Paleoecological methods for UVR reconstruction

Lake sediments for reconstruction of past UVR environments can be collected using any standard paleoecological technique including gravity, piston, percussion or freeze-coring (reviewed in [67]), although subsequent treatment of the sample must consider the type of analysis used for environmental reconstruction [68]. In general, sediment samples should not be exposed to direct light, warm temperatures (>4°C), extreme pH or active biological communities. While freeze-coring often minimizes degradation of biogeochemical fossils and provides the highest temporal resolution, this technique is limited to use in short-term analyses (<1 m sediment depth) and may be inappropriate for some delicate microfossils. In cases where millennium-scale UVR reconstructions are required, sediments are collected most often using Livingston, percussion or advanced piston-coring techniques [67]. Although many paleoecologists subsequently sample cores at relatively coarse intervals (> 1 cm), limits to temporal resolution are set by the rate of sediment deposition at the site, and the mechanical skill of the sampling process. Ages of individual sediment samples may be determined using a combination of naturally occurring radioisotopes, including 137Cs (post-1950 AD), 210Pb (post-1850 AD), 14C (pre-1850 dates) or U : Th ratio (pre-25 000 years), together with optically stimulated- or thermo-luminescence analyses where quartz grains are present [67]. Additional chronological control may be obtained from annual laminae (varves) and from quantification of natural (tephra, post-glacial clays, pollen spectra) or anthropogenic (Pb stable isotopes, soot particles, contaminants) sedimentary markers of known age. Further details of core collection and processing for individual fossils are presented within Last and Smol [67].

16.3.1 Microfossil indices of past UVR environments

Over the last decade, considerable progress has been made in the development of ecological calibration studies and development of statistical inference techniques to reconstruct past environmental variables from paleoecological data [69]. This widely used transfer function approach depends on strong statistical correlations between organismal abundance and measured chemical or biophysical variables in modern lakes, and assumes that these relationships remain valid equally in the past. In the case of UVR reconstructions, this approach is based on the striking changes in the chemical (nutrients, DOC) and physical (mixing regimes, water transparency) properties of lakes across arctic and alpine treelines, patterns which are reflected also in the zonal distribution of freshwater diatoms (class Bacillariophyceae) [22,70-72]. Multivariate statistical models describing variance in community composition as a function of measured environmental gradients are developed using survey data before being applied to historical reconstructions at individual sites.

In the circumpolar region of the Northern Hemisphere, diatom-based inference models have been developed for the quantitative reconstruction of DOC or CDOM [22,70,72], water color [73,74], and total organic carbon [74,75]. Because DOC is a limnological variable that is highly correlated with lake catchment vegetation and soils [e.g., 76-78], reconstructions of DOC based on the siliceous fossils of lacustrine diatoms can be used as a proxy for past vegetation shifts and climate [reviewed in 79]. When combined with statistical descriptions of the relationship between DOC content and UVR penetration [e.g., 15], these models provide the starting point for more detailed reconstructions of past UVR environments and the changes in terrestrial environments that regulate DOC flux.

To develop a UVR transfer function, between 25 and 100 lakes are selected to lie along obvious gradients of terrestrial vegetation, with care given to select lakes with similar morphology (depth) and hydrologic regime (closed basins, no peatlands) but with contrasting levels of the chemical parameter of interest, in this case UVR-absorbing DOC. Description of diatom response to predominant environmental gradients begins with concomitant collection of surface sediments (upper 0.25-0.5 cm) and a wide range of chemical, physical and biological parameters. Ideally, selection of environmental variables is based on prior knowledge of the main gradients in lake characteristics, particularly those related to DOC biogeochemistry. The best inference models result when there are relatively few strong gradients of environmental change, or when multiple gradients exhibit a high degree of linear correlation (e.g., DOC, CDOM, color). Extraction of diatom fossils from sediments follows standard protocols, beginning with digestion of the organic and carbonate sediment matrix using 30% H202 or mixtures of either nitric (HN03) and sulfuric acid (H2S04), or potassium dichromate (K2Cr207) and sulfuric acid techniques [80]. Normally, acidified sub-samples (~1 cm3 wet sediment) are heated for 2 h, at ca. 80°C, before repeated centrifugation and decanting with distilled water to neutralize the suspensions. Different concentrations of each diatom suspension are then deposited onto cover slips and left to dry before being mounted onto microscope slides using a permanent resin (Naphrax® or Hyrax®). Diatom enumeration is carried out using light microscopy (1000-1250 x magnification), with a minimum of ~ 500 valves counted in each sample to characterize fossil assemblage composition.

Development of diatom-based models for the reconstruction of DOC and other variables involves a three-step analytical approach [22,70]. First, multivariate statistics are used to identify the main environmental factors correlated with changes in diatom community composition. Common approaches include the use of canonical (direct gradient) ordinations that assume either a unimodal (Canonical Correspondence Analysis, CCA) or linear (Redundancy Analysis, RDA) change in species abundance along environment gradients (Figure 1A). Usually both species abundance (% or concentration) and environmental variables will require transformations to normalize variance prior to analysis. Additionally, abundances are often centred and standardized prior to ordination in order to improve the ease of biological interpretation. Species-environmental relationships are often summarized using an ordination bi-plot in which axes are constrained to be linear combinations of measured environmental variables, and species or lakes are plotted in the ordination space (Figure IB). This approach allows the relationship between species and environmental change to be modeled directly [81]. Environmental variables are added to the ordination using stepwise or forward selection and are retained only if they independently explain a significant (p < 0.05) amount of variance in fossil assemblage composition based on ordinations constrained to that variable alone. The significance of both ordination axes and individual variables is determined usually using Monte Carlo tests with 500 to 1000 iterations [82],

Assuming that DOC is identified as an important factor regulating variability in diatom community composition among survey lakes, the second step is to develop statistical descriptions of the mean (optimum) and variance (tolerance) in environmental conditions that regulate species abundance. In the case of UYR reconstructions, the responses of modern diatoms to a DOC gradient are

A tu rid ra a. * partly forested forest

CCA axe 1

io/Si

t>.

Prof i SÍ02

Sup 1

Y SorÈale

, active variables

Mg ^^

passive variables

CHlaN

S04

Figure 1. Canonical correspondence analysis of diatom community composition for lakes arranged along a gradient across treeline in the subartic of northern Québec, Canada [22], (a) Ordination biplot of sub-fossil diatom assemblages from 57 lakes lying in tundra (triangle), partly-forested (star) and forested catchments (solid), (b) Relations of environmental gradients in lake chemistry to changes in diatom community composition showing that changes in DOC concentrations are positively correlated to the main direction of community variance (i.e., CCA axis 1). (c) Diatom-based inference model (transfer function) describing the relationship between measured DOC and that inferred using weighted-average calibration and regression approaches. See text for additional detail.

[Figure modified from [15] with permission.]

Figure 1. Canonical correspondence analysis of diatom community composition for lakes arranged along a gradient across treeline in the subartic of northern Québec, Canada [22], (a) Ordination biplot of sub-fossil diatom assemblages from 57 lakes lying in tundra (triangle), partly-forested (star) and forested catchments (solid), (b) Relations of environmental gradients in lake chemistry to changes in diatom community composition showing that changes in DOC concentrations are positively correlated to the main direction of community variance (i.e., CCA axis 1). (c) Diatom-based inference model (transfer function) describing the relationship between measured DOC and that inferred using weighted-average calibration and regression approaches. See text for additional detail.

[Figure modified from [15] with permission.]

modeled using a weighted-average (WA) calibration. The WA approach assumes that each species exhibits a unimodal response to gradients of DOC, with highest relative abundance under conditions that are optimal for that taxon's growth [83]. The technique is insensitive to a poor fit of the unimodal model and is suited, therefore, to paleoecological studies in which sediments contain taxa produced in a variety of habitats throughout the year, while chemical data are obtained from less intensive sampling. In general, estimates of species optima and tolerance improve with the number of sites included in the survey, the length of the environmental gradient, and evenness of site distribution along the gradient. Further details on the WA calibration approach are given by Birks [69] and Hall and Smol [84],

In step three, past DOC concentrations are reconstructed from analyses of changes in fossil assemblages by applying modeled species-environment relationships using a WA regression approach (Figure 1C). Here, estimates of past DOC are inferred by multiple regression of species DOC optima, weighted by the relative abundance of taxa, but downweighted by the variability (tolerance) in the species-DOC relationship. The performance of these diatom-DOC transfer functions can be evaluated using randomization procedures (bootstrapping), while problems arising from poor fossil assemblage analogues within modern diatom communities can be assessed using dissimilarity indices (e.g., ANALOG; Line and Birks, unpublished program).

Finally, inference of past regimes of UVR and photosynthetically active radiation (PAR) regimes requires conversion of diatom-inferred DOC estimates into reconstructions of past irradiance environments. This goal is accomplished best by using bio-optical models that are based on DOC-irradiance relationships and on the response curves of algae for DNA damage and inhibition of photosynthesis [15,25]. Here, estimates of wavelength-weighted underwater UVR exposure (T*) are based on spectral attenuation and biological weighting to allow quantitative estimation of potential impacts of exposure to past UVR. Further, this powerful approach allows direct comparison of potential UVR impacts among causal factors including 03 loss and climate change [15,85]. However, in general, weighted transparency estimates based on DOC must be considered a lower boundary to variability in underwater UVR exposure, due to the influence of other physical factors (see above).

In the analysis of past spectral characteristics, T* is defined as J1 /K(X) e(A) £orel(A) F(X) da, where the integral is evaluated over 280-400 nm. K(a) is the diffuse attenuation coefficient at wavelength a calculated from statistical relationships with DOC in the survey lakes [85]. s(A) is the biological weighting factor for DNA damage (r*DNA) [86] or for inhibition of photosynthesis by UVR (r"PI) [87], and is expressed on a relative scale (e = 1.0 at 300 nm). £oreiU)ls the normalized surface irradiance at the study location (£0rei = 1.0 at 400 nm), while F(X) is the factor of enhancement in surface radiation flux for a given level of stratospheric ozone depletion (set to 1.0 for 330 Dobson Units). T* values are calculated at 1 nm intervals and are summed from 280 to 400 nm to give a total T* for UVR. The main advantage to the T* approach is that it does not require estimates of absolute UVR flux in order to evaluate the relative magnitude of

UVR impacts arising from past environmental change. Finally, although not explicitly attempted to date, such spectral irradiance models may be combined with diatom-inferred estimates of past lake depth to develop quantitative estimates of photon flux and biological exposure, assuming relatively small variations in past atmospheric transmission of UVR.

As with all inference model based approaches, reconstructions of past DOC and UVR levels using fossil diatoms are subject to problems arising from poor modern analogues for past diatom assemblages, taxonomic inconsistencies, insufficient range of past environmental variation, and multiple, co-linear environmental gradients [84]. Fortunately, inter-investigator harmonization of fossil taxonomy has reduced difficulties with species identification, while comparisons among multiple cores suggest that single, central sites capture the main components of historical variation, assuming little effect of riverine inputs, sediment redistribution or groundwater springs [e.g., 88]. As discussed in Pienitz et al. [89], the absence of modern analogues appears most problematic during the early Holocene, when terrestrial sources of DOC are poorly developed and benthic Fragilaria are the most common fossil diatoms. Similarly, inference models can perform poorly, despite strong environmental gradients, if variation in reconstructed variables (e.g., DOC) arise from multiple regulatory mechanisms (e.g., forest and peatland C sources). Although model performance can be improved through judicious site selection [e.g., 22], problems can arise during historical reconstructions unless multiple indicators are used to evaluate the relative importance of other potential forcing factors. In the case of DOC reconstructions, the use of peatland-specific microfossils [90], siliceous phytoliths from terrestrial plants [91], aquatic pollen [25] and near-infrared characterization of carbon sources [92] may provide insights into the sources and optical properties of terrestrial DOM. Similarly, use of multiple fossil indicators, each with different environmental sensitivity (e.g., invertebrates, pigments, diatoms) can help identify bias arising from confounding environmental variables. Finally, although within-lake variance in reconstructed variables is rarely as great as that expressed among lakes in surveys, multivariate evaluation of the main direction of variance in fossil assemblages can be used to identify historical eras in which environmental gradients are weak and reconstructions may be less reliable.

16.3.2 Other microfossil metrics of past UVR exposure

Recent observations show that some zooplankton deposit photo-protective pigments in their exoskeletons and resting eggs in response to high-energy irradiance [93,94]. Because these remains often are well preserved in lake sediments [95], analysis of the relative proportion of pigmented and hyaline remains in cores may be a valuable index of the relative exposure of taxa to UVR, particularly in response to long-term variations in UVR penetration arising from development of terrestrial DOM sources (D. Hessen, University of Olso, unpublished data). To date, no study has attempted to test this hypothesis. Similarly, because UVR impacts diminish with increased zooplankton size [58], analyses of fossil assemblage size structure may reveal changes in the average exposure of the population through time. However, we advise caution in both approaches, due to the potentially confounding influences of size-selective, visually-orienting predators such as fish. As shown elsewhere, colonization of Ashless habitats by vertebrate planktivores results in rapid elimination of both large-bodied and highly-coloured crustacean zooplankton, events which are clearly recorded in lake sediment archives [95]. Similarly, interpretation of fossil size structure may be complicated by the observation that many of the most sensitive organisms (rotifers, early instars) leave a poor fossil record. However, in spite of these caveats, we feel that there is significant unexploited potential for microfossil remains from invertebrates to be used as indices of population response to past UVR regimes.

16.3.3 Biogeochemical indices of past UVR exposure

Past UVR environments can be inferred from analyses of fossil photo-protective pigments produced by eukaryotic algae, phototrophic bacteria and other organisms [e.g., 24]. The basic principles underlying this approach are that many organisms produce sunscreen compounds in response to intense UVR, and that these compounds are deposited in sediments following the death of the organism. Potential fossil indicator compounds include scytonemin and its derivatives from cyanobacteria [24,96], mycosporin-like amino acids from algae and invertebrates (MAAs; [97]), and melanin from zooplankton, such as Daphnia [93]. However, because preservation of these pigments is rarely complete [see 98], absolute concentrations of pigment cannot be used to quantitatively estimate past UVR flux. Instead, most reconstructions have quantified the concentrations of UVR-absorbing pigments relative to those of ubiquitous compounds that measure total algal abundance (e.g., carotenoids, chlorophylls [chls]). This approach assumes that all measured pigments degrade at similar rates. Given this assumption, the index should record the past exposure of an "average" organism, and should be greatest when a high proportion of the primary producer population is exposed to damaging levels of UVR [24,26]. Naturally, exposure of other trophic levels may vary independently of algal exposure as a consequence of the adaptive strategies of individual animals (see above).

To date, most pigment-based reconstructions of irradiance have used water-insoluble compounds derived from cyanobacteria to estimate historical changes in the UVR exposure of phototrophic populations in response to climate change and human activities. Although other approaches are possible, high-performance liquid chromatography has been the main analytical method used to quantify both past UVR exposure and biotic responses. Leavitt and Hodgson [68] have provided a comprehensive review of the main methods used to isolate, identify and quantify fossil pigments. Here, we provide a brief overview of the main methods used in our laboratories to reconstruct past UVR environments.

Once removed from a lake, sediments should be frozen (< — 20 °C) in the dark under an inert atmosphere (N2, Ar, CO2) or vacuum until isolation and quantification of pigments. In order to improve the reproducibility of pigment extraction, well-mixed sediment sub-samples should be freeze-dried under a hard vacuum (<0.1 Pa) for 24-48 h. Lipid-soluble pigments are extracted from the bulk sediments by soaking powdered sediments in a mixture of degassed acetone:methanol:water (80:15:5, by volume) for 24 h in the dark and under an inert atmosphere at 0°C. Pigment concentrations are most often quantified by reversed-phase high-performance liquid chromatography (RP-HPLC), which separates complex mixtures according to the relative attraction of individual pigments for the non-polar stationary phase (both coating and support material) and the polar mobile solvent phase. So far, most UVR reconstructions have been based on polar pigments such as scytonemin or related compounds [24] that pass through the HPLC column rapidly and are among the first compounds to be detected. In most cases, efficient analysis of large sample numbers minimally requires an autosampler, high-pressure pumps (> 20 kPA), high-resolution column, and in-line photo-diode array spectrophotometer (300-800 nm range). Additional components may include in-line detectors of pigment fluorescence or mass-selective spectrometric detectors [e.g., 99]. Both the analytical system detailed by Mantoura and Llewellyn [100] modified by Leavitt and Findlay [101] and that of Wright et al. [102] have proven robust in isolating UVR-absorbing pigments from sediments of 500 lakes (see below), despite some limitations in resolving power [103, but see 104]. Analytical separation is achieved using either a two- or three-stage solvent system in which pigment extracts are introduced to the chromatographic column, and solvent polarity is systematically altered to sequentially isolate compounds of progressively decreasing polarity. At a minimum, accurate quantification of pigment abundance requires separation of marker compounds from contaminants, identification of a pigment's true identity, and calibration of the HPLC system with an authentic standard of known purity. Further details on HPLC calibration and pigment quantification are provided by Leavitt and Hodgson [68].

Past UVR penetration has been measured as a ratio of UVR-absorbing pigments : algal carotenoids, an index which is linearly related to the depth of UVR penetration in whole-lake experiments [24,26]. To date, most of our reconstructions have been based on the UVR-absorbing pigment, Ca, which has a mass of 635 according to mass spectrometric determinations using negative ion-atmospheric pressure chemical ionization techniques [68]. Visualization of this compound is improved substantially by first dissolving the whole extract into an injection solution containing an aged (3 months) solution of Sudan II dye before injection into the HPLC system [26,68]. Similar reconstructions can be achieved using the sum of scytonemin and its derivatives, compounds that preserve in lake sediments for over 100000 years (D. Hodgson, unpublished data). Abundance of UVR-absorbing pigments is expressed relative to total algal biomass in order to distinguish whether photo-protectant production arises from a unique population (e.g., surface dwelling) or represents a general response of the phototrophic community to UVR [24]. Total algal abundance can be measured as changes in the concentration (nmoles pigment g_1 dry sediment or nmoles pigment g_1 organic matter) or accumulation rate (nmoles pigment cm-2 yr~ of /^-carotene, a chemically-stable carotenoid ubiquitous in algae, chl a and its pheopigment derivatives, or the sum of individual algal group indicators including alloxanthin (cryptophytes), diatoxanthin (diatoms), colonial cyanobacteria (myxoxanthophyll or echinenone) and chlorophytes (lutein).

Surveys of modern lake communities suggest that pigment-based U VR indices are elevated only when a substantial portion of the algal community is exposed to potentially-damaging levels of irradiance. Indices range from 10-800% in clear lakes, depending on the presence of refugia, lake depth, circulation patterns and the distribution of algal biomass. Leavitt et al. [24] noted that UVR-absorbing compounds were abundant, and UVR indices elevated (>50%), in mountain lakes that were both shallow (< 5 m maximum) and of low DOM content (< 1.5 mg DOC 1~*). As recorded by Sommaruga et al. [38], these lakes often lie above treeline, have a high proportion of their volume exposed to > 1 % ambient UVR, and exhibit variable and often low C-specific attenuation of UVR by DOM [40]. Such a lack of physical refuge from UVR presumably requires phototrophic organisms to produce photo-protective compounds in order to reduce cellular oxidation from UVR-produced singlet oxygen and free-radicals [64],

Algal exposure to UVR may also be increased by the presence of fine particulates which, while reducing total irradiance, act to increase ratios of UVR: PAR because photon scattering is less wavelength dependent than is its absorbance by DOM (e.g., S in [43]). Exposure also increases because sedimentation of particulates scours DOM from the water column leading to lower absolute concentrations [e.g., 38], while adsorption of DOM to particles causes a "package effect" that decreases absorbance per unit pathlength of water without altering the specific absorbance of individual DOM molecules. Thus, for photosynthetic organisms, the need to remain in light necessitates exposure to high levels of UVR and the production of photo-protective compounds. In contrast, the presence of surface blooms (e.g., eutrophic lakes) does not seem to elevate UVR indices, probably because water column circulation reduces exposure and because productive lakes often have high levels of UVR-absorbing DOM [24,26].

Reconstruction of past UVR environments from sedimentary pigment profiles has a number of significant challenges before UVR indices can be quantitatively related to past photon flux. First, quantification of UVR-absorbing compounds is difficult because of considerable analytical requirements (HPLC, MS). To date, most reconstructions rely on partly characterized pigments produced by benthic cyanobacteria [24,26,27], but little is known of the precise structure of these compounds, or of their distribution among organisms [64]. Similarly, reconstruction of past UVR environments from scytonemin requires isolation and quantification of a series of derivatives using advanced mass spectrometric techniques [68,105]. Second, comparison of decadal historical records with annually resolved pigment profiles suggests that some post-depositional transformation occurs, and that very recent deposits (< 3 yr old) may not provide reliable indices of recent UVR environments [26]. Similarly, complete diagenesis of photo-protectant, chl and carotenoid pigments can occur, particularly in deep lakes (>100 m) with very low sedimentary organic matter contents (<2%) where degradation of pigments is often extensive [98], Finally, more research is required to standardize expressions of UVR flux. Presently, we favour a relative ratio of UVR-absorbing pigments to ubiquitous, chemically-stable carotenoids in order to correct for variation in fossil abundance arising from single populations (e.g., surface dwelling) or changes in total algal sedimentation [24,26]. However, while sensitivity analyses suggest that most historical changes in a UVR index arise from variation in deposition of photo-protectant compounds, the reciprocal nature of the ratio at least allows for the possibility of artifacts arising from changes in total algal abundance (e.g., trophic interactions, pH, etc.) independent of variations in past UVR exposure. As with most paleoecological analyses, this problem can be best alleviated through the use of multiple sedimentary proxies of DOC or UVR.

16.3.4 Sedimentary organic matter as an index of past UVR penetration

Reconstruction of UVR exposure from analyses of fossil diatoms and pigments requires levels of training or analytical expertise that are likely to preclude their widespread adoption as methods to evaluate UVR impacts on lake ecosystems. Consequently, recent research has begun to investigate the use of bulk sedimentary constituents as retrospective predictors of water-column DOC concentrations and UVR [27], In general, sediment organic content is determined as % mass loss-on-ignition (LOI) at 500°C for 1 h [106], To date, surveys of both shallow alpine [24] and subarctic [80] lakes distributed across treeline revealed strong linear correlations (Pearson r >0.75) between sediment organic matter (% LOI) and the dissolved organic matter content (mg DOC l-1) of overlying waters. In these cases, surveys spanned gradients of terrestrial vegetation development, from bare rock catchments to drainages with well-developed coniferous forests, and represent a wide range in supply of terrestrial-derived DOM [21,89]. Based on this relationship, past DOC concentrations can be reconstructed from organic matter (% LOI) profiles in sediment cores from regional lakes [27], while depths of UVR penetration (as 1 % surface irradiance) can be estimated using published optical models [11,14]. When this UVR depth is expressed as a proportion of area-weighted basin depth, an index of past UVR exposure and lake sensitivity can be calculated [cf., 38],

Accurate use of bulk sedimentary organic matter to reconstruct past UVR penetration requires strong correlations between dissolved and sedimentary organic matter in modern surveys, relatively consistent optical characteristics of DOM, and that these relations remain valid in the past. Positive correlations between water column and sedimentary organic matter may be reinforced by several mechanisms. First, because most water-column C resides in DOM, its sedimentation as colloids or precipitates is a major process increasing the organic matter content of bottom deposits. Second, while sedimentary organic matter is also a function of lake production and catchment erosion [107], these processes will tend to reinforce DOM-sediment relationships due to either

DOM production (algal, macrophyte sources of DOM) or enhanced removal (adsorption on fine particulates). Finally, in shallow lakes, sediments may act as a substantial source of DOM due to turbulent mixing and resuspension of DOMrich interstitial waters. Unfortunately, high variability in the specific attenuation characteristics of DOM has been documented from a number of lake surveys [40,43], and may limit the accuracy of sediment-based UVR reconstructions in some very clear lakes (<1 mg DOC l-1). As well, as with all paleoecological reconstructions, assumptions of constant regression relationships through time need to be viewed critically. However, the high agreement observed between pigment- and sediment-based UVR reconstructions (see below) suggests that this simple technique holds much promise, particularly in extreme environments where biochemical or microfossil preservation may be poor (e.g., early Holocene, saline lakes).

0 0

Post a comment