibo (5455}5360 m) and lies at ca 5030 m on "les, moraines, striations) were used by Potter (1976) to Mawenzi. ELA" gradients slope west}northwestward map the extent of a former ice cap ('140 km) atop across Kibo (4540}4575 m) and eastward across Mawenzi Mt. Badda (7352 N; published altitudes are inconsistent (4300}4240 m), leading to unequal values of !ELA on and range from 4350 to 4133 m), ca 160 km southeast di!erent sides of the mountain (Fig. 5c). These he at- of Addis Ababa (Fig. 5b). End moraines in W-trending tributed to complex meteorological in#uences. Osmaston valleys were noted as low as 3650 m. Lateral moraines, concluded that a !ELA of 770$60 m between the Main inferred to be of last-glacial age, reach as high as 4000 m. glaciation and a Recent ice advance (i.e., middle-to If a summit altitude of 4350 m is adopted, then minimum late-Neoglaciation) explained his results on most of ELA depression was 350 m. A comparable minimum Mazwenzi and Kibo. The calculated rise in ELA since !ELA results if the median altitude method is used. the Neoglacial maximum (a minimum of 60 m) increases the !ELA to at least 830$160 m (Table 1). 3.1.2. Kilimanjaro, Tanzania Mt. Kilimanjaro (3305S), the highest mountain in Afri- 3.1.3. Ruwenzori, Uganda ca, now supports (5 km of glacier cover. This volcanic The Ruwenzori Mountains (0320-25N) reach altitudes massif includes two summits; the highest, Kibo (5895 m), of more than 5000 m and contain many small glaciers lies west of a lower peak, Mawenzi (5147 m) (Fig. 5b). that collectively cover ca 4.5 km (Fig. 5d). During the During a succession of glaciations, glaciers on the vol- latest (Lake Mahoma) of at least three Pleistocene cano expanded to cover ca 153 km (Osmaston, 1989a). glaciations, ice covered ca 260 km and terminated as End moraines of the last (`Maina) glaciation on Kibo low as 2070 m on the eastern slope (Osmaston, 1989b). and Mawenzi reach as low as ca 3250 m. A date of 14,750$290 C yr BP[17,981 (17,647) To determine snowline depression on Kilimanjaro, 17,315 cal yr BP] from Mahoma Lake, provides a min- Osmaston (1989a) used a modi"cation of Kurowski's imum age for moraines that impound the lake (Living- (1891) method, which assumed that net accumulation is stone, 1962, 1975). a linear function of altitude and that the ELA lies at the Osmaston (1989b) calculated ELAs of present and mean altitude of the glacier area. Osmaston included an former glaciers using the Area}Height}Accumulation arbitrary weighting factor to take possible nonlinearity of method described above. He derived a su$cient number the accumulation trend into account, and concluded that of measurements to de"ne the regional ELA gradient, this approach, which he called the Altitude}Height}Ac- which descends to the east}southeast. The ELA along cumulation (Alt}Ht}Acc) method, was likely to give an approximately west}east transect across the range more reliable results. descends from ca 4720 to 4270 m (Osmaston, 1989b, Osmaston's (1989a) analysis disclosed an asymmetrical Fig. 7b), the estimated average ELA being ca 4600 m. distribution of glaciers, and an ELA that slopes gently During the Lake Mahoma glaciation the ELA" sloped 1074 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

Fig. 5. Sites in Africa where ELA" and !ELA estimates have been obtained. (a) Ras Dejen, Simien Mountain, Ethiopia; (b) Mt. Badda, Ethiopia; (c) Kilimanjaro, Tanzania; (d) Ruwenzori Mountains, Uganda; (e) Mt. Kenya, Kenya; (f) Mt. Elgon, Kenya}Uganda, and Aberdare Mountains, Kenya. See text for details. S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1075 eastward from ca 4100 to 3600 m (Osmaston, 1989b, and 12,000 yr BP [ca 39,800 and 14,111 (13,992) Table 10). Along this W}E transect, the !ELA increased 13,880 cal yr BP], but none that correlated with the mar- from ca 620 to 670 m. ine isotope stage 2 maximum. White (1981), using relative-age criteria, correlated the Hueyatlaco moraines 3.1.4. Mt. Kenya, Kenya (Diamantes Substage, Second Advance) on IztaccmH huatl and Santo TomaH s Substage moraines on Ajusco with The glaciers of Mt. Kenya (5202 m; 0309S) have been shrinking in area and now cover (1 km (Young and Pinedale moraines of the western United States that are Hastenrath, 1991). Modern ELAs are estimated to lie at generally regarded as correlatives of the isotope stage ca 4700}4725 m (Mahaney, 1990, Fig. 11.8) and glacier 2 maximum (e.g., Richmond, 1986). Subsequently, White termini at 4650$100 m. Moraines of the Liki II gla- and Valastro (1984) inferred that the Santo TomaH s drift is ciation (Osmaston, 1989b; Mahaney, 1990, Table 11) more than 25,000 C yr (ca 29,000 cal yr) old, based on extend as low as 3200 m, and are older than 15,000 a single radiocarbon date of a bulk sample from the C yr BP [18,003 (17,916) 17,830 cal yr BP]. Osmaston B horizon of a buried soil. The soil is developed on Santo (1989b, Table 11), using the median altitude method, TomaH s till and lies stratigraphically below tephra on calculated an ELA of 4200 m for the glaciers originating which another buried soil is developed that dates to " 15,090$150 C yr BP [18,186 (18,009) 17,832 cal yr BP]. on the highest summit (Batian), and a !ELA of 600 m (Osmaston, 1989b, Table 12). Mahaney (1990, Fig. 11.8), Recent Cl surface-exposure ages for the outer Hueyat- on the other hand, estimated that the full-glacial ELA, laco moraines on ItaccmH huatl indicate that the main Late based on lateral moraine altitudes, lay between ca 3680 Pleistocene advance probably culminated 19,000}18,000 (SW slope) and 4000 m (NW slope) (Fig. 5e); most likely it Cl yr ago, and that the inner moraines are 15,000} averaged close to ca 3700 m. Based on his data, the 14,000 Cl yr old (VaH squez-Selem, 1998). White (1981) calculated modern and past ELAs for !ELA ranged between ca 725 and 1020 m. The discrepancy between Mahaney's results (Fig. 5e, Table 1) and glaciers on the western slope of IztaccmH huatl and the Osmaston's may largely re#ect the di!erent methodolo- northern and eastern slopes of Ajusco. Based on mean gies used. altitudes and an AAR of 0.65, he determined an average ELA of ca 4880 m for glaciers on IztaccmH huatl and 3970 m for the Diamantes Second Advance, indicating an 3.1.5. Other glaciated African summits LGM snowline depression of 910 m (Fig. 6a). For Ajusco, Two additional low-latitude mountains, each lying be- which lies 65 km west of IztaccmH huatl, White calculated low the modern snowline, developed large glaciers during that the ELA" for the Santo TomaH s advance was 3270 m. the last glaciation (Osmaston, 1989b, Table 11). On This is ca 155}170 m below ELAs calculated for two Mt. Elgon (4320 m; 1330N), which had 75 km of ice at Neoglacial ice advances and ca 665 m lower than the the glacial maximum [prior to ca 11,000 C yr BP; present ice-free summit. Absence of glacier ice on Ajusco 12,966(12,917)12,865cal yr BP], moraines extend as low (White and Valastro, 1984, Fig. 2) implies an ELA of as 3350 m. Osmaston (1989b) calculated that the ELA " '3937 m, and a full-glacial !ELA of '665 m (Fig. 6b). lay at 3600}3900 m, which means a minimum !ELA of 420}720 m (Fig. 5f). In the Aberdare Mountains (4001 m; 0315}45S), which had 23 km of ice cover at the glacial 3.2.2. Altos de Cuchumatanes, Guatemala maximum, LGM moraines reach as low as 3200 m. Hastenrath (1974) reported evidence of glaciation in the Altos de Cuchumatanes (15330 N), a carbonate karst The calculated ELA" is 3700 m and !ELA was '300 m (Fig. 5f). upland that reaches altitudes of nearly 3800 m (Fig. 6c). An end-moraine complex, with up to 20 m of relief, descends to 3470}3600 m. Hastenrath's published recon- 3.2. Mexico and Central America naissance maps do not permit detailed topographic reconstruction of the former glaciers. He estimated that the 3.2.1. Mexican volcanoes associated ELA lay at ca 3650 m, which is close to the The high stratovolcanoes of Mexico's Cordillera median altitude of the glaciated terrain. The correspond- NeovolcaH nica display evidence of repeated Pleistocene ing minimum ELA is 150 m. Although no dates are glaciations. Two of the best-documented records are ! available, Hastenrath considers the moraine character- from Iztacc Hhuatl (5286 m; 19305}15 N) and Ajusco m istics comparable to those of presumed LGM age in the (3937 m; 19312.5 N) (White, 1962; Heine, 1976, 1978, mountains of Venezuela, Costa Rica, and Mexico. 1984; White and Valastro, 1984) (Fig. 10). Initially, chro- nologies of glaciation were based on relative-dating criteria, limiting radiocarbon ages (primarily of paleosols, 3.2.3. Sierra de Talamanca, Costa Rica wood fragments, peat, and calcrete), and correlation with In the Cordillera de Talamanca, culminating in Cerro glacial sequences elsewhere (Heine, 1984). Heine (1984) ChirripoH (9329N; 3819 m) and now below the regional reported evidence of glacier advances at ca 35,000 snowline, two groups of moraines delimit former small 1076

Table 1 Tropical snowline data

& ' ( Latitude Locality Altitude (m) Method ELA (m) ELA" (m) !ELA (m) ELA gradient Age control Reference(s) Figure assessed S.C. 23330N Taiwan Shan, Taiwan 3997 Cirque #oor 3350}3450 '400 No Ono (1988) 9a

19350N Mauna Kea, Hawaii 4206 AAR (0.60) 4715 3780 '425 (935) Yes 20,300$2300; Porter (1979); Dorn et al. 9b Porter 18,900$800 (1991) 19312.5 N Ajusco, Mexico 3937 AAR (0.65) '3800 3270 '665 No White and Valastro (1984) 6b / 19310N IztaccmH huatl, Mexico 5286 AAR (0.65) 4880 3970 910 No ca 21,000 ; White (1981) 6a Quaternary '15,090$150 15330N Altos de Cuchumatanes, 3798 Median '3800 3650 '150 No Hastenrath (1974) 6c Guatemala altitude 13314N Ras Dejen, Simen 4543 Median '4543 4250 '290 No Hurni (1989) 5a Science Mountains, Ethiopia altitude 9329N Cerro ChirripoH , Costa Rica 3819 Median '3819 3500}3550 '305 No '10,140$120 Hastenrath (1973) and 6d altitude Orvis and Horn (2000) Re v 8}93N Pico BolmH var, Venezuela &5000 Median &4700 3800 &900 No (12, 650$130 Schubert (1974, 1984) and 7a iews altitude (19,080$820 Clapperton (1993) 20 7352N Mt. Badda, Ethiopia 4350 Lateral '4350 4000 '350 No Potter (1976) 5b (2001) moraines, median altitude

6305N Mt. Kinabalu, Borneo 4101 Median 4570$150 3665 905$150 (?) No Koopmans and Stau!er 9c 1067 altitude (1967) }

4352N Nevado del RumHz, Colombia 5200 Median 4800}4900 3550}4000 1075 Yes '16,220$80 Herd (1974, 1982) Thouret 7b 1091 altitude ('19,500,(23,000) et al. (1996) 4348N Nevado de Santa Isabel, 4950 Median 4700}4800 3500}4000 1000 Yes Herd (1974, 1982) 7c Colombia altitude 1320N Mt. Elgon, Kenya}Uganda 4320 Alt}Ht}Acc '4320 3600}3900 '420}720 No '11,000 Osmaston (1989b) 5f 0320}25N Ruwenzori, Uganda 5109 Alt}Ht}Acc 4270}4720 3600}4100 620}670 Yes '14,750$290 Osmaston (1989b); 5d Livingstone (1975) 0310S Mt. Kenya, Kenya 5202 Alt}Ht}Acc 4700}4725 3680}4200 725}1020 Yes '15,000, (25,000 Osmaston, (1989b) and 5e Mahaney (1990) 0330S Antisana, Ecuador 5790 Median 4600}4830 3480}3860 970}1120 Yes '12,000, (30,000 Clapperton (1987, 1993) 6e altitude 0340S Aberdare Mountains, 4001 Alt}Ht}Acc '4001 3700 '300 Yes '12,200 Osmaston (1989b) 5f Kenya 1325S Chimborazo, Ecuador 6310 Median 4800}4900 3880}4090 810}920 Yes '12,000, (30,000 Clapperton (1987, 1993) 6e altitude 3305S Kilimanjaro (Kibo), 5895 Alt}Ht}Acc 5360}5455 4540}4575 830$60) Yes Osmaston (1989a) 5c Tanzania 3305S Kilimanjaro (Mawenzi), 5147 Alt}Ht}Acc 5030 4240}4300 830$60) Yes Osmaston (1989a) 5c Tanzania 5345S Mt. Wilhelm, Papua New 4509 Median &4600 3500}3600 '850}1010 No LoK %er (1972) 9d Guinea altitude, cirque #oor 6302S Mt. Giluwe, Papua New 4368 Median &4600 3500}3550 '820}870 No LoK %er (1972) 9d Guinea altitude, cirque #oor 6}6320S Sarawaged Range, Papua 4121 Median &4600 3650}3700 '470}520 No LoK %er (1972) 9d New Guinea altitude, cirque #oor 8325S Mt. Albert Edward, Papua 3990 Median &4600 3600}3650 '340}390 No LoK %er (1972) 9d New Guinea altitude, cirque #oor 8358S Mt. Victoria, Papua New 4036 Median &4600 3650}3700? '335}385 No LoK %er (1972) 9d S.C. Guinea altitude, cirque Porter #oor 7340S Cordillera Blanca, Peru '6000 THAR (0.2, 4985$120 4200}4400 440}970 Yes '13,280$190 Rodbell (1992) 6f /

0.4), lateral Quaternary moraines 7308}58S Co. Oriental (main divide &4500 THAR (0.2, 4620 3540}3640 900}1150 Yes '12,100$190 Rodbell (1992) 6f W), Peru 0.4), lateral

moraines Science 7308}58S Co. Oriental (main divide &4500 THAR (0.2, 4620 3150}3300 1100}1350 Yes '12,100$190 Rodbell (1992) 6f E), Peru 0.4), lateral

moraines Re v

7308}58S Co. Oriental (local divide, &4500 THAR (0.2, 4620 3850}3900 750}950 Yes (12,100$190 Rodbell (1992) 6f iews W), Peru 0.4), lateral moraines 20 5}173S Peruvian Andes Cirque #oors, &4700}5300 3200}4900 600}1400 Yes Fox and Bloom (1994) (2001) THAR (0.45)

103S Peruvian Andes Cirque #oors 4700}5000 3400}4200 900}1200 Yes Klein et al. (1999) 8 1067 123S Peruvian Andes Cirque #oors 4700}5000 3400}4600 700}1100 Yes Klein et al. (1999) 8

143S Peruvian Andes Cirque #oors 4500}5100 3600}4200 700}1000 Yes Klein et al. (1999) 8 } 1091 163S Peruvian}Bolivian Andes THAR (0.45) 4500}5200 3200}4400 550}1200 Yes Klein et al. (1999) 8 183S Chilean}Bolivian Andes THAR (0.45) 5100}5300 3200}4200 800}1100 Yes Klein et al. (1999) 8 203S Chilean}Bolivian Andes THAR (0.45) 5400}5600 3800}4400 950}1100 Yes Klein et al. (1999) 8 223S Chilean}Bolivian Andes THAR (0.45) 5400}5800 3800}4800 900}1100 Yes Klein et al. (1999) 8

&Estimated values in italics 'Minimum values in italics (Ages in C yr BP, except for Cl ages (in italics) )Based on Osmaston's (1989a) estimated average value of 770$60 m, and corrected for a minimum 60 m rise of ELA since the Neoglacial maximum. 1077 1078 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

Fig. 6. Sites in Mexico and Central America where ELA" and !ELA estimates have been obtained. (a, b) IztaccmH huatl and Ajusco, Mexico; (c) Altos de Cuchumatanes, Guatemala; (d) Cerro ChirripoH , Costa Rica. See text for details. valley glaciers (Weyl, 1956; Hastenrath, 1973) (Fig. 6d). which therefore is a possible minimum value for ELA The upper basins of the main glaciated valleys lie at depression. 3450}3550 m and contain several small moraine- and rock-dammed lakes. A C date of basal sediments 3.3. Andes of South America from moraine-impounded Lago de Morrenas (3480 m) indicates that the basin was deglaciated prior 3.3.1. Sierra Nevada de MeH rida, Venezuela to 10,140$120 C yr BP [12,112 (11,808) 11,123 Schubert (1974, 1984) and Schubert and Clapperton cal yr BP] (Horn, 1993). At a site near El Empalme (1990) reported evidence of multiple glacier advances in (2400 m), ca 60 km northwest, the paH ramo (alpine) pollen the Sierra Nevada de MeH rida (altitudes to 5000 m) be- zone dates to 20,750 C yr BP (ca 24,200 cal yr BP) and tween 8315 and 9300N latitude in the central Venezuelan represents a treeline depression of at least 650 m (Martin, Andes. He assigned the mapped drifts to several stades of 1964); possibly, this date is close to the time of maximum the MeH rida Glaciation, of presumed Late Pleistocene age. snowline depression as well. Assuming that glaciers The oldest recognized drift (Early Stade), which lacks headed close to 3700 m (i.e., just below most crest alti- dating control and has no clear morainal morphology, tudes) and terminated as low as ca 3300}3400 m (Hasten- may date to marine isotope stage 4 or 6 (Clapperton, rath, 1973; Orvis and Horn, 2000), the full-glacial ELA 1993, Table 14.1a). Well-developed moraines of the last (median-altitude method) lay at ca 3500}3550 m. glaciation (Late Stade) form a nested and overlapping This value is comparable to that recently derived by succession between 3000 and 3500 m altitude. A min- Orvis and Horn (2000) of 3506}3523 m, and represents imum age of 12,650$130 C yr BP [15,128 (14,874) a minimum ELA depression of ca 295}315 m (Fig. 6d). 14,621 cal yr BP] was obtained for peat upvalley from the They reported that the annual 03C isotherm lies at youngest of these moraines. Peat beneath and above 4900 m, or ca 1400 m above the late Pleistocene ELA, thick outwash just beyond the outer moraine limit is S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1079 dated 19,080$820 and 16,500$290 C yr BP [22,700 headward limits of individual ice tongues. The recon- and 19,736 (19,423) 19,153 cal yr BP], respectively, point- structed ELA on Nevado del RumHz, along a section ob- ing to an isotope stage 2 age for the Late Stade. A date lique to the modern ELA gradient, rose from ca 3550 m for basal peat on the #oor of a cirque demonstrates on the eastern slope to ca 4000 m on the western slope. deglaciation near the range crest by 11,470 C yr BP On Nevado de Santa Isabel, the Late Pleistocene ELA [13,463 (13,384) 13,317 cal yr BP] (Schubert and Clapper- along the same transect as the present ELA gradient rose ton, 1990). from ca 3500 to 4000 m (Herd, 1974). In both cases, Herd Schubert and Valastro (1974) mapped the glacial geol- calculated the !ELA as ca 950$50 m. However, if the ogy in the PaH ramo de la Culata (northern Venezuelan modern snowline values of Hoyos-Patin o (1998) are used, Andes), where end moraines of the last glaciation are the mean !ELA increases to 1075 for Nevado del RumH z traceable as low as ca 3150 m and cirque headwalls and 1000 m for Nevado de Santa Isabel (Figs. 7b, c). upvalley average 4450 m. Based on the median altitude method, the ELA" for Pico BolmH var (&5000 m) would 3.3.3. Ecuadorian Andes have been at ca 3800 m (Fig. 7a). This is about 900 m The Ecuadorian Andes comprise two parallel ranges, below the modern snowline [ca 4700 m according to the Cordillera Occidental (western range; 0322N}1329S) Schubert (1974), but more recently above 4700 m and the Cordillera Oriental (eastern range; 0301N} (Schubert, 1998)]. However, this !ELA value should be 2320S), separated by intermontane basins and forming considered only approximate, for the ELA gradient has dissected plateaulike surfaces at 3500}4000 m altitude not been considered and the modern snowline value is (Clapperton, 1987) (Fig. 7d). The ranges are surmounted only an estimate. by 14 glacierized stratovolcanoes more than 4600 m high (Jordan and Hastenrath, 1998). Clapperton (1987, 1993) 3.3.2. Cordillera Central, Colombia summarized the glacial geology of the mountains and Glaciers and perennial snow"elds on the high Andean noted that deposits of `full-glaciala age typically include volcanoes Nevado del RumH z (5200 m), Nevado de Santa 3}4 lateral and/or terminal moraines that date broadly to Isabel (4950 m), and Nevado del Tolima (5200 m) cover within the interval 34,000}12,000 C yr BP [38,700 and ca 36 km between latitudes 4335 and 5310N (Herd, 14,111 (13,992) 13,880 cal yr BP). In discussing his ap- 1974, 1982). According to Hoyos-Patin o (1998), the proach for determining ELA depression, he noted that snowline on Nevado del RuH iz lies at 4900 m on the some earlier studies used cirque-#oor altitudes to calcu- western #ank and 4800 m on the eastern #ank (Fig. 7b). late !ELAs. He cautioned that during the glacial max- On Nevado de Santa Isabel, the snowline lies at 4800 m imum, most cirques were buried under continuous on the western #ank and 4700 m on the eastern #ank ice"elds, implying that cirques likely originated during (Hoyos-Patin o, 1998) (Fig. 7c). These values are earlier intervals of reduced glacier cover. Therefore, such 60}170 m higher than those measured by Herd (1974, cirques cannot be used for constructing former snow- 1982) in 1972}73. Herd's values were based on average lines. Exceptions were mountain ridges in southern Ecua- transient summer snowline, which was estimated to lie ca dor that are too narrow to have supported ice"elds or ice 100 m below the actual snowline. caps. Herd (1974) mapped the approximate extent of late Clapperton (1987) estimated the ELA based on "eld Pleistocene glaciers on the massif and obtained a date of observations and aerial photographs, deriving values of 13,760$150 C yr BP [16,705(16,498)16,286cal yr BP] 4800}4000 m for Chimborazo (6310 m) in the Cordillera for peat directly beneath a tephra that overlies the outer- Occidental and 4600}4830 m for Antisana (5790 m) in the most drift of the last glaciation. His mapped glacier limit Cordillera Oriental (Fig. 7d). The values for ELA on approximates the closely nested limits of the early and Chimborazo volcano compared closely with those based late Murillo drifts of Thouret et al. (1996). The late on the upper limits of lateral moraines. He used the Murillo advance predates basal peat overlying the mo- median-altitude method to calculate ELA depression raines having an age of 16,220$80 yr BP [19,247 along transects across the two cordilleras, noting, how- (19,100) 18,970 cal yr BP], while the early Murillo mo- ever, that the results were `crude approximationsa be- raines likely date between 28,000 and 21,000 C yr BP cause the estimated modern ELAs were imprecise and [28,000 and 23,100 cal yr BP] based mainly on regional the reconstructions were based on topographic maps tephrochronology. In the Eastern Cordillera of Colom- with a large (40 m) contour interval. In both cordilleras, bia, the greatest advance of the last glaciation is believed eastward-sloping ELA and ELA" gradients are present, to have culminated between ca 23,500 and 19,500 and !ELA is greatest on the eastern #ank of each range C yr BP (ca 28,000 and 23,100 cal yr BP) (Helmens (Fig. 7d). Mean !ELA values increase from west to east et al., 1996). across the Andes, from 810}920 m in the Cordillera Occi- Herd (1974) determined the late Pleistocene ELA using dental to 970}1120 in the Cordillera Oriental (based on the median-altitude method, assuming that ice divides of values in Clapperton's Fig. 7; values in Clapperton's the full-glacial ice caps on the volcanoes represented the Table 2 are somewhat greater). 1080 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

Fig. 7. Sites in tropical South America where ELA" and !ELA estimates have been obtained. (a) Pico BolmH var, Venezuela; (b, c) Nevado del RumH z and Nevado de Santa Isabel, Colombia; (d) Chimborazo and Antisana, Ecuadorian Andes; (e) Cordillera Blanca and Cordillera Oriental, Peruvian Andes. See text for details.

3.3.4. Peruvian Andes Oriental (8308}9358S) (Fig. 7e). Rodbell (1992) used the Studies of snowline depression in Peru have centered THAR method and derived measurements for ca 20 mainly in the Cordillera Blanca (7340S) and Cordillera glaciers in each mountain range using 1 : 100,000-scale S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1081 topographic maps with 50 m contours, supplemented by Fox and Bloom (1994) made a regional study of Late measuring on 1 : 25,000-scale aerial photographs the up- Quaternary snowlines in the Peruvian Andes (5}173S). per limits of preserved lateral moraines. Results obtained They based their estimate of modern snowline altitude on using THAR values of 0.2 and 0.4 typically di!ered by ca the lower limit of snow depicted on 1 : 100,000-scale aer- 150}350 m. The accuracy of resulting ELAs was esti- ial photographs (1955}63). The data were transferred to mated to be $50 m. The age of full-glacial moraines is topographic maps (50-m contour interval), and an error unknown, but radiocarbon dates from moraine-dammed of $100 m was assumed. They showed that the snowline lakes and bogs provide minimum ages for deglaciation rises from ca 4700 m in the northern and eastern Andes in the Cordillera Oriental of 12,100$190 C yr BP to more than 5300 m in the south and west, and has [14,382 (14,114) 13,870 cal yr BP] and in the Cordillera a westward-rising gradient, especially in the north. The Blanca of 13,280$190 C yr BP [16,140 (15,859) overall pattern is similar to that derived by Nogami 15,559 cal yr BP]. (1976). Full-glacial snowline was reconstructed based on the cirque-#oor method, assuming that glaciers occupied 3.3.4.1. Cordillera Oriental. Rodbell (1992) obtained the lowest cirques contemporaneously. The time of this occupation, however, has not been dated. The recon- ELA" values for former glaciers lying east and west of the main divide, as well as west of a local divide lying ca struction indicates that snowline depression reached 20 km west of the main divide (Fig. 7e). No glaciers are a maximum of 1400$200 m on the eastern side of the Peruvian Andes but decreased westward to a minimum present in the study area, but ELA is based on a regional glaciation threshold estimate of 4620 m (Seltzer, of 600 m in the western ranges and on most of the 1987). Paleoglaciers on the western side of the local Peruvian Altiplano (Fig. 8). An average value was not divide had an average ELA of 3850-3900 m, represent- given, but based on their Fig. 7 it appears to be in the " range of 800}1000 m in the north, decreasing to ing a !ELA of 750}950 m. Corresponding ELA" values west and east of the main divide are 3540}3640 and 600}800 m in the south. Klein et al. (1999) subsequently expanded on the work 3150}3300 m, respectively. The !ELAs are 900}1150 and 1100}1350 m, respectively, with estimated mean values of of Fox and Bloom (1994) to include not only the Andes of ca 850}1000 and 1200 m for the two respective sides of southern Peru, but also Bolivia and northern Chile the range (Rodbell, 1992). These results demonstrate (Fig. 8). They adopted Fox and Bloom's estimate of the a W}E snowline gradient, with snowline depression in- modern snowline altitude in Peru (see above), and used creasing toward the Amazon Basin, which is, and appar- LANDSAT Thematic Mapper imagery to map the lower ently was, the primary source of precipitation. limit of snow cover in the southern part of their study region. Maps with scales of 1 : 50,000 (20 m contours) and 1 : 250,000 (250 m contours) were used. They noted the 3.3.4.2. Cordillera Blanca Rodbell's (1992) estimates of lack of adequate dating control for the last glaciation, but ELA in the Cordillera Blanca range from 4985$120 m assumed that moraines mapped as (local) LGM were west of the divide to 5030$110 m east of the divide, constructed contemporaneously. In Peru, cirque-#oor indicating no discernable gradient (Fig. 7e). The cal- altitudes were used to determine LGM snowlines; how- culated ELA" is 4400$100 to 4250$110 m ever, the age of the last cirque glaciation is unknown, and (THAR"0.2 and 0.4, respectively) west of the divide and so their assumed age equivalence is unproven. Further- 4200$170 to 4350$150 m east of the divide, giving an more, many cirques lay at the heads of valley glaciers. In average of ca 4300 m for the range as a whole. The the southern region, an adjusted THAR of 0.45 was used calculated !ELA is 440}900 and 530}970 m, respectively. to calculate speci"c snowlines, as well as a regional snow- Rodbell (1992) suggested that, based on comparison with line pattern for the last glaciation. In the area where data ELA and glaciation threshold values, the average from the two methods overlap, the di!erence in estimated !ELA for the range is ca 700 m. ELA between the methods is ca 175 m. The con"guration of the reconstructed glacial snowline is similar to that of 3.3.4.3. Regional Andean reconstructions. Nogami Fox and Bloom's (1994) modern snowline. Their cal- (1976) compared the modern snowline along the entire culated snowline depression over the tropical Andes Andean cordillera (103N}553S), based on the altitude of averaged 920$250 m. However, consistent with the con- existing glaciers shown on aerial photographs (pre-1976), clusions of Fox and Bloom (1994), the calculated !ELA with a Pleistocene snowline reconstructed using the cir- is ca 1200 m in the eastern cordillera of Peru and Bolivia, que-#oor method. The regional pattern disclosed that and 500}800 m to the west [see also Seltzer (1992, 1994), past and recent snowline surfaces rise westward between who calculated an ELA depression of only 300$100 m the northern equatorial Andes and northern Chile on the western slope of the Cordillera Real]. In the arid (ca 303S), at which latitude a shift occurs to northeast- ranges along the border of Bolivia and Chile, snowline ward-rising snowlines. The di!erence between Nogami's depression reached 1000}1200 m (Klein et al., 1999, two surfaces (!ELA) is less than 1000 m. Fig. 7). 1082 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

the LGM, an ice cap covered 70 km of the upper slopes above ca 3200 m altitude (Porter, 1979) (Fig. 9b). Two surface-exposure ages have been obtained for boulders at the surface of the youngest (Makanaka) drift and one for glacially abraded rock near the summit (20,300$2300, 18,900$800 yr, and 14,700$500 Cl yr, respectively; Dorn et al., 1991). These imply that the LGM occurred during marine oxygen-isotope stage 2 and that the summit was deglaciated by ca 15,000 yr ago. Based on the AAR method, the full-glacial ELA, corrected for ca 35 m of postglacial isostatic subsidence (2.5 m/10 yr; Porter, 1979), averaged ca 3780 m and had an eastward-sloping gradient. The minimum !ELA was ca 425 m. The July freezing isotherm now lies close to 4715 m, about 500 m above the summit, and likely approximates the ELA . Assuming a comparable relationship between July temperature and ELA" during the LGM, and using an AAR of 0.60$0.05, !ELA was 935$190 m.

3.4.3. Mt. Kinabalu, Borneo Glacial-erosional features below the summit of Mt. Kinabalu (4101 m; 6305N) on the island of Borneo were reported by Koopmans and Stau!er (1967) and Stau!er Fig. 8. Map of the Peruvian, Bolivian, and northern Chilean Andes showing pattern of regional ELAg surface (after Klein et al., 1999). Bold (1968). They estimated a glacial-age snowline of lines perpendicular to trend of isolines show arbitrary transects depic- 3735$75 m (12,000}12,500 ft) for the mountain based ted in Fig. 10 and included in Table 1. on the median-altitude method (Fig. 9c). The downvalley extent of ice was inferred from possible moraines at ca 3.4. Pacixc Islands 2835 and 3230 identi"ed on aerial photographs. The glacial deposits have not been dated. If the landforms are 3.4.1. Taiwan Shan, Taiwan correctly identi"ed as moraines and the higher one dates Along the Taiwan Shan, which forms the high crest of to the LGM, as inferred here, then based on the median- Taiwan, 62 peaks exceed 3000 m altitude; the highest, Yu altitude method, the average ELA" lay at ca 3665 m and Shan, reaches 3997 m. Kano (1934}35) identi"ed 35 cir- !ELA was at least 435 m. Koopmans and Stau!er esti- ques in the northern sector of these mountains, lying mated that the ELA lies at ca 4570$150 m, which mainly on the eastern side of the range. Moraines on would imply a !ELA of 905$150 m during the last cirque #oors and beyond cirque thresholds de"ne the glaciation. limits of former cirque and valley glaciers, as well as a small ice cap. Although no dates were available, Kano 3.4.4. Papua New Guinea assigned the landforms to the last glaciation and noted New Guinea is the only equatorial Paci"c island that cirque altitudes range from ca 3500 to 3730 m. He (5}93N) with numerous highlands (ca 3800}4500 m) that suggested that glaciers extended down to 3300 m on the generated a Late Pleistocene glacier cover. Most glacial- northern and eastern slopes of the highest peak, but were geologic studies have focused on the eastern half of the ca 300 m higher on the southern and western sides of the island (Papua New Guinea), which is more accessible range. than Irian Jaya to the west. Although no glaciers exist in Ono (1988, Fig. 1; Y. Ono, pers. comm. 2000) estimated the eastern highlands, LoK %er (1972) inferred that the the full-glacial snowline on Taiwan to lie at ca 3400 m, snowline lies at ca 4600 m, the reported altitude of the midway between an ELA" of 3350 m on Xue Shan and snowline in the glaciated areas of Irian Jaya (Verstappen, 3450 m on Yu Shan (to the south), derived using the 1964; Allison, 1976). The LGM snowline was determined cirque-#oor and glaciation threshold methods (Fig. 9a). by LoK %er (1972) using the arithmetic mean of the alti- Based on this "gure, and an inferred average crest altitudes of terminal moraines and the mean altitude of the tude of 3800 m, the !ELA was '400 m. catchment area, as well as the altitudes of the lowest cirque #oors (Fig. 9d), and is similar to estimates based 3.4.2. Mauna Kea, Hawaii on the glaciation threshold (LoK %er, 1971). Bowler et al. The summit of Mauna Kea (4206 m; 19350N) on the (1976) estimated the age of the LGM to be ca island of Hawaii lacks perennial glacier ice. However, at 18,000}16,000 C yr BP [21,648 (21,484) 21,313}18,972 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1083

(18,876) 18,784 cal yr BP] based on pollen-derived esti- 4.1. Estimates of modern ELAs mates and limiting C ages from two highland sites. Hope and Peterson (1976) reported that maximum de- Estimates of the modern snowline altitude probably pression of vegetation zones occurred 18,500}16,000 constitute the least-reliable component of the !ELA cal- C yr [22,000}18,972 (18,876) 18,784 cal yr] ago, and culations. In many cases, summit altitude is used as that in several areas substantial ice retreat had occurred a minimum altitude for the modern snowline, recogniz- by 14,500}14,000 yr [17,453 (17,369) 17,287}16,879 ing that the ELA must lie some unknown distance (16,792) 16,704 cal yr] ago. above a nonglacierized summit. In other cases, the alti- Mt. Giluwe (4368 m), a dome-shaped stratovolcano at tude of the modern snowline is estimated, either based on 63S latitude, was mantled by a large ice cap (188 km) limited "eld work (usually by noting the level of the during the last glaciation. Blake and LoK %er (1971) de- transient snowline during some part of the ablation sea- scribed end moraines that form concentric belts around son), by assuming a relationship between the snowline the mountain as low as 2750}3000 m, and alluvium, in- and the summer (July Northern Hemisphere) or annual terpreted as outwash, that overlies peat having an age of mean freezing isotherm based on radiosonde data, or by 23,600$1100 C yr (ca 27,400 cal yr) old. LoK %er (1972) using the mapped limits of snow and glacier cover depic- calculated an ELA" of 3500}3550 m (Fig. 9d), which ted on published topographic maps. In the latter ap- represents a minimum !ELA of ca 820}870 m. proach, generally it is assumed that snow/ice conditions For other areas of less-extensive glaciation lying be- shown on an array of maps that cover an area or region tween ca 53 and 93S latitude, LoK %er (1972) estimated the are essentially contemporaneous and represent a steady ELA" to lie between 3500 and 3700 m (Fig. 8d). However, state, and that the contour interval is suitable for relative- recent studies in the Sarawaged Range, using modern ly high-resolution interpolation between contours (e.g., topographic maps, suggest an ELA as low as 3400 m )30 m). (M. Prentice, pers. comm., 2000). If a snowline gradient The lack of a consistent and rigorous method of deter- existed across New Guinea, it was very gentle and cannot mining the modern snowline in the tropics could intro- be de"ned on the basis of available data. Adopting Ver- duce an error of up to several hundred meters in some of stappen's (1964) value of 4600 m for the modern snowline, the reported !ELA results. Furthermore, in a time when LoK %er (1972) concluded that snowline depression in global climate is warming, snowline values measured these areas during the last glaciation was ca 900}1100 m. several decades apart may di!er by tens of meters. Poten- tial errors may also occur when the ELA is inferred to lie at the level of the summer freezing isotherm, if this 4. Discussion assumed relationship is not always valid.

As summarized above, data bearing on snowline de- 4.2. Reconstructed Pleistocene ELAs pression in tropical latitudes are restricted to eastern Africa, Central and South America, and several Paci"c Data from 26 sites in the tropics permit reconstruc- islands. Di!erent methods have been used in deriving tions of full-glacial snowline (ELA") (Table 1; Fig. 10). ELA , ELA", and !ELA, but the results are not strictly However, for only 11 of these has the ELA" gradient been comparable. The AAR method, often regarded as the determined; nevertheless, in only two cases (IztaccmH huatl most reliable and consistent, has been used infrequently and Cordillera de MeH rida) may the lack of a calculated in tropical snowline studies. Altitude ratios (median alti- snowline gradient have introduced a signi"cant error tude, THAR, Alt}Ht}Acc) have been employed in most into the !ELA calculation. cases. In the few instances where lateral moraines were For most localities or regions, the age of the recon- used to de"ne former ELAs, the results were similar to structed ELA" is unknown, but it is commonly assumed those obtained using altitude ratios. Cirque-#oor alti- to equate with the `globala LGM (i.e., ca 21,000}15,000 tudes likely are the least-reliable method, especially with- C yr BP; ca 24,400}17,453 (17,369) 17,287 cal yr BP). out associated "eld studies and adequate dating control. Available radiocarbon dates (Table 1) generally are Although cirques may record some average regional level inadequate to verify whether the reconstruction rep- of glacial conditions (e.g., Porter, 1989), they do not resents full- or late-glacial (or even pre-LGM) condi- necessarily record a synchronous glacial event, such as tions. the LGM. Di!ering methodologies, therefore, introduce In South America, regional reconstructions for the potential variance to !ELA estimates, the magnitude of Andes of Peru, Bolivia, and Chile (Nogami, 1976; Fox which is di$cult to assess. In some temperate-latitude and Bloom, 1994; Klein et al., 1999), mainly using the studies (e.g., Meierding, 1982), methodological di!erences cirque-#oor method, add additional data for the tropics amounted to 100 m or more in calculated ELA"; compa- that supplement information from speci"c areas. Despite rable di!erences of 100}200 m probably should be ex- the acknowledged uncertainties and assumptions in- pected in tropical snowline studies. volved in these studies, including a potential error of 1084 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

Fig. 9. Sites in tropical Paci"c islands where ELA" and !ELA estimates have been obtained. (a) Taiwan Shan, Taiwan; (b) Mauna Kea, Hawaii; (c) Mt. Kinabalu, Borneo; (d) Papua New Guinea. See text for details.

$100}200 m at any site, they serve to emphasize the (1992). The second group (mainly south of 103S latitude) pattern, signi"cance, and overall consistency of regional includes regional reconstructions spanning segments of ELA gradients along and across mountain systems. the tropical (central) Andes between 73 and 223S latitude. The mean !ELA for the "rst set of data (n"12) is 4.3. Snowline depression 900$135 m. The Ruwenzori forms an outlier from the otherwise reasonably tight data set; if it is excluded, the For 12 of the tropical sites the modern ELA has been resulting mean is not statistically di!erent at 1" determined or estimated with su$cient con"dence that (925$115 m). !ELA at the LGM can be calculated. For the others, the The second group of data (n"8) is represented by one reported values are approximate, or only minimum esti- composite data set (Fox and Bloom, 1994; not plotted in mates (Fig. 10; Table 1). Fig. 10) and 7 transects parallel to ELA gradients at 23 To assess snowline depression, the data are divided latitude intervals, with values derived from Klein et al. into two groups. The "rst group (mainly north of 103S (1999, Fig. 7) (Fig. 8; Table 1). These data illustrate a sub- latitude) includes speci"c estimates at tropical sites in stantial regional range of !ELA values, especially perpen- Africa, the Americas, and Paci"c islands (Fig. 10). The dicular to the trend of the Andes. Through this sector of two summits of Kilimanjaro are considered a single lo- the cordillera, as far south as 183S latitude, the ELA" is cality, and average values are used for the Cordillera lowest, and !ELA values are greatest, on the Amazonian Blanca and the Cordillera Oriental reported by Rodbell slope. Although the !ELA varies regionally, Klein et al. S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1085

Fig. 10. (a) Modern and full-glacial ELAs for areas in the tropics (23.53N}23.53S latitude) that supported Pleistocene mountain glaciers. Where summits now lie below the snowline, minimum ELAs are shown as the summit altitude. In some cases ELA and (or) ELA" are shown with a range of values, primarily resulting from ELA gradients across mountains or mountain ranges. See text and Table 1 for details. (b) Full-glacial snowline depression (!ELA) for tropical mountains and mountain ranges that supported Pleistocene glaciers. Minimum values represent summits that lie below the modern snowline. In areas where ELA gradients exist, the number shown is the median of a range of values. Areas north of about 103S latitude have a mean !ELA of 900$135 m, whereas a regional study of Andes south of this latitude produced an estimated mean of 920$250 m (Klein et al., 1999).

(1999) calculated an average value of 920$250 m, the temperature depression, distance from precipitation mean being close to that of the "rst group of data dis- sources, or nonuniform changes in accumulation (i.e., cussed above (Fig. 10b). In the following discussion, the precipitation) and radiation (as in#uenced by cloudiness, value for the "rst data set will be used to represent global surface albedo, and topographic shading) in di!erent tropical snowline depression during the LGM. areas. The close similarity of derived !ELA values for A snowline depression of 900$135 m for tropical gla- tropical and temperate latitudes argues for a funda- ciers at the LGM is similar to estimates obtained for mental temperature control of !ELA and implies a reas- many temperate-latitude late Pleistocene mountain gla- onably consistent decline of air temperature during the ciers in both the Northern and Southern hemispheres glacial maximum in extra-polar alpine regions near mari- (e.g., Porter, 1975; Porter et al., 1983; Furrer, 1991). time sources of precipitation. Departures from the general average may be related to As discussed earlier, because sea level was ca 120 m any of a number of factors, including local variations in lower than today at the LGM (Fig. 3), a !ELA of 1086 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

900$135 m is equivalent to a !ELA of 780$135 m as precipitation falls as rain. In the Andes, for example, the a result of the changing ocean reference level. precipitation gradient is not uniform: rainfall reaches a maximum at ca 1000 m altitude, above which it de- creases. A drop in freezing level, therefore, may convert 5. Paleotemperature inferences from snowline data a larger percentage of the precipitation to snow and signi"cantly increase accumulation, without a change Paleotemperature values based on snowline depress- in net precipitation. In this way, a relatively uniform drop ion are frequently cited, but often uncritically. As Seltzer in air temperature might produce regionally variable (1994, p. 159) has emphasized, `climatic interpretations of glacier mass balances leading to nonuniform ELA ELA depression will always lack unique solutions be- depression. cause of the complexity of the problema (see also Porter, It is apparent that estimating paleotemperatures using 1977). The most common approach has been to calculate snowline data involves some substantial uncertainties. lowering of temperature (usually annual, summer, or Not only are there pitfalls in the use of di!erent method- July, but not always clearly stated) based on an assumed ologies, as well as signi"cant potential ranges of error, LGM atmospheric temperature lapse rate. Inferred lapse but the multiple factors that control glacier mass balance rates vary widely. For example, LoK %er (1970) applied and ELA do not permit an unequivocal and unique a lapse-rate range of !5 to !63C/km in the highlands paleotemperature solution. Nevertheless, the regional of New Guinea; Porter (1979) used a lapse rate of and global averages for tropical data suggest that the !5.33C/km for Mauna Kea, Hawaii; Clapperton (1987) simplistic lapse-rate approach may at least provide applied a lapse rate of !6.53C/km in Ecuador, a value a "rst-order approximation of regional and global tropi- also used by Rodbell (1992) and Seltzer (1987) for the cal temperature reduction at the LGM. Peruvian Andes; Wright (1983) and Osmaston (1989a) used a lapse rate of !73C/km in Peru and East Africa, respectively; and Fox and Bloom (1994) used a nonlinear 6. Other tropical LGM paleoclimate data and model lapse rate for the tropical Andes (!6.53C/km at simulations !3.5 km altitude to nearly !103C/km at 6 km). Hos- tetler and Mix (1999) adopted a `nominal tropical lapse The mean paleotemperature values based on snowline ratea of !5.5 3C/km. This range in lapse-rate values depression are in general accord with other paleotem- (!5.3 to !103C/km) by itself translates into a 4.53C perature estimates from tropical land areas that suggest range of values for temperature lowering, assuming full-glacial temperatures were substantially lower than a snowline depression of 1000 m. tropical warm-season SSTs (Fig. 11). Assuming a mean tropical lapse rate of !6$13C/km, The CLIMAP Project Members (1976, Fig. 3; 1981) and no change in precipitation, an average snowline derived a mean LGM tropical SST cooling of ca 1}33C, depression of 900$135 m translates into a full-glacial relative to modern (Fig. 11). More recent studies of the mean temperature depression of 5.4$0.83C. Using the tropical oceans report LGM SSTs that were 1.7$0.7 to full range (550}1400 m) of reported tropical !ELA 2.8$0.73C lower than present based on alkenone data (Table 1), the lapse-rate approach produces temperature (Lyle et al., 1992; Sikes and Keigwin, 1994; Bard et al., depressions ranging from 3.3 to 8.43C. Adjusted for sea- 1997) and Mg/Ca data (Lea et al., 2000), broadly consis- level lowering of 120 m, average temperature depression tent with mean CLIMAP estimates (Fig. 11). In contrast, is 4.7$0.83C. oxygen-isotope and Sr/Ca data from corals at Barbados In these simple, straightforward calculations, changes imply LGM SSTs 5 to 63C colder than now (Guilderson in the accumulation component of glacier mass balance et al., 1994). Crowley (2000), however, has questioned the have been ignored. Intuitively, it would seem to be an interpretation of the Sr/Ca data, noting that an SST important factor in some alpine regions. However, Sel- depression of this amount would leave ice-age corals at tzer (1994) assessed its importance and concluded that or below their limit of habitability. relatively large changes in precipitation would be re- A variety of terrestrial climate proxy data suggests that quired to a!ect ELA substantially. In tropical areas with LGM temperature lowering was greater over land than high precipitation, the limiting control on glacier extent over the ocean. Representative estimates (Fig. 11 and likely is the altitude of the 03C isotherm (Hostetler and Table 2), which are based on pollen data, noble-gas Clark, 2000). Of equal or greater importance may be the values in groundwater, and oxygen-isotope records in low seasonality of tropical climates, which leads to a rela- glacier ice, range from !5 to !123C (for additional tively constant height of the freezing isotherm (Klein data, see Farerra et al., 1999). Plotted with these temper- et al., 1999). At these latitudes, ELAs that lie above the ature estimates in Fig. 11 are values based on !ELA and level of the 03C isotherm are sensitive to accumulation !ELA. These average snowline-based estimates, at 1 changes, for above this level all precipitation falls as standard deviation, fall close to many of the other esti- snow. At lower altitudes, as temperature rises above 53C, mates of terrestrial temperature lowering, and therefore S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1087

Fig. 11. Estimates of temperature depression during the last glaciation (!Tg) at representative tropical sites, based on various climate proxies, compared with CLIMAP tropical sea-surface temperature (SST) di!erence between modern August and August 18,000 C yr BP (CLIMAP Project Members, 1976, Fig. 3). Temperature depression based on average tropical !ELA is shown by bold line and 1" range by light shading; value adjusted for sea-level lowering (!ELA) is shown by dashed bold line, and 1 " range by dark shading. Error estimates (1"), when reported, are shown by vertical dashed lines. 1 * eastern Atlantic Ocean alkenone data (Sikes and Keigwin, 1994); 2 * Indian Ocean alkenone data (Bard et al., 1997); 3 * central Paci"c Ocean alkenone data (Lyle et al., 1992); 4 * western and eastern Paci"c Ocean Mg/Ca data (Lea et al., 2000); 5 * Hawaiian foram !O data (Lee and Slowey, 1999); 6 * Barbados !O and Sr/Ca data (Guilderson et al., 1994); 7 * noble gases in Oman groundwaters (Weyhenmeyer et al., 2000); 8 * pollen in Guatemala lakes (Leyden et al., 1993); 9 * noble gases in Nigerian groundwaters (Edmunds et al., 1999); 10, 11, and 12 * pollen data from Panama, Brazil, and Ecuador, respectively (Colinvaux et al., 1996); 13 * noble gases in Brazil groundwaters (Stute et al., 1995); 14 * !O of Huascaran ice cap, Peru (Thompson et al., 1995); 15 * pollen data from Brazil (Colinvaux et al., 1996); 16 * noble gases in Nambia groundwaters (Stute and Talma, 1998).

Table 2 Representative tropical SST and terrestrial paleoclimatic data for the LGM

Data No. Latitude Location !SST (3C) Data Reference

SST data 1 03 E Atlantic Ocean !1.8 Alkenone Sikes and Keigwin (1994) 2 203N}203 S Indian Ocean !1.7$0.7 Alkenone Bard et al. (1997) 3 0357N Central Paci"c Ocean !1 Alkenone Lyle et al. (1992) 4 0319'}2348N W & E Paci"c Ocean !2.8$0.7 Mg/Ca Lea et al. (2000) 5 21.36N Hawaiian Islands !2 Foram !O Lee and Slowey (1999) 6 13315N Barbados !5 Sr/Ca Guilderson et al. (1994)

Terrestrial Data 7 23330N Oman !6.5$0.6 Noble gas Weyhenmeyer et al. (2000) 8 16355N Guatemala !6.5 to !8 Pollen Leyden et al. (1993) 9 11330'}13330N Nigeria !6 Noble gas Edmunds et al. (1999) 10 93N Panama *!5 Pollen Colinvaux et al. (1996) 11 03 Brazil !6 Pollen Colinvaux et al. (1996) 12 63S Ecuador !6 Pollen Colinvaux et al. (1996) 13 73S Brazil !5.4$0.6 Noble gas Stute et al. (1995) 14 93S Peru !8 to !12 !O of ice Thompson et al. (1995) 15 21}223 S Brazil !6 to !9 Pollen Colinvaux et al. (1996) 16 24330S Namibia !5.3$0.5 Noble gas Stute and Talma (1998)

support the conclusion that LGM surface land temper- coupled ocean}atmosphere model of intermediate com- atures typically were depressed more than SSTs of adjac- plexity (Ganopolski et al., 1998). The simulation showed ent oceans. Similar values have been reported in that tropical land areas cooled an average of 4.63C (com- a modeling study of LGM climate that used a global parable to ELA, Fig. 11), whereas SSTs between 203N 1088 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 and 203S cooled by 3.33C in the Atlantic, 2.43C in the information about his snowline studies in South Paci"c, and 1.33C in the Indian Ocean. America Helpful comments by reviewers Alan Gillespie, Farerra et al. (1999) used a variety of climate-proxy Michael Prentice, and Geo! Seltzer are greatly appreci- data to calculate average cold-season cooling at the ated. LGM that ranged from !2.5 to !3.03 at present sea level to ca !63 at 3000 m altitude, suggesting nonlinear lapse rates. Such a relationship is apparent in a recent rise in the tropical freezing level, which is closely linked to an References increase in SST (Diaz and Graham, 1996). Whereas the observed tropical SST change was ca 0.2}0.33C, temper- Allison, I., 1976. Glacier regimes and dynamics. In: Hope, G.S., Peter- ature change at the level of the freezing isotherm, based son, J.A., Radok, U., Allison, I. (Eds.), The Equatorial Glaciers of on a standard lapse rate of !63/km, was ca 0.63C, or New Guinea. Balkema, Rotterdam, A.A, pp. 39}59. two to three times as great (Crowley, 2000). Modi"cation Andrews, J.T., 1975. Glacial Systems. Duxbury Press, North Scituate, MA. of tropical lapse rates at the LGM could well be related Bard, E., Rostek, F., Sonzogni, C., 1997. Interhemispheric synchrony of to a change in the mean altitude of the tropical inversion, the last deglaciation inferred from alkenone palaeothermometry. in turn associated with a shift in the position and strength Nature 385, 707}710. of subtropical high-pressure cells (e.g., Hostetler and Bard, E., Hamelin, B. Fairbanks, R.G., 1990. U-Th ages obtained by Clark, 2000). mass spectrometry in corals from Barbados: sea level during the past 130,000 years. Nature 346, 456}458. Betts and Ridgway (1992) evaluated several factors Betts, A.K., Ridgway, W., 1992. Tropical boundary layer equilibrium that may be related to the discrepancy between tropical in the last ice age. Journal of Geophysical Research 97, snowline depression and SSTs. They computed that a de- 2529}2534. crease in average tropical SST by 23 at the LGM and an Blake, D.H., LoK %er, E., 1971. Volcanic and glacial landforms on Mount increase in tropical sea-surface pressure by 14 mbar (the Giluwe, territory of Papua and New Guinea. Geological Society of America Bulletin 82, 1605}1614. result of a 120 m fall in sea level) would account for an Bowler, J.M., Hope, G.S., Jennings, J.N., Singh, G., Walker, D., 1976. 800 m depression of the freezing isotherm. Signi"cantly, Late Quaternary climates of Australia and New Guinea. Quater- this result is consistent with recent (post-CLIMAP) esti- nary Research 6, 359}394. mates of SST lowering cited above (1.7}2.83C) and the Broecker, W.S., 1995. Cooling the tropics. Nature 376, 212}213. mean !ELA (780$135 m) derived in the present study. Broecker, W.S., 1997. Mountain glaciers: recorders of atmospheric water vapor content? Global Biogeochemical Cycles 11, 589}597. The mean !ELA value is also close to !ELAs (ca Bush, A.B.G., Philander, S.G.H., 1998. The role of ocean}atmosphere 800}870 m) reported by Hostetler and Clark (in press) interactions in tropical cooling during the last glacial maximum. based on mass-balance modeling of several tropical gla- Science 279, 1341}1344. ciers in New Guinea, Hawaii, Africa, and the Andes. Clapperton, C.M., 1987. Glacial geomorphology, Quaternary glacial This review has focussed on the derivation and assess- sequence and palaeoclimatic inferences in the Ecuadorian Andes. In: Gardiner, V. (Ed.), International Geomorphology 1986, Part II. ment of tropical paleosnowlines. The limited quantity Wiley, Chichester, pp. 843}870. and limitations of the existing information point to the Clapperton, C.M., 1993. Quaternary Geology and Geomorphology of need for additional studies to enlarge and improve the South America Elsevier, Amsterdam. data set. However, even with adequate data, translating Clark, D.H., Clark, M.M., Gillespie, A.R., 1994. Debris-covered glaciers snowline depression into estimates of land-surface tem- in the Sierra Nevada, California, and their implications for snowline reconstructions. Quaternary Research 41, 139}153. perature depression is likely to persist as a challenging CLIMAP Project Members, 1976. The surface of the ice-age Earth. problem. Among the important questions yet to be an- Science 191, 1131}1137. swered are: (1) what was the degree to which the mass CLIMAP Project Members, 1981. Seasonal reconstruction of the balance of tropical glaciers at the LGM was in#uenced Earth's surface at the last glacial maximum. Geological Society of by a change in precipitation? and (2) were LGM lapse America Map and Chart Series MC-36. Colinvaux, P.A., Liu, K.-B., de Olivera, P., Bush, M.B., Miller, M.C., rates di!erent than today, and were they linear or nonlin- Kannan, M.S., 1996. Temperature depression in the lowland tropics ear? At present, suitable evidence to answer these ques- in glacial times. Climatic Change 32, 19}33. tions remains elusive. Crowley, T.J., 2000. CLIMAP SSTs re-revisited. Climate Dynamics 16, 241}255. Diaz, H.F., Graham, N.E., 1996. Recent changes in tropical freezing heights and the role of sea surface temperature. Nature 383, Acknowledgements 152}155. Dorn, R.I., Phillips, F.M., Zreda, M.G., Wolfe, E.W., Jull, A.J.T., The initial draft of this paper was written while I enjoy- Donahue, D.J., Kubik, P.W., Sharma, P., 1991. Glacial chronology ed hospitality of Nick Shackleton in the Godwin Labor- of Mauna Kea, Hawaii, as constrained by surface-exposure dating. atory, Cambridge University, and Lloyd Keigwin in the National Geographic Research and Exploration 7, 456}471. Edmunds, W.M., Fellman, E., Goni, I.B., 1999. Lakes, groundwater and McLean Laboratory, Woods Hole Oceanographic Insti- palaeohydrology in the Sahel of NE Nigeria: evidence from hydro- tution. I thank Matsuo Tsukada for translation of a Ja- geochemistry. Journal of the Geological Society, London 156, panese article used in this review, and Geo! Seltzer for 345}355. S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1089

Fairbanks, R.G., 1989. A 17,000-year glacio-eustatic sea level record: Atlas of Glaciers of the World. U.S. Geological Survey Professional in#uence of glacial melting rates on the Younger Dryas event and Paper 1386-I, pp. I11}I30. deep-ocean circulation. Nature 342, 637}642. Hurni, H., 1989. Late Quaternary of Simen and other mountains in Farerra, I., Harrison, S.P., Prentice, I.C., Ramstein, G., Guiot, J., Bar- Ethiopia. In: Mahaney, W.C. (Ed.), Quaternary and Environmental tlein, P.J., Bonne"lle, R., Bush, M., Cramer, W., von Grafenstein, U., Research on East African Mountains. Balkema, Rotterdam, A.A., Holmgren, K., Hooghiemstra, H., Hope, G., Jolly, D., Lauritzen, S.- pp. 105}120. E., Ono, Y., Pinot, S., Stute, M., Yu, G., 1999. Tropical climates at Jordan, E., Hastenrath, S. L., 1998. Glaciers of Ecuador. In: Williams, the Last Glacial Maximum: a new synthesis of terrestrial palaeocli- R. S., Ferrigno, J. G. (Eds.), Glaciers of South America * Satellite mate data I. Vegetation, lake-levels and geochemistry. Climate Image Atlas of Glaciers of the World. U. S. Geological Survey Dynamics 15, 823}856. Professional Paper 1386-I, pp. I31}I50. Fox, A.N., Bloom, A.L., 1994. Snowline altitude and climate in the Kano, T., 1934-1935. Contribution to the glacial topography of the Peruvian Andes (5}173S) at present and during the latest Pleis- Tugitaka Mountains, Formosa. Chirigaku Hyoron (Geographical tocene glacial maximum. Journal of Geography 103, 867}885 Review of Japan, Association of Japanese Geographers) 10, (in Japanese). 606}623, 688}707, 816}835, 990}1017; 11, 244}263, 258-263 (in Furrer, G., 1991. 25,000 Jahre Gletschergeschichte dargestellt an Japanese, with English summary). einigen Beispilen aus den Schweizer Alpen. Vierteljahrsschrift der Klein, A.G., Seltzer, G.O., Isacks, B.L., 1999. Modern and last local Naturforschenden Gesellschaft in ZuK rich 5, 52 pp. glacial maximum snowlines in the central Andes of Pe9 ru, Bolivia, Ganopolski, A., Rahmstorf, S., Petoukhov, V., Claussen, M., 1998. and northern Chile. Quaternary Science Reviews 18, 63}84. Simulation of modern and glacial climates with a coupled global Koopmans, B.N., Stau!er, P.H., 1967. Glacial phenomena on Mount model of intermediate complexity. Nature 391, 351}356. Kinabalu, Sabah. Malaysia Geological Survey (Borneo Region) Gosse, J.C., Klein, J., Evenson, E.B., Lawn, B., Middleton, R., 1995. Bulletin 8, 25}35. Beryllium-10 dating of the duration and retreat of the last Pinedale Kurowski, L., 1891. Die HoK he der Schneegrenze. Geogra"sche Abbhan- glacial sequence. Science 268, 1329}1333. dlungen 5, 119}160. Guilderson, T.P., Fairbanks, R.G., Rubenstone, J.L., 1994. Tropical Lea, D.W., Pak, D.K., Spero, H.J., 2000. Climate impact of late Quater- temperature variations since 20,000 years ago: modulating inter- nary equatorial Paci"c sea surface temperature variations. Science hemispheric climate change. Science 263, 663}665. 289, 1719}1724. Hastenrath, S., 1973. On the Pleistocene glaciation of the Cordillera de Lee, K.E., Slowey, N.C., 1999. Cool surface waters of the sub- Talamanca, Costa Rica. Zeitschrift fuK r Gletscherkunde und Glazial- tropical North Paci"c Ocean during the last glacial. Nature 397, geologie 9, 105}121. 5412}5514. Hastenrath, S., 1974. Spuren pleistozaK ner Vereisung in den Altos Leyden, B., Brenner, M., Hodell, D.A., Curtis, J.H., 1993. Late Pleis- de Cuchumatanes, Guatemala. Eiszeitalter und Gegenwart 25, tocene climate in the Central American lowlands. Geophysical 25}34. Monograph 78, 165}178. Heine, K., 1976. Schneegrenzdepression, Klimaentwicklung, Bodenero- Livingstone, D.A., 1962. Age of deglaciation in the Ruwenzori range, sion und Mensch im zentralmexikanischen Hochland im juK ngerern Uganda. Nature 194, 859}860. PleistozaK n und HolozaK n. Zeitschrift fuK r Geomorphologie (Suppl.) Livingstone, D.A., 1975. Late Quaternary climatic change in Africa. 24, 160}176. Annual Review of Ecology and Systematics 6, 249}280. Heine, K., 1978. Neue Beobachtungen zur Chronostratigraphie der LoK %er, E., 1970. Evidence of Pleistocene glaciation in East Papua. mittelwisconsinzeitlichen Vergletcherungen und BoK den mexikanis- Australian Geographical Studies 8, 16}26. cher Vulkane. Eiszeitalter und Gegenwart 28, 139}147. LoK %er, E., 1971. The Pleistocene glaciation of the Saruwaged Range, Heine, K., 1984. Comment on `Pleistocene glaciation of Volcano Ajusco, central Mexico, and comparison with the standard Mexican Territory of New Guinea. The Australian Geographer 11, 463}472. LoK %er, E., 1972. Pleistocene glaciation in Papua and New Guinea. glacial sequencea by Sidney E. White and Salvatore Valastro, Jr. Quaternary Research 22, 242}246. Zeitschrift fuK r Geomorphologie (Suppl) 13, 32}58. Helmens, K.F., Kuhry, P., Rutter, N.W., Van Der Borg, K., De Jong, Lowell, T.V., Heusser, C.J., Anderson, B.G., Moreno, P.I., Hauser, A., F.M., 1996. Warming at 18,000 yr BP in the tropical Andes. Quater- Heusser, L.E., SchluK chter, C., Marchant, D.R., Denton, G.H., 1995. nary Research 45, 289}299. Interhemispheric correlation of Late Pleistocene glacial events. Herd, D.G., 1974. Glacial and volcanic geology of the RumH z-Tolima Science 269, 1541}1549. volcanic complex, Cordillera Central, Colombia. Ph.D. Disserta- Lyle, M.W., Prahl, F.G., Sparrow, M.A., 1992. Upwelling and produc- tion, University of Washington. tivity changes inferred from a temperature record in the central Herd, D.G., 1982. Glacial and volcanic geology of the RumH z-Tolima equatorial Paci"c. Nature 355, 812}815. volcanic complex, Cordillera Central, Colombia. Publicaciones Mahaney, W.C., 1990. Ice on the Equator. Quaternary Geology of GeoloH gicas Especiales del Ingeominas 8, 1}48. Mount Kenya, East Africa. William Caxton, Ltd., Sister Bay, WI. HoK fer, H. v., 1879. Gletscher und Eiszeitstudien. Sitzungberichte der Martin, P.S., 1964. Paleoclimatology and a tropical pollen pro"le. In: Academie der Wissenschaften in Wien, Mathematische-naturwis- Report on the VI International Congress on the Quaternary, War- senschaftliche Klasse 79. saw, 1961, Vol. II, pp. 319}323. Hope, G.S., Peterson, J.A., 1976. Palaeoenvironments. In: Hope, G.S., Meier, M.F., Post, A.S., 1962. Recent variations in mass net budgets of Peterson, J.A., Radok, U., Allison, I. (Eds.), The Equatorial Glaciers glaciers in western North America. International Association of of New Guinea. Balkema, Rotterdam, A.A, pp. 173}205. Scienti"c Hydrology 58, 63}77. Horn, S.P., 1993. Postglacial vegetation and "re history in the ChirripoH Meierding, T.C., 1982. Late Pleistocene equilibrium-line altitudes in the PaH ramo of Costa Rica. Quaternary Research 40, 107}116. Colorado Front Range: a comparison of methods. Quaternary Hostetler, S.W., Clark, P.U., 2000. Tropical climate at the last glacial Research 18, 289}310. maximum inferred from glacier mass-balance modeling. Science, Nogami, M., 1972. The snow line and climate during the last glacial 290, 1747}1750. period in the Andes mountains. Daiyonki-Kenkyu (The Quaternary Hostetler, S.W., Mix, A.C., 1999. Reassessment of ice-age cooling of the Research) 11, 71}80 ( in Japanese). tropical ocean and atmosphere. Nature 399, 673}676. Nogami, M., 1976. Altitude of the modern snowline and Pleistocene Hoyos-Patin o, F., 1998. Glaciers of Colombia. In: Williams, R.S., snowline in the Andes. Geographical Reports of Tokyo Metropoli- Ferrigno, J.G. (Eds.), Glaciers of South America * Satellite Image tan University 11, 71}86. 1090 S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091

Ono, Y., 1988. Last glacial snowline altitude and paleoclimate of the Schubert, C., Valastro, S., 1974. Late Pleistocene glaciation of PaH ramo eastern Asia. Daiyonki-Kenkyu (The Quaternary Research) 26, de La Culata, north-central Venezuelan Andes. Geologische Run- 271}280 ( in Japanese). dschau 63, 516}538. Orvis, K.H., Horn, S.P., 2000. Quaternary glaciers and climate on Seltzer, G.O., 1987. Glacial history and climatic change in the central Cerro ChirripoH , Costa Rica. Quaternary Research 54, 24}37. Peruvian Andes. M.S. Thesis, University of Minnesota. Osmaston, H., 1989a. Glaciers, glaciations and equilibrium line alti- Seltzer, G.O., 1992. Late Quaternary glaciation of the Cordillera Real, tudes on Kilimanjaro. In: Mahaney, W.C. (Ed.), Quaternary and Bolivia. Journal of Quaternary Science 7, 87}98. Environmental Research on East African Mountains. Balkema, Seltzer, G.O., 1994. Climatic interpretation of alpine snowline Rotterdam, A.A, pp. 7}30. variations on millennial time scales. Quaternary Research 41, Osmaston, H., 1989b. Glaciers, glaciations and equilibrium line alti- 154}159. tudes on the Ruwenzori. In: Mahaney, W.C. (Ed.), Quaternary and Sikes, E.L., Keigwin, L.D., 1994. Equatorial Atlantic sea surface tem- Environmental Research on East African Mountains. Balkema, perature for the last 30 kyr: a comparison of U Y , !0 and Rotterdam, A.A, pp. 31}104. foraminiferal assemblage temperature estimates. Paleoceanography "strem, G., 1966. The height of the glaciation limit in southern British 9, 31}45. Columbia and Alberta. Geogra"ska Annaler 48A, 126}138. Stau!er, P.H., 1968. Glaciation of Mount Kinabalu. Geological Society PeH weH , T.L., Reger, R.D., 1972. Modern and Wisconsinan snowlines in of Malaysia Bulletin 1, 63. Alaska. Proceedings of the 24th International Geologic Congress, Stuiver, M., Reimer, P.J., 1993. Extended C data based and Section 12, pp. 187}197. revised CALIB 3.0 C age calibration program. In: Stuiver, M., Phillips, F.M., Zreda, M.G., Smith, S.S., Elmore, D., Kubik, P.W., Long, A., Kra, R.S. (Eds.), Calibration 1993. Radiocarbon 35, Sharma, P., 1990. Cosmogenic chlorine-36 chronology for glacial 215}230. deposits at Bloody Canyon, eastern Sierra Nevada. Science 248, Stuiver, M., Reimer, P.J., Bard, E., Beck, J.W., Burr, G.S., Hughen, K.A., 1529}1532. Kromer, B., McCormac, G., Van Der Plicht, J., Spurk, M., 1998. Porter, S.C., 1975. Glaciation limit in New Zealand's Southern Alps. INTCAL 98 radiocarbon age calibration, 24,000}0 cal BP. Arctic and Alpine Research 7, 33}37. Radiocarbon 40, 1041}1083. Porter, S.C., 1977. Present and past glaciation threshold in the Stute, M., Forster, M., Frischkorn, H., Serejo, A., Clark, J.F., Schlosser, Cascade Range, Washington, U.S.A.: topographic and climatic P., Broecker, W.S., Bonani, G., 1995. Cooling of tropical Brazil (53C) controls, and paleoclimatic implications. Journal of Glaciology 18, during the last glacial maximum. Science 269, 379}383. 101}116. Stute, M., Talma, A.S., 1998. Glacial temperatures and moisture trans- Porter, S.C., 1979. Hawaiian glacial ages. Quaternary Research 12, port regimes reconstructed from noble gases and O-18, Stampriet 161}187. aquifer, Namibia. In: Isotope Techniques in the Study of Environ- Porter, S.C., 1981. Glaciological evidence of Holocene climatic mental Change. International Atomic Energy Agency, Vienna, change. In: Wigley, T.M.L., Ingram, M.J., Farmer, G. (Eds.), pp. 307}318. Climate and History. Cambridge University Press, Cambridge, Swanson, T. W., Porter, S. C., 1999. Surface-exposure ages for alpine pp. 82}110. glaciation in the southern North Cascade Range. American Geo- Porter, S.C., 1989. Some geological implications of average Quaternary physical Union Abstracts with Program, H12D-05. glacial conditions. Quaternary Research 32, 245}261. Thompson, L.G., Mosley-Thompson, E., Davis, M.E., Lin, P.N., Her- Porter, S.C., Pierce, K.L., Hamilton, T.D., 1983. Late Wisconsin moun- nderson, K.A., Cole-Dai, J., Bolzan, J.F., Liu, K.B., 1995. Late tain glaciation in the western United States. In: Porter, S.C. (Ed.), glacial stage and Holocene tropical ice core records from Huas- Late Quaternary Environments of the United States: the Late caraH n, Peru. Science 269, 46}50. Pleistocene. University of Minnesota Press, Minneapolis, pp. Thouret, J.-C., Van der Hammen, T., Salomons, B., JuvigneH , E., 1996. 71}111. Paleoenvironmental changes and glacial stades of the last 50,000 Potter, E.C., 1976. Pleistocene glaciation in Ethiopia: new evidence. years in the Cordillera Central, Colombia. Quaternary Research 46, Journal of Glaciology 17, 148}150. 1}18. Richmond, G.M., 1986. Stratigraphy and correlation of glacial Torsnes, I., Rye, N., Nesje, A., 1993. Modern and Little Ice Age equilib- deposits of the Rocky Mountains, the Colorado Plateau and rium-line altitudes on outlet valley glaciers from Jostedalsbreen, the ranges of the Great Basin. Quaternary Science Reviews 5, western Norway: an evaluation of di!erent approaches to their 99}127. calculation. Arctic and Alpine Research 25, 106}116. Rind, D., Peteet, D., 1985. Terrestrial conditions at the last glacial VaH squez-Selem, L., 1998. Glacial chronology of IztaccmHhuatl volcano, maximum and CLIMAP sea-surface temperature estimates: are central Mexico, based on cosmogenic Cl exposure ages and teph- they consistent? Quaternary Research 24, 1}22. rochronology. American Quaternary Association Program and Rodbell, D.T., 1992. Late Pleistocene equilibrium-line reconstructions Abstracts of the 15th Biennial Meeting, Puerto Vallarta, Mexico, in the northern Peruvian Andes. Boreas 21, 43}52. 5}7 September, p. 174. Rohling, E.J., Fenton, M., Jorissen, F.J., Bertrand, P., Ganssen, G., Verstappen, H.T.H., 1964. Geomorphology of the Star Mountains. Caulet, J.P., 1998. Magnitudes of sea-level lowstands of the past Nova Guinea, Geology 5, 101}158. 500,000 years. Nature 394, 162}165. Weyhenmeyer, C.E., Burns, S.J., Waber, H.N., Aeschbach-Hertig, W., Schubert, C., 1974. Late Pleistocene MeH rida Glaciation, Venezuelan Kipfer, R., Loosli, H.H., Matter, A., 2000. Cool glacial temperatures Andes. Boreas 3, 147}152. and changes in moisture source recorded in Oman groundwaters. Schubert, C., 1984. The Pleistocene and recent extent of the glaciers of Science 287, 842}845. the Sierra Nevada de MeH rida, Venezuela. Erdwissenschaftliche For- Weyl, R., 1956. Eiszeitlich Gletscherspuren in Costa Rica (Mit- schungen 18, 269}278. tleamerika). Zeitchsrift fuK r Gletscherkunde und Glazialgeologie 3, Schubert, C., 1998. Glaciers of South America}Glaciers of Venezuela. 317}325. In: Williams, R. S., Ferrigno, J. G. (Eds.), Satellite Image Atlas of White, S.E., 1962. Late Pleistocene glacial sequence for the west side of Glaciers of the World; Glaciers of South America. U.S. Geological IztaccmH huatl, Mexico. Geological Society of America Bulletin 73, Survey Professional Paper 1386-I, pp. I1}I10. 935}958. Schubert, C., Clapperton, C.M., 1990. Quaternary glaciations in the White, S.E., 1981. Equilibrium line altitudes of late Pleistocene northern Andes (Venezuela, Colombia and Ecuador). Quaternary and recent glaciers in central Mexico. Geogra"ska Annaler 63, Science Reviews 9, 123}135. 241}249. S.C. Porter / Quaternary Science Reviews 20 (2001) 1067}1091 1091

White, S.E., Valastro Jr., S., 1984. Pleistocene glaciation of Young, J.A.T., Hastenrath, S.L., 1991. Glaciers of Africa. In: Williams Volcano Ajusco, central Mexico, and comparison with the standard Jr., R.S., Ferrigno, J.G. (Eds.), Glaciers of the Middle East and Mexican glacial sequence. Quaternary Research 21, 21}35. Africa * Satellite Image Atlas of Glaciers of the World. U.S. Wright Jr., H.E., 1983. Late-Pleistocene glaciation and climate around Geological Survey Professional Paper 1386-G, pp. G49}G70. the JunmHn Plain, central Peruvian Highlands. Geogra"ska Annaler 65A, 35}43.