Sunteți pe pagina 1din 12
A Critical Evaluation of Late Tertiary Accelerated Uplift Rates for the Eastern Cordillera, Central Andes of Bolivia Mark H, Anders, Kathryn M. Gregory-Wodzicki, and Marc Spiegelman Lamoat-Doherty Earth Observatory of Columbia University, Palisades, New York 10964-8000, US.A. (e-mail: manders@ldeo.colurabia.edu) ABSTRACT. Many anicles have cited a study of apatite and aircon ssion-track ages from the Eastern Cordillera of the central Andes, Bolivia, as evidence for 2 marked acceleration of uplift rates over the last 40 myr. However, we show that an artifact of the graphical technique used to interpret the data, the common-variable problem, dives the apparent inerease. In order to determine what can he inferred about exhumation history from the dats, we developed 2 simple quantitative forward mode! to calculate the coupled apatite and zircon age profiles as a function of height and then used Markov Chain-Monte Carlo techniques to estimate the family of solutions that provide reasonable fits to the data. This model improves on linear regression analysis of clevation-age plots, which treats apatite and zircon data as independent systems. Results suggest that both exponential or stepunction velocity models ft the data well. We prefer the exponential model, however, because it appears to be more consistent with uplift history 2s inferred from paleoelevation data. In this model, the exhumation rate doubles every 20-35 m.yr. from about 120 m/Ma at 50 Ma to around 400-750 m/Ma for the present. Thus, we concur with the original interpretation of an acceleration in the exhumation rate, but our inference is based on a more physical model along with paleoclevation data. This interpretation differs from other reinterprecations of the data based on linear regression analysis that suggest that exhumation began at 40 Ma and then increased sharply sometime between 30 and 10 Ms, More fission track data, especially zircon data, are needed to dstinguish between these two scenarios. Introduction ‘The uplift history of the central Andes isof interest to both climatic and tectonic studies. The Andes affect global climate because they form the only barner to atmospheric citculation in the Southern Hemisphere and contain the second highest plateau on Earth, Their uplift has probably changed pat- tems of precipitation, seasonal heating, and upper atmosphere flow and increased rates of chemical ‘weathering (Ruddiman and Kutzbach 1989; Hay 1996, Broccoli and Manabe 1997}. Raymo et al. (1988) suggest that they may have contributed to dlobal cooling since the Late Miocene. The formation of the heart of the Andes, the ~4000-m-high, 350-km-wide central Andean Pla- teau, is somewhat enigmatic because it is associ sted with a noncollisional plate margin. Debate centers on the relative importance of crustal short- ening versus magmatic addition in thickening the crust and on the role of crustal processes versus mantle processes in producing uplift. Because ele- ‘Manuscript reeived March 17,2000, acepted july 17,2001 vation is a function of the thickness, density, and strength of the lithosphere, the uplift history of the Andes provides important constraints for these debates. However, few quantitative studies of Andean up: liftexist (for a review, see Gregory-Wodzicki 2000} Of these studies, one of the most influential is Ben- jamin et al. (1987), which measured apatiteand zir- con fission-track ages for the Huayna Powosi and Zongo plutons and associated metasediments from the Zongo River Valley in the Eastem Cordillera of Bolivia figs. 1, 2) Fission-track age estimates the time since a min: eral cooled through its closure temperature, Below the closure temperature, fission tracks are stable, while above this temperature, tracks are annealed, that is, healed. For apatite, the closure age is re: poreed as 110° 10°C. However, if a rock cools slowly, some annealing can occur between tem peratures of 60° and 100 (Gallagher et al. 1998} This temperature range is called the partial an- {The fumal of Geno, 2002, volume 110». $810} © 2002by The Univesity of Chea. AM ight reserved 00221374/2002/1001-005801.00 89 Py Mow Figure 1, map as processed by the Comell Andes Project. nealingzone. The partial annealing zoneis less well defined for zircon, most estimates range from be- tween 175° and 350°C (Gallagher et al. 1998). By providing information on cooling, fission- wack ages can be used to constiain the timing of exhumation and tectonic denudation in orogenic helts. For example, Beajamin et al. (1987) interpret the fission-track data from Bolivia to suggest that “uplift” rates of the Eastern Cordillera have in: creased “exponentially” since 40 Ma, beginning at rates of 100-200 m/Ma, then increasing signifi- cantly around 10-15 Ma to teach rates of 700 m/ Ma by 3 Ma (fig. 3. Many subsequent studies have secepted this model of accelerating exhumation. ‘However, there are some serious problems with the interpretation of this date besides those uncer- tainties typically associated with the analysis of fission-track ages |ie., Hurford 1991, Gallagher et al. 1998). First, as England and Molnar (1990) dis- cuss, fission-track data reflect the removal of over ying material, that is, exhumation, not surface up- lift. Benjamin et al. (1987) make the assumption that an increase in the exhumation rate reflects an increase in surface uplife, but thisis not necessarily true, Studies of drainage basins suggest that erosion rateincreases with relief (slope) (Ahnert 1970, Pinet and Souriaw 1988; Pazzaglia and Brandon 1996). In terms of climate, Ritter (1988} suggests that erosion ANDERS ET AL Location of Benjamin et al. (1987) fission-track stuly on base of USGS 30 are-second digital elevation rates reach @ maximum in semiarid climates (200-400 mm) and again in very wet climates {51000 mm}, as a function of vegetation type. Also, storm size may affect erosion rate (Gregory and Chase 1994). Thus, though uplift generally in creaseserosion rate, not all increases of erosion rate are due to uplift. Masek et al. (1994 further criticize the conclu: sions of Benjamin et al. (1987). They point out that the calculated uplift rates (more properly, exhu. ‘mation rates) in their figure 3 are correlated with the ages of the samples, resulting in a plot of 1/ time versus time. Thus, they suggest that the plot does not provide evidence of steadily increasing ex- humation rates. In this study, we first expand on the brief dis cussion of Masek et al. (1994) and show how figure 3 in Benjamin et al. (1987) is simply a representa: tion of the variance in determining closure age rather than an indication of accelerating echuma tion rates. We then explore what the data of Ben- jamin et al, (1987) do reveal about exhumation. We begin by discussing linear regression analysis of age-elevation plots, a standard analysis of apatite and zircon data. We then offer a new, and we will argue improved, approach: a forward model that uses both the apatite and zireon data together to constrain 4 single exhumation history. The results Joural of Geology from this model are compared with paleoelevation, data from the region to check for consistency. Assessing the Late Tertiary Exhumation Rate of the Cental Andes Determining Exhumation Rates. The wechniques used by Benjamin etal. (1987) to assess exhumation rates from fission-track ages, described in detail in Benjamin (1986), are complex enough to warrant some discussion hore. A suite of samples was col- lected between 1520 and 5100 m, and closure ages were determined for apatite and zircon separates. Wherever possible, apatite and zircon closure ages were analyzed from the same sample, thus provid- ing two control points. The data were then plotted 4s elevation versus age for each mineral and each rock type (ig. 2 of Benjamin etal. 1987; our fg. 2) Ttis the manipulation of these to primary data to produce figure 3 in Benjamin etal. (1987; our ig. 3) and the resulting interpretation of this plot that, ‘ve will show Ieads to erroneous conclusions about exhumation rates, Forthe 11 apatite ages, Benjamin et al. (1987) calculate exhumation rates using a spe- tal case of the mineral-pairs method. First, they subtract surface temperature specified as 10°C. from the closure temperature of apatite, specified as MOC, to get the toral temperature differential, They then divide the temperature differential by the geothermal gradient, which they assume to be # constant 30°C/km through time, to get the depth to the closure threshold. Thus, for each apatite sample, the amount of exhumation is 4 constant, 3.33 km, or 100°C/30°C/km, The X-axis of figure 3, the mean length of time for this amount of exhumation, is determined by dividing the closure age, t., by 2. The Y-axis, ex- humation rate, or “uplift” rate, R, is then deter- mined by dividing the depth to the closure thresh- old, 3.33 km, by the closure age ra 3.33. km ax x= R= a Note that a plot of rate versus age is simply a plot of L/X versus X (ie, a hyperbola], which is the basic problem with this technique (see below). For four samples with both apatite and zircon ages, the total amount of exhumation is calculated by subtracting the apatite annealing temperature from the zircon annealing temperature (specified as 210°C} to get the temperature difference, and then ACCELERATED UPLIFT RATES o oe ‘nat os) Figure 2. Apatite and zircon fission-tiack cooling ages plotted against elevation of samplingsite. Data from Ben: jamin et al, (1987). Solid lines are linear regression lines with age as the dependent variable, numbers are exhu mation rates calculated as the inverse of the regression slope. PAZ = partial annealing zone. Note datum 2380 ‘mwas adjusted irom Benjamin etal. (1987) for the correct mean age. dividing by the geothermal gradient, also yielding 3.33 km. Uplift rate is decermined by dividing 3.33 km by the difierence in the sample ages: yalatla (2) ‘The relief method was used for five pairs of sam- ples separated by at east 500m. In this calculation, uplift rate is calculated by dividing the difference between the two sample elevations, ht, and fh, by the difference in the sample ages: 3) For further information about these calculations, see Benjamin (1986). We do not discuss the five relief-method caleulations further, three of which appear as open boxes in Figure 3, because they do not bear on the crux of our analysis. Artifect in Assessing Exhumation Raies. Before we discuss in deail how the calculation deseribed above can lead to erroneous conclusions about ex- o Mow humation rates, we will present a generalized dis- cussion of what we call the common-variable prob- Jem, Pearson (1897) was the first to point out the spurious effects of a common variable in assessing the correlation coefficient; since then numerous other authors have discussed the effect (see Atchley ct al. 1976; Atchley and Anderson 1978, Kenney 1982; Jackson and Somers 1991; Schlager et al. 1998}. However, the problem associated with com- mon-variable plots is not restricted to just the cal- culation of a correlation coefficient [Schlager et al 1998} but also includes a distortion associated with calculating the regression slope. Anders et al. (1987) showed graphically and numerically that the slope ‘of a linear regression in log (¥/X) versus log. (X) space is dependent on the variance in the ¥ vari- able, As the variance in X increases for any given value of Y, the slope of « linear regression drives toward ~1. It follows from this that if Yin such a Jog plot is fixed toa single value, then any variance in the value of X will be plotted on a line whose slope is —1. This is an intrinsic property of plotting, «ny data in the form of log (¥/X] versus log (X}. In the case of a nonlog plot, the variance in X for 4 single value of ¥ will plot on a hyperbolic curve of 1/X. As with the case of the log plot, variance toward the low values of X will produce high values of ¥/X, which are a sole function of the degree of variance. For example, if X is the measurement of the age of something, any age determinations that underestimate the true age will result in higher val- ues of ¥/X. Any regression or interpretation of data plotted on this kind of graph will lend itself to in- terpreting high variance in X as an intrinsic prop- erty of an independent measurement of ¥/X. The plot of exhumation rate versus mean age as shown in figure 3 of Benjamin etal. (1987; our fi. 3)encounters the same difficulties when a common, value of time is used for both the independent var- fable and R, the dependent variable. The only dif ference from the generalized case is that for some of the estimates the age, or X term, is multiplied bby 2 when used tocaleulace the ¥ term, which only affects the shape of the resulting plot and not the distortion discussed above. Benjamin et al. (1987, p. 682) concluded from the shape of che daca in their figure 3 (our fig. 3) that the “curve indicates that uplift rates have been in- creasing exponentially for the past 40 m.y.” This is an erroneous conclusion that, unfortunately, has heen carried on in the literature concerning the up- lift history of the central Andes. Transforming a horizontal line in ¥ versus X space into ¥/X versus X space results in all values of ¥/X falling on a hyperbolic curve. In the case of Benjamin et al. ANDERS ET AL Figure 3. Apatite and zircon fission track data from Benjamin (1986) and Benjamin et al. (1987). See text or Benamin (1986) for technique used to convert Asion track data shown in figure 2 to data in this plot. Open circles represent apatite samples analyzed using the mod ified mineral-pair method. Solid circles represent paired apatite and zircon samples analyzed using the mineral: pair method. Open squares are apatite pairs or zircon pairs analyzed using the relief method. Numbers eorre spond to topographic elevation of the sampling site as listed in Benjamin et al. (1987). Datum 2380 m is as reported in Benjamin ot al. (1987}. Solid line curve is defined by equation y = 3.33 km/2x. See text for further discussion of the meaning of this curve. (1987), the hyperbolic curve is defined as y = 3.33/2x. Itfollows that all values of upliit rate must fall on this hyperbolic curve. Note that in figure 3 all the data for sampling elevation based on the mineral pairs technique except those with solid dots fall on this curve labeled y = 3.33/2x. These samples, taken from elevations 2040, 2680, and 3450 mare off the curve to the right. This is be cause their associated apatite ages were added to the calculated mineral-pair mean age. Subtracting the apatite age moves the data back on the curve, Datum 2380 m was simply miscalculated. Recal: culating this datum from Benjamin (1986) and sub- tracting the apatite age adjusts it to the 3.33/2x ‘Another way to look at these transformed dats is to suppose for a moment that the length of time for any sample to pass through 3.33 km to the sur face has a particular value. However, a measure ment of the closure age has, likeall measurements, Joural of Geology 4 variance associated with it. Any underestimation, of the closure age will result in a skew toward high, values of rate—the opposite is true for overesti- mates of age. The underestimated variance will ap- pear as an “exponential” increase in rate. The greater the variance, the greater the presumed rate of echumation. What the Data Reveal about Uplift History Linear Regression Models. One way to avoid the common-variable problem is to plot sample ele- vation versus age, as in figure 2, rather than ele- vation divided by age versus age. The typical ap- proach to interpreting these plotsis to fit rezression lines to the data using elevation as the independent (X) variable, The inverse of the slope can then be used to estimate the mean exhumation rate if sev. eral assumptions are made, including (1) the iso- therms remained parallel to the geoid; (2) the depth to the isotherms remained constant, that is, the geothermal gradient remained constant; and (3) the samples moved as a coherent block. We must assume that the isotherms were parallel to the geoid or the elevation data would not be proportional to the amount of material eroded and thus the slope of the elevation-age regression line would not represent the exhumation rate (fig 2) This assumption and the assumption of a constant geothermal gradient are problematic in areas with high relief and exhumation rates over 500 m/Ma, suchas the southern Alps, because the near-surface isotherms are influenced by the shape of surface topography (Stiwe et al. 1994; Brown and Sum- merfield 1997).In the Zongo Valley, these assump- tions are probably not a problem for the zircon ata because of the high closure temperature but could be a problem for the apatite data. “The Eastern Cordillera began to uplift as early as the Eocene, but the present high relief in the Zongo River Valley wes mostly created after3 Ma (Kennan etal, 1997) and thus postdates the apatite ages. Dur- ing the time period when the samples were passing, the apatite closure threshold, the mean surface to- pography probably dipped to the NE, as it does to- day. If the apatite-closing isotherm paralleled the mean surface, instead of the geoid, chen the apatite data would overestimate the true exhumation rate because the amount of exhumation represented by 2 given elevation would be overestimaced (Stuwe et al, 1994), In order to address this possibility, we corrected the elevation of each apatite data point of Benjamin ct al, (1987], assuming that the Miocene closure threshold followed the modem mean surface dip of ACCELERATED UPLIFT RATES 9 4 to the NE (Safran 1998), We could not correct, the data of Crough (1983] because sample locations were not given. In reality, if the apatice closure threshold did dip, it was probably at 2 shallower angle than the surface dip because isotherms follow topography in a dampened fashion (Stiwe et al 1994), By using a maximum dip value, our corree- tion will provide a lower limit on the mean ex- humation rate for the Zongo River Valley. ‘Another caveat is that the ssion-track data were collected over a horizontal distance of ~30 km, whereas ideally, ageelevation data are collected from a single vertical profile. Safran (1998) plotted fission-track age versus down-valley distance for theZongo data of Beniamin et al. (1987), along. with additional data from the nearby Challana and Ta ‘quesi River Valleys, and found that age tended to decrease downstream. Based on thermal modeling. shesuggesis that the best explanation for this trend is that the exhumation rate increases downstream. Thus, when estimating average exhumation rate from age-elevation plots, one must keep in mind that the value not only represents a temporal av- erage but also a spatial average. If we apply a lincar regression analysis to the Benjamin et al. (19871 data, the older zircons from between 63 and 101 Ma define a line with a slope of 30 m/Ma (r' = 21, s = 14 Ma, F = 2), and the younger zircon samples from between 25 and 46 Ma define a regression line with a slope of 150 mj Ma(r? = .65, 5 = 4 Ma, F = 12; fig. 2}. As Benjamin et al. (1987) point out, che interpretation of the older zircons is somewhat ambiguous. The ages could represent a time of slow exhumation and cooling, or they could represent a time of partial annealing in a paleopartial annealing zone, in which case the curve would not contain accurate information about the exhumation rate (although they would still provide lower bounds on the age) Benjamin etal. (1987) suggest that the low slope of the curve argues for the paleopartial annealing zone hypothesis. Whether the samples represent a pa Ieopartial annealing zone or not, the break in slope indicates che onset of significant unroofing at around 40 Ma. ‘The apatite samples from Benjamin et al. (1987), along with three apatite samples from the Huayns Potosi batholith from Crough (1983), define a re ‘gression line with a slope of 540 m/Ma (r? = 32, 5 =3 Ma, F = 7; fig. 2) If the elevations of Ben: jamin et als (1987) data are correeted for a4° slope ‘on the apatite closure isotherm, then the exhu- mation rate is reduced to 320 m/Ma. This mean exhumation rate is similar to the corrected exhu: mation rate calculated by Kennan (2000) of 300 m| Mow Ma; he also normalized the altitude of the apatite data of Benjamin et al. (1987} to reflect the mean slope of the modern topography. Thus, based on linear regression modeling, the mean exhumation rate for the Zongo Valley after ~15 Ma is 320-540 m/Ma. Kennan (2000) suggests that the difference in slope between the zircon and apatite samples imply that exhumation ratesinereased at 25 + 5 Ma. Ma- sek et al, (1994) derivea similar conclusion by plot- ting denudstion rate versus topography of the sam- pling area, although they suggest that the increase in rates occurred later, around 10-15 Ma. According, to the slope equality test of Sokal and Robli (1995, ». 498), the slope of the apatite age-elevation re- gression for the raw apatite data is significantly steeper than the younger zircon age-elevation re- gression, but the slope for the comrected data io sta tistically indistinguishable (P< 0.05) from the younger zircon age-elevation regression. This re- mains true even i, in the correction of the eleva- tion data, the dip of the apatite isotherm is damp- ened to 2°. Thus, the conclusion that the exhumation rote increased sometime during the Late Oligocene or Miecene depends on the validity of the assumption that the isotherms remained par- allel to the geoid, Forward Modeling of Coupled Zircoa-Apatite Data. Linear regression analysis of age-elevation plots is a valid approach to interpreting exhumation his- tory from fission-track data, but treating the apatite and zircon data as independent fails to exploit po- tentially useful information: the fact that the zir- cons mast record the same cooling history as the apatites. In response to this problem, we developed 4 simple quantitative forward model to calculate the coupled apatite and zircon age profiles asa func- tionof height and then used Markov Chain-Monte Carlo techniques w explore the family of solutions that provide reasonable fits to the data. Rather than provide a single best fit, these techniques calculate the logical consequences of model assumptions consistent with the dat To generate a one-dimensional forward model for coupled apatite and zircon ages in 2 coherent crustal block undergoing exhumation, we start by defining the function Zi) = J virwde, a which is simply the distance that the block moves, in time ¢ with velocity V(t). Because we assume that the velocity isalways positive (no oscillations), ANDERS ET AL Figure 4. Definitions of parameters in the forward ‘model fis the block with thicknessequal tothesampled fission-track transect. The block is uplifted through = distance (3 + + I)h, where gh is the distance between the apatite and zircon closure threshold, 2, and 2, re spectively, and 7h isthe distance between z,and the base ‘of the enustal block, then Zit) is a monotonic inereasing function that can be uniquely inverted such that given a known, distance Z we can inier a time ¢ by ©= Ti, (5) where T is the inverse function of Z. Given this framework, we can specify a general set of equa tions to describe the motions of a point in a co- herent block with a height b= 3580 m, equal to the thickness of the sampled fission-track transect, moving through a set of fixed closure depths z, for Zircon and z, for apatite (fig 4). Ifthe relative po- sition of the point within the block is I such that 1 =0 at the bottom of the block and J = | at the top, then we can define three specific times for this point: t, would be the total time it takes the point to move to the base of the zircon zone z,, tswould be the time it takes to move to the base of the apatite zone 2,, and t would be the total time i¢ takes to travel to its final position above the base of the apatite zone, Figure 4 shows the basic ge ‘ometry of the model and defines the distance be tween the zircon and apatite closure depths as 2,.72=Bh (6 and the distance between the apatite closure threshold and the base of the crustal block as yh. The model assumes that the distance between the apatite and zircon closure depths does not change with time but otherwise make no assumptions about the spacing between these horizons (e.z,, by Joural of Geology assuming the thermal gradient). Given these defi- nitions, it is clear that Zit) = ald, Zt) = hla + (UMN, "7 Zits) = hly +8+ 1), and therefore we can define the apatite age to be ty ~ tnd the zircon age tober, = ts ~ ty or wll: P) = Tihly + 6 + IN] — TIA[s + (1 2, ell; P) = Thhly + 6 + I~ TALL ~ I), (8) where Lis the relative height in the block and P is the set of adjustable parameters 6, 7, and any pa- rameters in the velocity model. To close the model requires inputting a specific function for Vit) and computing Z and T. Here, we have chosen two end-member velocity models for illustration (although many moreare possible}. The first is a simple exponential growth model with vie wer, ro} which yields 10) Te} = +n( +1} ay where « is a constant growth rate and V, is the initial veloeity, Both models assume that the zircon nd apatite closure thresholds occur at discrete depths, z. and z,, and the velocity Vit} is measured relative to these horizons. The second model assumes that the velocity is 4 step function with % ocicr Vil = j fy Vt AV tot where ¥V, is the initial velocity, which jumps to ¥¥ AV at some time ¢ (which is a free parameter in this model). Given Vit), we find ACCELERATED UPLIFT RATES 9s | Ye tee 21) = (3) We tave-e} oe o

S-ar putea să vă placă și