1.1.Introduction
Major changes occurred in Arctic climate during the 20th century, including an important increase of surface air temperature, particularly between 1970 and 2000 (Moritzet al. , 2002). Changes in environmental variables such as precipitation, sea-ice extent, snow cover, and permafrost have had important impacts on local communities and ecosystems. To better understand these changes beyond the available short instrumental record, long-term proxy climate records are needed from across a wide geographical area. In the western Canadian High Arctic, annually laminated sediments from lakes are the only available records with precise chronological control in an area that lacks tree-ring and ice-core records. Varve thicknesses have been mainly used to reconstruct past summer temperatures in Arctic regional syntheses such as those by Overpeck (1997)and more recently by Kaufman et al. (2009). However, recent studies showed that climate signal recorded by clastic varved sediments may be less straightforward than initially thought (e.g. Cockburn et Lamoureux, 2008a; Francuset al. , 2008a). These studies indicate that caution is needed when calibrating varve thickness (VT) using instrumental data, since sediment accumulation can result from different hydroclimatic and geomorphic processes, such as snowmelt, rain events, and landslides. In addition to VT, grain size measured at the annual scale is a sedimentary parameter with potential to reconstruct paleo-hydroclimatic conditions (Francuset al. , 2002; Kaufman et al., 2011). In this paper, we use the methods pioneered by Francus (1998) together with a new image acquisition and analysis software developed at INRS, to obtain grain-size measurements within each varve in a long Arctic lake sequence in order to demonstrate its potential for paleoenvironmental reconstructions and to produce a new long paleoclimate reconstruction with annual resolution.
1.2.Study site
Cape Bounty East Lake (hereafter, "East Lake") (74°55'N; 109°33'W; 5 m a.s.l.) is located on the south-central coast of Melville Island, Nunavut, in the western Canadian Arctic Archipelago (Figure 2.1). As part of a long-term research program (the Cape Bounty Arctic Watershed Observatory), hydrological, limnological, and sedimentary processes have been investigated and monitored in two adjacent lakes and watershed systems since 2003 (Lamoureuxet al. , 2006a; Cockburn et Lamoureux, 2008a; Cockburn et Lamoureux, 2008b; Dugan et al., 2009; McDonald et Lamoureux, 2009; Laurin, 2010; Pautler et al., 2010; Lewis et al., 2011; Stewart et Lamoureux, 2011). The East and West lakes (unofficial names) have similar morphologies: both are ~1,5 km2, and maximum depths are 30 m and 34 m, respectively. These monomictic lakes have near-isothermal water columns during most of the year. Each winter, a 1,7- to 2,5-m-thick ice forms, and usually persists as a pan until mid-July or early August. The watershed area of East Lake is 11,6 km2 and mainly composed of weathered Paleozoic sandstone and shale overlain by late glacial and Holocene regressive marine sediments (Hodgsonet al. , 1984). The glacioisostatic rebound in the area began after deglaciation ~ 10 ka 14C BP and became negligible since ~ 2 ka 14C BP (Hodgson, 1989). Temperatures at Cape Bounty are typically above freezing during June-August when snowmelt and occasional rainfall contribute to streamflow. The mean annual temperature is -17,9 °C and total precipitation is less than 60 mm per year at Rea Point, the nearest weather station located 105 km to the northeast (Hodgsonet al. , 1984).
1.3.1.Evolution of the watershed
Cuven et al. (2011) described four principal sedimentary units from a 737-cm-long sediment core from East Lake and developed a 4200-year-long varve thickness record. Basal unit 1 (which extends from an unknown age to 2192 BC) is massive, and corresponds to the marine stage of the isostatic depression. Unit 2 (2192 BC to AD 243) is finely laminated and interpreted as the transitional period from marine to freshwater conditions. Cuven et al.'s µ-XRF data suggest a decrease of marine influence concomitant with an increase in terrestrial runoff between 400 BC and AD 244. Once East Lake was fully isolated by glacioisostatic uplift from the ocean after AD 244, the sedimentation rate likely increased due to sediment focusing into the smaller lake basin and because of the increased erosion of the exposed unconsolidated marine sediments in the watershed (Cuvenet al. , 2011). Hence, after AD 244 conditions were more favorable for the formation and preservation of clearly defined clastic varves. In unit 3 (AD 244 to 1132), varves become clearly defined while the Fe/Mn ratio reaches maximum values, indicating more persistent water column anoxia (Cuven et al., 2011). Unit 4 (AD 1135 to present) better preserved and thicker laminae.
1.3.2.Sedimentary facies at East Lake
The mechanisms of varve formation at both lakes were inferred from the observation of thin-sections and the monitoring of tributaries and lake water column (Cuven et al., 2010; Cockburn and Lamoureux, 2008a,b). Lithozone A is the most common lithofacies and is composed of a fine silt layer deposited by nival melt runoff that is overlaid by a clay cap representing the settling of particles under the ice during winter (Figure 2.2). Lithozone B is graded and likely deposited by a turbidity current caused by runoff of higher intensity than the regular spring melt (Figure 2.2). Lithozone C is composed of poorly sorted sand layers occurring systematically above the initial snowmelt silt layer and below the clay cap (Figure 2.2). It is interpreted to be produced by high-discharge events, which are triggered by high-intensity rainfalls during summer after the snowmelt (Cuvenet al. , 2010).
1.4.Materials and methods
1.4.1.Sediment cores and thin-sections
In June 2006, two cores were retrieved from ice at the deepest (30 m) known location in East Lake (74°53'30.3 N; 109°32'85.5W): a short gravity corer (CB06G1: 41,5 cm) and a long vibra-corer (CBEV1: 737 cm). These cores were also studied by Cuven et al. (2011). An additional gravity core (07EL01: 42 cm) was retrieved in 2007 from the same location. We used the same thin-sections as those analyzed by Cuven et al. (2011), as well as new sections prepared from core 07EL01 using similar protocols.
1.4.2.Image and grain size analysis
Thin-section images were acquired using a transparency flatbed scanner at 2400 dpi resolution (1 pixel = 10.6 µm) in plain and cross-polarized light (De Keyser, 1999). Using the image analysis software developed at INRS-ETE (Francus et Nobert, 2007)and used by Cuven et al. (2011), regions of interest (ROIs) were identified on the flatbed scan digital images. The software then automatically acquired SEM images of the ROIs using a Zeiss Evo® 50 scanning electron microscope (SEM) in backscattered electron (BSE) mode. Eight-bit gray-scale BSE images with a resolution of 1024 x 768 pixels were acquired with an accelerating voltage of 20 kV, a tilt angle of 0° and a 9 mm working distance with a pixel size of 1 µm. These settings optimize contrast between clastic grains and clay matrix (Francus, 1998; Soreghan et Francus, 2005). Grains within the sedimentary facies appeared black following the transformation of the BSE images into black and white (Francus, 1998; Nederbragtet al. , 2004; Nederbragt et Thurow, 2005). Each sedimentary particle was measured for the position of the centre of gravity, area, length of the long and short axis of the best fitting ellipse, and angle of the long axis relative to horizontal (Annex 5) (Francus et Karabanov, 2000). Details of the algorithms used here are available in Annex 6. Finally, the size of each particle was calculated according to Francus et al. (2002) in order to determine the following particle-size distribution (PSD) indices for each varve: mD0 or the median disk apparent diameter (Francus, 1998), standard deviation (sD0), percentile 98% (P98D0), and maximum (maxD0) (Figure 2.2).
PSD indices were analyzed for the last 2845 years (or the upper 359,8 cm) from 6148 BSE images. When varves were thicker than the size of the image or when the number of grains identified within a single processed image was less than 900, results from several images were merged in order to obtain representative and robust measurements for each year of sedimentation (Francus et Pirard, 2005). Overall, an average of 2,2 images and 1164 particles were measured for each varve. For technical reasons, grain-size data were only extracted from vibra-core CB06EV1 for which the uppermost sedimentary layer is dated to AD 2000 by cross-dating with surface cores. Usable BSE images could not be acquired from cores 07EL01 and CB06EG1 because poor sediment impregnation that resulted in incorporation of sanding grit in the sedimentary facies.
1.4.3.Chronology
Radionuclides (210Pb, 137Cs) were measured on core CB06G1 by Cuven et al. (2011). A new varve count was performed using the same set of thin-sections as used by Cuven et al. (2011), except that we included the new 07EL01 surface core. For this new count, we analyzed BSE images in addition to the flatbed scan images (plain and cross-polarized light) of the thin-sections to manually record varve boundaries in the INRS-custom-software. Furthermore, additional BSE images (3072 x 2304 pixels wide and with a pixel size of 3,8 µm) in the uppermost 6 cm of both CB06G1 and 07EL01 cores were assembled in photo-mosaics to observe the lateral variation of sedimentary properties, and hence, to clearly identify each varve. The long core CB06EV1 was cross-correlated with the short surface core using distinctive marker beds (Lamoureux, 2001).
1.4.4.Weather station
Two meteorological stations were considered for comparison of climate parameters with sediment properties: Rea Point (Station ID: 2403450) and Mould Bay (Station ID: 2502700) located 100 km northeast and 320 km northwest of Cape Bounty, respectively. Temperature, rainfall, and snowfall were extracted from the Adjusted Historical Canadian Climate Data (AHCCD) http://www.ec.gc.ca/dccha-ahccd/. Additionally, cumulative melting degree-days (MDD) was calculated using the National Climatological Archive of Meteorological Service Canada.
1.4.5.Statistical methods
For the correlation analyses presented below, we used the Pearson coefficient, denoted r in the sequel. This coefficient is a measure of linear dependence and needs normality for inference purposes. However, extreme values resulting in heavily skewed underlying distributions prevent reliable results. For these reasons, we consider also two non-parametric (rank-based) coefficients, namely Spearman's r and Kendall's t. Spearman's r measures the strength of linear dependence between ranks of observations, whereas Kendall's t is based on concordant/discordant pairs and does not need any hypothesis on linear association (Abdi, 2007).
1.5.Results[1]
1.5.1.Chronology and error estimation
The uppermost surface of core 07EV1 allowed us to distinguish two additional varves between AD 2000 and 1998. These new varves require that the erosive turbidite recognized by Cuven et al. (2011) be shifted to AD 1971, and reduce the number of eroded years to five, but the rest of our rationale for the upper part of the chronology remains identical to Cuven et al. (2011).
Within the interval where image analysis was performed, we counted 2845 varves over the upper 359,8 cm, or 72 additional couplets compared to Cuven et al. (2011). However differences are larger in some intervals: between 166,7 and 155,5 cm (15,8%), between 259 and 246,7 cm depth (16,7%) where varves are more diffuse (Cuven et al., 2011), and between 352,5 and 341,9 cm depth (14,5%) where varves become very thin (average is 0,47 mm). This new chronology is referred hereafter as the CBEL12a chronology. Figure 2.3 shows the consistency between the two varve counts made by F.L. and S.C.
Comparison between the varve counts from Cape Bounty East Lake by Cuven et al. (2011) (SC) and this study (FL). Varve number 0 was formed in 2000
1.5.2.Grain-size variations
Sedimentary facies identified by Cuven et al. (2010) can be recognized by PSD indices measured by analysis of BSE images (Figure 2.2). Lithozone A, the classic simple nival melt deposit, occurs in AD 1807, 1819, 1826, and 1829, among others, and is characterized by low VT and relatively low values of all PSD indices (Figure 2.2; A). Lithozone B, caused by runoff of higher intensity than the regular spring melt, formed in AD 1804, 1825, and 1832 and is characterized by high VT and high PSD indices (Figure 2.2; B). Lithozone C, debris flows likely generated by high rainfall events, such as the one of AD 1811 and 1821 (Figure 2.2; C), is characterized by relatively low VT, average mD0 but high P98D0, and sD0 values with coarsest grain size of ~75 µm. Layers with high P98D0 and sD0values are poorly sorted and coarse skewed according to the Folk and Ward (1957) graphical method.
The different PSD indices show a number of common and specific trends along the laminated record (Figure 2.4). First, mD0 is relatively low in the lower part of the core but has a clear step increase at ~ AD 1350, and shows maximum (coarsest) values in the 20th century. Both P98D0 and sD0 exhibit similar variations through time, including the sharp increase at the beginning of the 20th century, with maximum values in the 1990's. These two indices are strongly correlated (r = 0,83, r = 0,84, t = 0,71, p < 0,0001 for all coefficients). Finally, maxD0 displays a steady, long-term increase toward the top of the core.
Figure 4 East Lake Varve Thickness (VT) and four PSD indices for the last 2845 years. Data are standardized relative to the mean and the standard deviation. A 15-year running mean is applied on the series.
1.5.3.Varve thickness and PSD indices
In addition to the test (r, r, and t correlations) on the original time series, we also performed the analysis on the decorrelated versions (Shumway et Stoffer, 2000)because the VT series (Figure 2.4) exhibit autocorrelations significant up to fourth order, and the PSD series show serial correlation with higher order. Results (Table 2.1) indicate that almost all correlations are weak but significant, except for Pearson's correlation between VT and maxD0. Removing temporal autocorrelation has a little impact on these results. The strongest, but still relatively weak, correlation between VT and PSD indices is with mD0. Despite the above-mentioned relatively weak correlation, VT time series (Figure 2.4) seems not to be in phase with any of the PSD indices, especially at the beginning of the 20th century, when VT remains stable, whereas P98D0, sD0, and mD0 reach their highest values. For the period spanning ~ AD 250 to 500, VT increases substantially while sD0 values decrease (Figure 2.4). This could be explained by the progressive stabilization of the sedimentary environment as outlined by Cuven et al. (2011).
Table 2.1 Pearson's (r), Spearman's (r) and Kendall's (t) correlations between varve thickness VT and the annually-resolved PSD indices, for the 2000-250 AD period, for original (top panel) and decorrelated (bottom panel) and their significance (p-values).
1.5.4.Calibration with the instrumental record
Data available for correlation are limited: only varves formed after 1971 (above the erosive turbidite) were considered to ensure chronological control. In addition, Mould Bay and Rea Point weather stations changed measurement protocols or stopped operating after 1996 and 1985, respectively. Table 2.2 summarizes correlations between instrumental meteorological data and PSD index P98D0. The correlations are evaluated from the raw time series because no serial correlation was detected. The strongest positive correlation (r = 0,85, n = 15, p < 0,001) is between the largest annual rainfall events at Rea Point and P98D0. Non-parametric correlation coefficients also indicate strong and significant correlations (Table 2.2).
Table 2 Pearson's (r), Spearman's (r) and Kendall's (t) correlations between PSD index P98D0 and instrumental data from Rea Point and Mould Bay and their significance (p-values).
The linear-regression plot suggests that some extreme values may influence this relationship (Figure 2.5). We used the Cook's distances method (Cook et Weisberg, 1982)to address this possibility. This is calculated with the following equation:
where MSE is the mean squared error, ŷi is the estimated value of the dependent variable yi for year i, and ŷi(t) is the fitted value when removing observation t. A value is considered to be influential if Dt> 1. As such, no data point is considered influential for the calibration record (Figure 2.5 B; 0,31). For linear regression, a threshold value 4/(T-2) is used, and Dt values > 4/(T-2) are usually considered outliers (Cook et Weisberg, 1982; Sheather, 2009). In our data series, two outliers are above this threshold (Figure 2.5). One way to address this problem is to remove the outliers; however, this approach is not attractive because of the small sample size. In addition, removing data points one by one yields 15 regression lines with slopes ranging from 0,669 to 0,889, compared to the slope for all data, which is 0,777. For the maximal P98D0 observation (50,55µm), these regressions give estimated rainfall from 19.13 to 23.5 mm. Finally, when years 1971, 1973, and 1980 are removed, the resulting regressions lead again to outliers in terms of Cook's distance. When data are contaminated with outliers, another methodology is to use robust rather than ordinary least squares estimation (Rousseeuwet al. , 1987); however, these two approaches yield approximately equal results when applied to our dataset.
These findings motivated the use of a nonlinear model coupled with robust estimation: the Box-Cox method that yields a log-log transformation (Box et Cox, 1964). Because no precipitation (zero value) was recorded in 1979, we added a constant (1 mm) to all rainfall observations to avoid negative log-value. The model then reduces to a linear model in log-transformed variables. For robust estimation, we use the bisquare weighting function (Maronnaet al. , 2006)(Figure 2.6), where
log (rainfall + 1) = -9,5337+ 3,2706 logP98D0 1)
Two other strong correlations between P98D0 and the instrumental record include those with (1) MDD at Rea Point (r = 0,59, p = 0,02), and (2) June temperatures at Mould Bay (r = 0,52, p = 0,007) (Table 2.2). The correlations with other PSD indices are weaker than those obtained for P98D0, except for sD0, which is itself strongly correlated with P98D0.
Figure 5 A: linear regression between P98D0 and largest rainfall event/year at Rea Point (1985–1971) with the Pearson's, Spearman's and Kendall's correlations. B: Cook's distances for linear regression between P98D0 and largest rainfall event/year at Rea Point (1985–1971). Horizontal line shows the critical value 4/(T-2).
Figure 6 Linear regression between log-transformed P98D0 and largest rainfall event/year at Rea Point (1985–1971) using robust estimation, and the 90% confidence interval (dotted lines).
1.5.5.Climate reconstruction
We use the log-log model (equation 4) to reconstruct rainfall amount for the last 2845 years. However, we present the reconstruction for only the period after ~AD 244, when the coring site was fully lacustrine, to avoid complications related to possible changes in sedimentological and geomorphological conditions (Figure 2.7). The reconstructed rainfall amount was quite steady from AD 244 until 1900 (Figure 2.7), whereas the 20th century appears as a strong positive anomaly. In detail, precipitation during the period between AD 400 and 1050 was generally below average, except for the 6th and the 7th centuries. Another period of high precipitation was centered around AD 1100. After AD 1200, precipitation amount was generally above or near average, with the exception of AD 1500-1550. The precipitation amounts were average during the second half of the 19th century, but increase substantially for the rest of the record.
Figure 7A Reconstructed largest annual rainfall amounts (mm) for the past 1750 years using the log-log transformed. The blue line is the raw data and the black line is the 15-years running median filter. B Reconstructed largest annual rainfall anomalies using the log-log transformation. The blue line is the raw data and the black line is the 15-years running mean.
1.6.Discussion
1.6.1.Chronology
Comparison between our chronology and the original from Cuven et al. (2011) show an overall difference of 2,6% (72 years) for the upper 360 cm (Figure 2.3). Discrepancies are more pronounced over intervals with very thin laminae (0,3 mm). These fine varves are difficult to delineate using thin-sections images compared with SEM images, which have better phase contrast and higher resolution. In these intervals, the counts based on SEM images were systematically and logically higher. We conclude that the new chronology is an improvement over the previous one. These results also highlight the value for SEM imagery for thin, fine-grained varves.
Cuven et al. (2011) identified eight likely erosive beds in the remainder of the 423-cm-long laminated sequence, and these have also been recognized in our BSE images. All of these units are thinner and finer-grained than the uppermost bed that is believed to have eroded five years of sediment, and they lack erosional features at the microscopic scale. We thus infer that our varve chronology is minimally affected by high-energy sedimentary events.
1.6.2.Climatic interpretation of the grain-size record at Cape Bounty
The significant and strong correlation between P98D0 at East Lake and rainfall events at Rea Point is consistent with process studies at Cape Bounty, a study area with the longest record of process-oriented observations in the Arctic. For instance, Dugan et al. (2009) demonstrated the important impact of two rainfall events (9,2 and 10,8 mm) that occurred in late June and July 2007 on the West River. They showed that these rain events generated peak discharge and suspended sediment transport two times larger than the nival melt in the same year. They also reported that downstream sediment traps recorded the rainfall as a hyperpycnal-flow deposit with the highest sediment deposition rate of the season and the coarsest mean grain size. Their results also demonstrated the importance of prior moisture conditions and the role of permafrost active-layer development as an important factor controlling rainfall runoff and sediment transport response to precipitation events.
Our observations and climate correlations with varve properties are in agreement with these monitoring results (Dugan et al., 2009; Lewis et al., 2011). The largest rainfall events at Rea Point were recorded late in the summer season in AD 1971 (25.9 mm) and 1973 (when 13 mm fell twice within two days) (Figure 2.8). In those years, warm conditions prevailed before the precipitation events (Figure 2.8). For instance, maximum temperatures reached uncommon values (> 13°C) during the 4 days before the major rainfall event that occurred on July 30, 1971 (Figure 2.8). Similarly, warm conditions were recorded in early august AD 1973 before the high rainfall events (Figure 2.8). During the calibration period, the highest P98D0 was measured in varves deposited during these two wet and warm years (Figure 2.9). Our record is also sensitive to summer rain events in more typical years, such as AD 1978. This was an average year in terms of MDD; it experienced an average rainfall event (7,6 mm) on 215 JD (Figure 2.8) that was sufficient to produce a sedimentary response, i.e. a small debris flow (Figure 2.9).
Figure 8 Examples of three three years with major daily rainfall events (1971, 1973 and 1978) at Rea Point, Nunavut, and the corresponding cumulative melting degree-days (MDD) for each year represented by the line.
Figure 9 Sedimentary structures corresponding to years AD 1971 to 1985, a period that encompasses major turbidites and debris flow in the late 20th century. The left panel presents a flat-bed scan (plain light) of a thin section with the location of ROIs (yellow squares) and the limits of the varves (horizontal black lines). The right panel presents three BSE images from years AD 1971, 1973 and 1977 (ROIs abx, aaz and aau respectively) that contain thick laminae with coarse grain size. Other years (AD 1975, 1977 and 1985) with rainfall signals are also characterized by coarse grain size. The turbidite-base from AD 1971 shows an erosive contact
Cape Bounty seems to be one of the few reported Arctic varved sequences to be sensitive to rainfall events. However, the overall paucity of long-term monitoring of river discharge and sediment yield in lakes, and the general focus on nival runoff and summer glacial melt may have led to a biased view of processes within lake catchments in the High Arctic. These precipitation events clearly trigger substantial suspended sediment yields in High Arctic lakes (Church, 1972; Church, 1974; Cogley et McCann, 1976; Church, 1988; Lewkowicz et Wolfe, 1994; Lewiset al. , 2005; Dugan et al., 2009; Lewis et al., 2011). In sum, we argue that the primary climate signal stored in Cape Bounty sediments are large summer rain events and are recorded by high P98D0, particularly when warm conditions predate the rain events.
The autocorrelation of VT and the PSD indices likely indicates some hysteresis in the sediment yield at Cape Bounty, as was described at Nicolay Lake, a similar Arctic watershed (Lamoureux, 2002). While it does not change the strength of the correlations established in this study (Table 2.2), this geomorphic signal should be considered carefully in similar studies.
1.6.3.High-energy facies characterization using image analysis
High-energy deposition events (turbidites and debris flow) identified by Cuven et al. (2010) can be distinguished by their coarse P98D0 and sD0 (Figure 2.10). The bases of turbidites have been similarly characterized by high P99D0 and sD0 values in the sedimentary sequence from South Sawtooth Lake, Ellesmere Island (Lewiset al. , 2010). The base of a turbidite is typically poorly sorted and coarse grained in the head and the body of the energetic turbiditic flow, whereas finer sediment is deposited afterwards, as the flow wanes and deposition occurs through particle settling (Reading, 1996). Lewis et al. (2010) interpreted this facies as deposited by hyperpycnal flow and likely generated by rainfall events. The coarse sD0and P98D0 values observed here appear to reflect this unsorted coarse sediment at the base of turbidite facies. On the other hand, debris flows are primarily composed of viscous mixtures of sediment and water in which the volume and mass of sediment exceeds that of water (Major, 2003). Because the absence of turbulence generates no dynamic sorting of material, the resulting deposit is poorly sorted (Nichols, 2009). The sharp increase of both P98D0 and sD0during the 20th century (Figure 2.4) is in agreement with observations from thin-sections in which turbidites and debris-flow deposits can be identified (Figure 2.9). In sum, high values of P98D0 and sD0 are observed within both turbidites and debris-flow deposits (Figure 2.10).
Figure 10 Time-series (AD 1695 to 1679) of normalized P98D0 and sD0; B: thin section of the time interval presented in A showing two prominent turbidites at AD 1689 and AD 1685; C: Blow-up of three thin varves from AD 1688 to 1686 : Varves formed during AD 1688 and 1686 have low P98D0 and sD0 values, whereas AD 1687 has high values. D Backscatter electron image of varve formed in AD 1687, showing coarse grain size interpreted as a debris-flow deposit
1.6.4.Particle-size versus varve-thickness variations
PSD and VT are significantly correlated, but the strength of correlations is weak (Table 2.1). The weak correlation was previously noted by Francus et al. (2002) who hypothesized that the decoupling of VT and grain size reflects the fact that VT is the sum of many different phenomena occurring throughout the year, whereas grain size can be related to single events, and hence better linked with the instrumental record. This was confirmed by the recent process study by Cockburn and Lamoureux (2008a) at Cape Bounty showing that the size of the coarser fraction collected in sediment traps tracks the high-energy flows better than the short-term mass accumulation.
The long-term evolution of VT and PSD indices at Cape Bounty is strikingly different (Figure 2.4). At Cape Bounty, several working hypotheses can be suggested to explain this decoupling. The geomorphic evolution of the lake and its watershed is controlled by the slow and progressive glacioisostatic rebound as inferred from geochemical and sedimentological properties of the 7.4-m-long core (Cuven et al., 2011). The increase in varve thickness between AD 200 and 500 cannot be due to a change in the tributary competence because there is no simultaneous change in our PSD indices. The clear step change in mD0 around AD 1350 does not correspond to a detected change in known boundary conditions in the watershed. This change could be climatically driven, although no particular trend in climate has been documented around this time in the region. Alternatively, the coarser sediment could result from progradation of the East Lake inflow delta during this period, which in turn would effectively increase the proximity of the core site, the sedimentation rate and the coarseness of the sediment (Cuven et al., 2011). However, such a change is expected to be progressive, which is not the case. An alternate mechanism entails the progressive flushing of the salt water from what is assumed to have been a meromictic lake, which could have switched sediment delivery from overflows to underflows, and resulted in the deposition of coarser sediment in the core locations.
1.6.5.Discussion of correlation with climate instrumental record
Rea Point is the closest weather station to Cape Bounty (100 km east). Despite its short duration, it offers one of the strongest, if not the strongest, correlation between an instrumental record and varve properties yet discovered in the Arctic. Our study is the first to demonstrate a correlation between rainfall events and clastic varves in this region. This linkage is not surprising because rainfall events are important quasi-annual events recorded in most Arctic varves records (Lamoureux, 2000; Lewiset al. , 2005). Soils in these cold regions support sparse vegetation and are frozen at depth, which limits water infiltration and can create a substantial runoff response when precipitation occurs (Church, 1988; Lamoureux, 2000). Furthermore, the western part of the Canadian High Arctic was more affected by the overall increase of precipitation during the second half of the 20th century as observed by the Coupled Model Intercomparison Project phase 3 (CMIP3) (Minet al. , 2008).
Because the Rea Point record is short (1969-1985), we investigated the possibility of using the longer record from Mould Bay instead. The largest rainfall events at Mould Bay and Rea Point are not correlated (r = -0,04, r = 0,01, t = -0,03, p > 0,9 for all coefficients) (not shown in Table 2.2), while annual rainfall shows a weak and insignificant relationship (r = 0,41, r = 0,17, t = 0,15, with p = 0,13, 0,55, 0,46, respectively). On the other hand, June temperatures and MDD significantly correlate (June temperatures: r = 0,87, r = 0,65, t = 0,46 and MDD: r = 0,70, r = 0,70, t = 0,47, with maximal p < 0,02) between Rea Point and Mould Bay. Similarly, mean air temperature at Cape Bounty is strongly correlated with Mould Bay for the monitoring period (2003 to 2009) (r2= 0,92; standard error = 3,5) (Lewis et al., 2011). Collectively, these observations suggest a common regional pattern for temperatures that can be used for correlation with sedimentary data. Additionally, P98D0positively and significantly correlates with June temperatures at Mould Bay (Table 2.2), although these correlations are weaker than for precipitation.
None of the sedimentary properties seems to be directly linked to snow accumulation instrumental record (Table 2.2). At Cape Bounty, Lewis et al. (2011) noted that the East Lake catchment has fewer concave surfaces favorable to accumulate snow (Figure 2.1). This geomorphological property implies that redistribution of snow by winds is important, and may impede a quantitative link between snow water equivalent and sediment transport.
In sum, sediment properties at Cape Bounty can mainly be explained by rain events in summer, with some influence of temperature and snow melt.
1.6.6.Climatic reconstruction
Few Arctic reconstructions of rainfall have been published to compare with our record. Of note is the amount of summer rainfall at Cape Bounty during the 20th century, which is unprecedented in the last 1750 years. As such, the precipitation of the 20th century in the western Arctic Archipelago is anomalous, analogous to summer temperature over the entire Arctic (Kaufman et al., 2009). Likewise, the sharp peak centered around AD 1100 (Figure 2.7) appears to be consistent with the increase of temperature that occurred around AD 1000 and 1100 according to the temperature reconstruction from Moberg et al. (2005). Moreover, the period spanning AD 950-1200 (known as the Medieval Climate Anomaly) experienced warmer conditions in many locations in the Arctic (e.g. Mann et al., 1999; Dahl-Jensen et al., 1998). The similarity between our precipitation record and the temperature record is not surprising because higher temperature allows for more moisture to be stored in the atmosphere. Moreover, warmer temperature may favor the decline of sea-ice cover over the nearby Arctic Ocean, allowing for more evaporation. Finally, as described above, warmer temperatures also have a positive feedback on the sensitivity of East Lake to summer precipitation by increasing the depth of the active layer (Lewis et al., 2011). Other prominent features of our record are linked with difficulty to climate forcing factors and other records.
1.6.7.Image analysis
This work highlights the potential of using annually resolved PSD to reconstruct past precipitation amounts. These parameters have been rarely investigated because of technical challenges that limited the generation of long time series from SEM image sets. The advantages of using this imaging technique is that very thin laminae, down to 0.3 mm, can be accurately measured. This type of measurement is not influenced by sediment compression with increasing depth, and individual laminations can be measured without the contamination from neighboring laminae, which is often the case with traditional subsampling techniques (Lotteret al. , 1997). This procedure is applicable to long records that take advantage of new software with substantial analytical efficiencies. In the case of East Lake, PSD indices are more strongly correlated with climate parameters than is VT. Revisiting key sites from the Arctic using this innovative technique may increase fidelity of long-term proxy climate records.
1.7.Conclusion
Annually resolved PSD obtained using image analysis provided a 1750-year-long quantitative reconstruction of summer rain events from East Lake in the Canadian High Arctic. Correlations of P98D0 with the instrumental record yielded one of the strongest correlations obtained to date in the Arctic (r = 0,85), which is also higher than varve thickness from the same site. Rain events in the western Canadian Arctic Archipelago increased to unprecedented levels in the 20th century compared to the last 1750 years.
[1] All of the data from East Lake presented in this study are available on-line through the World Data Center for Paleoclimatology http://www.ncdc.noaa.gov/paleo/pubs/jopl2012arctic/jopl2012arctic.html
1.Références
Bradley RS, Briffa KR, Cole J, Hughes MK, Osborn TJ (2003) The climate of the last millennium. Paleoclim Glob Change Future: 105-141.
Bradley RS, Jones PD (1993) Little Ice Age'summer temperature variations: their nature and relevance to recent global warming trends. The Holocene 3(4): 367-376.
Church M (1974) Hydrology and permafrost with reference to North America. Perma Hydrol: 7-20.
Cockburn J, Lamoureux S (2008a) Inflow and lake controls on short-term mass accumulation and sedimentary particle size in a High Arctic lake: implications for interpreting varved lacustrine sedimentary records. J Paleolimnol 40(3): 923-942.
Cockburn JMH, Lamoureux SF (2008b) Hydroclimate controls over seasonal sediment yield in two adjacent High Arctic watersheds. Hydrol Process 22(12): 2013-2027.
Cogley JG, McCann S (1976) An exceptional storm and its effects in the Canadian High Arctic. Arct Alp Res: 105-110.
Cuven S, Francus P, Lamoureux S (2011) Mid to Late Holocene hydroclimatic and geochemical records from the varved sediments of East Lake, Cape Bounty, Canadian High Arctic. Quat Sci Rev 30(19-20): 2651-2665.
Cuven S, Francus P, Lamoureux SF (2010) Estimation of grain size variability with micro X-ray fluorescence in laminated lacustrine sediments, Cape Bounty, Canadian High Arctic. J Paleolimnol 44(3): 803-817.
De Keyser TL (1999) Digital scanning of thin sections and peels. J Sediment Res 69(4): 962-964.
Dugan HA, Lamoureux SF, Lafrenière MJ, Lewis T (2009) Hydrological and sediment yield response to summer rainfall in a small high Arctic watershed. Hydrol Process 23(10): 1514-1526.
Folk RL, Ward WC (1957) Brazos River bar [Texas]; a study in the significance of grain size parameters. J Sediment Res 27(1): 3-26.
Francus P (1998) An image-analysis technique to measure grain-size variation in thin sections of soft clastic sediments. Sediment Geol 121(3-4): 289-298.
Francus P, Bradley R, Lewis T, Abbott M, Retelle M, Stoner J (2008) Limnological and sedimentary processes at Sawtooth Lake, Canadian High Arctic, and their influence on varve formation. J Paleolimnol 40(3): 963-985.
Francus P, Bradley RS, Abbott MB, Patridge W, Keimig F (2002) Paleoclimate studies of minerogenic sediments using annually resolved textural parameters. Geophys Res Lett 29(20): 1998.
Francus P, Karabanov E (2000) A computer-assisted thin-section study of Lake Baikal sediments: a tool for understanding sedimentary processes and deciphering their climatic signal. Int J Earth Sci 89(2): 260-267.
Francus P, Nobert P (2007) An integrated computer system to acquire, process, measure and store images of laminated sediments. 4th International Limnogeology Congress, Barcelona, 11-14th July.
Francus P, Pirard E (2004) Testing for sources of errors in quantitative image analysis. In: Francus P. (ed) Image analysis, sediments and paleoenvironments. Springer-science, Dordrecht, pp 87-102.
Hodgson D (1989) Le Quaternaire de la Reine-Élisabeth. In: Fulton, R.J. (Ed), Quaternary Geology of Canada and Greenland. Geological Survey of Canada, Ottawa, pp 477-514.
Hodgson D, Vincent JS, Fyles JG (1984) Quaternary geology of central Melville Island, Northwest Territories. Geological Survey of Canada, Geology of Canada 83-16, p 23.
Kaufman DS, Schneider DP, McKay NP, Ammann CM, Bradley RS, Briffa KR, Miller GH, Otto-Bliesner BL, Overpeck JT, Vinther BM (2009) Recent warming reverses long-term Arctic cooling. Science 325(5945): 1236.
Keatley BE, Douglas MSV, Smol JP (2006) Early-20th century environmental changes inferred using subfossil diatoms from a small pond on Melville Island, NWT, Canadian high Arctic. Hydrobiologia 553(1): 15-26.
Koerner R, Fisher D (1990) A record of Holocene summer climate from a Canadian high-Arctic ice core. Nature 343: 630-631.
Lamoureux S (2000) Five centuries of interannual sediment yield and rainfall-induced erosion in the Canadian High Arctic recorded in lacustrine varves. Water Resources Research 36(1): 309-318.
Lamoureux S (2002) Varve chronology techniques. Tracking environmental change using lake sediments: 247-260.
Lamoureux SF, McDonald DM, Cockburn JMH, Lafrenière MJ, Atkinson DM, Treitz P (2006) An incidence of multi-year sediment storage on channel snowpack in the Canadian High Arctic. Arctic: 381-390.
Laurin E (2010) The impact of experimental snow augmentation on soil thermal regimes and nutrient fluxes from High Arctic headwater catchments. Unpublished MSc. Thesis, Queen's University, Kingston. 148.
Lewis T, Braun C, Hardy DR, Francus P, Bradley RS (2005) An extreme sediment transfer event in a Canadian High Arctic stream. Arct Antarct Alp Res 37(4): 477-482.
Lewis T, Francus P, Bradley RS, Kanamaru K (2010) An automated system for the statistical analysis of sediment texture and structure at the micro scale. Computers & Geosci.
Lewis T, Lafrenière MJ, Lamoureux SF (2011) Hydrochemical and sedimentary responses of paired High Arctic watersheds to unusual climate and permafrost disturbance, Cape Bounty, Melville Island, Canada. Hydrol Process.
Lewkowicz AG, Wolfe PM (1994) Sediment transport in Hot Weather Creek, Ellesmere Island, NWT, Canada, 1990-1991. Arct Alp Research: 213-226.
Lotter A, Sturm M, Teranes J, Wehrli B (1997) Varve formation since 1885 and high-resolution varve analyses in hypertrophic Baldeggersee (Switzerland). Aquat Sci Res Across Bound 59(4): 304-325.
Major JJ (2003) Debris flow. In: Middleton GV. (ed) Encyclopedia of sediments and sedimentary rocks. Kluwer academic publisher, Dordrecht, pp 186-188.
McDonald DM, Lamoureux SF (2009) Hydroclimatic and channel snowpack controls over suspended sediment and grain size transport in a High Arctic catchment. Earth Surface Process Land 34(3): 424-436.
Min SK, Zhang X, Zwiers F (2008) Human-induced Arctic moistening. Science 320(5875): 518.
Moberg A, Sonechkin DM, Holmgren K, Datsenko NM, Karle ́n W (2005) Highly variable Northern Hemisphere tempera- tures reconstructed from low-and high-resolution proxy data. Nature 433(7026):613–617
Moritz RE, Bitz CM, Steig EJ (2002) Dynamics of recent climate change in the Arctic. Science 297(5586): 1497.
Nederbragt A, Francus P, Bollmann J, Soreghan M (2004) Digital sediment colour analysis as a method to obtain high resolution climate records. 105-124.
Nichols G (2009) Sedimentology and stratigraphy. Wiley-Blackwell, Chichester.
Overpeck J, Hughen K, Hardy D, Bradley R, Case R, Douglas M, Finney B, Gajewski K, Jacoby G, Jennings A (1997) Arctic environmental change of the last four centuries. Science 278(5341): 1251.
Pautler BG, Simpson AJ, Mcnally DJ, Lamoureux SF, Simpson MJ (2010) Arctic permafrost active layer detachments stimulate microbial activity and degradation of soil organic matter. Environ Sci Technol 44(11): 4076-4082.
Reading HG (1996) Sedimentary environments: processes, facies and stratigraphy, Wiley-Blackwell, Malden.
Soreghan M, Francus P (2005) Processing backscattered electron digital images of thin section. In: Francus P Image Analysis, Sediments and Paleoenvironments. Springer, Amherst, 203-225: 203-225.
Stewart KA, Lamoureux SF (2011) Connections between River Runoff and Limnological Conditions in Adjacent High Arctic Lakes: Cape Bounty, Melville Island, Nunavut. Arctic 64(2).