Radiocarbon dating of glacial dust layers and soils at Kilimanjaro’s Northern Ice Field

Abstract

As the African climate history recorded in Kilimanjaro’s Northern Ice Field (NIF) is rapidly lost, the age of the glacier remains disputed. Current age estimates from ice core data and glacial dynamics modeling disagree by an order of magnitude (11,700 vs ~1000 years old). We present radiocarbon dates of glacial dust and proximal soil samples collected from in and around the edge of the NIF that support the hypothesis that the peripheral portions of the NIF glacier are younger and more dynamic than the center. Samples of a dust-rich ice layer, 1.5 m above the base at the edge of the ~40 m high glacier ice cliff were found to be 1600–790 years old. These dates agree with the prediction of significant growth and shrinkage in glacial coverage over the last millennium. The variation in radiocarbon ages likely arises from heterogeneity in the summit volcanic soils, the main source of the dust recovered from the glacier, and cryoconite microbial activity. Sediment samples of a supraglacial pond had modern radiocarbon ages, further demonstrating the impact of cryoconite microbial communities on radiocarbon measurements of glacial samples. After accounting for the variability in the sample ages and the effects of microbial activity, we contend that the lower (i.e. younger) limit of the radiocarbon age range is the most reliable assessment of the age of the ice in which it was trapped.

Keywords

cryoconite glacier Kilimanjaro Northern Ice Field radiocarbon supraglacial

Introduction

Kilimanjaro’s retreating ice fields were originally popularized as an icon of global warming (Alverson et al., 2001; Irion, 2001; Thompson et al., 2002), and captured the attention of the general public and local communities (Mason, 2003). This popularization helped spur a debate in the scientific community over three overlapping factors: What has been the dominant forcing on these plateau glaciers during their formation and growth? What is currently responsible for the retreat of the Kilimanjaro’s glaciers? (Kaser et al., 2004) and How old are the current plateau glaciers?

Rising air temperatures are the dominant drivers of most glacial retreat cases worldwide (Oerlemans, 2005; Thompson et al., 2003; Zemp et al., 2009). For the equatorial glaciers on Kilimanjaro, however, the absence of measurable air temperature increases implicates radiative forcing in combination with decreased precipitation and cloud cover (Cullen et al., 2006; Hastenrath, 2010; Kaser et al., 2004; Molg and Hardy, 2004; Molg et al., 2003, 2008), although contributions of increasing air temperatures have also been invoked (Thompson et al., 2009), which highlights that each case should be treated individually (Molg et al., 2010). The root cause of glacial retreat has strong implications on glacial age, because variations in climate temperature and precipitation may operate on different timescales.

The original 11,700-year age assignment of the Kilimanjaro glacier by Thompson et al. (2002) was largely based on the δ¹⁸O profile recovered from ice cores in the Northern Ice Field (NIF), which were compared against those of the Soreq cave record (Bar-Matthews et al., 1999), where δ¹⁸O and δ¹³C records from calcite speleothems were assigned ages by ²³⁰Th–²³⁴U dating. While Thompson et al. (2002) recovered small amounts of organic matter from some sections of the NIF cores, they were only used in a limited fashion to support the age assignments of the layers because the radiocarbon ages of multiple samples from the same spot were not self-consistent.

The report by Kaser et al. (2010) derives a very different age for the glacier (≤800 years old) using measurements of glacial retreat on Kilimanjaro over the last century. From these measurements, they estimate an average lifespan of only 165 years because of radiation induced retreat, but they also suggest that glacial formation could occur rapidly (decades) during wet periods with high snowfall and greater cloud cover. The characteristic step pattern of the NIF glacier is then a signature of overlapping glacial cycles, leading to a vertical record with varying accumulation rates and discontinuities. By combining their linear decay model for the glacier lifespan with high stands from Lake Naivasha (Verschuren, 2001) as potential signatures of the start of a new glaciation cycle, they concluded that the current glacier most likely formed after ad 1200. While Kaser et al. (2010) do not comment on the correlation of the δ¹⁸O record with either air temperature or precipitation, they clearly express that precipitation level is the dominant mass balance forcing on the glacier itself.

One point of agreement on the history of the glacier is that the total glacial coverage of Kilimanjaro has changed significantly over the last ~1000 years (Kaser et al., 2010; Thompson et al., 2011). While at one point in time Kilimanjaro’s summit may have been host to an ice cap that covered the entire plateau, it shrank to an area smaller than the current day coverage of the NIF and then expanded again to encompass the three main sections there today: NIF, Southern Ice Field (SIF), and the Furtwängler glacier. The ice core analyses determined ages of ~200, ~1500, and ~4000 years old for the Furtwängler, SIF, and outer portions of the NIF, respectively (Thompson et al., 2002, 2011). The remaining contention over the age is for the central portion of the NIF, which the ice core work suggests dates back 11,700 years, but the Kaser et al. (2010) work places at closer to 1000 years old.

Similar to other studies of glacier age (Lowe and Walker, 2000; Matthews, 1984), Thompson et al. (2002) used radiocarbon dating of solid material in their ice cores at specific horizons. Improvements in the radiocarbon dating of solids in glaciers, especially temperate ones, may allow the determination of ages throughout future cores by collecting and measuring the age of carbonaceous particles (Sigl et al., 2009). However, in both cases, the influence of anthropogenic carbon as well as microbial processing by cryoconite communities on the measured radiocarbon ages needs to be addressed. Supraglacial measurements are especially influenced by the deposition of radiocarbon depleted anthropogenic aerosol (Stubbins et al., 2012), and conversely the incorporation of modern radiocarbon by photosynthetic, cryoconite microbial communities (Anesio et al., 2009; Cameron et al., 2012). While contamination from anthropogenic aerosol is no longer an issue at depths corresponding to <ad 1850, (Jenk et al., 2006), the basal layer of glaciers are also well known to contain an environmental niche for microbial activity (Anesio and Laybourn-Parry, 2012; Hodson et al., 2008). These factors need to be considered carefully when employing radiocarbon dating of carbonaceous particles in ice cores.

Here, we present radiocarbon dates of dust from ice layers close to the base of the NIF glacier and of proximal soils collected in 2008 and 2010 that further support the consensus of a younger more dynamic NIF at its edges. We discuss the potential causes for natural variation in measured radiocarbon ages from the glacier, seen in both our results and the previous ice core data, along with how that variation can be used to place age constraints on the glacier.

Methods

Site descriptions

We focused on collecting dust from layers in the glacier throughout the bottom ~2.5 m of the NIF along its southern face (the areas encompassed by Sites 1–6 in Figure 1). Only two of the sites (Sites 1 and 2) had the observed dust-rich layers. Site 1 is the closest to the NIF2 and NIF3 ice cores drilled by Thompson et al. (2002) at 104 and 120 m away, respectively. At the other sites, we collected proximal soil samples, sediment from supraglacial pools, insects pasted onto the glacier, and sediments in an ablation zone of the glacier for comparison. GPS coordinates and summary descriptions of the sites are provided in Table 1.

Figure 1.

Map of sampling sites. Uhuru peak (5895 m) is the high point on the Kibo peak of Kilimanjaro where the year round glaciers exist. Dust layers from the ice were collected at Sites 1 and 2. Soil samples were from Sites 3 and 7. Site 4 is a supraglacial site where sediment from a melt pool was collected. At Site 5, an insect was collected on the wall of the glacier. Sediment from an ablation zone of the glacier was collected at Site 6. The satellite image is from Google maps (^©2012 Cnes/Spot Image, DigitalGlobe, GeoEye) overlaid with a topographic map from http://www.OpenCycleMap.org, and the data points were plotted with assistance from GPSVisualizer.com.

Table 1.

Description of the sampling sites.

Site	GPS coordinates	Year	Description
1	S03 3.609′ E37 20.742′	2008 and 2010	Site with two visible dusty layers embedded in the glacier wall. Layers 1 and 2 were ~1.5 and 2.5 m off the ground, respectively. In 2010, multiple samples were taken from each layer.
2	S03 03.637′ E37 20.891′	2010	Site with one visible dusty layer embedded in the glacier wall ~1.5 m off the ground.
3	S03 03.589′ E37 20.998′	2010	Soil collected ~3 m from the base of the glacier.
4	S03 3.548′ E37 21.015′	2008	Site of a supraglacial pool where sediment was collected.
5	S03 03.590′ E37 21.236′	2010	Site at the base of the glacier where an insect was found encased in ice.
6	S03 03.428′ E37 21.261′	2010	Site at an ablation zone of the glacier. A large amount of sediment was collected that appeared to have been deposited because of internal melt networks within the glacier.
7	S03 03.894′ E37 20.858′	2010	Soil collected from the summit, but away from the glacier.

Site 1 was sampled in 2008 and 2010, and contained two visibly dusty layers embedded in the glacier. The layer heights varied along the length of the glacier, but layers 1 and 2 were ~1.5 and 2.5 m above ground level, respectively. The layers were part of the first step (~11 m high) in the two step (~40 m high) glacier wall (Figure 2). Samples were collected in both 2008 and 2010 for a total of four samples per layer.

Figure 2.

Site 1 layout. Layers 1 and 2 are marked. Exposed 2008 samples came from the right side of the face. In 2010, exposed samples were taken from both left and right sides, and unexposed material was collected and combined along the whole length. A sample from the refrozen meltwater coming down the left wall was also collected. The first ‘step’ of the 35 m ice wall was 10 m tall.

At Site 2, one sample was collected from a ~1.5 m high layer of dust in the glacier, similar to the layers at Site 1. A soil sample was taken at Site 3. Site 4 was one of a number of supraglacial melt pools and cryoconite holes observed with sediment collected at the bottom. Insects were observed adhered to the glacier wall at ~4–5 locations between Sites 1 and 6; at Site 5, one wasp was collected for radiocarbon dating to determine whether the insects were being revealed by ice cliff ablation, or had adhered to the wall during recent times. Site 6 was an ablation zone of the glacier where wet sediment was collected. The sediment appeared to have been trapped there because of pooling at that point from internal water flow as the glacier melted. Another soil sample was taken at Site 7, a location away from the glacier and close to, but not on top of, the actual peak.

Sample collection, transportation, and storage

Two types of samples were taken for the dust layers within the glacier (Sites 1 and 2): exposed (i.e. dust exposed to air) and unexposed (i.e. dust completely encased in ice). Exposed dust inclusions appeared extremely dried on the vertical surface of the glacier and presumably were revealed as the glacier sublimated. To ensure that exposure to air was not contaminating or changing the samples, dust encased in ice unexposed to air was also collected by chipping into the glacier. For exposed samples, new plastic centrifuge tubes (Dow Corning Centri-star cap, 50 mL) were used to capture dust nodules by nudging or scooping (sterile scoops, baked at 250°C overnight wrapped in foil) the dust inclusions into the tubes (Figure 3). In order to collect sufficient material for dating, multiple dust inclusions were added together to reach a total dust volume ≥5 mL. For the unexposed inclusions, the ice was chipped with a hammer and chisel (both cleaned with alcohol wipes) until the dust-rich ice broke free, which was collected into 55-oz Whirl-Pak bags. Approximately 250–350 g of ice was collected for each sample. The ice was allowed to melt and then the dust to settle for 1 h. The clear water was decanted, and the remaining dust material sealed in the Whirl-Pak with as little air as possible.

Figure 3.

Close up of dust layers 1 and 2 at Site 1. Circles highlight the visibly exposed dust in these layers that was collected and combined over either the left or right extents of the site.

At Site 1, multiple exposed dust inclusions were collected from along either the left or right portion of the glacier (Figure 2). Exposed samples were collected from Site 1 in 2008 (layer 1, right and layer 2, right) and in 2010 (layer 1, left/right and layer 2, left/right). Unexposed samples collected in 2010 were harder to access, so ice was combined from both the left and right sides. In addition, at Site 1, a sample was collected where refrozen meltwater was flowing down from above the layers with sediment in it (Figure 2). This flow originated above the layers, and the area where it did cross the layers was carefully avoided when sampling the layers. The ice was melted and decanted similar to the unexposed samples. Site 2 was visited in 2010, and only exposed dust was collected from the whole length of the layer.

Soil samples (Sites 3 and 7) were collected in 2010 with sterile scoops and stored in centrifuge tubes after digging down 2.5–7.5 cm into the topsoil with a spade. Care was taken to avoid the very top layers of the soil because of occasional traffic in the area from hikers who may have transported more modern soils on their boots from the forest below. The sample at Site 3 was taken ~3 m from the base of the glacier because it was clear of the ~1 m of soil closest to the glacier that appeared wet with meltwater. Water activity measurements determined that the soil was indeed below the threshold for growth at the time of sampling. For each soil sample, two centrifuge tubes were filled to the 40-mL mark.

Sediment samples from the supraglacial pond at Site 4 were collected in 2008 in Whirl-Paks, and then decanted similar to the melted ice samples. Insect samples taken in 2010 from Site 5 were collected using sterile scoops and stored in centrifuge tubes similar to the exposed dust samples. Sediment collected in 2010 in the melting edge of the glacier at Site 6 was removed by hand using a clean nitrile glove and stored in a Whirl-Pak after decanting. This was a large amount of sediment with a wet weight of ~200 g.

All dry samples (exposed dust and soils) were transported at room temperature, as the water activities measured were a_w < 0.5, which is below the microbial growth threshold (Grant, 2004). Dry dust samples acquired from the glacial layers were stored at 4°C, and soil samples were stored at room temperature prior to analysis. Wet samples, that is, the Whirl-Paks with sediment or dust from the glacier, were transported at 2–4°C using coolers and ice packs (Thermosafe #478, U-tek −1°C), and then stored in the lab at 4°C. Glacial samples (wet and dry) were sent for radiocarbon dating within 3 weeks of collection; soil samples were sent 3 months after collection.

Geologic characterization

Geologic characterization of all samples was made using a traditional reflected-light stereo-microscope. Verification of preliminary mineral identifications was accomplished using a combination of Raman spectroscopy and powder x-ray diffraction. Raman spectra were obtained with a Horiba Jobin Yvon LabRam HR confocal Raman microscope using a green (532 nm) excitation wavelength. Acquired spectra were compared with reference spectra from the RRUFF project database hosted by Caltech (Downs, 2006). X-ray diffraction patterns were collected on a Bruker AXS model D8 Discover diffractometer using CuKα radiation. Peak identifications were made using standard powder diffraction files from the International Centre for Diffraction Data (ICDD), 2000. To compare the finer-grained dust samples with the coarser soils, the <106 µm size fraction (U.S. Standard Sieve #140) of the soil samples was used for the comparison.

Radiocarbon dating

Radiocarbon dating was performed by Beta Analytic (Miami, FL). All samples were processed as described on their web site (Beta Analytic, 2010b). A bulk organic fraction age measurement was performed. Briefly, sample surface area was maximized, acid (HCl) was applied to eliminate carbonates, and samples were analyzed using accelerator mass spectrometry. Calendar calibration was performed as described by Beta Analytic (Beta Analytic, 2010a) using the IntCal04 calibration data (Reimer et al., 2004) and cubic spline fit mathematics as published by Talma and Vogel (Talma and Vogel, 1993).

Results

The radiocarbon dating results are compiled in Table 2, and dates are reported as conventional ages and 2σ calibrated age ranges in calendar years before 1950 (BP). All of the samples, except the insect, show similar δ¹³C, so no obvious difference in the origin of the organic carbon could be identified (Craig, 1954). Modern carbon was found in three locations: sediment from a supraglacial pond (Site 4), an insect frozen to the glacier wall (Site 5), and sediment collected in an ablation zone of the glacier (Site 6). Older carbon was found in dust layers (Sites 1 and 2) and in soil samples (Sites 3 and 7).

Table 2.

Calibrated and conventional ages for all samples.

Year	ID (beta)	Site	Description	% CO₂ by mass	C (mg)	Conventional age (BP) ± 40 years	2σ calibrated age (BP)	pMC	¹³C/¹²C (‰)
2008	260160	1	Layer 2, right	0.09	≥0.30	690	690–630; 600–560	91.7	−21.7
2010	287602	1	Layer 2, unexposed	0.05	2.04	470	540–490	94.3	−24.1
2010	287603	1	Layer 2, left	0.07	2.03	690	690–630; 600–560	91.8	−22.2
2010	287604	1	Layer 2, right	0.09	2.25	910	920–730	89.3	−22.1
2008	260159	1	Layer 1 right	0.04	≥0.30	810	790–670	90.4	−21.9
2010	287595	1	Layer 1, unexposed	0.11	0.64	1620	1600–1410	81.7	−22.2
2010	287596	1	Layer 1, left	0.11	0.98	1290	1290–1160	85.2	−22
2010	287597	1	Layer 1, right	0.06	0.31	1520	1520–1330	82.8	−22.7
2010	287599	1	Refrozen meltwater	0.5	0.31	930	930–740	89.1	−22.1
2010	287598	2	Layer by camp	0.15	1.04	830	800–680	90.2	−23.3
2010	292920	3	Soil by glacier	0.02	1.63	3460	3840–3630	65.0	−22.1
2008	278567	4	Melt pool on top	0.22	0.33	Modern	≥1950	109.5	−22
2010	287600	5	Insect	34.8	0.70	Modern	≥1950	105.2	−15.7
2010	287601	6	Melt sediment	0.12	9.84	Modern	≥1950	103.0	−21.5
2010	292921	7	Soil at the peak	0.02	1.22	1660	1620–1520	81.3	−25.1

At Site 1, both dust layers 1 and 2 were sampled multiple times and have radiocarbon measurements with different ages. In layer 2, two of the samples (2008 right and 2010 left) have identical ages, including their dual intercepts with the calibration curve leading to their two potential age ranges of 690–630 and 600–560 BP, despite the fact that they were taken from spatially separated areas on the separate visits in 2008 and 2010. The other two measurements in layer 2 have differing age ranges of 540–490 BP (2010 unexposed) and 920–730 BP (2010 right). Two of the measurements from layer 1 also have overlapping, but not identical, date ranges 1600–1410 (2010 unexposed) and 1520–1330 (2010 right) BP, and two have differing age ranges of 1290–1160 BP (2010 left) and 790–670 BP (2008 right). The age of the ice flow at Site 1 (Figure 2; 930–740 BP) is not distinct from the layer ages despite appearing to originate significantly above the layers at a step in the glacier ~11 m off the ground. At Site 2, the age of the layer is 800–680 BP and overlaps best with the ages from layer 1 at Site 1. Finally, the radiocarbon age ranges measured from the soil vary as well and are 1620–1520 BP (Site 7) and 3840–3630 BP (Site 3).

The absolute sample sizes and fractional carbon content were generally small (Table 2). The smallest sample sizes were just over 0.3 mg, and most were in the 1–2 mg range. The fractional carbon content was lowest for the soils samples (Site 3 and 7) at 0.02%, and highest for the sediment recovered from the melt pool (Site 4) at 0.22% and the ice flow (Site 1) at 0.5%. The dust layers were in between these values, ranging from 0.04% to 0.11%.

Geological characterization shows that the soil around the glacier is the primary inorganic component of the dust recovered from the ice. The organic component of the dust was not specifically sourced but must contain a mixture of atmospheric aerosol and organic material from the summit soil. In addition, no pollen or large organic fragments were observed in the dust from the glacial ice. Both the dust from the glacier and the surrounding soils are composed of dark, volcanic ash dominated by black to light gray, aphanitic, or glassy (locally vesicular or pumiceous) lithic shards, with white or transparent, rhombohedral phenocrysts of nepheline (Na₃KAl₄Si₄O₁₆). The matrix of the lithic shards consists primarily of amorphous glass, alkali feldspar (most likely sanidine), and local augite. Both dust and soil also contain minor accessory minerals such as apatite, anatase, and even isolated occurrences of natroalunite (NaAl₃(SO₄)₂(OH)₆; in layer 2 and Site 3). Dust samples typically comprise very fine sand-sized particles or smaller (generally <100 µm), with local grains ranging up to 200 µm (and rare grains up to 1 mm), while soil samples contain substantially less of the finer-grained size fraction and are typically composed of sand-sized particles ranging up to 2 mm or more in grain size. The alkali-rich nature of both dust and soil samples (alkali feldspar, nepheline, natroalunite) and their overall geologic character are consistent with descriptions of the summit geology as presented in the literature (Downie and Wilkinson, 1972). It should be noted that rocks with this alkali signature are rare in this region (Nonnette et al., 2008; Williams, 1968), occurring primarily at the summit of Kilimanjaro, making it unlikely that glacial dust was transported from a more distant location.

Discussion

There is a ~10,000-year discrepancy between the age of the glacier as derived from the ice core dating by Thompson et al. (2002) and that suggested by glacial dynamics modeling by Kaser et al. (2010). While both agree that there has been substantial change in the glacial coverage at the summit over the last 1000 years, the ice core work concluded that a core portion of the NIF glacier has existed for much longer. The radiocarbon dates presented here, sampled from the edge of the NIF, further support a dynamic glacier over the last millennia. Given the variability in the measured radiocarbon ages of our samples, interpreting them requires investigating the source of the organic carbon and the mechanism by which it became trapped in the glacier. Subsequently, it is possible to determine which radiocarbon age or age range is the most appropriate to make an age assignment of a given layer. We discuss how that approach might be applied to all of the radiocarbon dates, including those previously measured from the glacier ice cores.

Discussion of previous age assignments

In order to interpret the ice core data, Thompson et al. (2002) derived an estimated annual accumulation rate for the glacier (by aligning the early portions of the core against the ~1000 year record of lake level heights vs age determined for Lake Naivasha (Verschuren, 2001; Verschuren et al., 2000)), and assumed a correlation between δ¹⁸O and air temperature. This led to an interpretation of the ice core in terms of a chronological record of climate changes over the last 11,700 years that appeared in good agreement with other known records. However, from the beginning, this assessment included the caveat that the Soreq cave comparison was not conclusive because of the difficulty of fingerprint matching across different measurements and conflicting reports that δ¹⁸O was more closely controlled by precipitation than temperature (Blaauw, 2012; Gasse, 2002).

The correlation of δ¹⁸O and air temperature in the ice core record is an important area of debate to try and understand the major mass balance forcing on the glacier. A number of studies suggest that δ¹⁸O in the tropics may be more strongly governed by precipitation than air temperature (Araguas-Araguas et al., 1998; Barker et al., 2001; Gasse, 2002), in disagreement with Thompson et al. (2002). However, a δ¹⁸O record from diatoms in Lake Challa on Kilimanjaro (Barker et al., 2011), a record known to be precipitation/evaporation dependent, was found to be negatively correlated with the ice core record, suggesting that something other than precipitation is driving the ice core δ¹⁸O. This finding is consistent with, but not proof of, the air temperature assumption made by Thompson et al. (2002). The observed negative correlation is based on using the established 11,700-year timeline of the cores and does not investigate an alternative glacial age proposed by Kaser et al. (2010).

Glacial layer formation and the origin of the organic carbon

There are two general ways that material can be entrained in glaciers: by deposition on top or by interaction with the bed below (Benn and Evans, 1998). In a deposition scenario, small dust particles from the soil at the peak along with atmospheric aerosol would have been swept by winds and deposited onto the top of the glacier at a given point in time. Accumulation of snowfall followed by ice layer formation would trap the dust in a layer. Any subsequent dust layer formation would occur on top of the ice layer and come from a chronologically distinct period. Thus, the layers of dust would contain indicators of the age at which the ice layer it was trapped in formed. Thompson et al. (2002) interpreted their ice core data by assuming this type of deposition for the observed dust layers. Alternatively, basal interactions during a glacial advance could lead to a bed-parallel debris septum creating layered bands reminiscent to the ones observed in this work. We conclude that deposition is the mechanism by which the sampled dust layers formed, because of (a) the small grain size distribution, (b) the presence of a supraglacial pond with comparable sediment, and (c) an absence of significant slope required for basal interactions. Therefore, we will focus the rest of our discussion on the more likely eolian deposition scenario.

It is also important to note that the layers were not consistently observed everywhere on the glacier, only between Sites 1 and 2 (Figure 1), suggesting dynamics in this portion of the glacier not present elsewhere. Surficial melting of the glacier may have led to seasonal flows that could have accumulated and deposited the observed material in a supraglacial meltwater pool until a sufficiently large snow event froze the layer (i.e. the sampled dust layers may have formed as a deposit in a large, supraglacial meltwater pond). Indeed, we observed such a supraglacial meltwater pond with a substantial mud layer during the 2008 expedition, which was comparable in size with the length of the sampled dust layer (~35 m × 15 m). Moreover, Thompson et al. (2002) also saw discrepancies in dust layers between the three NIF ice cores they collected. The most significant dust layer in the central NIF3 core (17.5 m from the bottom) was not observed in the outer NIF1 and NIF2 cores and was used as evidence that the glacier had retreated past the current locations of the NIF1 and NIF2 cores at the time of layer formation.

No trend in how the samples were collected or stored can be found in the dates, suggesting that the variability of the measured radiocarbon ages cannot be accounted for through a collection bias or accidental sample contamination. In addition, carbonate was not observed in any of the samples, eliminating its presence as a source of carbon in this system. (As part of standard protocol, an HCl wash during radiocarbon pretreatment also eliminated any trace carbonate.) Ages for the right and left portions of the glacier are not more self-consistent, nor do the 2010 samples agree better within themselves than when compared with the samples from 2008. Sample exposure and/or wet versus dry storage biases are also not evident; the unexposed samples that were stored wet do not show a trend of being systematically older (because of leaching of dead carbon from the plastic) or younger (because of significant microbial activity during storage at 4°C) than the exposed samples that were stored dry. While it is difficult to completely rule out all possible bias in small sample sets dealing with small total amounts of dating material, it appears that in this case, the age variability is a true feature of the samples.

Given the range in measured dates within the dust layers, the organic carbon must have a heterogeneous source. It is well established that arid high altitude soils are open radiocarbon ecosystems that contain a mixture of living (e.g. photoautotrophic microbial communities (Freeman et al., 2009)) and dead materials that give rise to a heterogeneous radiocarbon distribution (Wang et al., 1996). The presence of microbial communities in these areas is not surprising as they are critical for the first steps of processing soils, starting with autotrophic organisms, but followed rapidly by communities which either feed on or coexist with the autotrophs (Esperschutz et al., 2011; Walker and DelMoral, 2003). Results in progress on bacterial DNA isolated from our glacier samples confirmed the presence of microbes as a source of organic material (Connon et al., in preparation). In addition, the measurement of modern radiocarbon ages from the supraglacial ponds also points to active microbial metabolism consistent with the cryoconite literature. The radiocarbon ages measured from current Kilimanjaro summit soils varied from ~1500 to 4000 BP consistent with the predicted age heterogeneity. However, they come with the caveat that contributions from radiocarbon dead anthropogenic aerosol could be influencing the age distribution along with the established heterogeneous soil age. The heterogeneity of ages in the dust layers, however, should not have been influenced by anthropogenic aerosol as it would have been deposited prior to ad 1850 when those effects became significant (Jenk et al., 2006).

Interpreting the variability in radiocarbon ages

The variability in radiocarbon age is likely controlled not only by the effects of input soil radiocarbon heterogeneity but also by well-known post-depositional incorporation of radiocarbon by microbial photosynthesis in cryoconite holes (Anesio et al., 2009; Cameron et al., 2012). If summit material were trapped there today without further processing, there would be a significant difference between the age of the glacial layer containing that material (present) and the radiocarbon age of the trapped material itself (1500–4000 BP). Similarly, soils trapped during the formation of the layers where we have recovered samples would have various radiocarbon dates, older than the actual date that the ice layer itself formed, leading to heterogeneity in the measured ages. However, our observation of modern radiocarbon levels in sediments from a supraglacial melt pool (Site 4) and an ablation zone of the glacier (Site 6) show that under present day conditions, cryoconite microbial activity can sometimes incorporate sufficient modern CO₂ to erase the radiocarbon signature of the older summit soil input material, and ‘reset the radiocarbon clock’. Cryoconite holes are heterogeneous features of the glacier, and there is no guarantee that microbial activity in each would be sufficient to totally reset the carbon clock under the prevailing conditions at the time of layer formation, and only a partial reset could have occurred. Thus, natural radiocarbon age heterogeneity in the input soils alone can lead to the range of radiocarbon ages measured, but there is also evidence that the ages could be further altered by varying amounts of microbial activity.

Understanding the variability in soil radiocarbon ages leads to the fact that the youngest age measurement of the soil is the best upper limit (i.e. oldest age) for the glacier layer it was recovered from if the layers have been formed by eolian deposition. This is best illustrated again by the example of what would happen if a layer were to form currently: Soils currently on the peak with radiocarbon ages of ~1500–4000 BP are deposited in a present day glacial layer; of the theoretically recovered ages, the 1500 BP measurement would be the closest to the true age, but still only an upper limit. If there was sufficient microbial activity to reset the carbon clock of the soils after they were deposited on top of the glacier, then the soil age would become the true layer age (present day). Unfortunately, there is no way to know what amount of carbon clock reset occurred, so the youngest soil measurement must be treated as an upper limit even if microbial activity is suspected.

Application to the glacier age

As discussed above, our upper limit to the age of each glacial layer is based on the youngest radiocarbon dates recovered from each layer. At Site 1, this leads to a 2σ date range for layer 1 of 790–670 BP, and for layer 2 of 540–490 BP. These layers are ~1.5 and 2.5 m above ground level, respectively. With only two layer ages, it is difficult to make an assessment of the age of the glacier in the area because any assumption about accumulation rates would be speculative. Nonetheless, they appear to be in general agreement with the ~800-year age for the glacier derived by Kaser et al. (2010). They are also in line with Thompson et al. (2002, 2011) measurements of the Furtwängler and SIF glaciers (~200 and 1500 years old, respectively), further reinforcing the idea of a dynamic ice cap on the Kilimanjaro summit throughout the Holocene. Even within the NIF, the ages could still be consistent with the Thompson et al. (2002) cores given that our samples come from the margins and they saw a progression in the age of their cores from ~4000 to ~11,700 years old as they moved inward. A final line of evidence suggesting that the area sampled may have been part of a more rapidly changing part of the NIF was the absence of other locations with dust-rich layers for sampling. Site 2 was the only other glacial dust layer sample collected (other areas for sampling were looked for between Sites 1 and 6, Figure 1), and its age and height are similar to layer 1 at Site 1. This suggests continuity in the age along the southwestern face of the glacier, but also suggests it may be a distinct portion of the glacier that was not formed at the same time as the central core. The step pattern of the glacier in this area is frequently cited as part of the evidence for dynamic changes in the glacial coverage (Kaser et al., 2010; Thompson et al., 2011).

The age assignment method from radiocarbon ages of heterogeneous material as discussed above can be applied to help interpret the puzzling ice core radiocarbon results from the Thompson et al. (2002) report. The radiocarbon dates for these cores were rejected as the primary basis for age assignment because of their inconsistency, and instead, certain data points were used as corroborating evidence for the broad timeline established by the δ¹⁸O profile comparison with the Lake Naivasha and Soreq cave records of Bar-Matthews et al. (1999) and Verschuren et al. (2000). The most notable corroborating radiocarbon age was obtained from a 48–49 m deep sample in the NIF3 core, where a small amount of material carbon dated to 9360 BP provided supporting evidence for the 11,700 BP basal age. However, two other samples measured from the same depth had ages of 4090 and 6700 BP (Figure 4). Thompson et al. (2011) point out that with the small amount of material (0.02–0.27 mg) they had for their measurements, it is much easier to contaminate this with modern material to lead to a younger age than to bias the sample older. However, the data shown here provide an alternate explanation by demonstrating that some variability in the radiocarbon ages is to be expected given the age heterogeneity of the input material and ensuing microbial activity. The data also demonstrate that in the absence of contamination, the youngest measured age rather than the oldest age would be the most reliable measurement and would be an upper limit on the age at that depth. Using this upper limit could lead to an acceptable alternate alignment of the δ¹⁸O ice core record against other established climate history proxies.

Figure 4.

A plot of all of radiocarbon dates from the Thompson et al. (2002) ice cores at their various depths. The age ranges measured in this work for the summit soils and layers 1 and 2 from Site 1 are also shaded out to show where there is overlap among the dates. Depths for the ice core are at an average value where a range was given.

While the data presented here are not sufficient to allow a definitive understanding of the radiocarbon dates from the ice cores, it is instructive to look at all of the dates taken from the cores along with the date ranges of the soil and ice layers presented here (Figure 4). There is a good degree of overlap between the ice core measurements and the combined soil and ice layer age ranges. As previously noted, the SIF ages are in agreement with the ice layer ages, and both are evidence of the recent changes in the glacial coverage at the summit. In addition, an interesting feature of the NIF radiocarbon ages from both cores also supports this idea; the sets of dates both start older toward the top of the core (~20 m), proceed to get younger and even modern values at quite a depth (~35 m), and then go back up to even older ages in the deepest sections (>45 m). The older material at the top could contain material from the summit with very little microbial activity that therefore maintains the imprint of mostly dead carbon (or somehow had been contaminated by material with anthropogenic radiocarbon dead material). The younger ages in the middle show that the glacier grew up from that point relatively recently and that right before accumulation started again, there was sufficient microbial activity to reset the carbon clock. The oldest material at the bottom was then trapped in the residual older core of the glacier, consistent with some amount of longer term glacial coverage.

Conclusion

The radiocarbon dates presented here from in and around the NIF at Mt Kilimanjaro are further evidence of recent (~1000 years) changes in the glacial coverage at the summit in agreement with both Thompson et al. (2011) and Kaser et al. (2010). The range in measured radiocarbon ages from the different dust samples collected is because of the heterogeneity of the organic carbon in the input summit soils combined with possible contributions from additional microbial growth after deposition in the glacier. The result of this range of ages is that the youngest radiocarbon ages measured are the most reliable as upper limit ages in the absence of sample contamination. Looking back at the Thompson et al. (2002) ice cores and using the youngest radiocarbon dates as upper limit constraints on the ice core age (in a scenario where the material was initially brought to the glacier by wind and deposited on the surface) might lead to a different alignment of its δ¹⁸O record with other established proxies. However, examination of all of the radiocarbon ages from the ice cores alongside those presented here shows evidence for both recent changes in the glacier along with an older core, suggesting more work will need to be done to reach a final conclusion on the glacier age.

Footnotes

Acknowledgements

The authors would like to dedicate this contribution to Emmanuel Mtui (1971–2012) from Mbahe, who served as senior guide for both expeditions to the Northern Ice Field. His good nature and gentle competence will be missed on the trails of Kilimanjaro. The authors would like to thank Morgan Cable, Stephanie Connon, and Christina Stam for help with sample collection, trip preparations, and good cheer on the mountain. Doug Hardy provided help with permitting and organizing the trips as well as useful discussions on Kilimanjaro. Ron Hatfield gave helpful insight on the collection, storage, and interpretation of the radiocarbon samples. And Ann Coppin and Mickey Honchell conducted literature searches on our behalf and assisted with procurement of reference materials.

Funding

Funding support was provided by the NASA Postdoctoral Program and Icy Worlds NASA Astrobiology Institute. The research described in this paper was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (^©2013 California Institute of Technology). Government sponsorship is acknowledged.

References

Alverson

Bradley

Briffa

. (2001) A global paleoclimate observing system. Science 293: 47–48.

Anesio

Laybourn-Parry

(2012) Glaciers and ice sheets as a biome. Trends in Ecology & Evolution 27: 219–225.

Anesio

Hodson

Fritz

. (2009) High microbial activity on glaciers: Importance to the global carbon cycle. Global Change Biology 15: 955–960.

Araguas-Araguas

Froehlich

Rozanski

(1998) Stable isotope composition of precipitation over southeast Asia. Journal of Geophysical Research: Atmospheres 103: 28721–28742.

Barker

Hurrell

Leng

. (2011) Seasonality in equatorial climate over the past 25 k.y. revealed by oxygen isotope records from Mount Kilimanjaro. Geology 39: 1111–1114.

Barker

Street-Perrott

Leng

. (2001) A 14,000-year oxygen isotope record from diatom silica in two alpine lakes on Mt. Kenya. Science 292: 2307–2310.

Bar-Matthews

Ayalon

Kaufman

. (1999) The Eastern Mediterranean paleoclimate as a reflection of regional events: Soreq cave, Israel. Earth and Planetary Science Letters 166: 85–95.

Benn

Evans

DJA

(1998) Glaciers and Glaciation. London: Arnold Publishers.

Beta Analytic (2010a) Calendar calibration. Available at: http://www.radiocarbon.com/calendar.htm.

10.

Beta Analytic (2010b) Pretreatment methods. Available at: http://www.radiocarbon.com/pretreatment.htm.

11.

Blaauw

(2012) Out of tune: The dangers of aligning proxy archives. Quaternary Science Reviews 36: 38–49.

12.

Cameron

Hodson

Osborn

(2012) Carbon and nitrogen biogeochemical cycling potentials of supraglacial cryoconite communities. Polar Biology 35: 1375–1393.

13.

Connon

Beaty

Ponce

(in preparation) Microbial diversity in glacial ice from Kilimanjaro.

14.

Craig

(1954) Carbon-13 in plants and the relationships between carbon-13 and carbon-14 variations in nature. The Journal of Geology 62: 115–149.

15.

Cullen

Molg

Kaser

. (2006) Kilimanjaro Glaciers: Recent areal extent from satellite data and new interpretation of observed 20th century retreat rates. Geophysical Research Letters 33: L16502.

16.

Downie

Wilkinson

(1972) The Geology of Kilimanjaro. Sheffield: Department of Geology, University of Sheffield.

17.

Downs

(2006) The RRUFF Project: An integrated study of the chemistry, crystallography, Raman and infrared spectroscopy of minerals. In: 19th General Meeting of the International Mineralogical Association, 23–28 July 2006. Kobe, O03–13.

18.

Esperschutz

Perez-de-Mora

Schreiner

. (2011) Microbial food web dynamics along a soil chronosequence of a glacier forefield. Biogeosciences 8: 3283–3294.

19.

Freeman

Pescador

Reed

. (2009) Soil CO2 flux and photoautotrophic community composition in high-elevation, ‘barren’ soil. Environmental Microbiology 11: 674–686.

20.

Gasse

(2002) Kilimanjaro’s secrets revealed. Science 298: 548–549.

21.

Grant

(2004) Life at low water activity. Philosophical Transactions of the Royal Society of London, Series B: Biological Sciences 359: 1249–1266.

22.

Hastenrath

(2010) Climatic forcing of glacier thinning on the mountains of equatorial East Africa. International Journal of Climatology 30: 146–152.

23.

Hodson

Anesio

Tranter

. (2008) Glacial ecosystems. Ecological Monographs 78: 41–67.

24.

International Centre for Diffraction Data (2000) Powder Diffraction File, PDF-2 Database - Release 2000. Joint Committee on Powder Diffraction Standards. Newton Square, PA: International Centre for Diffraction Data (JCPDS-ICDD).

25.

Irion

(2001) The melting snows of Kilimanjaro. Science 291: 1690–1691.

26.

Jenk

Szidat

Schwikowski

. (2006) Radiocarbon analysis in an Alpine ice core: Record of anthropogenic and biogenic contributions to carbonaceous aerosols in the past (1650–1940). Atmospheric Chemistry and Physics 6: 5381–5390.

27.

Kaser

Hardy

Molg

. (2004) Modern glacier retreat on Kilimanjaro as evidence of climate change: Observations and facts. International Journal of Climatology 24: 329–339.

28.

Kaser

Molg

Cullen

. (2010) Is the decline of ice on Kilimanjaro unprecedented in the Holocene? The Holocene 20: 1079–1091.

29.

Lowe

Walker

MJC

(2000) Radiocarbon dating the last glacial-interglacial transition (Ca. 14-9 C-14 ka BP) in terrestrial and marine records: The need for new quality assurance protocols. Radiocarbon 42: 53–68.

30.

Mason

(2003) African ice under wraps. Nature. Epub ahead of print 24 November. DOI: 10.1038/news031117-8.

31.

Matthews

(1984) Limitations of 14C dates from buried soils in reconstructing glacier variations and Holocene climate. In: Mörner

Karlén

(eds) Climatic Changes on a Yearly to Millennial Basis. Dordrecht: Springer, pp. 281–290.

32.

Molg

Hardy

(2004) Ablation and associated energy balance of a horizontal glacier surface on Kilimanjaro. Journal of Geophysical Research: Atmospheres 109. DOI: 10.1029/2003JD004338.

33.

Molg

Hardy

Kaser

(2003) Solar-radiation-maintained glacier recession on Kilimanjaro drawn from combined ice-radiation geometry modeling. Journal of Geophysical Research: Atmospheres 108: 4731.

34.

Molg

Kaser

Cullen

(2010) Glacier loss on Kilimanjaro is an exceptional case. Proceedings of the National Academy of Sciences of the United States of America 107: E68.

35.

Molg

Cullen

Hardy

. (2008) Mass balance of a slope glacier on Kilimanjaro and its sensitivity to climate. International Journal of Climatology 28: 881–892.

36.

Nonnette

Guillou

Gall

. (2008) New K-Ar age determinations of Kilimanjaro volcano in the North Tanzanian diverging rift, East Africa. Journal of Volcanology and Geothermal Research 173: 99–112.

37.

Oerlemans

(2005) Extracting a climate signal from 169 glacier records. Science 308: 675–677.

38.

Reimer

Baillie

MGL

Bard

. (2004) IntCal04 terrestrial radiocarbon age calibration, 0–26 cal kyr BP. Radiocarbon 46: 1029–1058.

39.

Sigl

Jenk

Kellerhals

. (2009) Towards radiocarbon dating of ice cores. Journal of Glaciology 55: 985–996.

40.

Stubbins

Hood

Raymond

. (2012) Anthropogenic aerosols as a source of ancient dissolved organic matter in glaciers. Nature Geoscience 5: 198–201.

41.

Talma

Vogel

(1993) A simplified approach to calibrating C14 dates. Radiocarbon 35: 317–322.

42.

Thompson

Brecher

Mosley-Thompson

. (2009) Glacier loss on Kilimanjaro continues unabated. Proceedings of the National Academy of Sciences of the United States of America 106: 19770–19775.

43.

Thompson

Mosley-Thompson

Davis

. (2002) Kilimanjaro ice core records: Evidence of Holocene climate change in tropical Africa. Science 298: 589–593.

44.

Thompson

Mosley-Thompson

Davis

. (2003) Tropical glacier and ice core evidence of climate change on annual to millennial time scales. Climatic Change 59: 137–155.

45.

Thompson

Mosley-Thompson

Davis

. (2011) A paleoclimatic perspective on the 21st-century glacier loss on Kilimanjaro, Tanzania. Annals of Glaciology 52: 60–68.

46.

Verschuren

(2001) Reconstructing fluctuations of a shallow East African lake during the past 1800 yrs from sediment stratigraphy in a submerged crater basin. Journal of Paleolimnology 25: 297–311.

47.

Verschuren

Laird

Cumming

(2000) Rainfall and drought in equatorial east Africa during the past 1,100 years. Nature 403: 410–414.

48.

Walker

DelMoral

(2003) Primary Succession and Ecosystem Rehabilitation. Cambridge: Cambridge University Press.

49.

Wang

Amundson

Trumbore

(1996) Radiocarbon dating of soil organic matter. Quaternary Research 45: 282–288.

50.

Williams

LAJ

(1968) Geochemistry and petrogenesis of the Kilimanjaro volcanic rocks of the Amboseli Area, Kenya. In: International Symposium on Volcanology, Canary Islands, pp. 862–888.

51.

Zemp

Hoelzle

Haeberli

(2009) Six decades of glacier mass-balance observations: A review of the worldwide monitoring network. Annals of Glaciology 50: 101–111.