Sedimentary Geology 148 (2002) 203 – 220 www.elsevier.com/locate/sedgeo

High-resolution saline lake sediments as enhanced tools for relating proxy paleolake records to recent climatic data series Xavier Rodo´ a, Santiago Giralt b,*, Francesc Burjachs c, Francisco A. Comı´n a, Rafael G. Tenorio d, Ramon Julia` c a

Climate Research Group, Center for Meteorology and Climatology, PCB-UB, and Department of Ecology, Faculty of Biology, University of Barcelona, Avgda. Diagonal 645, 08028 Barcelona, Catalunya, Spain b Royal Museum for Central Africa, Department of Geology, Leuvensesteenweg, 13, B-3080 Tervuren, Belgium c Institute of Earth Sciences ‘‘Jaume Almera’’ (CSIC), Lluı´s Sole´ i Sabarı´s, s/n, 08028 Barcelona, Catalunya, Spain d Department of Applied Physics, High-Tech School of Architecture, University of Sevilla, Avda. Reina Mercedes, 2, 41012 Seville, Spain Accepted 1 August 2001

Abstract Lake level in an endorheic saline lake in Southern Europe has been inferred for the last 105 years (1889 – 1994) at an annual level of resolution using two independent methods. First, ordination analyses (factor analysis (FA), correspondence analysis (CA) and detrended correspondence analysis (DCA)) have been used to point out the mineral successions of the sedimentary record. These successions are evidenced by the arch disposition of both samples and mineral phases in the plane defined by the first two eigenvectors calculated by these analyses. These temporal evolutions are the same as those obtained during the drying and refilling phases of saline lakes. Relay indices (RI) were obtained from distances to the first two eigenvectors, which accurately reconstructed the lake-level evolution during this period. Second, an inferred lake-level series was obtained using a multivariate time series model from the average maximum temperatures and total rainfall. Six main drought periods were found, which coincided with known droughts in the area. A high level of agreement was found between the two reconstructions, which offered the possibility of directly extending the instrumental record back into the past. Therefore, climatic changes could be reconstructed from saline lakes provided that an accurate chronological control of sedimentary processes is available. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Saline lakes; Lake level; Climatic changes; Paleoclimatic reconstruction

1. Introduction

* Corresponding author. Institute of Earth Sciences ‘‘Jaume Almera’’ (CSIC), C/ Lluı´s Sole´ i Sabarı´s, s/n, E-08028 Barcelona, Spain. E-mail address: [email protected] (S. Giralt).

Lake sediment records are useful tools that have long been used to reconstruct changes in lake water levels (Richardson, 1969; Bradbury et al., 1986; Rippey, 1982; Digerfeldt, 1986). Lakes can be important proxy indicators of climatic change in several ways

0037-0738/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 7 - 0 7 3 8 ( 0 1 ) 0 0 2 1 8 - 4

204

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

(Langbein, 1961; Bowler, 1981; Teller and Last, 1990; Mason et al., 1991; Guiot et al., 1993; Rodo´ et al., 1997), especially through the analysis of variations in levels and areas (Benson and Thompson, 1987; Benson and Paillet, 1989; Benson et al., 1991; Harrison et al., 1993). This is particularly true for closed lakes, which are known to be sensitive indicators of changes in climate (Flower and Foster, 1992; Mason et al., 1994). Interest in modern and ancient athalassic saline lakes has grown significantly in recent years (Hammer, 1986; Comı´n and Northcote, 1990; Teller and Last, 1990; Williams, 1991). These athalassic systems are also capable of recording very small variations in evaporation and precipitation, which can drastically affect physical and chemical conditions in lake waters (Pienitz et al., 1992; Last, 1994; Roca and Julia`, 1997). Paleoclimatic studies provide vital information on the long-term climatic trends, which are required to generate and test hypotheses about climate patterns and to validate climate models (Bender et al., 1994; Julia` et al., 1994; Kenneth and Ingram, 1995). Lake levels and areas integrate many physical variables and time scales and can be decomposed in terms of a sum of contributing components. Reconstruction of past water level changes and ecosystem functioning have usually been made using biological remains such as diatoms (Battarbee et al., 1988; Fritz and Battarbee, 1988; Fritz, 1990, 1996), ostracods (Engstrom and Nelson, 1991), and fossil pigments (Sanger, 1988; Leavitt et al., 1989), water content (Rippey, 1982) and sedimentology (Last and De Deckker, 1990; Flower and Foster, 1992; Almquist-Jacobson, 1995). All these provide clues to the understanding of lake dynamics, with external (climatic and catchment area uses by humans) and internal factors (biogeochemical processes) interacting, despite the difficulty of quantifying the relative contributions of the many influences. Nevertheless, there is a lack of reliable long-term series of direct lake-level records with the result that inferred climatic reconstructions are usually made using palaeolimnological data. The ultimate goal is thus to use historical (long-term) fluctuations in lake level as an indicator of regional climatic variations. When pursuing this approach, the key point is to know accurately how fluctuations in lake level will later be mirrored as differential depositions in the sedi-

mentary record. In semiarid environments, where evaporitic sequences record lake-level variations with great precision, the use of the sedimentary history is highly desirable. After having calibrated palaeodata in terms of lake level, it will be possible to infer palaeochanges in regional climate from lake-level fluctuations. Sedimentary facies in saline lakes record the biogeochemical changes taking place over time. In endorheic areas, the hydrological balance is related to rainfall (as the main input) and to evaporation (as the principal output). The surface area of the lake is a function of this balance (Langbein, 1961) with the result that oscillations in the lacustrine area are related to variations in this water balance, as evidenced by a number of saline lakes (LADWP, 1984; Vorster, 1985). Fluctuations in lake level produce concentrations or dilutions of the water, leading to changes in chemical composition (Eugster and Hardie, 1978). As the lake becomes dry, the saturation coefficient of the different mineral species is progressively reached. Carbonates, sulphates and chlorides are the most important authigenic salts precipitating during the drying phases of the lake, whereas, normally, carbonates precipitate during refilling phases, together with the input of siliciclastic minerals.

2. Site description Lake Gallocanta (40j50VN, 2j11VW) is a saline lake where all its components, including water level and mineral composition, fluctuate at different temporal scales driven by meteorological and climatic variations (Comı´n et al., 1991; Rodo´ et al., 1997). Lake Gallocanta is located in an endorheic watershed of the Iberian range (NE Spain) at 1000 m above sea level, between the Santa Cruz and Valdemadera Sierras. The first one is mainly constituted by Paleozoic quartzites, shales and sandstones forming the Hercinian basement, while Mesozoic marls, dolomites and sandstones forming the sedimentary cover compose the second one (Calvo et al., 1978). Between these Sierras and the lake, large Quaternary alluvial fans are developed. The climatological characteristics of this region are highly variable from year to year. The lake is located

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

in a semi-arid zone, with annual average rainfall below 600 mm (Comı´n et al., 1983). The maximum area of the lake is 20 km2 on average (this area being covered by water in 1975 when maximum depth was 2.5 m, maximum length 9.5 km and maximum width 3.4 km). The area of the watershed covers 520 km2. Most of the watershed (Mesozoic and Quaternary plains) is used for cereal cultivation, and native oak forest occupies a small area in the margins of the watershed, where a Paleozoic substrate dominates. The upper sediments in the central part of the lake are rich in organic material, mostly originated from cyanobacterial mats growing during periods of low water level (Comı´n et al., 1991), while top layers are rich in salt content in the margins of the lake (Comı´n et al., 1990). Eight temporary streams drain the watershed, which have water only sporadically. Only three of them discharge directly into the lake, one in the north shore, another one in the southwest shore and another one in the south shore of the lake, at least 3 km far from the sampling zone of this study. During the last 10 years, the maximum lake level fluctuated below 1 m and the lake remained dry for most of the time during this period (Rodo´ et al., 1997). Salinity fluctuates in Lake Gallocanta between 20 and 150 g/l in total dissolved solids, following the long-known model for this type of lakes (Langbein, 1961; Comı´n et al., 1990). It belongs to type IIIb, which is characterized by an ion dominance of Na – Mg –Cl –SO4 (Eugster and Hardie, 1978; Comı´n et al., 1990). These circumstances make Lake Gallocanta an excellent place to undertake paleoenvironmental studies. In this paper, lake-level evolution has been reconstructed using the mineralogical changes of two cores. The level of Lake Gallocanta has been also inferred for the last 105 years (1889 – 1994) from the relationships between climatic variables (rainfall and temperature) and direct measurements of the lake level. The good correspondence between both reconstructions enables future inferences to be made from the mineralogical records of saline lakes in other parts of the world, where the same conditions apply. Beyond this approach, regional climatic inferences could be made in remote areas where saline lakes are the only climatic ‘recorders’.

205

3. Materials and methods Four sediment cores were taken from two sites in the lake, two from the depocenter—labelled Gallocanta-63 —and two others from the geometric center of the lake —labelled Gallocanta-110 (points A and B in Fig. 1)— in order to account for possible variations due to spatial heterogeneity both in sedimentation rate and catchment runoff. Cores were taken with a piston corer from a raft, and, in all cases, at least 1.5 m of sediment were extracted. The cores were wrapped in aluminium foil and stored in a cool room at + 2 jC until sampling, which took place within a month. Cores were subsampled at resolutions as low as 1 to 2 mm with a Microtom in the first few centimeters to account for all the facies changes appearing in this section of the core. Cores were subsampled every centimeter below these few initial centimeters. Only the first 40 cm were studied because they parallel the available instrumental record. The samples for mineralogical study were weighed, dried at 60 jC for approximately 48 h, and weighed again in order to determine their water content. These samples were ground manually in a small agate mill and sieved to obtain a < 40-Am grain-size fraction. X-ray diffractions were performed with an automatic Siemens D500 X-ray diffractometer: Cu ka radiation, 40 kV, 30 mA and graphite monochromator. Quantification of different mineral phases was performed following standard procedures (Chung, 1974). Replicate analyses for contiguous samples indicated a precision of about F 1%, and, in extreme cases, a precision of about 3%. Sediment dating was performed by 210Pb, 226Ra and 137Cs by gamma spectrometry at the top of the core. Radiometric analyses were performed at the analytical service of the University of Sevilla (Spain) and at the Dept. of Physics of the University of Uppsala (Sweden). Radiometric measurements of 210 Pb, 226Ra and 137Cs were performed in 12 slices 2 cm thick, homogeneously distributed over the uppermost 55 cm of the core. These radiometric techniques have been used with a great degree of success in the establishment of chronologies in sediment cores mostly collected from soft-water lakes in temperate zones (Appleby et al., 1995; Noller, 2000). 210 Pb concentrations were determined in the different slices of the core through the measurement of its

206

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

Fig. 1. Geographical setting of Lake Gallocanta in Europe. Capital letters A and B indicate location of sediment extraction sites.

grand-daughter 210Po, in secular equilibrium, by alpha-particle spectrometry applying a total isotope dilution technique (El-Daoushy et al., 1991). 137Cs and 226Ra concentrations were determined by gammaray spectrometry (Bolivar et al., 1996) using, in the case of 137Cs, its gamma-ray emission at 661 keV. For 226 Ra, the gamma-ray emission at 352 keV of its daughter 214Pb was employed. In order to assure the secular equilibrium between the 226Ra and 214Pb in the samples, these were sealed in the measuring

containers and stored for at least 3 weeks before the measurements. A low-background spectrometer was used for the 226 Ra and 137Cs determinations. In background runs done during weekly intervals, clear and unambiguous traces of the 661 keV 137Cs peak were detected in the uppermost part of the cores. The counting rate of the 352 keV peak of 214Pb was carefully determined in order to be substracted from the sediment measurements. This background 214Pb peak has influence in

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

the 226Ra measurements, increasing slightly its uncertainty. However, this fact does not prevent a proper determination of this radionuclide in the sediment slices because its contribution to the peak obtained in the sediment measurements is low (always lower than 10%). Water depths were measured at the deepest point in the lake from a marked stick with a periodicity of 5 days for the period 1974 through 1994. Recorded instrumental climatic data date from the end of the last century. Total annual rainfall amounts and average annual maximum temperatures at the nearest meteorological station (Daroca, 20 km away) were available for the period 1910– 1994. In 1992, a meteorological station was placed on one shore of the lake and calibration with the Daroca series was then performed. Cross-correlations between the two series gave values of rxy = + 0.86 ( p < 0.001) for a total of 210 five-day data points. For the period from 1872 to 1910, when no data from Daroca were available, climatic series from Zaragoza (80 km away) were used. A good correspondence between the two rainfall series appears for the period in common and only a correction for the mean was applied to the Daroca series, for the interval 1872 to 1910. Fig. 2 shows the resulting linear lag-correlation plot of the two series for the period 1910– 1992. The last 2 years available (1992 –1994) were not included in the analysis, as they were not going to be used in successive analysis.

Fig. 2. Linear lag-correlation analysis of Daroca and Zaragoza rainfall series (1910 through 1992). Vertical bars denote correlation coefficients, and the 95% confidence limits are also shown.

207

Table 1 226 Ra concentrations determined in the different slices of the sediment of the Gallocanta-110 core Sample

Depth (cm)

226

G1 G1b G2 G2b G3 G3b G4 G4b G5 G5b G6 G7

0–2 4–6 8 – 10 14 – 16 18 – 20 23 – 25 28 – 30 33 – 35 38 – 40 43 – 45 48 – 50 53 – 55

21 F 2 24 F 2 27 F 3 21 F 3 15 F 2 23 F 3 11 F 1 11 F 2 24 F 2 23 F 2 25 F 3 33 F 3

Ra (Bq/kg)

Values shown are in Bq/kg.

Calibration with the Daroca series gave a value of rxy = + 0.6 ( p < 0.005) for rainfall and rxy = + 0.87 ( p < 0.001) for temperatures. Prior to computing linear lag-correlation analysis, serial correlation was inspected and, when necessary, removed. The longest temperature record goes back to 1889, and this year was therefore used as the starting date of the lakelevel reconstruction. Multivariate time series analysis, linear lag-correlation analysis and cross-spectra were computed using SPSSk and ITSMk. The mineralogical zones were obtained using the Edwards and Cavalli – Sforza’s chord distance method of TILIA (Grimm, 1987).

Fig. 3. Unsupported 210Pb profile with core depth of Gallocanta110. Vertical bars represent 95% confidence intervals.

208

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

Factor analysis (FA) was used to describe the relationships among the mineral phases (STATVIEWk SE+ Graphics) by means of relay indices (RI). Finally, unimodal analyses, such as correspondence analysis (CA) and detrended correspondence analysis (DCA), were employed to determine whether a mineralogical gradient existed throughout the core. CA proves to be useful for detecting, isolating and describing relays (Hennebert and Lees, 1991), while at the same time allowing us to observe the relative positions of both samples and mineral phases. Special use is made of the arch effect in which samples and components from data sets with a strong unidimensional structure (a relay) plot in the form of an arch in the plane of the first two eigenvectors. Thus, a RI can be plotted to reveal details of the sedimentary gradient. This gra-

dient is revealed when the RI is plotted for the different mineral phases that compose the lacustrine sediments, thereby allowing us to arrange them following the environmental gradient (Hennebert and Lees, 1991). CA (Hirschfeld, 1935; Fisher, 1940) is a method with early ecological applications (Goff and Cottam, 1967; Persson, 1981). Some of these reported the high effectiveness of CA in displaying predominant gradients within a community. However, CA also suffers from serious disadvantages (see Hill, 1973, 1979; Gauch, 1982). In these situations, detrending is intended to ensure that, at any point along the first eigenvector, the mean value of the site scores on the subsequent axes is about zero (Ter Braak, 1987). Regarding both the compression of the axis ends

Fig. 4. Mineralogical spectra from Gallocanta-110 (a) and from the upper part of Gallocanta-63 (b). The zones and subzones were established according to the cluster analysis (Grimm, 1987).

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

209

Fig. 4 (continued ).

relative to the axis middle, and to avoid underestimation of distances, CA analyses were recalculated regardless of the two portions of the record, previously separated by the cluster and factor analysis.

4. Results 4.1. Radiometric dating of the sediment. In Table 1, 226Ra concentrations determined in the different slices of the Gallocanta-110 core are shown. Equally, the unsupported 210Pb concentrations for every measured slice in this core were calculated (210Pb minus 226Ra concentrations). They are plotted against depth in Fig. 3. First of all, it is interesting to note the lack of constancy in 226Ra concentrations along the core. These concentrations suffer some fluctuations. Know-

ing that this radionuclide has a natural origin, being incorporated to the sediments associated to the material deposited and that its concentration in the material is strongly dependent on its composition, the variability in its concentration throughout the core points to the variability in the mineral composition of the analyzed samples (i.e. the low 226Ra value at 18 –20 cm deep of the Gallocanta-110 core could be attributed to the highest content of quartz). The unsupported 210Pb also shows an erratic behavior that differs from the exponential decrease of the unsupported 210Pb concentrations, known to occur with depth if the regime of sedimentation (amount of material deposited per year and composition of sediment material) is constant. In fact, it appears non-linear in Fig. 3. The non-uniform composition of the material linked to the likely variations in the sedimentation rate could be some of the reasons that explain the unsupported 210Pb profile obtained.

210

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

Deeper than 40– 45 cm, the unsupported 210Pb fraction begins to be undetectable. As a result of these fluctuations, it is not possible to define a proper and detailed chronology by means of 210 Pb alone, for the top 45 cm of the core. We can only infer an average sedimentation rate valid for these upper centimeters (0.31 F 0.04 cm/year). Unfortunately and due to the low spatial resolution of the available radiometric measurements, it has not been possible to obtain a detailed 137Cs profile along the core. Nevertheless, clear and unambiguous photopeaks at 661 keV were detected in the uppermost three layers measured (until 10 cm), whereas they were not detected in deeper layers. This fact allows us to assure that the slice G2b was formed before 1955 and to estimate a sedimentation rate in the range of 0.25– 0.33 cm/year for the upper 10 –15 cm of the core. In spite of the rough average sedimentation rate determined by the application of the 210Pb dating technique, the obtained rate for the uppermost layers agrees with the values obtained with the 137Cs dating procedure.

constant throughout the sequence. Additional details of the sedimentological processes within the lake and of the sedimentary sequences are discussed in Comı´n et al. (1990, 1991). Cluster analysis clearly separates two zones in Gallocanta-110 (Fig. 4a): the upper zone—to about 22 cm depth—and the lower zone to 40 cm. In the first zone, quartz, calcite and aragonite have higher percentage values. Clusters from the shorter core also show these two zones, with an upper zone to about 18 cm characterized by the dominance of aragonite and a lower zone with high percentages of gypsum. This mineralogical change observed in both cores also corresponds with the expansion of the Olea-Philyrea sp., and it has been interpreted as an increase in the regional humidity (Burjachs et al., 1996). A close observation of the mineral records makes clear of a common pattern of cycles between the two cores (Fig. 5), in spite of the sometimes divergent higher frequency mineral oscillations typical of local phenomena (Powell, 1995). In the most recent sediment cycle, corresponding approximately to the first 20 cm, there is a predominance of aragonite in the depo-

4.2. Geochemical lake-level evolution The mineralogical records are shown in Fig. 4a and b (for Gallocanta-110 and Gallocanta-63, respectively). Seven main mineral phases were recorded: six of them are common to the two cores (calcite, dolomite, aragonite, gypsum, quartz and illite). A new hydrous calcium magnesium carbonate was found in several muddy beds of the core Gallocanta-110 (Queralt et al., 1997). In Gallocanta-110, calcite is present throughout the sequence and ranges from 1 to 24 wt.%. Dolomite percentages are constant in the lower part of the core and progressively decrease upwards. The hydrous carbonate is present in the upper part of the sequence with percentages of about 2 wt.%. Aragonite is also present in the upper centimeters of the core. The gypsum percentages oscillate between 0.5 and 95 wt.% and the values for detrital minerals (quartz and illite) are roughly constant. In Gallocanta-63, gypsum values progressively decrease upwards while aragonite percentages increase. Calcite (15 wt.%), dolomite (6 wt.%), quartz (11 wt.%) and illite (24 wt.%) proportions are roughly

Fig. 5. Coevolution of the aragonite percentages from the Gallocanta-63 sequence and the gypsum values from the Gallocanta-110 core.

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

center, whereas there is a predominance of gypsum in the geometrical center of the lake. Despite these differences, Fig. 5 shows aragonite and gypsum in the two cores co-varying in those phases where both minerals coexist. Factorial analysis (FA) (Fig. 6) shows the opposition between detritic components (quartz and illite), typical of refilling phases (together with calcite) and gypsum as the most stable authigenic component in arid periods. This arrangement of variables is conservative, both for the whole Gallocanta-110 record and the Gallocanta-63 one, although it is in this last record that, when an FA is performed separately for the two main zones defined by cluster analysis, a substitution of aragonite for gypsum is seen. In fact,

211

both aragonite and gypsum are characteristic of the different stages in the lake’s evolution towards complete drying-out phases. Aragonite has its origin in the algal mats, which develop in big amounts during periods of low water level, occupying a large area of the lake (Comı´n et al., 1991). However, the closeness of siliciclastic components to both dolomite and calcite suggests these are closer in space and time. As a consequence of the one-dimensional evolution of water chemistry during the evaporitic sedimentation process, in the sense of progressive concentration at both short-term and long-term intervals, a succession of mineral phases would normally show a gradient. This gradient can be observed (Figs. 7 and 8)

Fig. 6. Factorial analysis of the two sequences. (Note the substitution between aragonite and gypsum in the Gallocanta-63 core).

212

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

in the arch disposition of samples in the plane defined by axes 1 and 2 of the CA. So, the position of samples with respect to the axis implies a particular mineral composition and, thus, a relative water volume. Due to the ‘‘distortion’’ effect of the second axis, detrended correspondence analysis were applied. It

must be pointed out that the arch disposition of both samples and mineral phases from Gallocanta-110 and Gallocanta-63 were not removed (Figs. 7 and 8). This fact implies that there are two gradients underlying in the mineralogical data structure of the cores. The RI for the two first eigenvectors of both cores allows an interpretation of these gradients (Fig.

Fig. 7. Mineral gradient expressed by the arch disposition of both samples (top left) and mineral phases (top right) for Gallocanta-63 derived from the correspondence analysis. Numbers in the top left diagram correspond to the sample number from the top to the bottom of the core. Note that the arch effect is not removed when detrended correspondence analysis is applied (bottom).

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

213

Fig. 8. Idem for Gallocanta-110.

9). In Gallocanta-110, the first eigenvector represents the infilling process, while the second one displays the retraction phase. In the case of Gallocanta-63, this interpretation is reverse; the first eigenvector displays the retraction processes while the second eigenvector represents the infilling phase. This inverse behaviour has been interpreted as a consequence of the core location in the lacustrine system.

Then, position on the second axis, in Gallocanta110, and to the first axis, in Gallocanta-63, is used as the relay index (RI) (Hennebert and Lees, 1991), being an indicator of relative water level oscillations. The two relative water level reconstructions obtained from Gallocanta-63 and Gallocanta-110 display the same pattern but differ in the intensity of events (Fig. 10). RI values were obtained on a percentage scale

214

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

Fig. 9. Relay indices (RI) of the mineral phases from the first two eigenvectors (EV) obtained for Gallocanta-63 and Gallocanta-110. See text for details.

and were adjusted to the absolute scale of real lakelevel values with a non-linear regression function (Comı´n et al., 1991; Sokal and Rohlf, 1995). 4.3. Lake-level reconstruction from 1889 through 1994 using climatic data The period covered for the reconstruction extended back to 1889, when both temperature and rainfall data

were recorded. Rainfall and temperature series are shown in Fig. 11. Even though a more accurate model optimization of the relationship between climatic variables and lake level, which uses evaporation data, wind and relative humidity was already built (Rodo´, 1997), the present model is being developed for data with different resolution and varying temporal coverage (Meentemeyer, 1989; O’Neill et al., 1989; Turner et al., 1989). Due to the lack of accurate and long

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

215

Table 2 Results of the multivariate time series model fitted to the water depth of Lake Gallocanta Estimates a b a

0.268 62.487 741.1 df

Fig. 10. Relative lacustrine water level obtained by the measurement of the distance between samples and axis 1. Thick line refers to Gallocanta-63 and thin line to Gallocanta-110.

enough data series of these additional variables for the period 1889 through 1994, a relationship between three variables (average maximum temperature, total rainfall and lake level) was used, even though the percentage of variability accounted for was lower than when using the six variables. In spite of this, the fitted multivariate time series model was adequate and highly significant (Table 2). These two variables (rainfall and temperature) account for a 62.5% of the total variance in modern lake level. The multiple least squares regression model was performed on stationary series, and only a first-order difference was applied to lake level due to the moderate autocorrelation present in this series (Berthouex and Brown, 1994). Results obtained with this simple model with only two inde-

Fig. 11. Total annual rainfall and annual average temperatures from Daroca for the period 1889 – 1992. Series are smoothed with a centered 5-year moving average.

S.E.

t-value

0.101 20.097 262.4

2.658 3.109 2.825

SS

MS

38075 22820

19038 1756

Model Residual

2 13

F = 10.845

Significance of F = 0.0017

p-value 0.02 0.008 0.014

Independent variables are total rainfall (a) and average maximum temperature (b). a denotes a constant, df = degrees of freedom, SS = sum of squares and MS = mean square. Significances of both estimates and the model are highlighted in bold letters.

pendent variables accounted for such a large part of the total variability, and were so significant both for the parameters and the model that it seemed unnecessary for the geochemically based approach of this study to address here the full complexity of a more exact water-balance model for the lake. The simple time series model fitted is shown in Table 2 with the details. Higher order parameters for the combination of the two variables were also tested, but other explanatory variables than the ones there shown were found not to be significant at the 5% level.

Fig. 12. Inferred lake level for Lake Gallocanta for the period 1890 – 1995 using a multivariate time series model. A bandpass filter of 5 years resulting from this reconstruction is shown. Black line refers to the modeled series and the broken grey line to the instrumentally measured lake-level from 1970 to 1995.

216

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

A model-simulated series for lake level in Lake Gallocanta appears in Fig. 12. The original annual series resulting from the model was filtered with a bandpass filter of 5 years to show longer-term trends. Inferred lake level matches the instrumental record quite accurately. Low level phases in Lake Gallocanta coincide with periods of decreased rainfall, though this relationship is modulated by temperature, which expands the series maxima and minima. In general, periods of enhanced rainfall are associated with low temperatures. Since 1889, there have been five main droughts (the 1992 –1994 one makes a total of 6), recorded as level minima, which on most occasions led to the complete drying-out of the lake. When observing Fig. 12, one must take into account the temporal resolution of the data that makes it difficult to see the complete drying-out phases because, to have a zero value in the data, the lake must be empty for a long period of time. A thin layer of water can still remain in the depocenter, whereas the rest of the lake is dry as has been stated for other saline lakes (De Deckker, 1988). In spite of this, dates for lake minima coincide with historical droughts (Guiral, 1981). The main drought periods occurred in 1895, 1926, 1947– 1950, 1966, 1982 –1984 and 1992 –1994, with the ones in 1895 and 1982 – 1984 being the most intense. Drought periods are characterized by being at one or more standard deviation below the long-term mean. These periods have recently been shown to be related to El Nin˜o-Southern Oscillation extratropical teleconnection patterns (Rodo´ et al., 1997). Lake maxima occurred in 1889, 1917, 1931 –1934, 1958 and 1971 – 1973, with the highest maximum in 1917 at 2.47 m. Periods between droughts are 31, 21 –24, 16 – 19, 17 and 9 to 10 years, which show an increase in the frequency of dry periods. Otherwise, there is no longterm trend in the mean of lake level for this period, even though there has been a drastic decrease in the last 15 to 20 years, linked both to a regional decrease in rainfall (Rodo´ et al., 1997) and to an increased rate of water extraction from the watershed. For the specific case of Lake Gallocanta, and basing on the measured lake-level record (1974 – 1994), we can hypothesize the behaviour of a hypothetical lake having an extreme response time of < 1 year, in opposition to slower ( > 10 years) response times. A lake with < 1 year response time would in fact provide the actual re-

sponse time in its level shown in Fig. 12: a fall from 1974 to 1980, then a rapid drop to a constant low level from 1980 to 1986, then a rapid rise to a constant higher level from 1986 to 1992, then a rapid decrease until 1994. On the time scale of the data set, the variations would be largely in the low-frequency (LF) range, and so a true physical correlation between level and rainfall and temperature should be feasible.

5. Discussion Lakes, particularly those with no outlet (like Lake Gallocanta), act as low-pass filters of climatic variables (Mason et al., 1994), filtering out high-frequency components, and thus tending to emphasize the longterm evolution of climate. When spectral coherences between the real-measured-level series and the modelsimulated series are analyzed, values are fairly low, in spite of the few data points. However, when lower frequency components of simulated series are also considered (for instance with the aid of bandpass filters of different span), coherences grow considerably in the case of 5 years, where the maximum similarity occurs. This fact confirms the low-pass filter nature of Lake Gallocanta, as it has been shown for other lakes (Mason et al., 1991). Furthermore, the correlation between the stationary 5-year inferred series and the real lake-level series is higher (rxy = 0.71; S.E.: 0.28; p < 0.001) and more significant than the one obtained between the annual inferred series and the lake level (rxy = 0.46; S.E.: 0.17; p < 0.01) once autocorrelation has been removed. Variable smoothing factors are common to lakes of different size, morphometry and lake catchment areas, and, consequently, are also characteristic of the resolution achieved by each particular lake. However, in any case, the present approach showed how the response time for Lake Gallocanta is centered in the low-frequency (LF) range of variations (i.e. on a time scale much longer than the equilibrium response time of the lake, meaning that the lake is always approximately ‘‘in equilibrium’’ with the climate). Therefore, as Mason et al. (1994) state, true physical correlations between the lake levels and the water flow rates will be possible since the variations are mainly made up of frequency components in the LF range, and the lake area (and hence level) is directly related to the current values of input and

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

output flow rates. In spite of this assertion, it must be noted that the response time is not generally constant for any lake, but depends on lake area, the rate of change of level with area, and the net evaporation from the lake surface. Therefore, it is likely to have different values at different times and levels, and so the frequency ranges represented by the LF and the high frequency (HF) will change with time. Comparison between the two (geochemical and climatic) inferred series is quite good, according to the range of values and the situation of maxima and minima. The accordance is even better for the last part of the record (from the beginnings of the seventies). However, a slight delay is evident in Fig. 13 in some occasions (at approximately 1926 and 1935). In addition, sometimes, the two series are completely out of phase (such as in 1900 and in 1948) or there is a remarkable delay between the two records. This could be due to two major factors. (1) When the lake dries out, or when the level is extremely low, there is no — or might not be any — sedimentation in the record and even a deflation process due to wind action may cause sediment losses, whereas chronological interpolation does not take these phenomena into account. This delay problem is much more important the longer the lake remains dry.

217

(2) The sampling scale used in: (a) the radiometric method. In spite of post-sedimentary processes occurring in these environments, such as deflation and bioturbation, a linear relationship appears between the two inferred lake-level series. However, errors might happen due to the uncertainty associated to any radiochronological control. (b) The mineralogical data. Samples have been taken at most at 2-mm intervals. Even sampling at this high resolution and due to low sedimentation rates, 2 mm may well represent more time than the average 5-year value employed. The good agreement in both low and high water level phases can also be seen, though perhaps those periods following durable major oscillations are the ones better reflected in the sediment. Recent history thus seems to be important for future depositions. The geochemistry of lake water, both in closed and open lakes, evolves in time towards a progressive salinization, with the more soluble salts remaining in the water (chlorides and sulphates). For short temporal periods (of about a millenium) and for samples that are close in time, differences can be treated as comparable despite droughts and partly homogenized by bioturbation processes. In spite of this, concordance between the two series is high, which makes the geochemistry of this kind of widely fluctuating lake an excellent indicator of lakelevel evolution.

6. Conclusions

Fig. 13. Sequential time series for the mineral-inferred level series (dotted) and for the inferred water depth (continous) of Lake Gallocanta, for the period covering 1889 through 1992.

Saline lakes that undergo strong fluctuations in their water levels and pass through a wide range of salinity conditions record lake-level changes as distinctive depositional phases in their sediments. The temporal succession of different mineral phases shows a gradient formed by the progressive concentration in lake water. In spite of mineral spatial heterogeneity in the sedimentary record and of the sampling density, there is a co-varying dynamic in the two core zones. Aragonite and gypsum are the two main mineral phases reflecting critical drought periods, whereas detritals and calcite sediment reflect refilling periods. Relay indices obtained from the correspondence analysis are a good indicator of relative lake level. Annual rainfall and temperature series in Lake Gallocanta explain the 62% of the total variability of

218

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

lake level, which has been reconstructed for the period from 1889 through 1994. Six main drought periods can be identified between 1889 and 1994: 1895, 1926, 1947– 1950, 1966, 1982 –1984 and 1992 –1994, with the one in 1982– 1984 being the most intense. Equally, five other periods corresponding to lake water maxima can also be identified, namely 1889, 1917, 1931 – 1934, 1958 and 1971 – 1973, with the highest maximum in 1917 when the lake reached a depth of 2.47 m. Saline lakes such as Lake Gallocanta act as lowpass filters of climatic variables, attenuating most of the high-frequency components and thus showing the long-term evolution of climate with a high degree of accuracy. It is also shown that these high-frequency components do not represent a noticeable part of the variability, which appears to be dominated by lower frequency components. This paper demonstrates how these geochemical records are very useful tools when trying to infer proxy climatic changes and can serve as detailed tracers if sampled at adequate levels of resolution. In these cases, the mineral depositional sequence in endorheic lakes, based on a good chronological control of sedimentary processes (by 137Cs, 210Pb, 14C), becomes an outstanding tool for reconstructing regional climate dynamics and tracing back historical lake-level changes. The current view that sediment records in these lakes are of very limited resolution compared with lake varved sediments (Ralska-Jasiewiczowa et al., 1992; Zolitschka et al., 1992) should change, as should the view that they are strongly affected by postsedimentary processes such as seepage (Merkt, 1971). These findings offer new insights and enable climate changes to be reconstructed in climatic transition regions (Hammer, 1986), like the Iberian Peninsula (CLIMAP, 1984), which makes them particularly useful when monitoring regional climatic changes. Their usefulness lies in the fact that there is normally a lack of both meteorological stations or other lacustrine environments with high temporal resolutive sediments located in semiarid areas.

Acknowledgements This work was partially funded by CICYT (AMB95-0935-E and CLI95-1905). We thank Dr. F. Plana and J. Elvira from the ICT ‘‘Jaume Almera’’

(CSIC) for X-ray diffractions, the Instituto Meteorolo´gico Nacional (serv. Arago´n) for supplying climatic data series, and the rangers in L. Gallocanta for unvaluable field assistance. We also thank I. Mason, J.M. Rouchy and an anonymous reviewer for their useful comments to a former version of this manuscript. S.G. is in receipt of a grant from the Spanish Ministry of Education (Becas de Perfeccionamiento de Tecno´logos y Doctoress en el Extranjero).

References Almquist-Jacobson, H., 1995. Lake-level fluctuations at Ljustja¨rnen, central Sweden and their implications for the Holocene climate of Scandinavia. Palaeogeogr., Palaeoclimatol., Palaeoecol. 118, 269 – 290. Appleby, P.G., Jones, V.J., Ellis-Evans, J.C., 1995. Radiometric dating of lake sediments from Signy Island (maritime Antarctic): evidence of recent climatic change. J. Paleolimnol. 13, 179 – 191. Battarbee, R.W., Anderson, N.J., Appleby, P.G., Flower, R.J., Fritz, S.C., Haworth, E.Y., Higgitts, A., Jones, V.J., Kreiser, A., Munro, M.A., Natkanski, J., Oldfield, F., Patrick, S.T., Richardson, N.G., Rippey, B., Stevenson, A.C., 1988. Lake Acidification in the United Kingdom 1800 – 1986: Evidence from Analysis of Lake Sediments. Ensis, London, 68 pp. Bender, M., Sowers, T., Dickson, M-L., Orchardo, J., Grootes, P., Mayewski, P.A., Meese, D.A., 1994. Climate correlations between Greenland and Antarctica during the past 100.000 years. Nature 372, 663 – 666. Benson, L.V., Paillet, F.L., 1989. The use of total lake-surface as an indicator of climatic change: examples from the Lahotan Basin. Quat. Res. 32, 262 – 275. Benson, L.V., Thompson, R.S., 1987. Lake-level variation in the Lahontan Basin for the past 50,000 years. Quat. Res. 28, 69 – 85. Benson, L.V., Meyers, P.A., Spencer, R.J., 1991. Change in the size of Walker Lake during the past 5000 years. Palaeogeogr., Palaeoclimatol., Palaeoecol. 81, 189 – 214. Berthouex, P.M., Brown, L.C., 1994. Statistics for Environmental Engineers. Lewis Publishers, Boca Raton, Florida, 352 pp. Bolivar, J.P., Garcı´a-Tenorio, R., Garcı´a-Leo´n, M., 1996. A method for the determination of counting efficiencies in gamma-spectrometric measurements using HPGe detectors. Nucl. Instrum. Methods A 382, 495 – 502. Bowler, J.M., 1981. Australian salt lakes: a palaeohydrologic approach. Hydrobiologia 82, 431 – 444. Bradbury, J.P., Leyden, B., Salgado-Laborian, M., Lewis, W.M., Schubert, C., Digerfeldt, G., 1986. Studies on past lake level fluctuations. In: Berglund, B.E. (Ed.), Handbook of Holocene Palaeoecology and Palaeohydrology. Wiley, Chichester, pp. 127 – 143. Burjachs, F., Rodo´, X., Comı´n, F.A., 1996. Gallocanta: ejemplo de secuencia palinolo´gica en una laguna efı´mera. Estud. Palinol., 25 – 29. Calvo, J., Gonzalez, J.M., Villena, J., 1978. Primeros datos sobre la

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220 sedimentacio´n de dolomı´a en la Laguna de Gallocanta. Tecniterrae 21, 1 – 10. Chung, F.H., 1974. Quantitative interpretation of X-ray diffraction patterns of mixtures: II. Adiabatic principles of X-ray diffraction analysis of mixtures. J. Appl. Crystallogr. 7, 526 – 531. CLIMAP, 1984. The last interglacial ocean. Quat. Res. 21, 123 – 224. Comı´n, F.A., Northcote, T. (Eds.), 1990. Saline Lakes IV. Hydrobiologia, vol. 197. Dr. Junk Publishers, The Hague, Netherlands, 308 pp. Comı´n, F.A., Alonso, M., Lo´pez, P., Comelles, M., 1983. Limnology of Gallocanta Lake, Aragon, northeastern Spain. Hydrobiologia 105, 207 – 221. Comı´n, F.A., Julia`, R., Comı´n, M.P., Plana, F., 1990. Hydrogeochemistry of Lake Gallocanta (Arago´n, NE Spain). Hydrobiologia 197, 51 – 66. Comı´n, F.A., Julia`, R., Comı´n, M.P., 1991. Fluctuations, the key aspect for the ecological interpretation of salines lake ecosystems. In: Ros, J.D., Prat, N. (Eds.), Homage to Ramon Margalef; or, Why There is Such Pleasure in Studying Nature. Oecologia Aquatica, vol. 10. University of Barcelona, Barcelona, pp. 127 – 135. De Deckker, P., 1988. Biological and sedimentology facies of Australian salt lakes. Palaeogeogr., Palaeoclimatol., Palaeoecol. 62, 237 – 270. Digerfeldt, G., 1986. Studies on past lake-level fluctuations. In: Berlung, B. (Ed.), Handbook of Holocene Palaeoecology and Palaeohydrology. Wiley, New York, pp. 127 – 144. El-Daoushy, F., Olsson, K., Garcı´a-Tenorio, R., 1991. Accuracies in Po-210 determination for lead-210 studies. Hydrobiologia 214, 43 – 52. Engstrom, D.R., Nelson, S.R., 1991. Paleosalinity from trace metals in fossil ostracodes compared with observational records at Devils Lake, North Dakota, USA. Palaeogeogr., Palaeoclimatol., Palaeoecol. 83, 295 – 312. Eugster, H.P., Hardie, L.A., 1978. Saline lakes. In: Lerman, A. (Ed.), Lakes: Chemistry, Geology, Physics. Springer-Verlag, New York, pp. 237 – 294. Fisher, R.A., 1940. The precision of discriminant functions. Ann. Eugen. 10, 422 – 429. Flower, R., Foster, I.D.L., 1992. Climatic implications of recent changes in the lake level at Lac Azigza (Morocco). Bull. Soc. Geol. Fr. 163, 91 – 96. Fritz, S.C., 1990. Twentieth-century salinity and water-level fluctuations in Devils Lake, North Dakota: test of a diatom-based transfer function. Limnol. Oceanogr. 35, 1771 – 1781. Fritz, S.C., 1996. Paleolimnological records of climate change in North America. Limnol. Oceanogr. 41, 882 – 889. Fritz, S.C., Battarbee, R.W., 1988. Sedimentary diatom assemblages in saline and freshwater lakes of the Northern Great Plains, North America: preliminary results. In: Round, F. (Ed.), Proceedings of the Ninth Symposium on Living and Fossil Diatoms, Bristol, England. Biopress, Bristol, pp. 265 – 271. Gauch, H.G., 1982. Multivariate analysis in community ecology. Cambridge Studies in Ecology. Cambridge University Press, Cambridge, New York, 298 pp. Goff, F.G., Cottam, G., 1967. Gradient analysis: the use of species and synthetic indices. Ecology 48, 793 – 806.

219

Grimm, E., 1987. A Fortran 77 program for stratigraphically constrained cluster analysis by the method of incremental sum of squares. Comput. Geosci. 13, 13 – 35. Guiot, J., Harrison, S.P., Prentice, I.C., 1993. Reconstruction of Holocene precipitation patterns in Europe using pollen and lake-level data. Quat. Res. 40, 139 – 149. Guiral, J.J., 1981. Introduccio´n al conocimiento del clima de la cuenca de la laguna de Gallocanta. In: Proyex, S.A. (Ed.), Estudio de la biocenosis de la laguna de Gallocanta y su cuenca. CEOTMA. Minist. Obras Pu´blicas Urbanismo, Madrid, pp. 119 – 138. Hammer, U.T., 1986. Saline lake ecosystems of the world. Monographiae Biologicae, vol. 59. Dr. W. Junk Pub., Dordrecht, 616 pp. Harrison, S.P., Prentice, I.C., Guiot, J., 1993. Climatic controls on Holocene lake-level changes in Europe. Clim. Dyn. 8, 189 – 200. Hennebert, M., Lees, A., 1991. Environmental gradients in carbonate sediments and rocks detected by correspondence analysis: examples from the Recent of Norway and the Dinantian of southwest England. Sedimentology 38, 623 – 642. Hill, M.O., 1973. Reciprocal averaging: an eigenvector method of ordination. J. Ecol. 61, 237 – 249. Hill, M.O., 1979. DECORANA-A FORTRAN Program for Detrended Correspondence Analysis and Reciprocal Averaging. Cornell University, Ithaca, NY, 52 pp. Hirschfeld, H.O., 1935. A connection between correlation and contingency. Proc. Cambridge Philos. Soc. 31, 520 – 524. Julia`, R., Negendank, J.F.W., Seret, G., Brauer, A., Burjachs, F., Endres, Ch., Giralt, S., Pare´s, J.M., Roca, J.R., 1994. Holocene climatic change and desertification in the Western Mediterranean region. Terra Nostra 1, 81 – 84. Kenneth, J.P., Ingram, B.L., 1995. A 20.000-year record of ocean circulation and climate change from the Santa Barbara basin. Nature 377, 510 – 513. Los Angeles Department of Water and Power (LADWP), 1984. Background report on Geology and Hydrology of Mono Basin. Report of the Aqueduct Division, Hydrology Section, Los Angeles, CA. Langbein, W.B., 1961. Salinity and hydrology of closed lakes. U.S.G.S. Prof. Paper, vol. 412. U.S. Gov. Print Office, Washington, 220 pp. Last, W.M., 1994. Paleohydrology of playas in the northern Great Plains: perspectives from the Palliser’s Triangle. In: Rosen, M.R. (Ed.), Paleoclimate and Basin Evolution of Playa Systems. Geological Society of America Special Paper, vol. 289. Boulder, Colorado, pp. 69 – 80. Last, W.M., De Deckker, P., 1990. Modern and Holocene carbonate sedimentology of two saline volcanic maar lakes, southern Australia. Sedimentology 37, 967 – 981. Leavitt, P.R., Carpenter, S.R., Kitchell, J.F., 1989. Whole-lake experiments: the annual record of fossil pigments and zooplankton. Limnol. Oceanogr. 34, 700 – 717. Mason, I.M., Harris, A.R., Birkett, C.M., Cudlip, W., Rapley, C.E., 1991. Remote sensing of lakes for the proxy monitoring of Climatic Change. Proc. 16th Ann. Conf. Remote Sensing Society, 314 – 324.

220

X. Rodo´ et al. / Sedimentary Geology 148 (2002) 203–220

Mason, I.M., Guzkowska, M.A.J., Rapley, C.G., Street-Perrot, F.A., 1994. The response of lake levels and areas to climatic change. Clim. Change 27, 161 – 197. Meentemeyer, V., 1989. Geographical perspectives of space, time and scale. Landscape Ecol. 3, 163 – 173. Merkt, J., 1971. Zuverla¨ssige Ausza¨hlungen von Jahressschichten in Seesedimenten mit Hilfe von Gross-Du¨nnschliffen. Arch. Hydrobiol. 69, 145 – 154. Noller, J.S., 2000. Lead-210 geochronology. In: Noller, J.S., Sowers, J.M., Lettis, W.R. (Eds.), Quaternary Geochronology: Methods and Applications. AGU Reference Shelf, vol. 4. American Geophysical Union, Washington, pp. 115 – 120. O’Neill, R.V., Johnson, A.R., King, A.W., 1989. A hierarchical framework for the analysis of scale. Landscape Ecol. 3, 193 – 205. Persson, S., 1981. Ecological indicator values as an aid in the interpretation of ordination diagrams. J. Ecol. 69, 71 – 84. Pienitz, R., Walker, I.R., Zeeb, B.A., Smool, J.P., Leavitt, P.R., 1992. Biomonitoring past salinity changes in an athalassic subartic lake. Int. J. Salt Lake Res. 1, 91 – 123. Powell, T.M., 1995. Physical and biological scales of variability in lakes, estuaries and the coastal ocean. In: Powell, T.H., Steele, J.H. (Eds.), Ecological Time Series. Chapman & Hall, New York, pp. 119 – 134. Queralt, I., Julia`, R., Plana, F., Bischoff, J.L., 1997. An hydrous Cabearing magnesium carbonate in recent saline lake sediments, Salines Lake, Spain. Am. Mineral. 82, 812 – 819. Ralska-Jasiewiczowa, B., van Geel, B., Goslar, T., Kuc, T., 1992. The record of the Late Glacial/Holocene transition in the varved sediments of the lake Gosciaz, central Poland. Sver. Geol. Unders. 81, 257 – 268. Richardson, J.L., 1969. Former lake-level fluctuations—their recognition and interpretation. Mitt. Int. Ver. Limnol. 17, 78 – 93. Rippey, B., 1982. Sedimentary record of rainfall variations in a subhumid lake. Nature 296, 434 – 436. Roca, J.R., Julia`, R., 1997. Late-glacial and Holocene lacustrine sequence evolution based on ostracode assemblages in southeastern Spain. Geobios 30, 823 – 830.

Rodo´, X., 1997. Ecological time scales in an aquatic ecosystem: the Gallocanta Lake (NE Spain) (in catalan). PhD Thesis, University of Barcelona, 409 pp. Rodo´, X., Baert, E., Comı´n, F.A., 1997. Variations in seasonal rainfall during the present century: relationships with the North Atlantic Oscillation and the El Nin˜o-Southern Oscillation. Clim. Dyn. 13, 275 – 284. Sanger, J.E., 1988. Fossil pigments in paleoecology and paleolimnology. Palaeogeogr., Palaeoclimatol., Palaeoecol. 62, 343 – 359. Sokal, R.R., Rohlf, F.J., 1995. Biometry. The Principles and Practice of Statistics in Biological Research. Freeman, New York, p. 850. Teller, J.T., Last, W.M., 1990. Paleohydrological indicators in playas and salt lakes, with examples from Canada, Australia and Africa. Palaeogeogr., Palaeoclimatol., Palaeoecol. 76, 215 – 240. Ter Braak, C.J.F., 1987. Ondination. In: Jongman, R.H.G., Ter Braak, C.J.F., van Tongeren, O.R.F. (Eds.), Data Analysis in Community and Landscape Ecology. Pudoc, Wageningen, pp. 91 – 173. Turner, M.G., Dale, V.H., Gardner, R.H., 1989. Predicting across scales: theory development and testing. Landscape Ecol. 3, 245 – 252. Vorster, P., 1985. A water balance forecast model for Mono Lake, California. Master’s Thesis, California State University, Hayward. Earth Resources Monograph, 10. San Francisco, California, U.S. Forest Service, Region 5. Williams, W.D. (Ed.), 1991. Salt Lakes and Salinity. Hydrobiologia, Dr. Junk Publishers, The Hague, Netherlands, p. 210. Zolitschka, B., Haverkamp, B., Negendank, J.F.W, 1992. Younger Dryas oscillation—Varve microstratigraphic, palynological and palaeomagnetic records from lake Holzmaar, Germany. In: Bard, E., Broecker, W.S. (Eds.), The Last Deglaciation: Absolute and Radiocarbon Chronologies. NATO ASI Series I, vol. 2. Springer-Verlag, Berlin, pp. 81 – 101.