A geophysical strategy for measuring the thickness of the critical zone developed over basalt lavas

Tropical volcanic island chains are exposed to severe chemical weathering, especially in areas of high rainfall, as is generally true for parts of Oahu, Hawaii (for review, see Nelson et al., 2013). Laterites, or lateritic weathering profiles (LWPs), typically develop in warm, wet regions where chemical weathering is rapid. Warm, moist conditions promote plant growth, leading to high CO₂ concentrations in the soil, where carbonic acid is consumed in weathering reactions of thermodynamically unstable primary igneous minerals. Chemical weathering reaction rates strongly depend on temperature as well (Dixon et al., 2009), with high mean temperatures promoting laterite development. These conditions abound on Oahu, which has strong rainfall variations across the island (Fig. 1), making this an ideal location to study LWPs. Further, the presence of essentially a single type of bedrock, tholeiitic basalt, simplifies the interpretation of weathering rates.

Chemical weathering has a significant impact on geological materials and thus on human life. Laterite development influences ground stability, atmospheric CO₂ concentrations (Beaulieu et al., 2012), and ore deposits (i.e., bauxites) (Meyer et al., 2002), making the study of the critical zone (the highly variable near-surface zone in the earth of complex chemical, biological, and mechanical interactions) in areas like Oahu important. Some have argued that weathering is the most important, yet least-studied geologic process (Wilson, 2004; West, 2012). For engineering considerations (i.e., ground stability), one of the most important features of the critical zone is its thickness and average shear-wave velocity. Many models require this parameter in order to understand denudation and weathering rates and the accompanying consumption of CO₂ (Hilley et al., 2010; West, 2012). However, Hilley et al. (2010) stressed that the relationship between denudation and weathering zone thickness is poorly understood. More data from various climatic and tectonic environments are needed in order to better understand this relationship.

This purpose of our study is to assess chemical weathering rates in Hawaiian basalts and more generally the weathering of intermediate to mafic igneous rocks in tropical environments. This was addressed by testing the ability of multichannel analysis of surface waves (MASW; Park et al., 1999) methods to create a velocity model of the subsurface that defines the base of the critical zone, shows variations within the overlying LWP, and provides downward rates of advance of the weathering front. The new data collection and processing procedures described herein were needed to more accurately model the weathering profile, especially at LWP thicknesses encountered in Oahu. An allied focus was to calibrate the MASW models with geological information from logged wells and limited outcrops, which defined the depth of the LWP-bedrock interface.

We investigated several sites on Schofield Barracks in the central plain of Oahu (HI-1, HI-2, HI-3, HI-4, HI-5; Fig. 2) and one site at an Oahu regional park (HI-6). Schofield Barracks is ideal for our study due to the locally well-developed and exposed LWPs (Fig. 3), numerous boreholes with detailed geologic logs, and our ability to gain general access to most areas of the military base away from cultural interference. Each site is overlain by a thick LWP above relatively unaltered basalt. In order to provide a baseline case, we have also applied the same set of techniques to an area of shallowly buried, young, and mostly unaltered basalt flows with compositions similar to those in Oahu in a semiarid part of central Utah (Ice Springs in the Sevier Desert) (Fig. 4).

MASW (Park et al., 1999) has become a widely used geophysical technique to estimate shear-wave velocities in the near-surface environment (Zeng et al., 2012; Park, 2013). In any seismic method, surface-wave energy dominates the seismic shot record with Rayleigh waves (or ground roll) making up more than two-thirds of the total energy generated, which is considered noise in traditional seismic techniques (Xia et al., 2002). However, the dispersive properties of surface waves allow for elastic properties of soil and rock to be inferred, and from which a shear-wave velocity profile can be derived (Park et. al, 1999). Many applications have dealt with measurements of the shear-wave velocity of the upper 30 m (i.e., Vs30). Other applications have mapped the bedrock-soil interface (Carnevale et al., 2005; Mahajan and Rai, 2011; Boiero et al., 2013; Park, 2013; Sirles et al., 2013).

The island of Oahu is dominated by two large shield volcanoes, the 3.2–1.8 Ma Koolau volcano to the east and the 4.0–2.6 Ma Waianae volcano to the west. The western side of the Waianae volcano and the eastern side of the Koolau volcano have collapsed into the ocean; the remaining island is constructed of their overlapping remnants (e.g., Sherrod et al., 2007). Our survey sites are located on the central part of the island on smooth, flat plains between the two volcanoes, which are deeply incised by numerous streams.

Tholeiitic basalt is the dominant rock type, and the high variability of rainfall across the island makes it ideal to study weathering rates reflected in LWPs of variable thickness. Whereas bedrock is chemically homogeneous, large variations occur in the primary textures between ‘a‘a and pahoehoe flows, as well as in pyroclastic materials. Annual precipitation varies by more than an order of magnitude across Oahu (Lau and Mink, 2006). All but one of our profiles are within a mesic climate zone, averaging ∼1500 mm of annual rainfall (Fig. 1). Monitoring wells with geologic logs located near our survey areas suggest LWPs ranging in thickness from ∼30 to 50 m.

The LWP or critical zone profiles we investigated consist of relatively thin, 1–2-m-thick soils overlying saprolite, which composes the majority of the LWP. The inorganic constituents of soil and saprolite consist chiefly of kaolin-group clays (halloysite ± kaolinite) and iron oxides and hydroxides (maghemite, magnetite, goethite, ferrihydrite) (Nelson et al., 2013; Yaede, 2014) developed from the hydrolysis and oxidation of primary igneous minerals. Locally, kaolin-group clays have been partially leached to form gibbsite. LWPs reflect mutually encrusting and fine-grained mineral masses. Although soil formation involves dilation and collapse (e.g., Vitousek et al., 1997), textural and other evidence suggests that saprolite composes nearly the same volume as the parent rock, supporting the assumption of isovolumetric weathering in determining the rates used in this study (Patterson, 1964, 1971; Gardner, 1980; Michel, 1984; Hill et al., 2000; Chadwick and Chorover, 2001; Patino et al., 2003; Chabaux et al., 2011). Alkaline and alkaline earth cations exhibit nearly complete leaching and silicon is often very strongly depleted (Nelson et al., 2013; Yaede, 2014).

The Ice Springs survey site in the Sevier Desert of Utah (Fig. 4) is composed of Pleistocene basalt flows overlain by a thin veneer of Pleistocene to Holocene fine-grained basin-fill sediments. These basalts are mostly unweathered due to the semiarid climate; mean precipitation is <230 mm/a. A driller’s log from a well near our site reveals ∼5 m of valley fill sediments overlying ∼26 m of Quaternary basalt lava. We confirmed the depth to the top of lava with a hand-augered hole to be 4.7 m.

Data acquisition involved multichannel shot gathers similar to conventional common depth point (CDP) reflection surveys (Telford et al., 1990). An innovative active-source survey method and a combined active and passive source method were used in order to establish a best practice to improve depth of investigation and overall resolution of velocity variations. Both experimental setups used a 7.25-kg sledge hammer source with 24 or 48 active receiver stations, with vertically spiked, low-frequency geophones (4.5 Hz), which are standard for MASW recording (Park et al., 1999). Our surveys used fixed-source offsets moving through the receiver spread 24 times by increments of 5 ft (1.5 m) to generate a 24 or 48 channel record (Figs. 5 and 6). The modeling results are presented as two-dimensional (2D) shear-wave velocity profiles, based on interpolations between 1D inversion results using dispersion spectra (e.g., Fig. 7) (Park et al., 1999). The depth below ground surface to the top of unweathered basalt, where available from lithologic logs, is shown correlated to the velocity model.

For the standard active-source MASW survey, the initial source-to-receiver offset (X1 in Fig. 5) was set at 35 ft (10.7 m) with three records stacked at each station. The source was then advanced at 5 ft (1.5 m) (dx in Fig. 5) increments for a total of 25 stations. The recording time was set at 2 ms with no acquisition filters. After initial experiments failed to provide the needed depth of investigation, a new strategy of using increased multiple offsets and combining their respective overtone images (i.e., phase dispersion or Rayleigh wave phase velocity expressed as a function of frequency) was explored. The derived overtone image showed improved energy accumulation in the frequency-phase velocity domain that was used to better identify dispersion trends. This allowed for dispersion curves to be derived in the lower frequency domain with less subjectivity and greater confidence. For three of the survey areas (HI lines 3 and 6 and the Utah Ice Springs line; Figs. 2 and 4), three additional source-nearest receiver offsets were employed at 70 ft (21.3 m), 105 ft (32 m), and 140 ft (42.7 m). The other two areas (HI lines 4 and 5) required only two additional offsets, 70 ft (21.3 m) and 105 ft (32 m).

The surface wave records were processed using SurfSeis (Park, 2006). The data processing and inverse modeling included encoding the field geometry into the shot records, derivation of dispersion images (overtone images), extraction of a dispersion curve, inversion into a 1D shear-wave velocity function for each record, and interpolation into a 2D shear-wave velocity profile. Some alterations to the traditional processing steps were needed in order to improve the depth of investigation. Each source-nearest receiver offset record was initially treated as a separate experiment until the dispersion curve extraction step. The dispersion image was generated by transforming the shot gather into the phase-velocity frequency domain (Park et al., 1998). The several dispersion images (as many as four), one from each offset record (with the same active phone geometry; Fig. 7), were then merged. This final image was used to determine the appropriate extraction of the fundamental mode of the Rayleigh wave by tracing the high-energy concentrations on the enhanced dispersion images (Fig. 7). The shear-wave velocity profile was then calculated using a normal iterative inversion (Xia et al., 1999).

For some MASW surveys, we repeated the survey along the line of receiver locations, using a conventional CDP configuration for reflection and/or refraction recording using compressional waves. In such cases, we switched from 4.5 Hz to 28 Hz geophones, imposed a low-cut frequency filter of 15 Hz, and recorded with a 0.25 ms sample rate. Data processing followed a routine series of steps, including CDP sorting and stacking, muting of direct and refracted arrivals, and normal move-out velocity analysis, from which we derived a simple RMS (root-mean square) velocity function for time to depth conversion of the CDP stack.

In an attempt to find a more effective strategy for increasing the depth of investigation, a mixed passive and active source technique was developed. Passive source MASW (Park et al., 2005; Park and Miller, 2008) is useful for increasing the depth of investigation, as well as for simplifying the field procedure; however, a significant loss can occur in the accuracy of the resulting velocity-depth model (e.g., 30%), relative to an active source survey. This is because two parameters remain completely unknown: the location and the excitation time of the seismic source. Passive source methods try to solve these two uncertainties by assuming a perfect randomness in both location and excitation time of surface waves. True randomness cannot be achieved in reality, which is why the accuracy always drops below that of the active survey regardless of the type of receiver array used. A compromise solution is to design a survey that reduces the uncertainty in the source location and excitation time, while also increasing the source-nearest receiver offset. Such an approach would not only increase the depth of investigation, but also widen the bandwidth for surface-wave recording, especially into low frequencies. Thus, combining aspects of active and passive source approaches provides advantages from each.

The active survey uses one predefined source point that delivers only one impact that triggers one recording of seismic waves. In this way, generation of surface waves is clearly defined in both space and time. However, the passive survey uses neither predefined source location nor the predefined impact generation time. Recording is started at an arbitrary time that listens to ambient surface waves generated at arbitrary places. In this way, both location and generation time of surface waves are regarded as unknowns in the processing algorithm. The survey that combines aspects from both types is defined as an active/passive survey. For example, the multisource offset survey used in this study had a predefined first impact location that triggered the recording. In this sense, it was an active survey. In addition, during the long recording period, several impacts were delivered at different locations at different times. Although these impacts were somewhat controlled in space with the same spacing between successive impact points and also in time with a fairly even interval between impacts, the main purpose was to introduce as many impacts at as many source offsets as possible so that surface waves with a broad range in wavelengths could be generated. The information for the location and time of these additional impacts was not used during the processing, but the processing algorithm (Park et al., 2005; Park and Miller, 2008) simply assumed more surface waves (than the ones generated from the first impact) could be generated at arbitrary times and at arbitrary locations. In this sense, the survey included aspects of a passive survey, and therefore is termed an active/passive survey.

The experimental setup for the passive/active source survey was similar to the active survey design. However, in this case the initial source-nearest receiver distance was set at 60 ft (18.3 m) ( = 12 × receiver spacing) at which a 60 s record (Figs. 8 and 9) was generated by moving the source backward (downline) 5 ft (1.5 m) and striking the moving plate every ∼5 s (12 strikes total) (Fig. 9). The recording was made over 48 receiver stations (e.g., Fig. 8). A second record was similarly generated with the same source-nearest receiver distance after repositioning the source 5 ft (1.5 m) upline and moving the 48-channel receiver array upline by 1 station (1.5 m). We thus acquired a total of 25 records.

The processing scheme was based on the utilization of the several MASW techniques developed for both active and passive surveys and basically treats the 60 s records like active source data. Our approach adapts a technique (Park and Shawver, 2009) that can process surface waves generated at multiple source offsets while minimizing the chaotic near-field (i.e., closer to the source) effects of surface waves. This multisource offset technique is known to increase the accuracy of surface-wave analysis in a broad depth range of investigation (Park and Carnevale, 2010). The processing scheme can dynamically detect the source characteristics (i.e., angle and distance) from the continuous scanning of surface waves for a wide range of azimuth (Park, 2008). The technique, however, uses only two opposite azimuth angles of ±180° based on the inline surface wave propagation (as suggested by Park and Miller, 2008) that is valid for the linear receiver array used.

High-quality well logs were available for many areas of Schofield Barracks as well as for the other sites in Utah and Hawaii (Figs. 2 and 4). The well data provided relatively detailed stratigraphic information from drill cuttings and engineering data (e.g., drilling rate) and provided critical constraints on interpreting the variability in LWP thickness. Identification of the top of the unweathered basalt was based on a description of rigidity and a transition from reddish-orange to olive-gray to black in the cuttings. Limited basalt bedrock outcrop control was also available in the vicinity of Patsy T. Mink Central Oahu Regional Park (herein Mink Park; Fig. 1).

We present MASW velocity models for each of our sites, comparing the results for single offset, combined offsets, and the new passive/active strategy. Geologic constraints from well data and/or measured sections are discussed where applicable, and the geochemistry of the sampled lateritic material in the weathered zone (LWP) is discussed. Rates of the downward advance of the weathering front are also presented. The overall seismic data quality was good. Signal to noise ratios generally were 80% to about 95%. Shot records showed clear and nearly ideal surface wave dispersion (Fig. 6). Because we used 4.5 Hz phones to preferentially record surface waves, we have limited our analysis in the frequency domain to that frequency.

The combination of multiple source-receiver offsets provided dramatic improvement on the dispersion images, especially in the low frequencies, and thus resulted in a greater investigation depth for the velocity models. Dispersion images from HI line 3 exemplify this improvement (Fig. 7), with other sites exhibiting similar results (see also the reference site at Ice Springs, Utah). This improvement was as much as 20 m additional depth. No single source-receiver offset provided a model sufficiently deep for the investigation of the full zone of interest.

The active/passive source method also increased the depth of the velocity models, but to a greater degree than just using multiple offsets with a purely active source. This method was employed at two sites, HI lines 3 and 6 (Fig. 2). Comparison of dispersion images (Fig. 10) based on an active source only and the active/passive source method demonstrates the improvement in resolving the low-frequency portion of the spectrum (e.g., <15 Hz). The results of the active/passive source surveys were used more for confirming the average thickness of the LWP and less for resolving fine-scale velocity variations.

The site in Utah (Fig. 4) tested the ability of the MASW method to provide velocity images of geologically recent basalt overlain and underlain by lower velocity material, and to provide a geological analog for comparison with the Oahu sites. In contrast to eastern Oahu, the dry climate of Utah has resulted in little to no chemical weathering of the bedrock such that the interface between basalt and valley fill sediments is sharp. Geochemically, the basalts from this location and Oahu are similar and are expected to display similar shear-wave velocities.

A driller’s log from a nearby water well shows a depth to bedrock of 4.6 m while a hand-augered hole right on the survey line showed silt, marl, and a layer of thin volcanic ash over bedrock at 4.7 m (also the depth at which refusal was encountered) (Fig. 11). These data correlate well with our seismic velocity model derived from the combination of four offsets. The velocity for this boundary occurs at ∼500 m/s (Fig. 12A). The water well further indicates a second, deeper bedrock-marl boundary at 38 m, which matches a downward velocity inversion from higher velocity basalt to interpreted lower velocity sediments.

Using only a single source-nearest receiver offset of 35 ft (10.7 m), we were able to image the depth to bedrock, and only a few meters below that. With the combination of the 35 ft (10.7 m), 70 ft (21.3 m), and 105 ft (32 m) offsets, the experiment was able to increase the depth of the model, but it still did not reveal the bottom of the lava flow on the southwest end of the profile. Adding a fourth offset at 140 ft (42.7 m) permitted the picking of a high-confidence dispersion curve that enabled our model to image beneath the lava flow and into underlying valley fill sediments.

At the reference site we also acquired a conventional shallow, high-resolution compressional wave CDP profile coincident with the MASW profile (Fig. 12B) in order to compare the utility of one method versus another. Although the CDP profile does not provide quantitative velocity information to the same degree as the MASW results, it does verify the base of the 4.7-m-thick upper layer of unconsolidated sediments, which matches both the well data and the shear-wave velocity profile. The top of the buried basalt layer, as defined by the reference shear-wave velocity of 500 m/s and by the well data, corresponds to a mostly unreflective zone on the CDP profile, as would be expected from a lack of sharp internal acoustic impedance contrasts within the basalt. However, the sediments underlying the basalt layer (as observed in the nearby deep well and inferred from the MASW profile) are expressed as a zone of layered low-frequency reflectivity matching a zone of shear-wave velocities <500 m/s (Fig. 12B). This exercise provides an integrated geophysical and geological interpretation, while demonstrating that a CDP approach to image boundaries within a basalt layer can be challenging, which indicates the need for a model-based approach.

The location of HI line 3 was chosen for its smooth ground surface and ready access, as well as for its proximity to a logged well. Initial experiments with limited offsets proved to be insufficient at this site, probing a depth of only 18 m. This resulted in no velocity gradients or velocities comparable to our reference site, suggesting that the depth of penetration was poor. The combination of the four offsets produced significant improvements and gave enough depth to image mostly unaltered bedrock below ∼25 m (Fig. 13). The apparent bedrock boundary varies between 26 m and 34 m at the 500 m/s velocity boundary, as inferred from the reference site (Figs. 4 and 12A). This agrees with a nearby well (1-1) and is broadly consistent with well data in the vicinity (2-1) (Fig. 11). A subtle velocity inversion appears at ∼15 m; this may be a result of original textural differences between lava flow types where original mineralogical or textural variations can give difference seismic velocities. For example, the core of an ‘a‘a flow should contain fewer discontinuities (vesicles and cooling joints) even when completely converted to saprolite. Relict preservation of primary flow textures is readily apparent in road and stream cuts at Schofield Barracks and elsewhere on Oahu (Fig. 3).

We also implemented the active/passive source method at this site along the same line of receivers (Fig. 14A). This dramatically improved the depth of investigation to ∼76 m. Using the velocity reference of 500 m/s, bedrock appears at about the same depth as determined from the multiple offset model described here (Fig. 13).

Well data from across Schofield Barracks show a similar overall internal variability (Fig. 11). The nearby well (1-1, within 30 m of the profile) indicates depth to bedrock at ∼33 m, consistent with our velocity model. Another well (2-1) located ∼200 m away from the profile shows a greater depth of ∼42 m.

In order to examine variability in velocity structure, we acquired two lines located ∼3.5 km to the east of HI line 3 (Fig. 2). HI lines 4 and 5 were located ∼100 m apart and surveyed combining only three source-nearest receiver offsets (35, 70, and 105 ft; i.e., 10.7, 21.3, and 32.0 m), which provided a depth of investigation sufficient to reach basalt bedrock. The velocity models show a similar bedrock depth of ∼36 m, but contain variable internal velocities within the LWP (Fig. 15). This suggests that large velocity contrasts in the critical zone can occur over rather short distances (e.g., 100 m). These profiles are also within the range of expected depths to basalt, as suggested by the well data in the greater vicinity (Fig. 2).

We chose a third study area, located ∼7.8 km south of our sites on Schofield Barracks (Fig. 2), in order to consider larger-scale variability on Oahu. This experiment was conducted on a drier part of the island, with ∼800 mm/a of rainfall (Figs. 1 and 2). The critical zone should be significantly thinned due to climate differences reducing weathering rates. Because the survey was conducted at a large park (Mink Park) near a logged well (Fig. 2), the field conditions were ideal. The acquisition and processing and modeling parameters were identical to those for HI line 3. Interpreted bedrock appears at ∼15 m depth, which is about half the depth in the profile for HI line 3 for this boundary. Both the active/passive source and the combination of four different offsets show the top of basalt bedrock at about the same depth (Figs. 14B and 16).

The park site provided an opportunity to directly measure the depth to unweathered bedrock with two independent geologic constraints: nearby well data and a measured section from a nearby road cut. The U.S. Geological Survey well, located 160 m north of our seismic profile, showed the depth-to-bedrock at 20 m, which closely matches our velocity model, considering our survey line was ∼5 m below the top of the hill on which the well is located (Figs. 1 and 2). A measured section along a road cut located ∼600 m to the north along Highway 99 (in a valley eroded by Kipapa Stream) gave a similar thickness of ∼20 m, also starting 5 m higher than the survey location in the park. Both MASW experiments thus correlated well with the two geologic constraints.

Having derived the depth below ground surface to the top of basalt bedrock, we used these data to compute simple weathering rates. Weathering rates were calculated based on observed depth to bedrock from wells, the one measured section along near Kipapa Stream, and our shear-wave velocity models by dividing LWP thickness by the youngest age for Koolau basalts (Table 1). The youngest end of the age range was used because profiles were acquired on the flat, uneroded surfaces on the interior part of the island. The areas of our surveys and the well locations are not dissected by streams nor are they accumulating sediment. Thus, we can infer that the underlying bedrock should be relatively young Koolau basalt. The weathering rate calculations are also based on the well-founded assumption that below the soil zone proper, saprolite formation does not result in a significant change in volume. The range for rate of advancement of the weathering front on Schofield Barracks is 0.010–0.026 m/ka. Weathering rates calculated from Mink Park show a slower rate of advancement of the weathering front between 0.005 and 0.011 m/ka. The Mink Park location is drier (∼800 mm/a) whereas the Schofield Barracks locality is wetter (∼1500 mm/a) (Figs. 1 and 2); accordingly, the depth to rigid basalt bedrock for the former is roughly half that of the latter.

The results of our experimentation with different seismic methods show that an accurate shear-wave velocity model of the subsurface within a thick LWP can be obtained with improved field procedures. Geologic constraints confirm interpretations from the MASW velocity models. The Utah reference site for relatively unweathered basalt with shallow burial guided interpretations in Oahu, including establishing a shear-wave velocity threshold for picking the depth of unaltered basalt. A small crew of three to four individuals can set up an experiment and collect the data in 4–6 h.

The velocity models and well data in this study imply that large changes in the shear-wave velocity of the critical zone can occur over short vertical and horizontal distances (Figs. 11–16). Velocity gradients are a proxy for stiffness or shear strength, which in turn represent changes in cohesiveness (i.e., mineralogy), porosity, and other discontinuities. Wells, if carefully logged or cored, provide one-dimensional descriptions; however, no lateral information is obtained unless a drilling campaign employs closely spaced wells. The modified versions of MASW methods presented herein can be used to overcome this limitation, and are especially useful where conventional seismic methods (e.g., CDP) may not be effective (Xia et al., 1999; Hunter et al., 2002). For example, during site evaluation for road construction, Sirles et al. (2013) found that the variations of depth to bedrock on Oahu were larger than expected, and detected significant lateral heterogeneity in weathering. From initial borings, engineers were surprised by (1) the existence of soft soils, (2) a highly variable depth to bedrock, and (3) depth to bedrock exceeding the predicted 30 m. As a result, a continuous geophysical evaluation of the subsurface was needed.

We note that the 500 m/s reference shear-wave velocity used for basalt bedrock is lower than the expected velocity of deeply buried unweathered basalt as typically measured by direct methods (e.g., Redpath, 2007). Other researchers have similarly measured relatively low velocities for shallow basalt bedrock. For example, Sundararajan and Seshunarayana (2014, p. 3798) used MASW combined with active source methods to measure the shear-wave velocity of shallowly buried Deccan basalts in the Jabalpur area of central India; they found the weathered zone above a basaltic bedrock to be 350–750 m/s, whereas the basalt bedrock was “about 750 m/s or more.” As part of a study to assess the shear-wave velocity models used as input for the National Earthquake Hazards Reduction Program seismic hazard maps for the Big Island of Hawaii, Wong et al. (2011) reported shear wave velocities for approximately 20 sites ranging from 271 m/s to 580 m/s with a log mean value of 400 m/s (Vs30, i.e., averaged over upper 30 m of the subsurface). In the very driest portions of the island, weathering zones are thin (ca. 1 m), yet Vs30 is consistently less than 580 m/s, producing velocities that are comparable to our sites in Oahu and Utah and for our reference value of 500 m/s.

Because the bedrock in the study sites is relatively shallow and well above the depth below which fractures and porosity would be expected to be sealed (thus increasing velocity), the values we find are a minimum. Further, the strongly dispersive nature of Rayleigh waves in shallow basalt bedrock makes exact correlation to direct measurements of velocity (e.g., from a drill hole) difficult. In spite of this, there is internal consistency within the MASW models that give similar velocities for each area. Casto et al. (2010) noted that higher mode surface-wave analysis is required in order to obtain accurate shear-wave velocity measures of bedrock. Although we have not attempted to use them in our profiles, dispersion trends at the high frequencies in some of our field records (e.g., 100–125 Hz) indicate that deep bedrock may have a shear-wave velocity as high as 3650 m/s, comparable to expected velocities for deep, non-fractured, and unweathered basalt. Thus, future improvements in data reduction routines may permit more accurate representations of shear-wave velocity values for deep bedrock from MASW. In the meantime, internally consistent use of the fundamental mode of the surface wave permits recognition of the critical zone–bedrock boundary even if inferred bedrock velocities are somewhat underestimated.

We compiled a small set of well log data in combination with MASW results in order to assess the thickness of LWPs as a function of mean annual rainfall for Oahu (Fig. 17). All of the data shown are for locations on flat, interfluvial surfaces where the underlying basalt is from the Koolau volcano. Although the data are not well correlated, they suggest a possible concave-upward trend where LWP thickness may reach a maximum thickness despite increasing rainfall. The data also suggest a minimum annual rainfall between 500 and 1000 mm/a, below which saprolite development is very slow, presumably due to low plant cover, depressed soil CO₂, and the unavailability of water to drive weathering reactions. Figure 17 also suggests that, for a given mean annual rainfall, LWPs along an isohyet may exhibit wide variability in thickness, consistent with our observations as well as with those of Sirles et al. (2013). At 1000 mm/a, it appears possible for LWP thickness to vary from <10 m to >40 m. Thus, simple predictions concerning LWP thickness cannot be made on the basis of rainfall for bedrock of a given lithology. We speculate that heterogeneity in permeability produces differential flux rates through developing LWPs. Whatever the origin, it seems necessary to actually measure LWP thickness, whether the motivation is for engineering considerations or evaluating the nature and variability of chemical weathering rates.

There remains much uncertainty about the relationship between denudation rates and critical zone thickness, and how thickness varies in different climate settings (Hilley et al., 2010). One attempt to address the issue of determining weathering zone thickness is to use a derived relationship between erosion rates and relief (Montgomery and Brandon, 2002; Hilley and Porder, 2008), assuming the local groundwater system may restrict the thickness of the critical zone. While this relationship works well in most environments (Hilley et al., 2010), areas of high relief and extreme chemical weathering rates as found on Oahu were shown to be incompatible with field measurements.

Chemical weathering rates in young rocks in a tropical climate may temporarily outpace erosion rates and lead to a thicker than expected weathering zone. Another complicating factor in ocean island settings is the vertical compartmentalization of shallow circulating groundwater systems that sustain perennial stream flows and much deeper groundwater systems of the Ghyben-Herzberg lens. In Oahu, these two systems may be vertically separated by hundreds of meters of unaltered rock, where shallow perched systems are more closely related to saprolite development (e.g., Nelson et al., 2013).

The rates of advancement of the weathering front at Schofield Barracks (0.010–0.026 m/ka) are similar to, but lower than those determined by Nelson et al. (2013; 0.05 m/ka), who employed calculations based on leaching of SiO₂ and averaged over large areas. However, that study accounts for more than just the downward rate of advance of the weathering front. In addition to shallow weathering, the calculations of Nelson et al. (2013) account for weathering reactions in the deep aquifer due to extended contact times with unweathered basalt within the Ghyben-Herzberg lens and an assumption of steady-state removal of soil. Thus, the weathering rates inferred in this study are consistent with other work, yet can help discriminate the various mechanisms of denudation.

The drier climate conditions from Mink Park result in a slower rate of weathering compared to the wetter climate of Schofield Barracks. This correlation between weathering rates, weathering zone thickness, and climate is likely to exist across the island and may be confirmed with further work. However, an examination of measured LWP thickness versus rainfall shows considerable variability.

The velocity models show lateral and horizontal variability that is not readily interpretable based on the available geological constraints from nearby well data, and some of the fine-scale variability seems likely to be an artifact of the inversion. Nevertheless, the occasional large outcrops exposed in deep stream cuts (e.g., Fig. 3D) bear testimony to the strong variation in texture that can occur even though the basic rock type is essentially the same. Figure 18A provides a generalized conceptual model of how velocity data, well data, and saprolite alteration may interrelate. Given the range of textures observed in basalt outcrops on Oahu, a broad spectrum of geological explanations is possible for the velocity variations. For example, the shallow low-velocity zone is easily interpreted as part of the soil zone. A high-velocity zone within an LWP could represent the core of an ‘a‘a flow, and the velocity inversion beneath it the transition back to altered pahoehoe (Fig. 18B). Figure 15 also shows examples of deeper velocity inversions, including a low-velocity zone embedded in faster interpreted saprolite. The core of an ‘a‘a flow, with no vesicles and widely spaced cooling joints, would be expected to have different acoustic properties compared to a sequence of pahoehoe flows with vesicles, closely spaced fractures, and flow contacts. This would also be true of tephra or rubble zones at the top and bottom of ‘a‘a flows. These differences may be retained even after strong chemical weathering.

Our study provides much needed research on the determination of volcanic rock weathering zone thickness and variability at a local scale, which can lead to a greater understanding of chemical weathering rates as a function of climate. Our study expands the use of MASW beyond typical geotechnical and engineering applications to investigate geologic problems not amenable to conventional seismic methods. MASW models show potential variations possibly caused by textural differences between flow types and degrees of chemical weathering. The results provide constraints for deriving the rate of advancement of the weathering front and how it relates to climate variations. This study also furthered the technology of MASW with innovative experimental designs that improve the quality and confidence of the shear-wave velocity models and simplify the field procedures. The method outlined herein can be expanded to the study of weathering in other basaltic regions and potentially to other geologic environments where velocity gradients dominate over discrete impedance contrasts.

This research was supported in part by funding from the College of Physical and Mathematical Sciences at Brigham Young University. Data processing and visualization were made possible by a generous software grant from the Landmark (Halliburton) University Grant Program. We thank Michael F. Weber and students from Brigham Young University-Hawaii for their assistance in the field work.