The Antarctic Drilling Program (ANDRILL) successfully drilled and cored a borehole, AND-1B, beneath the McMurdo Ice Shelf and into a flexural moat basin that surrounds Ross Island. Total drilling depth reached 1285 m below seafloor (mbsf) with 98 percent core recovery for the detailed study of glacier dynamics. With the goal of obtaining complementary information regarding heat flow and permeability, which is vital to understanding the nature of marine hydrogeologic systems, a succession of three temperature logs was recorded over a five-day span to monitor the gradual thermal recovery toward equilibrium conditions. These data were extrapolated to true, undisturbed temperatures, and they define a linear geothermal gradient of 76.7 K/km from the seafloor to 647 mbsf. Bulk thermal conductivities of the sedimentary rocks were derived from empirical mixing models and density measurements performed on core, and an average value of 1.5 W/mK ± 10 percent was determined. The corresponding estimate of heat flow at this site is 115 mW/m2. This value is relatively high but is consistent with other elevated heat-flow data associated with the Erebus Volcanic Province. Information regarding the origin and frequency of pathways for subsurface fluid flow is gleaned from drillers' records, complementary geophysical logs, and core descriptions. Only two prominent permeable zones are identified and these correspond to two markedly different features within the rift basin; one is a distinct lithostratigraphic subunit consisting of a thin lava flow and the other is a heavily fractured interval within a single thick subunit.


During the austral summer of 2006–2007, the Antarctic Drilling Program (ANDRILL) successfully drilled a borehole, AND-1B, beneath the McMurdo Ice Shelf (Naish et al., 2007a). The ice cover was ∼82 m thick and water depth was 855 m. Total depth of the borehole reached 1285 m below seafloor (mbsf) with 98 percent core recovery of sedimentary rocks for detailed study of climate and ice-sheet dynamics (Naish et al., 2007b). The drillsite was situated on the northwest corner of the Ross Ice Shelf, ∼9 km southeast of Hut Point Peninsula (Fig. 1), and above a flexural moat basin that surrounds Ross Island (Horgan et al., 2005). This basin was formed by loading of the crust from the basaltic volcanoes that constitute Ross Island within the tectonic framework of the Victoria Land Basin, a region of late Cenozoic crustal extension of the West Antarctic Rift System (e.g., Wilson, 1999; Hall et al., 2007; Henrys et al., 2007). Characterizing the marine hydrogeology at this site requires an understanding of fine-scale heat-flow patterns and structural details in order to evaluate moat-wide circulation processes, thermal refraction along the volcanic edifice, and channelized fluid flow (Harris et al., 2000a, 2000b).

A comprehensive downhole measurements program was devised for this borehole to complement ANDRILL's scientific objectives (Naish et al., 2006), and activities included the acquisition of numerous geophysical logs and vertical seismic profiles. Morin et al. (2007) present a detailed summary of logging operations that includes the technical specifications for logging tools. As part of the effort to contribute to the sparse heat-flow database for the region and gain additional insight into fluid-transport processes associated with these sub-basin environments, several of the geophysical logs are analyzed to quantify heat flow at this site and to investigate the hydrologic properties of the sedimentary rocks. Specifically, these measurements consist of three sequential temperature profiles as well as caliper, formation resistivity, acoustic televiewer and natural-gamma logs. Selected physical properties measured on core (Niessen et al., 2007) and detailed descriptions of core (Krissek et al., 2007) are also utilized to aid in the analyses.


The ambitious ANDRILL program of drilling and coring lasted for ∼60 d at the AND-1 site. A PQ drillpipe (85.0 mm inside diameter – I.D.) was used to a depth of 238 mbsf, followed by a HQ drillpipe (63.5 mm I.D.) to 702 mbsf and a NQ drillpipe (47.6 mm I.D.) to the total depth of 1285 mbsf (Falconer et al., 2007). Drilling fluid consisted primarily of seawater mixed with potassium chloride and polymers. Once these operations were concluded, the borehole was made available for downhole measurements. However, these in situ measurements could only reach a depth of ∼1000 mbsf because of the length limitations of the logging cable.

Geophysical logs included a succession of three temperature logs recorded over a span of 5 d; these are illustrated in Figure 2 with an expanded view across a 100-m section shown in Figure 3. A natural-gamma log, stacked and corrected for signal attenuation through the drillpipe, is also presented for lithologic reference (Williams et al., 2008). The temperature sensor housed within the logging tool has a resolution of 0.007 °C. Due to extensive bridging in an unstable borehole at the start of logging operations, a NQ drillpipe was lowered into the open hole below the base of the HQ drillpipe (set at 690 mbsf from the original 702 mbsf) in order to permit access with logging tools. The first temperature log collected on 28 December 2006 was run inside this pipe. The second log was recorded ∼11 h later after the NQ pipe had been retracted and the open hole remained stable. The last temperature log was recorded 3 d after this second one on 1 January 2007, but the logging tool could only be lowered to a depth of 816 mbsf before encountering an obstruction that prevented it from descending further. All three logs consistently registered a temperature at the seafloor of −1.68 °C, a value that is in good agreement with hydrographic data obtained from the water column at nearby locations that indicate bottom-water temperatures of about −1.89 °C (Barrett et al., 2005).

In general, the three successive temperature logs (early, intermediate, and late time) depict a fairly rapid rebound toward equilibrium geothermal conditions, possibly an effect of relatively small borehole diameters that restrict initial cooling (Langseth, 1990). The profiles are interspersed with sharp anomalies that typically dissipate over time and may no longer be apparent in the late temperature record (Fig. 3). These small drifts in temperature may represent instances of transient fluid exchange between the borehole and the formation (e.g., Drury et al., 1984; Ge, 1998; Šafanda et al., 2007) or localized intervals of fluid convection within the borehole (e.g., Diment, 1967; Fisher and Becker, 1991). Substantial thermal anomalies appear at 647 mbsf, a consequence of circulation loss and drilling fluid entering the surrounding formation within a fracture zone delineated by a thin lava flow (Falconer et al., 2007; Krissek et al., 2007). Below this depth, the temperature profiles do not recover sufficiently from this fluid invasion to follow a clear and systematic return to thermal equilibrium over a 5-d period. Moreover, the borehole is open below 690 mbsf and temperatures are more susceptible to disturbances caused by subtle fluid movement between permeable zones that is not hindered by the presence of drillpipe. Therefore, the following examination of temperature re-equilibration with time is confined to the uppermost part of the borehole between the seafloor and 647 mbsf.

The three temperature records obtained across this interval are analyzed to arrive at values of the true, undisturbed temperature at selected depths in the borehole from which an overall estimate for the background geothermal gradient can be determined for this site. Specific depths were chosen where the temperatures displayed a relatively smooth transition between successive profiles and where extraneous thermal disturbances appeared to be minimal; two examples of these locations are depicted as horizontal dashed lines in Figure 3. There are numerous techniques for extrapolating a series of transient temperature records to infinite recovery time, thereby computing the true formation temperature from in situ measurements. Hermanrud et al. (1990), for example, present an overview of 22 different methods along with their underlying assumptions. Of these, the line-source method of Horner (1951), as implemented by Lachenbruch and Brewer (1959), is relatively precise even though it does contain an inherent bias to lower temperatures that reduces accuracy slightly (Beardsmore and Cull, 2001). A detailed study describing the proper application of this method has recently been reported by He et al. (2008).

In this method, a Horner plot is constructed from the following expression using temperature data obtained at any particular depth z: 
where Tlog = downhole temperature obtained from temperature log (°C), Ttrue = true formation temperature (°C), Q = rate at which heat is being supplied to the borehole (W/m), λ = thermal conductivity of surrounding rocks at depth z (W/mK), Δt = time elapsed from cessation of mud circulation to actual measurement (hr), and tdc = time elapsed between end of drilling and end of mud circulation (hr). Data from the three temperature profiles shown in Figure 2 and corresponding to three depths are presented in Figure 4 in the form of a Horner plot. Temperature is plotted versus the logarithmic term in Equation (1), and it is not required that the parameters Q and λ be known. A linear best-fit is drawn through the points and its intercept at the y-axis represents the true formation temperature at infinite recovery time (Δt = ∞).

The true temperature at each depth is based upon the spread of three points from which a regression line is constructed and a correlation coefficient determined. In this ANDRILL case, having three points (three temperature logs) instead of only two greatly reduces the uncertainty of this technique (e.g., Kreyszig, 1983). However, estimating values for Δt and tdc is problematic because some drillsite operations that continued after drilling and coring had formally ceased may have produced disturbances to the borehole fluid and the thermal environment. These activities included intermittent circulation of drilling fluid to wash the borehole wall in preparation for acoustic televiewer logging, efforts to clear obstructions in the open hole below the base of the HQ drillpipe, and attempts at cutting the HQ drillpipe (Falconer et al., 2007). Consequently, values of Δt and tdc are approximate and contribute to any error in estimating Ttrue. However, their placement in the logarithmic term of Equation (1) and their eventual use in the best-fit approximation of the Horner method tend to diminish this uncertainty.

A total of 19 depths were selected in which the three temperature records displayed a smooth transition from one profile to the next, as illustrated in Figure 3; the extrapolated values of the true temperatures are plotted as solid points in Figure 2. A regression line drawn through all points intercepts the seafloor at a temperature of −2.4 °C and represents the temperature distribution as a function of depth in this borehole (correlation coefficient R = 0.988). This exercise defines a linear geothermal gradient of 76.7 K/km at this site. An estimation of the temperature gradient between each two adjacent points and across the depth interval between them is presented in Figure 5A, where deviations from the average value (vertical dashed line) reflect variations in the bulk thermal conductivity of the surrounding rocks. Coincidently, a natural-gamma log is presented in Figure 5C to illustrate lithologic variability. For comparison, Bücker et al. (2001) present a compilation of geothermal-gradient data for the Transantarctic Mountains and Victoria Land Basin, with values ranging from 24 to 80 K/km.


Thermal conductivity was not directly measured on core samples from the AND-1B borehole. Referring to Equation (1) and the Horner plots (Fig. 4), λ is indeterminable from the slope of the regression line because Q is unknown. As an alternative method, Lee et al. (2003) proposed a finite-element solution for deriving the thermal conductivity of rocks from borehole thermal-recovery data. Jones and Pascal (1994) developed a numerical model that considers the effects of grain size and orientation on the thermal conductivity of composites. For our purposes, however, values of thermal conductivity, λ, are more effectively estimated from empirical mixing models of components and evaluated against data reported for comparable sedimentary rocks.

Numerous mixing models have been proposed to compute the thermal conductivity of sediment mixtures based upon packing arrangements and particle geometries among matrix components (e.g., Birch and Clark, 1940; Hashin and Shtrikman, 1962; Crane and Vachon, 1977; Pribnow and Sass, 1995). These can be broadly separated into a few categories (e.g., Beardsmore and Cull, 2001): (1) harmonic mean or series combination, (2) arithmetic mean or parallel combination, (3) geometric mean or random mixture, and (4) square-root mean. We prefer to use either category (3) or (4) because of the nature of the rocks encountered in this borehole, where particles are most likely to be randomly orientated and distributed (Pompilio et al., 2007). Some investigators (e.g., Woodside and Messmer, 1961; Lovell, 1985; Vasseur et al., 1995) have concluded that the geometric-mean model (3) is adequate for predicting experimental data. Roy et al. (1981), on the other hand, preferred the square-root mean model (4), arguing that it was derived from stronger physical principles than the others.

The geometric-mean model, in general terms, is: 
where λb = bulk thermal conductivity of mixture (W/mK), λi = thermal conductivity of component i (W/mK), and vi = volume fraction of component i (dimensionless). In order to accurately apply Equation (2) to compute λb, the mineral composition of the sedimentary rocks must be known. Core descriptions (Krissek et al., 2007) identify 16 separate lithostratigraphic subunits that extend from the seafloor to 647 mbsf and consist of varying amounts of diamictite, mudstone, sandstone, and diatomite, among other constituents. Rather than assigning a value of thermal conductivity for each mineral associated with these deposits, we assume that the solid fraction of the sediment mixture can be represented by a single value for the thermal conductivity of the grains.
Thus, the geometric-mean expression for a saturated and isotropic binary mixture of grains and seawater reduces to: 
where λg = thermal conductivity of solid grains (W/mK), λf = thermal conductivity of fluid, in this case seawater (W/mK), and Φ = porosity (dimensionless). Also, under these assumptions, the square-root mean relation is: 

A representative value for λg can be constrained by examining similar analyses of thermal conductivity performed for comparable sedimentary rocks and marine sediments. As part of the Cape Roberts Drilling Project based at the western edge of the Victoria Land Basin, a heat-flow study conducted by Bücker et al. (2001) consisted of recording borehole temperatures in Hole CRP-3 and measuring the thermal conductivity of core samples. An average bulk thermal conductivity was determined to be 2.1 W/mK. Many of the lithologic units identified in the Hole CRP-3 (Powell et al., 2001) are similar to those encountered in the AND-1B borehole, though porosities appear to be generally lower and average ∼0.25 (Jarrard, 2001). Substituting these values of λb and Φ into Equations (3) and (4) yields estimates for the grain thermal conductivity, λg, that range from 2.8 to 3.2 W/mK and are mainly a function of quartz content. These values are in good agreement with those reported by Villinger et al. (1994) for marine sediments, where values of λg between 2.6 and 3.2 W/mK were found to predict temperature profiles accurately. Consequently, we use this range of values for the grain thermal conductivity (2.8 - 3.2 W/mK) in our computation of bulk thermal conductivity for AND-1 sedimentary rocks.

Sediment porosities, as illustrated in Figure 5B, are computed from density measurements performed on core (Niessen et al., 2007) and have an average value of ∼0.50 across the uppermost 650 m. The sedimentary rocks having the highest porosities are the diatomites, due to the hollow nature of diatom frustules. These are clearly delineated in the natural-gamma log (Fig. 5C), where high radiation values generally correspond to sub-glacial diamictite and low values to open-water diatomites (Williams et al., 2008). The diatomites also have the lowest quartz content.

Assuming that the vertical flow of heat through the ocean floor remains constant, variations in porosity with respect to depth tend to mimic the vertical distribution of the temperature gradient (Fig. 5). Conversely, the profile of bulk thermal conductivity, λb, which is negatively correlated to porosity, mirrors that of the temperature gradient. With Φ equal to 0.50 and λg ranging between 2.8 and 3.2 W/mK, the geometric-mean model (Equation 3) and the square-root mean model (Equation 4) yield an average value for λb of 1.5 W/mK ± 10 percent. The estimates of bulk thermal conductivity for sandstones and siltstones in this borehole are slightly lower than those reported by Bücker et al. (2001) for similar sedimentary rocks because of generally lower porosities encountered at Hole CRP-3 (Jarrard, 2001). Correspondingly, estimates of thermal conductivity for diatomites are generally higher than those reported for marine siliceous oozes (Morin and Silva, 1984; Lovell, 1985) because diatomite porosities are substantially lower. Therefore, once adjustments are made for differences in porosity, values of λb in the AND-1B borehole are comparable to other reported values.

It should be noted that the thermal conductivities of the grains and of seawater are also dependent upon temperature. Since in situ temperatures reached 70 °C at a depth of ∼1000 mbsf (Fig. 2), changes in these constituent conductivities are expected. However, a slight increase in the thermal conductivity of seawater (Lawson et al., 1959; Yusufova et al., 1978) with increasing temperature is counteracted by a slight decrease in the value of λg (Birch and Clark, 1940; Sass et al., 1992). Given the relative error in the estimation of λg, the uncertainty related to temperature effects is negligible.

The product of this average value of bulk thermal conductivity and the best-fit linear geothermal gradient yields an estimate of 115 mW/m2 for the heat flow at this site (Fig. 1B). This value is higher than most heat-flow measurements reported along the margin of the Victoria Land Basin (Blackman et al., 1987; Bücker et al., 2001), but is consistent with the elevated heat flow (Sato et al., 1984; Della Vedova et al., 1992) associated with the McMurdo Volcanic Group, Erebus Volcanic Province (Kyle, 1990), and Terror Rift (Cooper et al., 1987; Salvini et al., 1997). The spatial distribution of heat-flow values for the region is displayed in Figure 1A and summarized in Table 1.

The high heat flow may be a direct manifestation of a thermal anomaly within the upper mantle and beneath Ross Island (Watson et al., 2006), and also is typical of the effects of crustal extension and thinning on temperatures within rift basins (Jaupart and Mareschal, 2007). For example, high heat flow values greater than 110 mW/m2 have been reported for the Basin and Range Province in southwestern United States (Sass et al., 1994). These are not uncommon and are directly attributable to rapid rifting processes. The heat flow measured at the AND-1B borehole may also be directly affected by hydrothermal circulation sustained through recharge and discharge along the nearby volcanic edifice; this effect would not be prominent along the margin of the Victoria Land Basin.


Because plans to run straddle-packer tests (Naish et al., 2006) were not implemented during this project, no in situ measurements of hydraulic conductivity (e.g., Becker and Davis, 2004) were made in this borehole and no values of this property can be accurately assigned to various lithostratigraphic units. Nevertheless, information regarding the nature and frequency of permeable zones within the sedimentary moat may be gleaned from drillers' records of lost circulation of drilling fluid into the formation combined with supplemental downhole data. Falconer et al. (2007) noted only two instances of significant loss of circulation, and these were later confirmed from reviewing detailed records of fluid pumping and corresponding return rates at the drill rig. These zones appear (1) at 647 mbsf corresponding to a thin lava flow, and (2) at 815 mbsf where return of drilling fluid was completely lost and never totally regained for the remainder of drilling and coring operations. Both of these permeable zones are delineated by conspicuous shifts in the temperature logs (Fig. 2).

The upper permeable zone (647 mbsf) was cased with HQ drillpipe and remained so during all phases of geophysical logging. Consequently, no acoustic televiewer logs (Zemanek et al., 1970) could be obtained across this section to inspect the borehole wall visually. However, this depth corresponds to a major lithostratigraphic subunit LSU 5.2 (Krissek et al., 2007) that is also reflected in a marked shift in the natural gamma log (Morin et al., 2007). LSU 5.2 is composed of a 2.8-m-thick phonolitic lava with tiny shear fractures related to the early stages of cooling and larger voids related to viscous flow and ductile-brittle transition during the later stages of lava formation (M. Pompilio, 2009, personal commun.); a split-core photo of this rock is shown in Figure 6A. These fractures and voids likely account for the loss of circulation and high permeability at this depth.

The lower permeable zone (815 mbsf) is within the 139-m thick lithostratigraphic subunit (LSU 6.1) that extends from 759.32 to 897.95 mbsf, and consists of a range of different lithologies including mudstone, diamictite, sandstone and conglomerate/sandy breccia (Krissek et al., 2007). It is located in the openhole section below the base of the HQ drillpipe and was accessible for logging with the acoustic televiewer. The magnetically oriented image of the borehole wall at this depth, corrected for local magnetic declination, is presented in Figure 7 alongside coincident profiles of caliper, early temperature, formation resistivity (induction), and natural-gamma activity. A large fracture, washed out to a width of almost 1 m at the borehole, is detected in both the televiewer and caliper logs that strikes approximately east-west and dips roughly 60° to the north. A split-core photo of the rock just below this interval is shown in Figure 6B. The rock is highly fractured, with a missing section from 816.48 to 816.64 mbsf, and no core recovery from 816.82 to 819.19 mbsf. Considering uncertainties in the depth correlation between downhole measurements and drillers' depths (roughly +/− 1 m) due to a combination of factors such as cable stretch, pipe stretch, and a possible error in the circumference of the encoder wheel from which cable depth is determined (Morin et al., 2007), the marked absence of core likely corresponds to the large gap in the televiewer image and identifies the source of circulation loss. Slight decreases in the natural gamma and resistivity logs at a depth corresponding to this prominent fracture (Fig. 7) indicate a lack of material (caliper log) needed to generate a signal.

Regrettably, neither permeable zone can be clearly recognized from the surface seismic records; the upper lava flow because it is too thin and the lower fractured interval because of its lack of contrast in physical properties within a thick lithostratigraphic subunit. Consequently, their lateral extents cannot be extrapolated to a larger scale to help develop a conceptual model of moat hydrology and basin fluid-transport. Based upon depositional processes and conditions, the lava flow is probably a local feature (not basin wide) that extends for perhaps several km in length but narrower in width (M. Pompilio, 2009, personal commun.). However, the deeper permeable zone is characterized by a section of heavily fractured rocks that may represent a fault and thereby imply the presence of a broader hydrologic system. Several investigators have reported evidence of channelized water circulation in the oceanic lithosphere (Fisher and Becker, 2000; Géli et al., 2008), where hydrogeological processes may be affected by a variety of factors such as distribution and thickness of sediment cover, location of outcrops, bedrock topography, and spatial distribution of permeability (Davis and Elderfield, 2004). Closely spaced heat-flow data may also help delineate wavelengths and constrain the geometry of circulation cells (e.g., Fisher and Becker, 1995; Harris and McNutt, 2007). However, reliable heat-flow measurements are too sparse within the southern Victoria Land Basin (see Figure 1 and Table 1) to provide the type of fine-scale resolution required to construct realistic submarine hydrogeological models (e.g., Harris et al., 2000b).


Data from three temperature logs recorded over a span of 5 d in the AND-1B borehole were extrapolated to true, undisturbed temperatures; the corresponding thermal gradient from the seafloor to 647 mbsf was determined to be 76.7 K/km at this site. An average value of thermal conductivity for the sedimentary rocks was computed from empirical mixing models constrained to a narrow range of possible constituent thermal conductivities. Profiles of temperature gradient and sediment porosity tend to parallel each other as the thermal conductivity responds inversely to porosity. The corresponding heat flow is estimated to be 115 mW/m2. Although this magnitude of heat flow is higher than typically reported for the Victoria Land Basin on a regional scale, it is in good agreement with more local measurements associated with the Erebus Volcanic Province and its elevated thermal anomaly. Heat-flow values that exceed roughly 100 mW/m2 are common for the Terror Rift and appear to be a direct manifestation of its most recent deformation with crustal thinning and extension.

Two dominant pathways for subsurface fluid flow are identified and these correspond to two markedly different features within the rift sub-basin: (1) a distinct lithostratigraphic subunit consisting of a thin lava flow, and (2) a heavily fractured interval within a single thick subunit. Additional, but less prominent, transmissive zones may also exist as perhaps suggested by several subtle disturbances to the temperature profiles (e.g., Ge, 1998; Birkholzer, 2006) and by drillers' records that indicate slight and sporadic loss of circulation. However, these two zones probably represent the major pathways for fluid flow intercepted by this borehole. Regrettably, their spatial extents are not well constrained, and insight into general hydrologic characteristics cannot be extrapolated to a basin scale that includes a flexural moat and a volcanic edifice.

Coupled heat-transport and fluid-flow processes are expected to be vigorous near Ross Island due to crustal extension and thinning combined with permeable zones located both at lithologic contacts and within lithostratigraphic subunits. However, reliable heat-flow data are sparse in this area and hydrogeologic simulations that provide important insight into the thermal regime require fine-scale, high-resolution measurements. Nevertheless, these new results determined from geophysical logs collected in the AND-1B drillhole provide high-quality data that may be integrated into this regional database.

The authors are grateful to C. Bücker, R. Saltus, Guest Associate Editor T. Paulsen, and an anonymous reviewer for their thorough and thoughtful reviews that improved this manuscript considerably. ANDRILL is a multinational collaboration among the Antarctic programs of Germany, Italy, New Zealand, and the United States. The authors would also like to thank the project operator, Antarctica New Zealand, and Alex Pyne at Victoria University of Wellington, along with the drilling crew of Webster Drilling and Enterprises Ltd. for their tireless and conscientious support at the drillsite that enabled geophysical logs to be collected. This material is based upon work supported by the National Science Foundation under Cooperative Agreement No. 0342484 through sub-awards administered by the ANDRILL Science Management Office at the University of Nebraska–Lincoln, and issued through Northern Illinois University, as part of the ANDRILL U.S. Science Support Program. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. Support for SAH and DH was provided by New Zealand Foundation for Research Science and Technology and the Royal Society of New Zealand Marsden.