Climate, , and the morphology of the

David R. Montgomery Greg Balco Sean D. Willett Department of Geological Sciences, University of Washington, Seattle 98195-1310, USA

ABSTRACT Large-scale topographic analyses show that hemisphere-scale climate variations are a ®rst-order control on the morphology of the Andes. Zonal atmospheric circulation in the creates strong latitudinal precipitation gradients that, when incor- porated in a generalized index of intensity, predict strong gradients in erosion rates both along and across the Andes. Cross-range asymmetry, width, hypsometry, and maximum elevation re¯ect gradients in both the erosion index and the relative dominance of ¯uvial, glacial, and tectonic processes, and show that major morphologic features cor- relate with climatic regimes. Latitudinal gradients in inferred crustal thickening and struc- tural shortening correspond to variations in predicted erosion potential, indicating that, like tectonics, nonuniform erosion due to large-scale climate patterns is a ®rst-order con- trol on the topographic evolution of the Andes.

Keywords: , erosion, tectonics, climate, Andes.

INTRODUCTION we argue for the ®rst-order importance of cycle. Some studies have attribut- The presence or absence of rang- large-scale climate zonations and resulting dif- ed local variations in structural, metamorphic, es at the global scale is determined by the lo- ferences in geomorphic processes to the mor- and geomorphic characteristics of the central cation and type of plate boundaries. Other fac- phology of mountain ranges. Andes to erosion (Gephart, 1994; Masek et al., tors become important in the evolution of 1994; Horton, 1999), but none has considered individual mountain systems. In particular, TECTONIC AND CLIMATIC SETTING variations in erosional mass removal at the spatially variable erosion resulting from cli- OF THE ANDES scale of the entire . mate gradients may localize exhumation and The in¯uences of climate, erosional pro- The highly variable climate of the Andes deformation in orogens and thereby in¯uence cesses, and tectonics on orogen morphology re¯ects its position transverse to hemisphere- the geologic structure and morphology of may be deconvolved in the Andean orogen be- scale, Hadley cell-driven precipitation regimes mountain ranges (Beaumont et al., 1991; Zei- cause it is a hemisphere-scale, north-south± (Fig. 1). In the Intertropical convergence zone tler et al., 1993; Avouac and Burov, 1996). oriented range with large gradients in temper- (10ЊN±3ЊS), both sides of the range receive Ϫ1 Earlier studies of climatic geomorphology ature and rainfall across a single convergent annual rainfall exceeding 2 m´yr . In the sub- have limited relevance to this issue because margin. Uplift of the Andes began ca. 25 Ma, equatorial northern Andes (3ЊS±15ЊS), oro- they simply classify into normal (¯uvi- concomitant with accelerated convergence be- graphic interception of the trade winds deliv- Ϫ1 al), glacial, and arid zones and generally de- tween the Nazca and plates ers Ͼ2 m´yr of rainfall to the Amazon side Ϫ1 pict an alpine area as a single morphoclimatic (Allmendinger et al., 1997). Early theories of of the range and Ͻ0.2 m´yr to the Paci®c zone that crosscuts multiple low-elevation formation of the Andes emphasized crustal side, and westerly winds produce the opposite morphoclimatic zones (Tricart and Cailleux, growth by magmatic processes, but estimates relationship in the temperate south of 1972). Even though the large-scale morphol- of structural shortening and evidence for sym- 33ЊS. The central part of the range (15ЊS± ogy of mountain belts must record the com- metric paleomagnetically de®ned rotation on 33ЊS) is in the subtropical belt of , bined effects of climatic and tectonic process- the northern and southern ¯anks of the Alti- where there is little precipitation on either side es, only a few studies explore climatic factors plano gave rise to the hypothesis that the var- of the range, or on the high of the (Willett et al., 1993; Brozovic et al., 1997). iable size and thickness of the range result . These major climate boundaries in Here we show that geomorphometric pa- from nonuniform crustal shortening, with the Andes are not dependent upon orographic rameters such as cross-range asymmetry, hyp- maximum shortening and consequent thick- effects, but are robust features of the general sometry, and maximum elevation of the An- ening at the center of the Andean (Is- circulation in the Southern Hemisphere, and des re¯ect the in¯uence of zonal climate acks, 1988; Gregory-Wodzicki, 2000). How- therefore may be considered a priori condi- regimes on the nature and intensity of ero- ever, direct structural shortening estimates are tions under which the mountain range sional processes. In addition, we show that limited to the Eastern and Suban- developed. consequent latitudinal gradients in erosion po- dean and thrust belt. In the Altiplano and tential are correlated with the crustal mass dis- Western Cordillera, crustal structures are - TOPOGRAPHIC ANALYSIS tribution and inferred orogenic shortening of scured by sedimentation or volcanism, and We focus on four aspects of the large-scale the range, suggesting an ambiguity in the cur- global positioning system measurements (Nor- geomorphology of the Andes: (1) a general- rent interpretation of crustal mass distribution abuena et al., 1998; Kendrick et al., 1999) ized index of erosion intensity based on re- as the result of variations in the tectonic en- may be in¯uenced by short-term strain accu- gional slope and ¯uvial discharge, (2) cross- vironment. On the basis of these observations mulation associated with the -zone range asymmetry, (3) regional hypsometry

᭧ 2001 Geological Society of America. For permission to copy, contact Copyright Clearance Center at or (978) 750-8400. Geology; July 2001; v. 29; no. 7; p. 579±582; 3 ®gures. 579 Figure 1. A: Maximum (dark line) and mean (gray area) elevation in 1؇ bins. Red circles are elevations of modern perennial snowline and blue circles are lowest elevation of extent, both from Schwertfelder (1976). B: Topography and convergence velocity. Vectors headed in open circles denote long-term velocity of Nazca and plates relative to (DeMets et al., 1994); those headed in closed circles denote global positioning system (GPS) velocities at coastal sites, relative to stable South America (Norabuena et al., 1998; Kendrick et al., 1999). C: Mean annual precipitation, overlain on shaded-relief map of western South America. D: False-color image of South America showing areas with steep slope in yellow, high precipitation in blue. Red pixels have

calculated IE above 90th percentile relative to all pixels in image. E: Cross-range asymmetry, de®ned to be fraction of range volume above level that drains to west: values greater than 0.5 (lighter shade of gray) indicate that bulk of range is west of divide.

(the elevation distribution of the topography), determined by summing the annual precipita- IE values shows that the zone of maximum and (4) the relationship between the maximum tion (P) over the matrix of upslope grid cells predicted erosion is on the eastern side of the elevation and the perennial snowline. We used each of drainage area A: range in the northern Andes and on the west- topography from the global 30 s GTOPO30 ern side in the southern Andes. Only small, digital elevation model; topography, slope, IE ϭ ΄΅͸ PAii S. (1) localized areas of high IE are predicted in the and ¯ow direction from the 1 km HYDRO1K central Andes (Fig. 1D). DEM; and mean annual precipitation digitized We used this simple approach because (1) it from Hoffmann (1975). For purposes of our is not clear which process formulation is most Cross-Range Asymmetry analysis, we de®ned the eastern boundary of appropriate for modeling landscape-scale ero- We de®ned a cross-range asymmetry index the Andes as the approximate limit of sion rates across 1 km grid cells in which net as the ratio of the volume of the topography or older units mapped on continental-scale erosion re¯ects an aggregation of ®ner scale above sea level on the west side of the divide geologic maps (UNESCO, 1978). effects from multiple, interacting processes; to that of the entire range within a given lat- (2) vegetation and land use, which cannot be itude band (Fig. 1E). Between 2ЊS and 42ЊS Erosion Index predicted from digital elevation models, com- most of the range is to the east of the drainage Rates of ¯uvial and hillslope erosion are plicate simple relationships between precipi- divide, whereas south of 42ЊS most of the governed by processes characterized by dif- tation and erosion rate; (3) erosion models at range is west of the drainage divide. North of ferent erosion laws, but the net large-scale this scale inherently require calibration be- 2ЊS, the inclusion of the areas draining to the erosional potential of a landscape increases cause slopes calculated from coarse-resolution Sea with areas draining to the Pa- with precipitation, drainage area, and slope. grids are gentler than actual gradients (Zhang ci®c places most of the range on the Thus, we evaluated large-scale patterns in ero- and Montgomery, 1994); and (4) data on dif- west side of the drainage divide. Cross-range sion potential by using a simple parametric ferences in erosivity due to soil type and par- asymmetry tracks latitudinal variations in measure of erosional intensity (IE) based on ent lithology generally are not available at the moisture delivery due to prevailing wind the product of local slope (S) and discharge scale of interest. In the Andes, the pattern of directions.

580 GEOLOGY, July 2001 1957), here the aggregate pattern is geograph- ically consistent with variations in erosional processes. In the northern Andes, concave-up hypsometric curves, which are characteristic of ¯uvially dissected landscapes, re¯ect the dom- inance of ¯uvial erosion in a wet tropical cli- mate. In the southern part of the range, have selectively eroded high elevations, creat- ing a shoulder in the hypsometric curves. In the central Andes, the hypsometric curves are nearly linear, with a convexity imposed by the relatively ¯at hypsometry at elevation of the Altiplano. This form describes a weakly incised tectonic wedge and mechanically limited pla- teau, implying that ¯uvial incision is ineffective relative to . This strong associa- tion of hypsometry with climatically driven variations in geomorphologic processes dem- onstrates that both the nature of the dominant erosional mechanism and its rate relative to tec- tonic uplift are fundamental to the overall to- pographic expression of the Andes.

Figure 2. Normalized hypsometric curves for 3؇ latitude bins of Andes; curve color Maximum Elevation corresponds to location in northern (red), central (yellow), and southern (blue) Andes. The tendency for the elevation of the pe- rennial snowline to track mountaintops is well Hypsometry variations suggest that ¯uvial, tectonic, and gla- known (Mill, 1892), but the causal basis for Hypsometric curves, which show the pro- cial processes, respectively, dominate the mor- this relationship and the relative ef®ciency of portions of a landscape at different normalized phology of the range in these different zones. glacial erosion remain more controversial. In elevations, have strikingly different, but - Although individually these hypsometric the Andes, the maximum elevation and the ally consistent, shapes in the northern, central, curves could re¯ect different developmental snowline are greater than 5 km north of 30ЊS, and southern Andes (Fig. 2). These latitudinal stages in a classical interpretation (Strahler, and both decrease toward the pole thereafter, such that only a small fraction of the topog- raphy remains above the snowline at any lat- itude (Fig. 1A). The distinct shoulder to the hypsometry of the southern Andes also de- scends with the perennial snowline. The cor- respondence betweeen total relief and snow- line elevation supports the hypothesis that higher rates of erosion in glacial and perigla- cial environments effectively limit the relief of mountain ranges (Brozovic et al., 1997). This implies that high topography cannot persist at high latitudes and that the high Andes termi- nate at 35ЊS in part because they intersect the perennial snowline at this latitude.

DISCUSSION The observation that topographic changes along the Andes correspond with large-scale variations in climate suggests that zonal climate patterns affect the orogen-scale morphology of the Andes. This conclusion has implications both for general understanding of landscape evolution and for speci®c large-scale tectonic Figure 3. A: Volume of Andes above sea level calculated from 1؇ latitude bins. B: Excess erosion -rate, relative to largest 1؇ bin, is required to explain volume difference under uniform tectonic interpretations for the Andes. For example, Is convergence. We calculated required latitudinal variation in erosion rates under constant tec- acks (1988) neglected the effect of mass re- tonic convergence by calculating missing mass above sea level in each 1؇ latitude bin as VXS␣/ moval by erosion when inferring latitudinal At, where VXS is excess volume in given bin compared to largest bin (14؇±15؇S), A is bin area, variations in convergence from a crustal mass .t is time (taken to be 25 m.y.), and /( ), where 2.7 g´cm؊3 and 3.3 g´cm؊3 balance. However, the latitudinal variations in ؍ ␳m ؍ ␳c ␳m ؊␳c ␳c؍␣ Note that because of selection of strictly east-trending bins for analysis, region between 13؇S and 17؇S, where range trends northwest rather than north, has anomalously large volume in mean IE also track variations in present excess each bin. C: Mean annual precipitation. D: Mean erosion intensity index value. crustal volume in the Andes (Fig. 3). An ex-

GEOLOGY, July 2001 581 treme interpretation of this correlation, taking Speci®cally, we see three archetypes of climatic Gephart, J.W., 1994, Topography and subduction ge- the opposite assumptions to the analysis of Is- control on large-scale landscape form: (1) nor- ometry in the central Andes: Clues to the me- chanics of a noncollisional orogen: Journal of acks (1988), would hold net convergence con- mal ¯uvial erosion in the northern Andes where Geophysical Research, v. 99, p. 12 279±12 288. stant from 45ЊSto5ЊN and explain the current high precipitation rates maintain a narrow Gregory-Wodzicki, K.M., 2000, Uplift history of width of, and crustal volume in, the Andes as mountain range; (2) tectonic dominance of the central and northern Andes: A review: the result of latitudinal variations in erosion landscape form in the central Andes, where Geological Society of America Bulletin, v. 112, p. 1091±1105. rate. In this case, the observed distribution of there is little erosion except in big valleys, Hoffmann, J.A.J., 1975, Atlas climatico de America crustal volume requires latitudinal variations in leading to crustal thickening by tectonic wedge del Sur: Ginebra, World Meteorological Organi- average erosion rate during the lifetime of the propagation, the formation of a mechanically zation, scales 1:10 000 000 and 1:5 000 000, Andes of Ͻ2 mm´yrϪ1 (excluding the glaciated limited plateau, and linear hypsometry; and (3) 28 p. Horton, B.K., 1999, Erosional control on the ge- southern Andes). These required variations are glacial land sculpting that preferentially erodes ometry and kinematics of thrust belt devel- within the range of reasonable erosion rates and the highest ground in the southern Andes, re- opment in the central Andes: Tectonics, v. 18, broadly correlate with the independently deter- sulting in an excess of elevation at the glacial p. 1292±1304. mined I values. Hence, it is reasonable to sug- limit and a systematic decline in maximum el- Isacks, B.L., 1988, Uplift of the central Andean plateau E and bending of the Bolivian orocline: Journal of gest that climatically in¯uenced gradients in evation toward the pole. The coincidence of Geophysical Research, v. 93, p. 3211±3231. erosion rates contribute to the latitudinal vari- low inferred erosion rates (on the basis of cal- Jorden, T.E., Isacks, B.L., Allmendinger, R.W., ation in range width and crustal volume. culated IE values) in the latitudes and the Brewer, J.A., Ramos, V.A., and Ando, C.J., Other local structural variations may be the greatest width of the Andes suggests that lack 1983, Andean tectonics related to geometry of result of variable erosion. Range-wide changes of erosion plays an important role in mass ac- subducted : Geological Society of America Bulletin, v. 94, p. 341±361. in geology are broadly consistent with this idea: cumulation in the mountain belt. If the devel- Kendrick, E.C., Bevis, M., Smalley, R.F., Jr., Cifuen- the crystalline rocks of the northern and south- opment of the Altiplano re¯ects the mechanical tes, C., and Galban, F., 1999, Current rates of ern Andes re¯ect deeper exhumation, and the limit to crustal thickening (Pope and Willett, convergence across the central Andes: Esti- preserved sedimentary and volcanic cover of 1998), then its existence implies that tectonic mates from continuous GPS observations: Geo- physical Research Letters, v. 26, p. 541±544. the central Andes indicates that exhumation thickening has outpaced erosional mass remov- Masek, J.G., Isacks, B.L., Gubbels, T.L., and Field- there has been minimal. For example, the East- al; its position in the global desert belt suggests ing, E.J., 1994, Erosion and tectonics at the ern Cordillera and Subandean zone of that this dominance of tectonic shortening was margins of continental : Journal of have undergone 2±6 km of exhumation since possible, at least in part, because of the arid Geophysical Research, v. 99, p. 13 941±13 956. Mill, H.R., 1892, The realm of nature: An outline 10 Ma north of 19ЊS (Benjamin et al., 1987), climate of this latitudinal band. We conclude of physiography: , Charles Scrib- but Ͻ1 km since that time to the south (Masek that the large-scale distribution of crustal mass ner's Sons, 366 p. et al., 1994; Gregory-Wodzicki, 2000). This in a mountain belt is controlled by not only Norabuena, E., Lef¯er-Grif®n, L., Mao, A., Dixon, difference is immediately apparent in the trun- tectonic shortening, but also by the type and T., Stein, S., Sacks, I.S., Ocola, L., and Ellis, M., 1998, Space geodetic observations of cation of the prominent fold-and-thrust belt by intensity of erosional processes. Nazca±South America convergence across the the apparent erosional ``bite'' in the area with central Andes: Science, v. 279, p. 358±362. ACKNOWLEDGMENTS high rainfall to the north. Pope, D.C., and Willett, S.D., 1998, A thermal- Supported by a Hertz Foundation Graduate Fel- mechanical model for crustal thickening in the We are not arguing that tectonic variations lowship (to Balco) and in part by National Science central Andes driven by ablative subduction: are unimportant in the evolution of the Andes. Foundation grant EAR-9903157. We thank Peter Geology, v. 26, p. 511±514. In fact, the major changes in the topography Zeitler and Bryan Isacks for their constructive cri- Schwertfelder, W., editor, 1976, Climates of Central and mass distribution in the Andes also corre- tiques of the manuscript. and South America: New York, Elsevier, 532 p. Strahler, A.N., 1957, Quantitative analysis of wa- late with tectonic parameters such as the ori- REFERENCES CITED tershed geomorphology: American Geophysi- entation and dip of the subducting slab (Jorden Allmendinger, R.W., Jordan, T.E., Kay, S.M., and cal Union Transactions, v. 38, p. 913±920. Isacks, B.L., 1997, The evolution of the Alti- et al., 1983; Gephart, 1994) and major geologic Tricart, J., and Cailleux, A., 1972, Introduction to plano-Puna Plateau of the central Andes: An- provinces (Gansser, 1973). For example, the climatic geomorphology: New York, St. Mar- nual Review of Earth and Planetary Sciences, 's Press, 274 p. high volume segment of the central Andes be- v. 25, p. 139±174. UNESCO, 1978, Tectonic map of South America: tween 10ЊS and 30ЊS corresponds to the steeply Avouac, J.-P., and Burov, E.B., 1996, Erosion as a Brasilia, Commission for the Geological Map driving mechanism of intracontinental moun- dipping segment of the Nazca slab, particularly of the World, scale 1:5 000 000. tain growth: Journal of Geophysical Research, when one recognizes that our analysis likely Willett, S.D., 1999, and orography: The v. 101, p. 17 747±17 769. effects of erosion on the structure of mountain overestimates the mass between 13ЊS and 17ЊS Beaumont, C., Fulsack, P., and Hamilton, J., 1991, belts: Journal of Geophysical Research, (Fig. 3). This situation complicates the inter- Erosional control of active compressional or- v. 104, p. 28 957±28 981. ogens, in McClay, K.R., ed., : pretation of crustal mass distribution as the re- Willett, S., Beaumont, C., and Fullsack, P., 1993, New York, Chapman and Hall, p. 1±18. sult of one or the other of these two seemingly Mechanical model for the tectonics of doubly Benjamin, M.T., Johnson, N.M., and Naeser, C.W., covariant forcings. We view tectonics and ero- vergent compressional orogens: Geology, 1987, Recent rapid uplift in the Bolivian An- v. 21, p. 371±374. sion as a coupled system, with potential for des: Evidence from ®ssion-track dating: Ge- Zeitler, P.K., Chamberlain, C.P., and Smith, H.A., feedback between climate-driven erosion and ology, v. 15, p. 680±683. 1993, Synchronous anatexis, metamorphism, Brozovic, N., Burbank, D.W., and Meigs, A.J., tectonic forcing on shallow crustal processes and rapid denudation at Nanga Parbat (Paki- 1997, Climatic limits on landscape develop- (Willett, 1999), or even deep, mantle processes. stan Himalaya): Geology, v. 21, p. 347±350. ment in the northwestern Himalaya: Science, Zhang, W., and Montgomery, D.R., 1994, Digital Could the large accumulation of mass in the v. 276, p. 571±574. Altiplano, which seems to have been permitted elevation model grid size, landscape represen- DeMets, C., Gordon, R.G., Argus, D.F., and Stein, tation, and hydrologic simulations: Water Re- by slow surface erosion, also affect the dynam- S., 1994, Effect of recent revisions to the geo- sources Research, v. 30, p. 1019±1028. ics of the subduction zone? magnetic reversal time scale on estimates of current plate motions: Geophysical Research Manuscript received November 7, 2000 Letters, v. 21, p. 2191±2194. CONCLUSIONS Revised manuscript received March 7, 2001 Gansser, A., 1973, Facts and theories on the Andes: Manuscript accepted March 15, 2001 Our results support the idea that global cli- Geological Society of London Journal, v. 129, mate patterns in¯uence orogen morphology. p. 93±131. Printed in USA

582 GEOLOGY, July 2001