Complex speleothem growth processes revealed by trace element mapping and scanning electron microscopy of annual layers

Complex speleothem growth processes revealed by trace element mapping and scanning electron microscopy of annual layers

Geochimica et Cosmochimica Acta, Vol. 69, No. 20, pp. 4855– 4863, 2005 Copyright © 2005 Elsevier Ltd Printed in the USA. All rights reserved 0016-7037...

608KB Sizes 0 Downloads 48 Views

Geochimica et Cosmochimica Acta, Vol. 69, No. 20, pp. 4855– 4863, 2005 Copyright © 2005 Elsevier Ltd Printed in the USA. All rights reserved 0016-7037/05 $30.00 ⫹ .00

doi:10.1016/j.gca.2005.06.008

Complex speleothem growth processes revealed by trace element mapping and scanning electron microscopy of annual layers P. C. TREBLE,†,* JOHN CHAPPELL, and J. M. G. SHELLEY Research School of Earth Sciences, The Australian National University, Canberra ACT 0200 Australia (Received October 19, 2004; accepted in revised form June 2, 2005)

Abstract—Closely-spaced transects measured by excimer laser ablation inductively-coupled plasma mass spectrometry (ELA-ICPMS) at 5 and 32 ␮m spatial resolution are used to generate trace element composition maps (Ba, Sr, Mg, U, Na, P and Al) from MND-S1, a previously studied modern stalagmite from southwest Australia (Treble et al., 2003: EPSL 216: 141). Rainfall at the site is highly seasonal, and trace elements in MND-S1 show strong seasonal variation. Trace element maps show that Ba, Sr, U and Na concentrations coherently follow annual growth layers identified from Scanning Electron Microscopy (SEM) images. The SEM images also reveal that stalagmite growth did not proceed uniformly: growth layers vary in thickness and locally pinch out. Highly preferential crystal growth, determined by nucleation sites left by the previous year’s growth, may be responsible for this uneven growth layering. Differential crystal growth apparently causes variability of trace element concentrations along each annual layer, although additional disequilibrium processes affect Mg, which is less distinctly banded than Ba, Sr, U and Na. Uneven and discontinuous growth layers influence the number of annual cycles, their wavelengths and seasonal amplitudes measured in any one transect. This has clear implications for studies that use annual trace element cycles as chronological markers, growth rate or seasonality proxies. Copyright © 2005 Elsevier Ltd duced. However, studies have predominantly been based upon 1 or 2 linear transects that do not critically examine the reproducibility of the measurements. There are a few exceptions: Roberts et al. (1998), using a 0.15 ⫻ 0.15 mm SIMS imaging aperture, found Sr concentrations cycled consistently with annual layers in a Scottish speleothem while Finch et al. (2003) found more complex relationships between parallel tracks for a South African speleothem. Finch et al. constructed a 0.3 ⫻ 0.3 mm map of Sr concentrations from interpolated SIMS data and showed that Sr concentrations did not form the expected consistent banding that would represent a systematic incorporation of a seasonal Sr signal into the speleothem annual growth layers. They suggested that disequilibrium growth zoning of aragonite crystallites was responsible for the observed heterogeneity but a detailed comparison between the Sr map and the crystal structure was not presented. In this study, we find two-dimensional structure in trace element composition in an annually-layered speleothem mapped using ELA-ICPMS. This has obvious implications for the accuracy of climate proxies constructed from one-dimensional linear transects. Moreover, the processes responsible for such two-dimensional structure remain to be established. Hence, we have examined the geochemical relationship between trace element variations, visible banding, and fine-scale crystal structure in a well-characterised modern stalagmite, MND-S1, from Moondyne Cave in southwest Australia. The trace element, O and C isotope records from MND-S1 were published by Treble et al. (2003, 2005). Treble et al. (2003) compared the trace element data from MND-S1 with the instrumental rainfall record by amalgamating eleven parallel transects, measured by ELA-ICPMS, each 32 ␮m wide and approximately 15 ␮m apart, along the growth axis, into a composite record. Distinct cycles were evident in each of the eleven individual transects and these cycles could generally be

1. INTRODUCTION

1.1. Background The use of geochemical and other characteristics of speleothem annual layers as proxies for climate variation, growth rate and chronology is well established (Genty and Quinif, 1996; Genty and Massault, 1997, 1999; Proctor et al., 2000; Genty et al., 2001; Proctor et al., 2002, Frisia et al., 2003). Annual layers are characterised by colour, luminescence and/or petrographic variations produced by seasonal changes in drip water chemistry including foreign ion particles (Baker et al., 1993; Genty and Quinif, 1996; Frisia et al., 2000). Measurement of annual geochemical cycles has greatly expanded the detailed multi-proxy information available from these layers. Potentially, annual geochemical cycles record climate-sensitive processes such as seasonal variations in soil/rock weathering (Roberts et al., 1998; Fairchild et al., 2001), ground surface bio-productivity (Huang et al., 2001; Baldini et al., 2002; Treble et al., 2003) and rainfall (Finch et al., 2003; Treble et al., 2003; Treble et al., 2005). The interest in speleothem annual geochemical cycles is largely attributable to the wider availability of high-resolution in situ measurement techniques such as secondary ionization mass spectrometry (SIMS), excimer laser ablation inductivelycoupled plasma mass spectrometry (ELA-ICPMS) and synchrotron radiation (Kuczumow et al., 2003). Far smaller volumes of material are analysed using in situ techniques and correspondingly, very detailed geochemical records are pro-

* Author to whom correspondence should be addressed (Pauline. [email protected]). † Present address: Research School of Earth Sciences, Mills Road Building 61, The Australian National University, Canberra ACT 0200 Australia. 4855

4856

P. C. Treble, J. Chappell, and J. M. G. Shelley

Fig. 1. Two transects of Ba concentrations measured over the upper 3.5 mm of MND-S1 using the 32 ␮m circular laser mask with 15 ␮m gaps between transects. Cycles may generally be matched between transects, but arrows indicate missing or double cycles where reproducibility is poor.

matched between transects (Fig. 1 shows two transects of Ba concentration). However, some cycles could not be matched between transects, because of double or missing peaks (Fig. 1), which suggests that the stalagmite growth process may interfere with the preservation of the trace element signal. To investigate this, we mapped trace element concentrations (Ba, Sr, U, Mg, P, Na and Al) using interpolated ELA-ICPMS data. In this article we compare the trace element composition maps with detailed scanning electron microscopy (SEM) images of growth layers in the stalagmite, to characterise the nature of these mis-matched cycles. 1.2. Study Site and Sample Description Stalagmite MND-S1 was taken from a boardwalk in Moondyne Cave in southwestern Australia. The boardwalk was in place from 1911 to 1992 and the age of MND-S1 thus is precisely known. MND-S1 is an ideal stalagmite for investigating the consistency of trace element banding because rainfall in southwestern Australia is strongly seasonal, with over 80% falling between May and October. Mean annual rainfall is approximately 1000 mm, and annual evapotranspiration is 750 mm, with winter monthly evapotranspiration rates of 30 mm/month and summer rates of 100 mm/month (Wang et al., 2001). The transfer of this winter rainfall through the relatively thin roof of Moondyne Cave (15–20 m thick) produces distinct trace element cycles in Ba, Sr, U, Mg, P and Na of approximately 0.2– 0.4 mm wavelength, the total being equivalent to the number of years the boardwalk upon which MND-S1 grew was in the cave. Treble et al. (2003) described the annual trace element cycles and demonstrated that annually averaged P, U and Mg concentrations in MND-S1 responded to a sustained decrease in regional rainfall that began in the late 1960s and which has persisted to the present. Of these elements, P most systematically recorded the rainfall decrease and closely mimicked both the sub-decadal and seasonal rainfall variations throughout the record. Groundwater P may have also influenced the positive relationship between U and rainfall owing to the strong affinity between phosphate and uranyl ions. Mg showed an inverse relationship with rainfall, both on the intraand inter-annual scale. Higher Mg in summer and dry years was attributed to larger amounts of calcite precipitated from solu-

Fig. 2. Cross section of MND-S1 with solid black lines indicating the area shown in Fig. 3 which details the location of analyses. Dashed black lines indicate discarded calcite off-cuts and white spacing indicates material lost due to saw cuts.

tion prior to seepage water reaching MND-S1 resulting in relatively Mg enriched drip water. Ba, Sr and Na were argued to be driven by speleothem growth rate due to the similar trends between the inter-annual concentrations of these ions and growth rate. It was not possible to collect regular drip water from Moondyne Cave apart from one occasion in June 2002. Moondyne Cave is developed in Pleistocene calcarenite limestone that forms a ridge between Cape Leeuwin and Cape Naturaliste in southwestern Australia. As with other karst caverns in the area, Moondyne Cave is considered to have formed syngenetically; i.e. cavern development commenced when the host calcarenite was becoming cemented (Jennings, 1968). The porous calcarenite is overlain by humus-rich sandy soil and dense Eucalyptus forest. MND-S1 is 33 mm tall, 59 mm wide and grew draped over the corner of a wooden boardwalk, 60 m from the cave entrance. The morphology of MND-S1 is less regular than common ‘candle-shaped’ stalagmites owing to its draped growth form (Fig. 2). In cross-section, visible white layers alternate with clear layers, but there are too few visible bands to be annual. A 5 mm wide section for trace element analyses was cut along the axis of maximum growth, which comprises the most regular and thickest layering (Fig. 2). This was the same section examined in Treble et al. (2003, 2005). The location of the mapped section relative to the eleven transects used to construct the 1911–1992 record is shown in Figure 3. The upper 5– 6 mm of MND-S1 corresponding to the early 1970s to 1992, is translucent and contains no colour banding, but wavy annual layers were detected from crystallite patterns, described below. Examination of a thin section under crossed-polarized light shows that stalagmite MND-S1 consists of several large crystals with columnar palisade fabric, formed from the growth and coalescence of numerous smaller crystallites that make-up the

Trace element mapping of speleothem annual layers

Fig. 3. Schematic of central growth axis region of MND-S1. The location of the trace element map is shaded in the upper left corner. The patterned region indicates where measurements for the 1911–1992 composite trace element record were made. The solid line along the youngest growth surface indicates the exposed broken edge and the dashed line is the true final surface. The length of the arrow provides the scale.

columnar crystals, as described by Kendall and Broughton (1978) and Frisia et al. (2000). Crystallites are defined as the smallest building blocks of the crystal structure and may be as small as a unit cell. When the orientations of closely-stacked crystallites are aligned, these crystallites essentially behave as and determine the properties of a single large composite crystal (Kendall and Broughton, 1978). The columnar palisade fabric of MND-S1 is ‘closed’ between growth-years 1911 to ⬃1930, ‘open’ between ⬃1930 to ⬃1970 and ‘closed’ between ⬃1970 and 1992, where ‘open’ and ‘closed’ fabric refers to the lateral coalescence between columnar crystals: closed fabric consists of columnar crystals that are laterally fused, providing a smooth almost featureless surface, whilst the spaces between columnar crystals in open fabric are preserved and often fluid-filled (Kendall and Broughton, 1978). Figure 4 shows closed columnar fabric in the youngest portion of MND-S1. The transition between open and closed fabric is shown in Treble et al. (2005; their Fig. 2). The calcite formed from ⬃1970 onwards is fully fused and featureless apart from cleavage planes, which are visible be-

4857

cause of micro-tearing and plucking of the crystal surface when it was cut and polished. Along the youngest growth surface, adjacent columnar crystal terminations have coalesced into rhombohedral terminations protruding approximately 50 –100 ␮m above the surface of the fused calcite crystal, visible in thin section (Fig. 4). The columnar palisade structure is almost featureless below the youngest cleavage plane (indicated by grey circles in Fig. 4) suggesting virtually complete fusion of column boundaries (Fig. 4). The edges of the thick section analysed for trace elements by Treble et al. (2003) and used in this study were damaged during cutting giving rise to a final surface of larger rhombohedral shapes where the crystal tips were broken at cleavage planes. The youngest cycle was lost in three transects measured by Treble et al. (2003) owing to cutting damage, but this did not compromise the comparison with the instrumental climate record because eleven transects were measured overall. The calcite formed after the rainfall decrease in the late 1960s contains regular wavy lines of small holes, described by Treble et al. (2003), which coincide with troughs in Ba, Sr and U and peaks in Mg annual cycles. The prominence of these wavy lines after the rainfall decrease suggests that they reflect reduced water availability. Drip water flux at the site is observed to be lowest from late summer through autumn (February–May) when effective precipitation is very low, and summer drip water activity has probably become lower since about 1970, owing to the reduction in rainfall. Hence, it seems highly likely that these wavy lines are depositional hiatuses resulting from lower summer drip water. Regardless of their precise nature or exact timing, the hiatuses provide a fortuitous visual template of the annual growth layers across the calcite surface. We present findings concerning the structure of these layers, and further investigate their relationship with trace element concentrations. 2. METHODS

Fig. 4. Crystallites merging on youngest (unbroken) growth surface resulting in reduction and simplification of crystallite terminations. Small dark linear features in lower half of image are fluid inclusions formed between remnant crystallite boundaries in closed columnar fabric. The linear feature running between the grey circles is a cleavage plane. Photograph is of a thin section under cross-polarized light. The thin section was cut from calcite near to the analysis area indicated in Figures 2 and 3 but from the stalagmite flank rather than the apex.

The polished surface of MND-S1 was photographed under reflected light before laser ablation and a template was made of the wavy annual layers. SEM images were created after ELA-ICPMS analysis of the mapped area. SEM imaging of the uncoated specimen was conducted in backscatter mode with the sample chamber at variable air pressures between 10 –20 Pa. Prior to SEM imaging, the section was washed in a solution of 0.4 M KI and 3mM I2 in an ultrasonicated bath, to remove a gold coating used in 18O/16O analyses by SIMS (Treble et al., 2005). The fine structure of the crystallite fabric was revealed after washing with the KI-I2 solution, which is thought to have also scavenged Ca ions. Mg, Sr, Ba and U concentrations were initially measured by ELAICPMS in twelve, adjacent transects approximately parallel to the crystallographic c-axis, using a 5 ⫻ 50 ␮m rectangular slit aperture providing 5 ␮m resolution in the direction of the c-axis. The transects were each 1.4 mm long and the total width of the ablated area was 0.5 mm wide. The laser ablation and quadrupole ICP system at the Research School of Earth Sciences, The Australian National University are described in Eggins et al. (1998) and Sinclair et al. (1998). In this study, masses 26Mg, 43Ca, 88Sr, 138Ba and 238U were measured, each with an integration time of 0.01 s, pulsing the laser at 20 Hz over the surface moving at 1 mm/min on a motorised stage. The 5 ⫻ 50 ␮m laser mask was chosen to maximise sampling in the direction of speleothem growth (200 –250 measurements/cycle). However, the combination of ablation volume, motor speed and laser pulse frequency required in order to obtain satisfactory counting statistics is somewhat arbitrary and similar results can be achieved with various combinations. The same area was later re-mapped using a larger 32 ␮m circular spot aperture, 15 Hz and 1 mm/min (18 transects). The same features in Ba,

4858

P. C. Treble, J. Chappell, and J. M. G. Shelley

Sr, U and Mg were reproduced and masses 23Na, 31P and 27Al were included in the second round of measurements. Each acquisition was performed after pre-cleaning by ablating with a larger laser spot. Background counts were measured at the beginning and end of the mapping and subtracted from sample counts. Instrumental drift was not present over the short measurement duration. No suitable homogeneous carbonate standards were available for highresolution trace element measurements, and we use National Institute of Standards and Technology SRM612 glass (Pearce, 1996) which was measured at the beginning and end of transects in the usual standardsample-standard bracketing technique. The relative standard deviation (RSD; 2␴/mean) calculated from the SRM612 data were: 23Na 3.2%, 26 Mg 13.0%, 27Al 3.2%, 31P 5.4%, 43Ca 3.2%, 88Sr 6.0%, 138Ba 2.8%, 232 Th 4.0% and 238U 3.6%. The different matrices of our standard and sample means that the analytical precision for the carbonate will differ from that of the glass. Signal intensity was higher on the stalagmite for Mg, P, Ca and Sr and consequently the analytical error (RSD) will be lower than for SRM612 for these masses, but is higher for Na, Al, Ba, Th and U, for which the signal intensity was lower. The linear transects were smoothed with a 3-point moving average to reduce the influence of transient spikes in signal intensity before being divided by Ca. The metal/Ca ratio of the unknown is then corrected to concentrations using the known concentrations of SRM612 (Sr: 78.4, Ba: 41.0, U: 37.88, Mg: 85.09, P: 39.1 and Na: 103870 ppm; Pearce, 1996). The high spatial resolution data obtained with the 5 ⫻ 50 ␮m slit were smoothed with a 25-point box smooth, preserving the main features of the annual cycle. The linear transects were then gridded using a linear spline interpolation in a spatial contouring package. Care was taken to not interpolate more data than originally produced, by selecting the same or fewer number of grid nodes as data points. 3. RESULTS

3.1. SEM Images of Annual Growth Layers Annual growth layers, separated by dark wavy stripes, are clearly evident in SEM images of the etched surface (Figs. 5a, b). When polished, the dark stripes appear as lines of small holes (Fig. 5c) and indicate that growth interruption became an annual feature in calcite formed after rainfall decreased in the late 1960s. The growth layers are uneven and semi-continuous (Fig. 5a), and some layers pinch out altogether (e.g. above and to the lower right of the white box in Fig. 5a). Growth layers with uneven lateral thicknesses are observed throughout MND-S1 when the entire section is viewed under an optical microscope. Occasional pinched or discontinuous layers vary in length from 0.25 mm to ⬎5 mm overall, but are far more frequent after the rainfall decrease. The thickness variations and discontinuities in growth layers indicate crystal growth in MND-S1 is more complicated than precipitation of simple layers of uniform thickness. Uneven layer thickness may arise from partial erosion of uniformly-thick layers or from variation of growth across the surface. We cannot rule out the possibility of partial erosion but two observations argue for point to point variability of growth. Firstly, each annual layer appears to be composed of a couplet of crystallites: the earlier phase is comprised of elongated, distinct crystallites while the later phase is comprised of shorter or more completely fused crystallites (Fig. 5b). The change in crystallite morphology mid-way through each band may represent a change in drip water characteristics such as drip-rate, fluid calcite saturation state and concentrations of foreign ions (c.f. Frisia et al., 2000), but we cannot confirm this as we do not have detailed drip water measurements of these variables from Moondyne Cave. In any case, complete couplets would not be preserved if erosion were significant. We have not observed any instances where the first half of the layer appears to

Fig. 5a– c. SEM images showing annual layers marked by growth hiatuses (wavy dark lines running top to bottom). 5a shows an overview with laser ablation tracks (running right to left) indicating mapped area. The two rhombohedral terminations along the left side of 5a are due to breakage and tearing along cleavage planes. 5b is an enlargement of the white box in 5a showing the crystallites in detail and the two holes in 5c (polished section) indicate the physical discontinuity between each growth year. The uneven growth layers demonstrate that annual growth does not proceed by simple even-thickness layering in MND-S1.

have been eroded back. Secondly, as is shown later, the annual cycles of Ba, Sr, U and Na cycles are symmetrical within each layer and show no sign of truncation, which erosion would induce. Growth that varies from point to point may be due to precipitation along channels of drip water that switch as the surface topography of the stalagmite evolves, but the presence of annual couplets suggests the location of crystal growth does not switch until the end of each year’s growth is complete. This in turn implies that the location of each year’s growth proceeds

Trace element mapping of speleothem annual layers

in a preferential manner depending on the crystal substrate. Processes that influence nucleation may be the surface shaped during the previous year’s growth or foreign ions at the surface deposited from drip water or settled from aerosols when driprates are at their lowest (late summer). No Th, Al or Si was detected, indicating that mineral detritus was absent or below ELA-ICPMS detection levels. For more than 5 years preceding as well as during the period when these annual layers regularly appear in MND-S1, Moondyne Cave was closed to tourists and thus their formation could not have been influenced by deposition of lint or lamp smoke. The presence of organic ions cannot be ruled-out as we did not search for these. 3.2. Trace Element Concentration Maps 3.2.1. Ba, Sr, U and Na concentration maps Ba, Sr, U and Na concentration maps are shown in Figures 6a, b, c and d, respectively and show clear patterns of seasonally varying trace element concentrations. With the exception of U, these ions were argued previously to be controlled by MND-S1 growth rate (Treble et al., 2003). A template of the dark stripes separating the physical annual layers described in section 3.1. is overlain on each map and confirms that low Ba, Sr, U and Na concentrations coincide in most, but not all instances, with the annual depositional hiatuses. The general agreement between the physical annual layers and the Ba, Sr, U and Na concentrations confirms that the variability between individual transects seen in Figure 1 results from uneven and discontinuous growth layers. This uneven growth significantly impacts on the wavelengths, amplitudes and number of annual cycles of trace element concentrations in individual transects and clearly shows that a single transect is insufficient to faithfully capture the information contained in MND-S1 trace element concentration cycles. Overall, there is strong coherent seasonal variation in Ba, Sr and U, and to a lesser degree Na, with the highest concentrations occurring midway between the annual growth hiatuses and the lowest concentrations at the hiatuses. Annual cycles of these trace elements within each growth layer are distinct but there are noticeable variations along individual growth layers, amounting to approximately 1 ppm for Ba, 5–10 ppm for Sr, 0.02 ppm for U, and 10 ppm for Na, or approximately 20 –30% of the typical seasonal variation for each element. This variability along each growth layer reduces the reproducibility between closely-spaced ELA-ICPMS measurement transects. 3.2.2. Mg, P and Al concentration maps The Mg concentration map (Fig. 6e) shows that maximum Mg concentrations straddles the annual growth interruption and lower concentrations occur approximately mid-way in each growth layer. This relationship is the inverse of that for Ba, Sr, U and Na (Figs. 6a– d) and was reported previously (Treble et al., 2003). Overall, Mg concentrations in MND-S1 lack the coherent seasonal variation observed for Ba, Sr, U and Na, but this is not due to measurement sensitivity because Mg signal intensity was ⬃25 times higher than Ba intensity. Later investigation confirmed the same results when masses 24Mg, 25Mg and 26Mg were re-measured along one transect. Moreover Mg shows more variation along each growth layer than Ba, Sr, U

4859

and Na, described above, and is approximately 50% of the seasonal Mg concentration variation, compared with approximately 20 –30% for Ba, Sr, U or Na. Aluminium appears only as a stripe (Fig. 6f) that follows a cleavage tear in MND-S1, and probably originated as embedded polishing residue. Phosphorus shows weak banding although not with any clear relationship to the physical annual layering (Fig. 6g). A stripe of high P concentration (55– 60 ppm) coincides with the high-Al stripe and may also reflect contamination (this cleavage tear is outside the section measured by Treble et al. 2003 for the composite 1911–1992 trace element record and does not affect the reported P concentrations). The lack of banding in the P map appears to be due to the smoothing effects of the interpolation method. Phosphorus cycles in the individual transects are more variable than the cycles of Ba, Sr and U. This is consistent with the interpretation made by Treble et al. (2003) that speleothem P is flushed from the soil biomass during high infiltration events. 4. DISCUSSION

4.1. Complex Growth Processes: Implications for Micro-Analyses Seasonal variation in trace element concentration coincides with physical annual growth layers in MND-S1 but point to point variations of crystal growth or possibly crystal erosion have affected the layering stratigraphy and the preservation of annual trace element cycles. These processes, as well as lateral variation in trace element concentrations along each layer, are responsible for the trace element variability between closely spaced transects observed in Figure 1. Complex growth impacts significantly on annual cycle wavelength, number of cycles, and seasonal amplitude in trace element concentrations. Figure 7 illustrates the potential impact by comparing Ba measurements from transects 1 and 12: the two transects are only 0.5 mm apart, but 5 cycles are seen in transect 1 compared with 7 cycles in transect 12. Two growth years appear to be missing in transect 1. This exercise clearly demonstrates that uneven physical layering and lateral variation in trace element concentrations will affect sub-annual measurements. Single measurement transects may not be representative and interpretations of chronology, growth rate, and seasonal concentration range based on a single transect may be flawed. Thus, it is important to examine the reproducibility of annual geochemical cycles on parallel tracks, in speleothem studies. In cases where speleothems clearly contain perfectly parallel annual layering, a demonstration that trace element cycles coincide with these layers may be sufficient but we emphasise that variation along a layer must be investigated also. A thorough investigation would include a comparison of multiple closely-spaced transects of trace element data or mapped sections in order to demonstrate the reproducibility of speleothem annual trace element cycles. 4.2. Growth Processes Affecting Layer Thickness and Trace Element Concentrations Variation in trace element concentrations along each growth layer may be due to the same process that causes layer thickness to vary. Layers in MND-S1 vary in thickness but regularly

4860

P. C. Treble, J. Chappell, and J. M. G. Shelley

Fig. 6. Mapped concentrations of Ba (a), Sr (b), U (c), Na (d), Mg (e), Al (f) and P (g). Black lines are physical layers seen in Fig. 5a. Scales on map edges are mm. Transects measured using the 32 ␮m spot (d, f, g) are shorter than those measured using the 5 ⫻ 50 ␮m slit (a– c, e). Ba, Sr and U show distinct annual concentration cycles between physical layers while Mg concentrations shows significant variability along the growth layers.

consist of a couplet of elongate and shorter crystallites, suggesting that growth processes result in the same pattern each hydrologic year, although varying in rate from point to point. The pinching-out of annual layers (Fig. 5a) suggests that growth can be completely suppressed for a year at some points

on the surface. The growth of MND-S1 over the edge of the boardwalk no doubt contributes to a more complex stratigraphy than that of a regular candle-shaped stalagmite. However, we emphasize that while layers are observed to vary in thickness throughout MND-S1, the pinching of layers only becomes

Trace element mapping of speleothem annual layers

Fig. 7. Ba concentrations along transects 1 and 12 smoothed with a 25-point box filter. Transects 1 and 12 are the top and bottom transects used to construct the two-dimensional Ba composition map, corresponding to 1.1 and 0.6 mm on the y-axis in Fig. 6a. These transects highlight the discrepancy between the number of cycles that would be recorded if only transect 1 (5 cycles) or 12 (7 cycles) were measured.

pronounced following the rainfall decrease suggesting preferential crystal growth mechanisms are influenced by seasonal growth quiescence rather than the original substrate morphology. That growth can be completely suppressed at some points on the surface hints that the location of each year’s growth depends on the surface conditions for nucleation created at the end of the previous year’s growth. Crystal growth may be enhanced at nucleation sites provided by accidents (dislocations) in crystal structure, due to incorporation of foreign ions or mis-match of coalescing crystals (Wenk et al., 1983). Crystal defects in MND-S1 could be generated during summer, when drip rates are at their lowest or cease completely. Furthermore, defects could also be initiated if foreign matter such as clay was deposited on the stalagmite during the summer period of quiescence. However, clay appears to be ruled-out because the clay indicators: Al, Si or Th were not detected. Organic ions, which we did not search for, could coat crystal surfaces, although no luminescence banding was visible under UV light in the mapped section. Another possible mechanism that could influence crystal nucleation is a change in crystallite morphology towards the end of summer, possibly due to drip water chemistry (e.g. Frisia et al., 2000) when drip rate becomes very low or ceases. We have observed that the crystallite tips coalesce or merge near the final surface of MND-S1 (Fig. 4), reducing the number of crystallite terminations from which growth might commence. We suggest a combination of crystal defect density and/or crystallite morphology along the growth surface at the end of a hydrological year may determine the nucleation sites of the following year’s growth. Uneven distribution of these sites across the surface could be responsible for the uneven layer thickness and such processes could be relevant for any speleothem containing annual layers and in particular, annual hiatuses. 4.3. Equilibrium and Non-Equilibrium Trace Element Uptake Growth that varies from point to point across the speleothem surface could contribute to the variable trace element concen-

4861

trations along each growth layer, but presumably all trace elements would show similar patterns of heterogeneity if this were the only process. Mg (Fig. 6e) shows greater variability along each growth layer than either Ba, Sr or U. Percentage wise, seasonal Mg concentrations are relatively subdued (approximately 15% of the mean value as opposed to almost 50% for Ba), however, this does not explain the heterogeneity along the layer. The processes of Mg incorporation into calcite are well studied. Mg2⫹ has a smaller ionic radius than Ca2⫹ and is thought to readily substitute for Ca in calcite (Oomori et al., 1987). Past research indicates that Mg uptake in calcite is proportional to the fluid Mg/Ca ratio when the Mg/Ca ratio is above 7.5 (Mucci and Morse, 1983). Regardless of whether the drip water Mg concentration in Moondyne Cave varies seasonally with residence time of drip waters in the aquifer or otherwise e.g. flushing of Mg salts after major infiltration events, if Mg uptake in MND-S1 was proportional to the Mg content of its drip water, then the MND-S1 Mg map should show little variation parallel to the advancing growth surface. The high degree of variability in Mg concentration across the growth surface suggests that other processes interfere with equilibrium uptake of Mg. We discuss two mechanisms for Mg disequilibrium put forward by other authors. Huang and Fairchild (2001) raise the possibility that disequilibrium partitioning of trace elements between crystal micro-zones, as proposed by Paquette and Reeder (1995), could affect high-resolution measurements of speleothem Mg concentrations. However, as the concentrations of trace elements in drip waters can vary significantly through the year (e.g. Baker et al. 2000), we expect that variations in drip water concentrations would dominate, and any variability created by crystal micro-zoning would be too small to detect. Mucci and Morse (1983) reported a positive linear relationship between fluid and calcite Mg/Ca ratios, resulting in a constant partition coefficient (DMg) of 0.0123 when fluid Mg/Ca ratios exceed 7.5. Thus when fluid Mg/Ca ⬎7.5, the calcite Mg/Ca ratio varies in proportion with that of its fluid. A speleothem growing from such a solution should have relatively constant Mg concentrations at the crystal-fluid contact even if the rate of growth across this surface varied, since Mg uptake is not sensitive to growth rate (Mucci and Morse, 1983). However, when the fluid Mg/Ca ratio is ⬍ 7.5, which is typical for cave drip waters, Mucci and Morse found that DMg increased exponentially between 0.013 to ⬎0.03 as the Mg/Ca ratio decreased. This relative enrichment in the calcite Mg/Ca ratio is attributed to the relative increase of Mg adsorbed onto non-lattice sites compared with Mg incorporated into lattice sites. This effect is not noticed in calcites grown from solutions with Mg/Ca ratios above 7.5 as in this situation, the amount of Mg adsorbed onto non-lattice sites is constant (sites are saturated) and small compared with the amount of Mg taken up into lattice sites which is proportional to the fluid Mg/Ca ratio (Mucci and Morse, 1983). Subsequent studies (Howsen et al., 1987; Huang and Fairchild, 2001) also found elevated DMg values (⬃0.02– 0.03) for solutions with low Mg/Ca ratios, close to those typical of cave waters, although the experimental range investigated by Howson et al. (1987) and Huang and Fairchild (2001) was not broad

4862

P. C. Treble, J. Chappell, and J. M. G. Shelley

enough to confirm the exponential increase in DMg at lower fluid Mg/Ca ratios. A Mg/Ca ratio of 0.2 was measured in a spot sample of winter drip water collected in Moondyne Cave, which is well below 7.5 and thus Mg uptake by adsorption may be significant in MND-S1. The resulting Mg concentration across the growth surface would thus depend on the density of adsorption (nonlattice) sites across this surface. We have previously argued (section 4.2) that the crystal surface properties may be altered when drip rates are low, such that crystal nucleation is favoured at some points and suppressed at others on MND-S1’s surface. The density of adsorption sites may vary also, resulting in variable amounts of absorbed Mg across the surface. We believe that this non-equilibrium process of Mg adsorption is important and may have significantly contributed to Mg heterogeneity along MND-S1’s layers. 4.4. Growth Rate Sensitive Trace Elements and Their Compositional Maps Treble et al. (2003) argued that Ba, Na and Sr in MND-S1 are kinetically controlled by growth rate. These elements have the clearest annual trace element cycles and show coherent seasonal variation in the compositional maps. Fairchild et al. (2001) report annual Na cycles in speleothem calcite grown from drip waters with near-constant Na/Ca ratios, confirming that growth rate can strongly influence at least some speleothem trace elements. Drip rates in southwest Australian caves are observed to be lowest during late summer and an annual hiatus regularly appears after the late-1960s rainfall decrease. We infer from these observations that there is a seasonal difference in growth rate i.e. nil or very low in summer and higher in winter, and thus that an annual growth cycle exists. We believe this annual growth cycle is the primary driver of growth rate sensitive ions and gives rise to the clear annual banding in the compositional maps of Ba, Na and Sr. Secondary to this, smaller variations in growth rate across the growing surface as described in section 4.2, may account for the observed variation along the layers that is approximately 20 –30% of the seasonal variation. Uranium is the only ion that shows clear banding, and that was not identified in Treble et al. (2003) to be growth rate sensitive. However, it was pointed out in Treble et al. (2003) that variation in inter-annual growth rate is probably more subtle than intra-annual growth rate giving rise to the possibility that growth rate may dictate the annual cycle of an element but other processes affecting the mean concentration of that ion in drip water from one year to the next may control longer-term trends. Trace element concentrations driven directly by climatic factors rather than growth rate, which may be a complex function of several variables (Dreybrodt, 1988; Baker et al., 1998), are obviously of greatest interest but interpreting the signal of climate sensitive ions such as Mg and P that may not preserve clear annual cycles becomes difficult. The advantage of the coincidence of growth rate sensitive ions with climatically sensitive ions is that the former provide clear annual chronological markers, at least at sites such as southwest Australia that have a strong and distinct annual rainfall season. The matching of growth rate controlled cycles e.g. Ba between

multiple transects allows seasonal information to be extracted for climate sensitive elements (e.g. P and Mg), which can then be averaged to obtain representative climate signals preserved in stalagmites such as MND-S1 where the annual growth layers are complex (Treble et al., 2003). 5. CONCLUSIONS

Trace element mapping of a stalagmite from southwest Australia, MND-S1, by ELA-ICPMS revealed patterns of Ba, Sr, U, Mg and Na concentrations corresponding to the annual growth layering visible across the stalagmite surface. Trace elements whose uptake is dependent on growth rate produce the clearest coherent seasonal variation within each layer. Uneven and pinched annual layers reveal that calcite crystal growth in MND-S1 is complex and that this significantly impacts on the preservation of annual geochemical cycles. This complex growth is argued to be the result of preferential nucleation that varies from point to point across the surface. Preferential growth may be influenced by crystallite fusing or by foreign ions deposited over the surface during summer, rather than stalagmite morphology. Preferential growth may also contribute to variability in trace element concentrations along the advancing growth front. This may be relevant for any speleothem especially those containing annual growth hiatuses. In addition, Mg displays greater along layer heterogeneity due to non-equilibrium uptake of Mg. This may be because of the relative importance of adsorption sites for calcite precipitated from solutions with low Mg/Ca ratios. Variable crystal growth processes that govern uptake of different trace elements have strong implications for the reproducibility of high-resolution in situ measurement techniques. This study demonstrates that single linear transects analysed by these methods may be insufficient to capture the true trace element variability. Trace element maps or numerous replicate transects are recommended to adequately establish trace element behaviour in speleothems. Acknowledgments—We thank Silvia Frisia, Ian Fairchild, Andy Christy, Henry Schwarcz and Associate Editor Miryam Bar-Matthews for discussion and comments on the physical and trace element banding; Juan Pablo Bernal for the KI etching process; and Bruce Railsback for pointing to Mucci and Morse’s 1983 study. P.T. acknowledges the financial support from an Australian Postgraduate Award and RSES, ANU during analyses and from UCLA during the preparation of this manuscript. Associate editor: M. Bar-Matthews REFERENCES Baker A., Smart P. L., Edwards R. L., and Richards D. A. (1993) Annual growth banding in a cave stalagmite. Nature 364, 518 –520. Baker A., Genty G., Dreybrodt W., Barnes W. L., Mockler N. J., and Grapes J. (1998) Testing theoretically predicted stalagmite growth rate with recent annually laminated samples: Implications for past stalagmite deposition. Geochim. Cosmochim. Acta 62, 393– 404. Baker A., Genty D., and Fairchild I. J. (2000) Hydrological characterisation of stalagmite drip waters at grotte de Villars, Dordogne, by the analysis of inorganic species and luminescent organic matter. Hydrol. Earth Sys. Sci. 4, 439 – 450. Baldini J. U. L., McDermott F., and Fairchild I. J. (2002) Structure of the 8200 year cold event revealed by a speleothem trace element record. Science 296, 2203–2206.

Trace element mapping of speleothem annual layers Dreybrodt, W. (1988) Processes in karst systems. Springer-Verlag, New York. Eggins S. M., Kinsley L. P. J., and Shelley J. M. G. (1998) Deposition and element fractionation processes during atmospheric pressure laser sampling for analysis by ICP-MS. Appl. Surf. Sci. 127, 278 – 286. Fairchild I. J., Baker A., Borsato A., Frisia S., Hinton R. W., McDermott F., and Tooth A. F. (2001) Annual to sub-annual resolution of multiple trace-element trends in speleothems. J. Geol. Soc., Lond. 158, 831– 841. Finch A. A., Shaw P. A., Holmgren K., and Lee-Thorp J. (2003) Corroborated rainfall records from aragonitic stalagmites. Earth Planet. Sci. Lett. 215, 265–273. Frisia S., Borsato A., Fairchild I. J., and McDermott F. (2000) Calcite fabrics, growth mechanisms and environments of formation in speleothems (Italian Alps and SW Ireland). J. Sediment. Res. 70, 1183–1196. Frisia S., Borsato A., Preto N., and McDermott F. (2003) Late Holocene annual growth in three Alpine stalagmites records the influence of solar activity and the North Atlantic Oscillation on winter climate. Earth Planet. Sci. Lett. 216, 411– 424. Genty D. and Massault M. (1997) Bomb 14C recorded in laminated speleothems: calculation of dead carbon proportion. Radiocarb. 39, 33– 48. Genty, D. and Massault, M. (1999) Carbon transfer dynamics from bomb-14C and ␦13C time series of a laminated stalagmite from SW France-modelling and comparison with other stalagmite records. Geochim. Cosmochim. Acta 63, 1537–1548. Genty D. and Quinif Y. (1996) Annually laminated sequences in the internal structure of some Belgium stalagmites—importance for paleoclimatology. J. Sediment. Res. 66, 275–288. Genty D., Baker A., Massault M., Proctor C., Gilmour M., PonsBranchu E., and Hamelin B. (2001) Dead carbon in stalagmites: Carbonate bedrock paleodissolution vs. ageing of soil organic matter. Implications for 13C variations in speleothems. Geochim. Cosmochim. Acta 65, 3443–3457. Howson M. R., Pethybridge A. D., and House W. A. (1987) Synthesis and distribution coefficient of low-magnesium calcites. Chem. Geol. 64, 79 – 87. Huang Y. and Fairchild I. J. (2001) Partitioning of Sr2⫹ and Mg2⫹ into calcite under karst-analogue experimental conditions. Geochim. Cosmochim. Acta 65, 47– 62. Huang, Y., Fairchild, I. J., Borsato, A., Frisia, S., Cassidy, N. J., McDermott, F., and Hawkesworth, C. J. (2001) Seasonal variations Sr, Mg and P in modern speleothems (Grotta di Ernesto, Italy). Chem. Geol. 175, 429 – 448. Jennings J. N. (1968) Syngenetic karst in Australia. In Contributions to the study of karst (eds. P. W. Williams and J. N. Jennings) pp.

4863

41–110. Department of Geography Publication no. G5, Australian National University, Canberra. Kendall A. C. and Broughton P. L. (1978) Origin of fabrics in speleothems composed of columnar calcite crystals. J. Sediment. Petrol. 48, 519 –538. Kuczumow A., Genty D., Chevallier P., Nowak J., and Ro C. (2003) Annual resolution analysis of a SW-France stalagmite by X-ray synchrotron microprobe analysis. Spectrochim. Acta B. 58, 851– 865. Mucci A. and Morse J. W. (1983) The incorporation of calcite overgrowths: influences of growth rate and solution composition. Geochim. Cosmochim. Acta 47, 217–233. Oomori T., Kaneshima H., Maezato Y., and Kitano Y. (1987) Distribution coefficient of Mg2⫹ ions between calcite and solution at 10 –50°C. Mar. Chem. 20, 327–336. Paquette J. and Reeder R. J. (1995) Relationship between surface structure, growth mechanism, and trace element incorporation in calcite. Geochim. Cosmochim. Acta 59, 735–749. Pearce N. J. G., Perkins W. T., Westgate J. A., Gorton M. P., Jackson S. E., Neal C. R., and Chenery S. P. (1996) A compilation of new and published major and trace element data for NIST SRM 610 and NIST SRM 612 glass reference materials. Geostand. News. 21, 115–144. Proctor C. J., Baker A., Barnes W. L., and Gilmour M. A. (2000) A thousand year speleothem proxy record of North Atlantic climate from Scotland. Clim. Dynam. 16, 815– 820. Proctor C. J., Baker A., and Barnes W. L. (2002) A three thousand year record of north Atlantic climate change. Clim. Dynam. 19, 449 – 454. Roberts M. S., Smart P. L., and Baker A. (1998) Annual trace element variations in a Holocene speleothem. Earth Planet. Sci. Lett. 154, 237–246. Sinclair D. J., Kinsley L. P. J., and McCulloch M. T. (1998) High resolution analysis of trace elements in corals by laser ablation ICP-MS. Geochim. Cosmochim. Acta 62, 1889 –1901. Treble P. C., Shelley J. M. G., and Chappell J. (2003) Comparison of high-resolution sub-annual records of trace elements in a modern (1911–1992) speleothem with instrumental climate data from southwest Australia. Earth Planet. Sci. Lett. 216, 141–153. Treble P. C., Chappell J., Gagan M. K., McKeegan K. D., and Harrison T. M. (2005) In situ measurement of seasonal ␦18O variations and analysis of isotopic trends in a modern speleothem from southwest Australia. Earth Planet. Sci. Lett. 233, 17–32. Wang Q. J., McConachy F. L. N. and Chiew F. H. S. (2001) Climatic Atlas of Australia: Evapotranspiration. Bureau of Meteorology Publication, Melbourne. Wenk H. R., Barber D. J., and Reeder R. J. (1983) Microstructures in carbonates. In Carbonates: Mineralogy and Chemistry (ed. R. J. Reeder), pp.301–367. Vol. 11. Mineralogical Society of America, Washington D. C.