Rates of weathering rind formation on Costa Rican basalt

Transcription

1 Pergamon doi: /j.gca Geochimica et Cosmochimica Acta, Vol. 68, No. 7, pp , 2004 Copyright 2004 Elsevier Ltd Printed in the USA. All rights reserved /04 $ Rates of weathering rind formation on Costa Rican basalt PETER B. SAK, 1, *, DONALD M. FISHER, 1 THOMAS W. GARDNER, 2 KATHERINE MURPHY, 2 and SUSAN L. BRANTLEY 1 1 Department of Geosciences, Pennsylvania State University, University Park, PA 16802, USA 2 Department of Geosciences, Trinity University, San Antonio, TX 78212, USA (Received July 23, 2002; accepted in revised form September 22, 2003) Abstract Weathering rind thicknesses were measured on 200 basaltic clasts collected from three regionally extensive alluvial fill terraces (Qt 1, Qt 2, and Qt 3) preserved along the Pacific coast of Costa Rica. Mass balance calculations suggest that conversion of unweathered basaltic core minerals (plagioclase and augite) to authigenic minerals in the porous rind (kaolinite, allophane, gibbsite, Fe oxyhydroxides) is iso-volumetric and Ti and Zr are relatively immobile. The hierarchy of cation mobility (Ca Na K Mg Si Al Fe P) is similar to other tropical weathering profiles and is indicative of differential rates of mineral weathering (anorthite albite hypersthene orthoclase apatite). Alteration profiles across the cm-thick rinds document dissolution of plagioclase and augite and the growth of kaolinite, with subsequent dissolution of kaolinite and precipitation of gibbsite as weathering rinds age. The rate of weathering rind advance is evaluated using a diffusion-limited model which predicts a parabolic rate law for weathering rind thickness, r r, as a function of time, t(r r t), and an interface-limited model which predicts a linear rate law for weathering rind thickness as a function of time (r r k app t). In these rate laws, is a diffusion parameter and k app is an apparent rate constant. The rate of advance is best fit by the interface model. Terrace exposures are confined to the lower reaches of streams draining the Pacific slope near the coast where the stream gradient is less than 3 m/km, and terrace deposition is influenced by eustatic sea level fluctuations. Geomorphological evidence is consistent with terrace deposition coincident with sea level maxima when the stream gradient would be lowest. Assigning the most weathered regionally extensive terrace Qt 1 (mean rind thickness cm) to oxygen isotope stage (OIS) 7 (ca. 240 ka), and assuming that at time 0 rind thickness 0, it is inferred that terrace Qt 2 (r r cm) is coincident with stage 5e (ca. 125 ka) and that Qt 3 (r r cm) is consistent with OIS 3 (ca. 37 ka). These assignments yield a value of k app of cm s 1 (R ). Only this value satisfies both the existing age controls and yields ages coincident with sea level maxima. Using this value, elemental weathering release fluxes across a weathering rind from Qt 2 range from mol Si m 2 s 1 to molkm 2 s 1. The rate of rind advance for the Costa Rican terraces is myr 1. Basalt rind formation rates in lower temperature settings described in the literature are also consistent with interface-controlled weathering with an apparent activation energy of about 50 kj mol 1. Rates of rind formation in Costa Rica are an order of magnitude slower than reported for global averages of soil formation rates. Copyright 2004 Elsevier Ltd 1. INTRODUCTION Pedogenic maturity and weathering rind thickness are widely used as indicators of landscape age (i.e., Cernohouz and Solc, 1966; Chinn, 1981; Colman and Pierce, 1981; Knuepfer, 1988; White et al., 1996; Meyer and Leidecker, 1999). Weathering rinds are defined as discolored and permeable crusts enriched in immobile oxides (i.e., Fe 2 O 3, TiO 2, and Al 2 O 3 ) relative to unweathered cores (Colman, 1982a, 1982b). Weathering rind thickness is easily measured in the field making rinds an effective tool for determining the relative degree of surficial deposit weathering. Measured rind thickness has been used as an effective tool in terrace correlation among adjacent basins (see e.g., Ricker et al., 1993; Fisher et al., 1998; Marshall, 2000; Murphy, 2002). Where independent absolute age constraints are available (i.e., from radiocarbon dating of organic debris, cosmogenic dating, tephra chronologies, etc.) empirical best-fit models have been developed relating rind thickness to age for clasts of constant lithology (e.g., Colman and Pierce, 1981; Colman, 1986; * Author to whom correspondence should be addressed (psak@bucknell.edu). Present address: Geology Department, Bucknell University, Lewisburg, PA 17837, USA Knuepfer, 1994). Empirical models (Table 1) suggest rates of increasing rind thickness may be expressed by a power law (Chinn, 1981; Knuepfer, 1988), a relaxation law (Whitehouse et al., 1986) or a logarithmic law (Cernohouz and Solc, 1966; Colman, 1977; Colman and Pierce, 1992). Although none of these studies based their modeling on physico-chemical models of weathering, empirical models have yielded age estimates for landforms with weathering rinds of different thickness (e.g., Colman and Pierce, 1992). All of these studies have been undertaken in temperate climates such as the western United States (i.e., Colman, 1977; Colman and Pierce, 1981; Meyer and Leidecker, 1999), New Zealand (i.e., Chinn, 1981; Gellatly, 1984; Ricker et al., 1993), Japan (Oguchi and Matsukura, 1999) and Bohemia (Cernohouz and Solc, 1966) where rind thickness clast size. In tropical climates, surface and near-surface weathering rates are significantly faster than in temperate climates (e.g., Strakhov, 1967; White, 1995; White and Blum, 1995), resulting in accelerated rates of both pedogenesis and rind genesis. Therefore, thicker weathering rinds in tropical climates might yield age estimates with greater temporal resolution. However, previous models for rind development lack physical basis and are not thought to predict accurate ages for rinds when rind thickness is large (Knuepfer, 1994).

2 1454 P. B. Sak et al. Table 1. Empirical models for rind development as a function of time. Relationship Equation Parent Material Study Area Reference Logarithmic r r 4.64 log t Basalt Bohemia Cernohouz and Solc (1966) Logarithmic r r log t Basalt W. Yellowstone, Colman and Pierce (1981) r r log t Basalt MT r r log t Basalt Yakima, WA r r log t Basalt McCall, ID r r log t Andesite Puget Lowland, WA r r log t Andesite Truckee, CA r r log t Andesite Lassen CA Rainier, WA Power Law r r t Sandstone New Zealand Chinn (1981) Power Law r r t 0.75 Sandstone New Zealand Knuepfer (1988) r r t 0.86 Relaxation Law r r e t/ Sandstone New Zealand Whitehouse et al. (1986) Diffusion Limited r r x t 0.5 Submarine Basalt Hawaii Moore (1966) Measured rind thickness (r r ). Age of deposit (t). x constant. One special case of weathering rind formation, obsidian hydration, has been modeled using a physico-chemically based, rather than empirical, model of rind formation (Friedman and Smith, 1960; Friedman and Long, 1976). Obsidian hydration has been fit to a parabolic rate law (see Eqn. 6 and discussion below). In this paper, weathering rinds on basaltic clasts from the B horizon (zone of illuvial clay accumulation) of Costa Rican river terraces are evaluated assuming rind growth is interface-limited (linear rate law) or diffusion-limited (parabolic rate law). The diffusion-limited model assumes that the conversion of basaltic parent material to weathering rind is rate-limited by diffusion through the rind (Shen and Smith, 1965; Moore, 1966; Luce et al., 1972; Friedman and Long, 1976; Smith, 1981). Such end member models have previously been rejected for other weathering rind systems (e.g., Colman, 1982a) although theoretical derivations of a parabolic rate law have been suggested (Colman, 1986). To constrain these physico-chemical models, chemical composition variations are characterized by bulk chemical analyses of core and rind material, a detailed electron microprobe transect across the core rind boundary, and X-ray diffraction analysis of weathering rinds. Riebe et al. (2001) emphasize the rarity of age-dated soils wherein erosion is insignificant. Such soils could provide unambiguously determined weathering rates. In effect, the weathering rinds investigated herein are a well-constrained example of just such a soil: each rind develops in situ on a known parent lithology at relatively constant climatic conditions, and no erosion of outer surface occurs. Understanding weathering rates of these end member cases will thus contribute to the ability to decouple the contribution of physical and chemical processes in erosion, a goal of much current interest in landscape evolution. In particular, most quantified estimates of chemical weathering rates have heretofore implicitly or explicitly assumed interface-, rather than transportlimited weathering (see discussion in Kump et al., 2000). In this study we attempt to document which regime is operative during one case of tropical weathering. 2. GEOLOGIC SETTING Weathering rinds are observed on clasts from three fluvial terraces (Qt 1, Qt 2, Qt 3) that are distinguished on the basis of pedogenic maturity and elevation. These terraces are preserved in the lower reaches of the Rios Parrita and Barranca across the central Pacific coast of Costa Rica (Fig. 1). Terrace exposures are confined to the lower reaches of both rivers where the modern stream gradient is 3 m/km and range in elevation from 5 to 210 m above mean sea level (Pazzaglia et al., 1998; Sak, 1999; Marshall, 2000). Where absolute age constraints are scarce and geomorphic surfaces are highly dissected, terrace deposits may be correlated using locally calibrated chronosequences. The use of soils as a reliable indicator of landscape maturity requires wellconstrained boundary conditions (Jenny, 1980). The essential uniformity of all soil-forming factors aside from time makes this study area a suitable setting for analysis using a chronosequence. Detailed annual climatic data is available for various sites within Costa Rica including a site at Finca Los Angeles (9 30 N84 24 W) at an elevation of 5m(Fig. 1a). The mean annual temperature (MAT) and mean annual precipitation (MAP) for the period extending from are 27.3 C and 3085 mm, respectively (Instituto Meteorológico Nacional de Costa Rica, 1992, unpublished data). The temperature remains constant throughout the year ranging from minimum of 26.5 C in October to a maximum of 28.5 C in March. In contrast, the precipitation is not evenly distributed throughout the year. The majority of the precipitation falls between the months of July and November with a maximum of 570 mm in October. On average only 28.5 mm of precipitation falls in March (Instituto Meteorológico Nacional de Costa Rica, 1992, unpublished data). Paleoclimatic studies of the Costa Rican coastal zone indicate relative stability throughout the Quaternary (Gómez, 1986). The region is characterized by a transitional ecosystem from tropical dry forest to tropical moist forest (Sawyer and Lindsey, 1971). The local topography of the study area is uniform with all soil description sites from flat-topped terrace deposits confined to 250 m above sea level on the southfacing piedmont and coastal plain. Parent material consists of gravel clasts with limited textural and compositional variability within thick ( 5 m) alluvial fill

3 Rates of weathering rind formation 1455 Fig. 1. A) Simplified geologic map of the central Pacific coast region. Geologic data are modified from Tournon and Alvarado (1997) and Marshall et al. (2000) and supplemented by our mapping efforts. Symbols: W location of Finca Los Angeles weather station; location of dated woody debris at the base of Qt 2. terrace deposits. All analyzed samples are basaltic clasts collected from the B horizons of 5 m thick alluvial fill terrace deposits. 3. MATERIAL AND METHODS 3.1. Field Sampling and Measurements Complete site descriptions including location, aspect, vegetation, slope, land use, elevation, and soil properties were made at each terrace exposure (Sak, 1999; Marshall, 2000). Surface elevation was determined from topographic maps with 20 m contour interval and a Sokkia AIR-HB-1L hand-held digital barometer. Soils were described for matrix color for both moist and wet conditions (Munsell, 1990), texture, consistency, abundance and type of clay films, structure, and horizonation (number of horizons, thickness of each horizon, and type of horizon boundary). These characteristics were noted because they are widely accepted to display progressive development with increasing time (e.g., Bilzi and Ciolkosz, 1977; Harden, 1982; Harden and Taylor, 1983; Birkeland, 1999). At each soil location (including several exposures from each terrace unit) clasts were sampled from the B horizon. Rind thickness was determined by splitting the clast with a rock hammer and measuring the thickness of the discolored crust (Fig. 2). If a portion of the rind remained in the soil upon removal, then rind thickness was reported as the sum of the rind on the clast and remaining in the outcrop. If rinds broke while being removed from the exposure they generally separated in the outermost portion of the rind but never at the core rind boundary. Rind was distinguished from soil matrix by color (Fig. 2a). To determine the characteristic clast shape, the three mutually perpendicular axes of 83 individual clasts from Qt 2 were measured in the field. In the field, several basalt and basaltic andesite clasts and samples of soil matrix were collected for further petrographic and geochemical analysis. Soil samples consist of grab samples of matrix material from the B horizon of Qt 2, and Qt 3. In all instances, soil samples were collected from depths of 0.75 m below the top of the B horizon. In one instance, a single clast from an exposure of Qt 2 (sample 6, Fig. 2c) was impregnated in molten wax to preserve the rind for further study. The clast was removed from the protective wax casing by heating it to a maximum temperature of 55 C for 12 h in the laboratory 4 months later. Then the clast was cut along the intermediate (B) axis with a rock-saw and analyzed as described below Chemical and Mineral Analyses Mineralogic and textural variability across the core rind boundary of five clasts from Qt 2 (samples 3,7, 8, 9, and 10) were determined using a petrographic microscope. X-ray diffraction (XRD) was performed on rind samples from all three terrace units, using a Rigaku Geigerflex X-ray diffractometer operating at 40 kv and 45 ma with a Dmax-B controller and Cu K radiation. Samples were analyzed from 3 to 75 2 at 1 2 /min. In all instances, analyzed rinds were sampled from the midrind at 0.5 to 1 cm from the observed core rind boundary. Samples from four clasts from Qt 2 (samples 2, 6, 7, and 9) and soil samples from Qt 2 and Qt 3 were digested and analyzed using X-ray

4 1456 P. B. Sak et al. Fig. 2. Photographs of weathered clasts illustrating decreasing rind thicknesses with decreasing terrace elevation. A) clast from Qt 1 in an outcrop; B) weathered clast from Qt 2, note the abrupt boundary at the core rind interface; C) sample 6 from Qt 2; D) clast from Qt 3. All clasts are from basaltic parent material. In A, note that the matrix color differs from the rind color. fluorescence (XRF) and inductively-coupled plasma mass spectrophotometry (ICP-MS). All the samples were analyzed for major (Si, Al, Ca, Mg, Na, K, Fe, Mn, Ti, P) and some for trace (Cr, Sr, Y, Zr, and Ba) elements. Nine samples were obtained by drilling cores 4 mm in diameter and 1.5 cm long across a 31 mm transect from the center of the core to the outer rind boundary of a single clast from Qt 2 (sample 6, Fig. 2c). These samples were digested and analyzed using ICP-MS for the same major and trace elements described above. High-resolution spatial variability across the core rind boundary was determined by probing a polished thin-section of sample 10 from Qt 2. Using a Cameca SX-50 electron microprobe, forty equally spaced points were probed for Si, Ca, Al, Ti, Fe, Na, and Mg along a 13.6 mm transect. Probe spot size was 100 m; depth of analysis was 1 m. Analysis of rind material was semiquantitative because of porosity and hydrated phases. Weight percent of rind analyses totaled 40 to 73%, as compared to 83 to 92% for core analyses. Phenocryst compositions in the core of samples 2 and 10 (Qt 2) were determined using a Cameca SX-50 electron microprobe with a 2 m spot size. Phenocrysts were quantitatively analyzed for Fe, Mn, Ti, Ca, K, Si, Na, Al, and Mg content. Samples of rind material and adjacent soil matrix from five sites in Qt 2 were examined using a Philips XL-20 scanning electron microscope (SEM) to examine primary mineral surface morphologies. Goldcoated samples were scanned using a 25 kv beam. 4. RESULTS 4.1. Field Measurements The three alluvial terrace deposits consist of 5 m thick, clast-supported accumulations of well-rounded clasts ranging from 4 to 60 cm in diameter with sandy matrix. The terraces are distinguished by pedogenic characteristics, mean rind thickness, and elevation. Surface soil properties range from red (10 R) matrix, 540 cm thick B horizon for Qt 1 (Marshall, 2000) to brown (10YR) matrix, 200 cm thick B horizon for Qt 3 (Table 2). Terraces display decreasing pedogenic maturity with decreasing elevation (Table 2). From highest to lowest, we generally observed decreases in B-horizon and clay film thickness, redness of the matrix, and rind thickness enveloping unweathered basaltic cores (Table 2). Weathering rinds are observed on all rock types (lithic sandstones, claystones, mudstones, volcaniclastics, and basaltic andesites) within all terraces. The thickness, color, and the nature of the rinds vary as a function of parent material. For porous clast lithologies, a diffuse zone marks the core rind boundary. For all nonporous lithologies such as the basaltic compositions, a sharp boundary clearly separates pristine rock and rind material (Fig. 2). Rinds on basaltic clasts are thinner than rinds developed on other lithologies. Fine- to mediumgrain lithic sandstone clasts are characterized by gray-colored rinds, whereas basaltic clasts have a diagnostic rust-colored rind (Fig. 2a). All of the clasts are characterized by low aspect ratios. The mean long axis (A) to intermediate axis (B) ratio is 1.3 and the mean intermediate axis (B) to short axis (C) ratio is 1.4 (Fig. 3). Clast shape does not vary measurably from one terrace unit to another (Fig. 3). Additionally, twenty one ( n) measurements of rind thickness were made around one clast (sample 6) from Qt 2 to estimate precision. Rind thickness ranged from cm with a mean and median of 1.7 cm. In the field, rind thickness does not appear to vary appreciably with orientation around individual clasts. Furthermore, we observe no evidence suggesting that rind thickness varies with clast shape. Mean rind thicknesses range from thin ( cm) to

5 Rates of weathering rind formation 1457 Table 2. Pedogenic properties and isotopic age control for the flight of alluvial terraces. Terrace Number Radiometric age Maximum height above base level (m) Maximum B horizon thickness (cm) Matrix Clay Color Texture 1 Structure 2 Consistency 3 Films 4 Rind Thickness 5 (cm) YR/10R c abk s,p 4p Qt ka (Beta-79380) 352 ka (Ar/Ar) Qt ka (Beta ) YR 5YR c, cl c,sbk s,p 2 3,pf Qt YR 4/4 sicl m,abk s,sp 2,pf c clay, si silt, s sand, l loam. 2 m medium, c coarse, sbk subangular blocky, abk angular blocky. 3 s sticky, p plastic, sp slightly plastic. 4 2 common, 3 many, 4 continuous; pf on ped faces, po pore linings. 5 Uncertainty reported as standard error. Radiocarbon ages determined by Beta Analytical, Inc (Maimi, FL). Data from Fisher et al. (1998). Data from Marshall (2000). Data from Marshall et al. (2003). thick ( cm) from Qt 3 to Qt 1 (Table 2, Fig. 2). Field measurements of rind thickness show no significant correlation with either clast size (Fig. 4) or depth beneath the top of the B horizon (based on field inspection). The one exception to these findings is in the case of ghost clasts where the entire clast has been converted to rind material. We observe an increase in the size of clasts that have been converted entirely to rind material (ghost clasts) within Qt 1 and Qt 2, while no ghost clasts were observed in Qt 3, the pedogenically least mature unit (Fig. 4). Absolute age control is scarce along the central Pacific coast of Costa Rica. In the absence of such controls for Rio Parrita terraces, ages are inferred from correlation to pedogenically similar terraces dated in adjacent basins (Marshall, 2000). We infer on the basis of horizonation, clay film abundance, and soil structure and texture (Table 2) a relative hierarchy of terrace ages. A ka Ar/Ar age date for a pyroclastic flow beneath Qt 1 on the Rio Barranca provides a maximum age for the terrace (Marshall et al., 2003). Twenty five km east of the Rio Parrita, woody debris from a mangrove deposit at the base of a terrace deposit correlative with Qt1(Marshall, 2000) yielded a radiocarbon date of 46 ka (Fisher et al., 1998). Thus, the regionally extensive Qt 1 gravels across the central Pacific coast region were deposited between 352 ka and 46 ka. Mangrove debris at the base of Qt 2 in the vicinity of Esterillos Centro (Fig. 1a) yield a 40.3 ka radiocarbon age (Sak, 1999). Thus, based upon regional correlation of pedogenically similar fluvial terrace deposits along the Pacific coast of Costa Rica, the two most mature terraces (Qt 1 and Qt 2) lie within the range of to 40.3 ka Chemical and Mineralogical Analyses Petrographic analysis of four clasts identified as basalt in the field from Qt 2 show similar core textures and mineralogies. All Fig. 3. Flinn diagram of axial ratios from 83 clasts from Qt 2. Vertical axis is the ratio of the long (A) to intermediate (B) axes of clast oblates and the horizontal axis is the ratio of the intermediate (B) to short (C) axes. The point (1,1) represents a sphere on this diagram. Fig. 4. Plots of rind thickness as a function of clast size (radius measured along the intermediate (B) axis) for individual clasts from Qt 2. Symbols represent field measured thickness. Uncertainty of 0.5 cm for both clast size and rind thickness is incorporated in symbol size. Note that rind thickness shows no clear dependence on clast size. Solid gray line synthetic curve predicted by the interface limited model (Eqn. 8) and k app cm s 1. Symbols: black dots rinds measured on clasts where the unweathered core remains.

6 1458 P. B. Sak et al. Table 3. Chemical analyses from electron microscopy. Sample MnO FeO TiO 2 CaO K 2 O SiO 2 Na 2 O Al 2 O 3 MgO Total Mineral 7 Core Augite Quartz Quartz K-spar Ilmenite 10-Core Augite Augite Augite Epidote 10-Rind Augite Kaolinite Plagioclase(Ab 86 An 13 ) Epidote Epidote Epidote Quartz Sphene Ilmenite Gibbsite Note: oxide abundances reported as weight percent contain abundant unaltered plagioclase phenocrysts, lesser abundances of unaltered augite and apatite, and unaltered quartz phenocrysts in a very fine-aphanitic matrix (Table 3). The dark-colored matrix consists mainly of plagioclase microlite and lesser abundances of augite microlite quartz, epidote and glass and is too fine-grained to point count. Bulk chemical analysis (XRF and ICP-MS) of four analyzed cores also document relatively uniform composition (Tables 4, 5). In the rind, undistorted ghosts of plagioclase phenocrysts are porous and highly altered with extensive iron oxide coatings. The matrix material in the rind, like the phenocrysts, is highly porous and contains extensive iron oxide coatings. XRD spectra for rind materials sampled from midrind on clasts from Qt 2 and Qt 3 show evidence of increased weathering intensity in pedogenically more mature units. In all samples, peaks consistent with quartz and iron oxyhydroxides are recognized. Both short broad and well-defined kaolinite peaks are recognized in Qt 3 samples. No kaolinite peaks are recognized in samples of the outer portions of the rind from Qt 1 although kaolinite peaks are recognized in the inner-rind near the core rind boundary (Murphy, 2002). Gibbsite is present in all samples from Qt 1 and Qt 2 and half of the analyzed Qt 3 samples. Minerals observed in rinds on samples from each terrace are summarized in Figure 5. Bulk chemical analyses along a nine point transect (samples ) across a single clast (sample 6) show a uniform core composition and variability in rind composition (Table 5). Weight percent abundances of all major oxides besides Al 2 O 3, Fe 2 O 3, and TiO 2 decrease coincident with the observed core rind boundary. The weight percent of Al 2 O 3,Fe 2 O 3, and TiO 2 are greater in the rind than in the core (Tables 4, 5). Within the rind, the abundance of Si, Ca, Mg, Na, K, Mn, Sr, and Ba all decrease away from the core rind boundary (Table 5). The results of bulk chemical analysis (ICP-MS) of soil matrix material are markedly different from the bulk chemistry of the rind and core material (Table 4). For example, concentrations in the soil range from 42 to 45.7 weight percent SiO 2 and from 14.4 to 18.1 weight percent Fe 2 O 3, compared to 10.2 weight percent SiO 2 and 20.3 weight percent Fe 2 O 3 in the rind Unit Sample Number Table 4. Bulk chemistry of samples from B horizons of fluvial terraces. Sample Type SiO 2 Al 2 O 3 CaO MgO Na 2 O K 2 O Fe 2 O 3 MnO TiO 2 P 2 O 5 Cr 2 O 3 Rb Sr Y Zr Nb Ba Qt 2 1 b Soil Qt 2 4 b Soil b Soil b Core b Soil b Core a Core b * Core b * Rind Qt 3 13 b Soil Weight percent of specified oxide. All analyses (major and trace) determined by XRAL Laboratories (Ontario, Canada). ppm of specified trace element. * Mean chemical composition of sample 6 core or rind (Table 5). a Bulk chemistry determined using XRF. b Bulk chemistry determined using ICP-MS.

7 Rates of weathering rind formation 1459 Table 5. Bulk chemical analyses from a transect across sample 6 from Qt 2. Sample Number Distance (mm)* Type SiO 2 Al 2 O 3 CaO MgO Na 2 O K 2 O Fe 2 O 3 MnO TiO 2 P 2 O 5 Cr Sr Y Zr Ba Core Core Core Core Core C** Core Rind Rind Rind Rind R** Rind * Distance measured from the center of the core outward. Weight percent of specified oxide. All analyses determined by XRAL Laboratories (Ontario, Canada) using ICP-MS. ppm of specified trace element. ** Mean core (6.1 through 6.5) and rind (6.6 through 6.9) compositions. (Table 4). Concentration of K 2 O is higher in soil as compared to rind (Table 4). High SiO 2 and K 2 O concentrations (compared to rinds) may reflect a more felsic soil protolith, consistent with faster weathering clasts and sand of more felsic composition like that observed in the modern river system (Tables 4, 5). However, the composition of this limited data set of soils does not show systematic compositional variability with increasing pedogenic maturity. Detailed microprobe analysis across the core rind boundary of sample 10 (Qt 2) records decreasing weight percent abundances of Si, Ca, Na, and Mg in a narrow ( 500 m wide) zone defining the core rind interface (Fig. 6). In thin-section, the core rind boundary is defined by the first appearance of pore space. The Fe abundance increases in this same zone while the weight percent Al and Ti show little change (Fig. 6). Coreward of the core rind boundary, Na, and possibly Ca and Mg concentrations increase while Al, Fe, and Ti remain constant (Fig. 6). SEM analysis of fractured surfaces of rind and soil grains reveals that, in the rinds, mineral surfaces are generally devoid of etch pits, while in the soil matrix, etch pits are sometimes observed (Fig. 7). The mineral grain shown in Figure 7a was analyzed by energy dispersive X-ray (EDX) and revealed a chemical composition that is qualitatively consistent with either pyroxene or Fe-stained plagioclase. No etch pits were observed. In contrast, pitting on the primary mineral soil grain documented in Figure 7b is consistent with etching on pyroxenes reported in the literature (Berner et al., 1980). However, most surfaces were Fe-rich making interpretation of primary mineral composition and identification of etch pits difficult. In addition, mineral faces in the rind and soil are commonly mantled with light-colored, irregular sponge-likemasses and globules. Energy dispersive X-ray analysis (EDX) of these irregular masses reveals abundant Al, Si, and Fe. Based on similarities with published pictures (Colman, 1982b) we interpret this material as allophane Sources of Error Potential sources of rind thickness variability include: 1) measurement of rinds along oblique sections, 2) variable parent material composition, 3) variable parent material texture, 4) failure to locate the core rind boundary accurately, and 5) failure to locate the rind soil boundary accurately. These problems are discussed in the next paragraph. Measurement of rinds along oblique sections would result in apparent rind thickness measurements greater than the true thickness. To minimize this potential source of error, when necessary, clasts where split along two axes to ensure that data represents measurement of true rind thickness. Given the threedimensional exposure of the core rind boundary in many cases, this error is presumed 0.5 mm. Bulk chemical analysis of core material measured on representative clasts from Qt 2 suggests that the parent material is essentially uniform between the terrace units (Table 4). Examination of thin-sectioned core material from Qt 2 also suggests uniform texture and mineralogy. The other three identified sources of error do not systematically produce thicker or thinner rind values. Furthermore, given the distinct color (Fig. 2) and composition of the rind as compared to the soil and core (Table 4), it is unlikely that either the core rind or rind soil boundaries were misidentified. One test of measurement error is reproducibility at a given site. Repeated visits to field sites yielded the same mean rind thickness. For example, a single exposure of Qt 2 was visited four times and measurements of 8 clasts made by a total of nine workers. Mean rind thickness (2.8 cm) varied by 4mm between the four visits. 5. DISCUSSION These results suggest chemically-based models for weathering rind development. First, mass balance calculations are used to quantify volumetric strain, relative cation mobility, and elemental loss. These data are combined with analysis of mineral surface morphologies within the rind and adjacent soil matrix to assess the rate-limiting step in rind development. Second, we present interface- and diffusion-limited models for rind formation. We then evaluate the predicted ages from the two models to determine which model provides the best correlation between terrace age and the timing of major sea level high stands within the age constraints of 40 to 350 ka. Finally, diffusion- and interface-limited models are also both applied to

8 1460 P. B. Sak et al. published weathering rind data for basaltic clasts from diverse climatic regimes, and the apparent activation energy is calculated Mass Balance Calculations Two requirements must be satisfied when using mass balance calculations to estimate the extent of open-system mass transport (Brimhall and Dietrich, 1987; Brimhall et al., 1991). The weathering product must form in a homogenous parent material of uniform age and weatherable elements must be ratioed against an inert component present in both parent and product (Chadwick et al., 1990). Since the rind is enveloping a single basaltic clast of known composition (Table 5), the first requirement is satisfied. Volumetric changes are estimated from the strain, i,w, where i,w is related to the ratio of weathered volume to the initial volume. Where an immobile element, i, is present in the system, strain i,w can be calculated as i,w pc i,p w C i,w 1, (1) where w is dry bulk density of the weathered rind (measured as 1.4 g cm 3 ), p is dry bulk density of the parent material (the core, measured as 2.8 g cm 3 ), c i,w is mass concentration (weight percent) of element i in the rind, and c i,p is mass concentration of element i in the core (see oxide wt % in Table 5). Bulk density was measured three times on three separate samples from sample 6 by standard immersion techniques for both the core and rind and in both cases varied by 0.1 g cm 3. The volumetric strain for sample 6 is therefore calculated below assuming constant core and rind densities of 2.8 and 1.4 g cm 3, respectively. Positive strain represents dilation of the rind, negative strain represents collapse, and zero strain represents isovolumetric weathering (Fig. 8a). Here, calculations of the strain factor, i,w, are shown based on both Zr and Ti since these Fig. 5. Growth rate-curves for weathering rinds on basaltic clasts as a function of time. Minerals observed in rinds at midrind are also noted. (A) Fluvial terraces along the central Pacific coast of Costa Rica. Age control for terraces (horizontal arrows) are based on radiometric ages and the relative pedogenic maturity of the host deposits (Table 2). Rind thicknesses are mean values /n. Symbols: Shaded curve area range of ages that satisfy independent age constraints assuming transport-limited rind advance: Bold line growth-rate curve calculated using Eqn. 8 and k app cm s 1 (see text for explanation); Dashed gray line rate of rind advance assuming far-from-equilib- rium labradorite dissolution rate (measured in the laboratory) is ratelimiting; Solid gray line rate of rind advance calculated using Eqn. 6 and cm 2 s 1 ; Q, Fe ox, kaolin, gibb-indicate presence of quartz, iron oxides, kaolinite, and gibbsite, respectively in XRD spectra of rind material. (B) Growth rate curve for fluvial terraces from Nasuno-ga-hara, Japan (Oguchi and Matsukura, 1999). Age control for deposits (horizontal lines) based on radiometric, tephra chronologies, and cosmogenic dating of host deposits. Symbols: Bold line best-fit growth rate curve calculated using Eqn. 8 (k app cm s 1 ), Dashed gray line rate of rind advance assuming far-from-equilibrium labradorite dissolution rate (measured in the laboratory) is rate-limiting; Gray line rate of rind advance calculated assuming transport-limited rind growth (Eqn. 6) and cm 2 s 1 (see text for explanation); (C) Growth rate curve for moraines from West Yellowstone, MT (Colman and Pierce, 1992). Age control for deposits (horizontal lines) based on radiometric, tephra chronologies, and cosmogenic dating of host deposits. Rind thicknesses are mean values ( ). Symbols: Solid black line best-fit growth-rate curve calculated using Eqn. 6 (k app cm s 1 ); Dashed gray line rate of rind advance assuming far-from-equilibrium labradorite dissolution rate (measured in the laboratory) is rate-limiting; Gray line best-fit rate advance assuming diffusion control (Eqn. 8) and cm 2 s 1 ; Idd, chlor and Al indicate iddingsite, chlorite, and allophane, respectively in XRD spectra of rind material.

9 Rates of weathering rind formation 1461 Fig. 6. Electron microprobe transect across the core rind boundary of sample 10 collected from Qt 2. X-axis represents distance from an arbitrary point (x 0) in the core out toward the outer surface of the rind, and y-axis represents weight percent of specified oxide. Core rind boundary defined by visual inspection of the thin section and hand specimen. Variations in oxide abundances within the core are attributed to repeated probing of large plagioclase phenocrysts. To dampen the variability due to phenocryst position a three point moving average was used to smooth the data. Plots are arranged in order of decreasing cation mobility as defined by j,w (Eqn. 2). elements are often highly immobile in soil (Barshad, 1964; Marshall, 1977; Milnes and FitzPatrick, 1989) and were shown to be immobile in a tropical watershed in Puerto Rico (i.e., White, 1995; White et al., 1998). For sample 6, the calculated strain values, i,w, based on Zr and Ti are near zero across the rind (Fig. 8a), consistent with isovolumetric weathering and Zr and Ti immobility. Additionally, the presence of undistorted ghosts of plagioclase phenocrysts in the rind is consistent with isovolumetric weathering. Given that Zr mobility and loss increase with mean annual precipitation (Kurtz et al., 2000) we use Ti as our inert component in the mass balance calculations. In fact, Ti varies more across the inner core (samples 6.1 to 6.4, 0.10 Ti,p 0.03) (Appendix 2) than across the rind (samples 6.6 to Ti,w 0.03) (Appendix 2), consistent with the assumption of a constant density across the rind (Fig. 8a). Variability in the core is attributed to localized compositional heterogeneity. Similarly, strains calculated for saprolites in other localities based upon mass balance (Cleaves, 1993; White, 1995) have shown isovolumetric weathering.

10 1462 P. B. Sak et al. Fig. 7. SEM photomicrographs of fractured primary mineral surfaces from (A) the rind of sample 10 (Qt 2) and (B) the adjacent soil matrix Qt 2. Both minerals imaged are probably augite (see text). The slightly negative strain value ( Ti,p 0.12) calculated for a sample taken at a point 2 mm coreward of the visually defined core rind boundary of sample 6 (sample 6.5, Fig. 8a) is consistent with a calculated 0.3 g cm 3 decrease in the core density at this point assuming Ti is immobile (Fig. 8a is calculated assuming constant rind and core density). This observation is consistent with some alteration occurring coreward of the observed core rind boundary, as already documented in Figure 6. For sample 6 the concentration of Ti doubles from the core to the rind due to residual enrichment of this inert component (Table 5). If weathering of the core is isovolumetric as suggested by the presence of undistorted phenocryst ghosts in the rind, then Ti,w 0 throughout the rind and the isovolumetric alteration corresponds to 50% rind porosity. The core contains no recognizable pore space (i.e., porosity 1%). Only in the outer-most portion of the core at sample 6.5 (sample location 2 mm from the visually defined core rind boundary, Fig. 2c) is an intermediate porosity (10%) calculated from the observed Ti concentration, assuming Ti,w 0, and Eqn. 1. The extent of loss or gain of an element j from the rind, j,w, is calculated as j,w wc j,w p C Ti,w 1 1, (2) j,p where, Ti,w is the volumetric strain for Ti, as determined in Eqn. 1 (Fig. 8a). Mass transfer coefficients, j,w are calculated assuming p 2.8gcm 3 and w 1.4gcm 3 for samples except 6.5 where w is assumed to equal 2.5 g cm 3 (see earlier discussion). When j,w 0, the element j is immobile. If j,w 1, element j is completely lost during weathering. The more negative the value of j,w, the more mobile the element (Fig. 8b,c). Consistent with our previous discussion, j,p values for some cations decrease from the core values of 0 at a point 1.5 mm coreward of the visually observed core rind boundary, indicating that alteration occurs ahead of the boundary observed visually (Fig. 8b,c). The sequence of mobility, based on the relative loss of each element documented at the outer edge of the weathering rind compared with the unweathered core, is Ca Na K Mg Si Al Fe P(Fig. 8b,c, Appendix 2). This mobility is similar to the sequence (Ca Na Mg Si Al K Fe) reported for thick saprolite accumulations developed on granitic bedrock in the Rio Icacos watershed of Puerto Rico (White et al., 1998) and weathering rinds on andesitic and basaltic clasts from the western United States (Ca Na Mg Si Al K Ti) as reported by Colman (1982b). The one difference is that K mobility differs in the Costa Rican data as compared to the North American or Puerto Rican data sets. The relative immobility of K in the Rio Icacos (Puerto Rico) watershed is related to high regolith concentrations (16 21 weight percent) of slowly weathered K-rich biotite (Murphy et al., 1998), a phase not present in the Costa Rican rinds. Colman (1982b) attributes the relative immobility of K in weathering rinds of the Western United States to adsorption on halloysite, a phase not observed in Costa Rican rinds. In basaltic corestones from Costa Rica, the trace K is probably found predominantly in feldspar Mineralogic Gradient Mineralogic variability of core samples is determined from bulk chemical compositions (Tables 4, 5) using a CIPW norm (Table 6) because point counting of the fine-grained groundmass is impossible. Although glass is present in the samples, no attempt was made to assess the total glass content. Assuming a typical FeO : Fe 2 O 3 ratio for unaltered basaltic andesite of 10:1 (Irvine and Baragar, 1971), normalized modal abundances are calculated (Table 6). Modal abundances of minerals in the mean core composition of sample 6, 37% anorthite, 21% albite, 20% hypersthene, 10% diopside, 3% ilmenite, quartz, orthoclase, magnetite, and apatite (Table 6), are similar to samples 2 (Qt 2), 7 (Qt 3) and 9 (Qt 3) (Table 6). Similar calculations can be completed for the rind. However, the rind is composed primarily of secondary minerals and so CIPW calculations are impossible. Modal abundances of rind samples (Table 5), are therefore determined as follows. All Na and the requisite Si and Al are allocated to albite. All P and the requisite Ca are allocated to apatite. Remaining Ca and the requisite Si and Al are assigned to anorthite. Ca is allocated to apatite before anorthite because apatite is the only phosphate-bearing phase considered in the CIPW norm used to calculate modal abundances in core samples. All Mg, an equal molar amount of Fe 2, and the requisite Si are assigned to hypersthene. Although diopside is present in the core (Table 6), allocation of Fe and Mg to both hypersthene and diopside in the rind is under-determined and hypersthene is therefore used as a proxy for all pyroxenes in the rind (hypersthene, diopside, and

11 Rates of weathering rind formation 1463 Fig. 8. Weathering characteristics as a function of distance from the center of sample 6 from Qt 2: (A) volumetric strain i,w (Eqn. 1), (B) (C) elemental mass transfer j,w calculated from Eqn. 2, (D) modal abundance of primary and secondary minerals across sample 6 (see text), (E) mineralogic mass transfer j,w (calculated from Eqn. 4 assuming ilmenite immobility). Note: i,w 0 in the core near the core rind boundary (sample 6.5) is consistent with a 0.3 g cm 3 density decrease within 2 mm of the visible core rind boundary. Mass balance calculations assume constant core and rind densities of 2.8 and 1.4 g cm 3, respectively, except at sample 6.5 where was set equal to 2.5 g cm 3 based on the density decrease inferred in (A). Negative values of j,w imply loss of the element j in the weathering product relative to the parent material. Reported j,w 0 within the core may reflect localized composition heterogeneity or analytical errors. Dashed lines document the visually defined core rind boundary: the reaction front lies coreward of the visible core rind boundary. augite, Table 3). All the Ti and the requisite Fe are assigned to ilmenite. Assuming quartz is inert in the soil environment, we calculate quartz abundance in the rind by maintaining a constant quartz : ilmenite ratio of 0.81 (the ratio calculated for mean core composition). The remaining Si and requisite Al are assigned to a kaolinite (Si 2 Al 2 O 5 (OH) 4 ) phase. This allocation implicitly includes allophane (Si 3 Al 4 O 12 nh 2 O). The remaining Al and Fe are assigned to gibbsite and iron oxyhydroxides, respectively. Differences in Si : Al ratios between kaolinite and allophane will lead to increasing

12 1464 P. B. Sak et al. Table 6. Modal abundances of samples from Qt 2. Sample Number Distance (mm)* Type Q or ab an di hy il Mt ap kaol gibb Fe ox 2 N/A Core N/A Core N/A Core Core Core Core Core Core C** Core Rind Rind Rind Rind R** Rind * Distance measured from the center of the sample 6 core outward across rind. Weight percent normative of secondary phases (see text). kaol-kaolinite allophane; gibb-gibbsite; Fe ox-iron oxyhydroxides. ** Mean composition. Assuming a constant quartz: ilmenite ratio (see text). overestimation of gibbsite in samples with significant allophane. Across the rind of sample 6, rind-ward of the core rind boundary, we observe an abrupt decrease in anorthite and a systematic decrease in the weight percent of normative albite, orthoclase, and hypersthene (Table 6, Fig. 8d). Kaolinite/allophane increases in concentration across the rind and then decreases. Across the same interval, the modal abundance of ilmenite and apatite remain constant (Table 6, Fig. 8d). Gibbsite and Fe oxyhydroxides show increased modal abundances away from the core rind boundary (Table 6, Fig. 8d). The pattern of increasing gibbsite abundance across the rind (i.e., with increasing extent of weathering) of sample 6 is consistent with XRD analyses documenting increasing gibbsite with increasing weathering extent in rinds from terraces Qt 3 to Qt1(Fig. 5). The decreasing kaolinite abundance across rind sample 6 (i.e., with increasing extent of weathering, see Fig. 8d) parallels the trend of decreasing kaolinite crystallinity in pedogenically more mature deposits. Although no kaolinite was recognized in XRD analyses of midrind samples from Qt 2, kaolinite and/or allophane could be present but confined to the immediate vicinity of the core rind boundary of these more mature samples. Consistent with this, extensive examination of the inner rind of samples from Qt1 revealed the presence of kaolinite under X-ray diffraction (Murphy, 2002). The extent of loss or gain of a mineral, j, from the rind, j,w is calculated j,w wu j,w p u i,w 1 1, (3) j,p where, i,w is the volumetric strain for ilmenite ( i,w 0), u j,p is mass concentration (weight percent) of mineral j in the core and u j,w is mass concentration (weight percent) of mineral j in the rind, When j,w 0 mineral j is immobile. When j,w 1, mineral j is completely lost during weathering. The sequence of mineral mobility, based on the relative loss of each mineral as documented in the outer edge of the rind compared to the unweathered core, is anorthite albite hypersthene orthoclase apatite (Fig. 8d). As noted previously, differential mineral dissolution occurs within the core, before the visually defined core rind boundary (Fig. 8d,e). For example, an,p ( 0.22) and ab,p ( 0.17), where an anorthite and ab albite, in sample 6.5 lie outside the range defined by the 1 ( ; , respectively) of the inner core (samples ), while hy,p ( 0.05, hy hypersthene) at 6.5 lies within the range of the inner core ( , Fig. 8e, Appendix 2). These observations suggest that alteration of plagioclase, occurring before hypersthene, starts 2mm coreward of the optically defined core rind boundary. Using the CIPW norm for the mean core composition of sample 6 (Table 6) and the rind composition of sample 6.6 (Table 6), balanced equations for the conversion of basaltic minerals to rind material can be written: Na 0.4 Ca 0.6 Al 1.6 Si 2.4 O 8(s) 4.8H 2 O (l) 3.2H (aq) 3 0.4Na (aq) 0.6Ca 2 (aq) 2.4H 4 SiO 4(aq) 1.6Al(OH) 2(aq) (4a) KAlSi 3 O 8(s) H (aq) 4.5H 2 O (l) 3 K (aq) 2H 4 SiO 4(aq) 0.5Al 2 Si 2 O 5 (OH) 4(s) (4b) Ca 0.8 Mg 0.8 Fe(II) 0.2 Fe(III) 0.1 Al 0.3 Si 1.8 O 6(s) 1.95H 2 O (l) 3.2H (aq) 3 0.3FeOOH (s) 0.15Al 2 Si 2 O 5 (OH) 4(s) 1.5H 4 SiO 4(aq) 0.8Ca 2 (aq) 0.8Mg 2 (aq) 0.1 H 2(g) (4c) 0.5Al 2 Si 2 O 5 (OH) 4(s) 2.5H 2 O (l) 3 Al(OH) 3(s) H 4 SiO 4(aq) (4d) In these weathering reactions, we assume core pyroxenes are augite in composition, and plagioclase has uniform (Ab 37 An 64 ) composition. Pyroxenes of augite composition are generally consistent with phenocryst compositions (Qt 2, Table 3). The