Fluctuating asymmetry of birch leaves did not increase with pollution and drought stress in a controlled experiment

Ecological Indicators 84 (2018) 283–289

Contents lists available at ScienceDirect

Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind

Research paper

Fluctuating asymmetry of birch leaves did not increase with pollution and drought stress in a controlled experiment Vitali Zvereva, Ang Dawa Lamab, Mikhail V. Kozlova, a b

MARK

⁎

Section of Ecology, Department of Biology, University of Turku, 20014 Turku, Finland Section of Biodiversity and Environmental Science, Department of Biology, University of Turku, 20014 Turku, Finland

A R T I C L E I N F O

A B S T R A C T

Keywords: Betula pubescens subsp. czerepanovii Bioindication Blinded measurements Heavy metal resistance Research methodology

Fluctuating asymmetry (FA), defined as small, non-directional deviations from perfect symmetry in morphological characters, is often recommended as a handy indicator of environmental stress. A reliance on observational data to provide empirical evidence for the effects of different stressors on FA, in combination with increasing attempts to use FA in environmental research, underscore the need for careful examination of the relationships between environmental stress and FA. We experimentally tested the hypotheses that (i) heavy metal and drought stress increase the leaf FA in plants, and that (ii) plants persisting in heavily polluted sites possess greater stress tolerance and therefore show smaller increases in leaf FA in response to heavy metals than do plants from unpolluted sites. We collected mountain birch, Betula pubescens subsp. czerepanovii, seeds from eight polluted and ten unpolluted sites and reared the seedlings in a sophisticated greenhouse (phytotron). We compared leaf FA between control seedlings, seedlings irrigated with water containing copper and nickel sulphates, and seedlings exposed to drought. Leaf FA showed no response to either heavy metals or drought, despite significant impacts of these treatments on seedling height, leaf size and photosynthetic efficiency. This FA result was independent of the level of pollution at the site of seed origin and consistent for FA values based on low and high accuracy measurements of leaf width, as well as for FA values based on measurements of the widths and areas of leaf halves. Our findings add to accumulating evidence regarding inconsistent relationships between FA and abiotic stress, thereby questioning the indicatory value of FA. We strongly recommend that the use of FA as an indicator of environmental stress be limited to study systems for which the existence of cause-and-effect relationship between the stressing impact and the changes in FA is confirmed by controlled, blinded experiments.

1. Introduction Ecological management requires timely prediction of emerging environmental problems. This requirement gives special value to ecological indicators that could be used as early warning systems to signal problems not yet apparent (Dale and Beyeler, 2001). One such indicator, introduced in the early 1990s, is fluctuating asymmetry (FA), which is defined as small, non-directional deviations from perfect symmetry in morphological characters of plants and animals. FA has been advertised as a universal and easy to measure stress indicator (Zakharov, 1990; Clarke, 1992; Parsons, 1992; Graham et al., 1993; Freeman et al., 1993). In line with previous studies, one of us (MVK) once suggested that the FA of birch leaves could serve as a convenient indicator for rapid assessment of environmental quality (Kozlov et al., 1996). However, this conclusion was contested a few years later, when an analysis of multiyear data demonstrated no relationship between the leaf FA of mountain birch Betula pubescens subsp. czerepanovii (Orlova)

⁎

Hämet-Ahti and heavy metal and sulphur dioxide loads (Valkama and Kozlov, 2001). Further analysis of data collected from multiple plant species in the impact zones of 18 industrial enterprises in the Northern hemisphere also failed to reveal any significant differences in FA between heavily polluted and unpolluted sites (Kozlov et al., 2009), potentially due to a rapid development of evolutionary adaptations to pollution (Kozlov, 2005; Eränen et al., 2009). The continued accumulation of negative and inconclusive results dampened the initial optimism regarding the use of FA in environmental studies (Lajus et al., 2009), and sceptical reviews (Palmer, 1996; Clarke, 1998; Bjorksten et al., 2000; Rasmuson, 2002) began to point out a general inconsistency in the relationships between FA, stress and fitness. Nevertheless, despite the justified concerns expressed by evolutionary biologists, the number of applied studies substantiating the indicatory value of FA or deriving conclusions on relative levels of environmental disturbance based on FA measurements in plants and animals continues to increase (reviewed by Kozlov, 2017). These

Corresponding author. E-mail address: mikoz@utu.fi (M.V. Kozlov).

http://dx.doi.org/10.1016/j.ecolind.2017.08.058 Received 27 February 2017; Received in revised form 21 August 2017; Accepted 23 August 2017 1470-160X/ © 2017 Elsevier Ltd. All rights reserved.

Ecological Indicators 84 (2018) 283–289

V. Zverev et al.

with altitude of the study site and with birch hybridisation (Wilsey et al., 1998). By contrast, other studies have indicated no change in birch FA after heavy grazing (Berteaux et al., 2007) or in response to nutrient stress (Black-Samuelsson and Andersson, 2003) and no association of birch FA with either leaf growth rate (Kozlov, 2003) or the date of leaf fall (Kozlov, 2004).

ongoing attempts to use FA in applied ecological research underscore the need for careful examination of the relationships between environmental stress and FA. The reliability of FA as a bioindicator can be explored by controlled experiments that focus on the cause-and-effect and dose-and-effect relationships between FA and the stressor(s) of interest. Unfortunately, experiments of this kind remain scarce (e.g. Mal et al., 2002; Savelieva et al., 2017; Kolbas et al., 2014; Nishizaki et al., 2015; Sandner and Matthies, 2017), and the shortage of experimental data is in no way compensated by the wealth of observational studies which demonstrate statistically significant associations between FA and potentially stressing impacts. Furthermore, these correlative studies, due to their great number, outweigh experimental studies in narrative reviews and metaanalyses (e.g. Knierim et al., 2007; Allenbach, 2011; Beasley et al., 2013), thereby leading to an impression of a consistently high indicatory value for FA. For example, Beasley et al. (2013) concluded that FA is a legitimate biomarker of environmental stress. This shortage of experimental data prompted us to publish the outcomes of an experiment that explicitly addressed the effects of two abiotic stressors, heavy metals and drought, on the leaf FA in mountain birch. This experiment, conducted in 2003, was at that time regarded sceptically by the reviewers, as our conclusion was that stress had no effect on FA. This obliged us to exclude the larger part of the data on FA from our previous manuscript (Eränen et al., 2009) in its final published form. Based on this experience, we strongly agree with the suggestion by Diaz-Gil et al. (2015) that a strong bias may exist in the published literature regarding positive relationships between stress and FA. This gives special importance to the ‘negative’ results, as publication of these would advance the understanding of the value of FA for environmental ecology and management. The goal of the present study was to conduct an experimental exploration of the impacts of heavy metals (nickel and copper) and drought on the FA of mountain birch leaves. We tested two specific predictions derived from the theory of developmental stability (Freeman et al., 1993; Møller and Swaddle, 1997) and from earlier observational studies (Kozlov et al., 1996; Hodar, 2002; Kozlov and Niemelä, 2003; Fair and Breshears, 2005; Kozlov, 2005). We predicted that (i) the exposure of growing birches to heavy metals and drought will increase the FA of their leaves, and that (ii) the increases in leaf FA in response to heavy metals will be smaller in birches from heavily polluted sites than from unpolluted sites.

2.2. Experimental design

2. Materials and methods

Mountain birch seeds were collected in October 2002 from five mother trees in each of 18 study sites around the Kola Peninsula (for locality data consult Eränen et al., 2009). Eight of these sites were located in heavily polluted areas near the smelters in Monchegorsk and Nikel, and ten sites were located in pristine (unpolluted) habitats. Seeds were stored at +3 °C until 15 January 2003, and then stratified for one month, followed by germination at +21 °C under a 24 h day length. Three weeks after sowing, the seedlings were individually replanted into pots in a mixture of 70% standardized peat soil and 30% perlite and grown at +15 °C until the beginning of the experiments. The experiment was conducted using a sophisticated greenhouse (phytotron) at the University of Tromsø, Norway. The air temperature was controlled at an accuracy of ± 0.5 °C, and air humidity of ± 5% RH, with a constant water vapour saturated deficit of 530 Pa. We used nine seedlings from each of the 79 mother trees. These seedlings were evenly distributed among three growth chambers with identical climatic characteristics, and among three treatments within each chamber: control, heavy metal stress and drought stress. Thus, each chamber contained three seedlings from each mother tree, and these three seedlings were subjected to different treatments. The experiment was initiated on 13 May 2003 at +15 °C in natural light conditions. Five days a week, the seedlings in the control and heavy metal stress treatments were watered with 30 mL of tap water, whereas seedlings in the drought stress treatment received 15 mL. The water used to irrigate the seedlings in the heavy metal treatment included added nickel and copper sulphates to give concentrations of 5 mg L−1 copper and 10 mg L−1 nickel. On the last 2 days of every week, all plants were irrigated with clean tap water in the same quantities as mentioned above. After three months, seedlings exposed to heavy metals and drought demonstrated smaller height, leaf size and leaf photosynthetic efficiency than control seedlings; seedlings from polluted sites showed lower accumulation of foliar nickel and lower stress in heavy metal treatments than seedlings from clean sites (for more details, consult Eränen et al., 2009).

2.1. Study area and study object

2.3. Leaf sampling and measurements

The Kola Peninsula is located in the north-western part of European Russia, next to Finland and Norway, to the north of Polar Circle. The mountain ranges and heavy industry in this region create multiple stress gradients, thereby offering unique possibilities for studies on environmental and evolutionary ecology. The presence of two large non-ferrous smelters, in Monchegorsk and Nikel, with similar compositions of pollutants and similar histories of environmental impacts (Kozlov et al., 2009), offers replication of heavily contaminated areas within a similar natural environment (Eränen et al., 2009; Ruotsalainen et al., 2009). Mountain birch is the tree-line species and one of the main forestforming trees in subarctic Europe, and it is the only tree species that is still relatively abundant in the extremely contaminated habitats surrounding the smelters at Monchegorsk and Nikel (Kozlov et al., 2009). Its high ecological importance in subarctic forests has led to intensive study of mountain birch in both pristine and disturbed environments (Wiegolaski, 2005; Eränen et al., 2009; Zverev, 2009). Leaf FA of different birch species was reported to increase in industrially polluted areas (Kozlov et al., 1996; Ivanov et al., 2015) and in the years with low early summer temperatures (Valkama and Kozlov, 2001), as well as

On 8–9 August 2003, we collected two fully expanded leaves from the top of each seedling, press-dried these leaves between sheets of filter paper, and then mounted them as ordinary herbarium specimens. Some seedlings had died prior to sampling, or had only a single green leaf remaining by the time of sampling, so we obtained a total of 1331 leaves. For each leaf, one author (VZ) measured the width of the left and right halves at the midpoint between the base and the apex of leaf lamina. The measurements were conducted with a ruler to the nearest 0.5 mm (low accuracy measurements hereafter); the perpendicularity of the measurement line to the midrib was controlled visually. This measurement protocol was identical to one used in earlier studies by our team (Kozlov et al., 1996; Valkama and Kozlov, 2001). The measurements were conducted twice, with a two-month interval between measurements. We excluded the possibility that the low accuracy of our measurements was the primary reason for our failure to detect a stress effect on leaf FA by scanning a random subsample of 150 leaves (50 leaves per treatment) at 600 dpi and re-measuring these leaves in 2016. We used ImageJ and one author (ADL) measured the width of the left and right leaf halves at the midpoint between the base and the apex of leaf lamina 284

Ecological Indicators 84 (2018) 283–289

V. Zverev et al.

Table 1 Sources of variation in the results of leaf measurements. Source of variation

Side (S) Individual (I) S×I

Width, low accuracy data

Width, high accuracy data

Area, high accuracy data

F

df

P

F

df

P

F

df

P

40.8 19.0 3.35

1,1330 1330,1330 1330,2662

< 0.0001 < 0.0001 < 0.0001

0.02 20.6 36.1

1,149 149,149 149,300

0.90 < 0.0001 < 0.0001

2.45 74.4 736.1

1,149 149,149 149,300

0.12 < 0.0001 < 0.0001

Finally, we correlated the FA values obtained using different measurement protocols to explore the impacts of the measured leaf trait and of the accuracy of measurements on the recorded FA levels.

and area of left and right leaf halves (to the nearest 0.1 mm and 1.0 mm2, respectively; high accuracy measurements hereafter). Again, all measurements were conducted twice, with a two-week interval between measurements. The perpendicularity of the measurement line to the midrib was controlled by instrumentation. This second measurement protocol was identical to one currently recommended for studies of the FA of plant leaves (Kozlov et al., 2017). Both VZ and ADL performed all measurements blindly with respect to the treatment.

3. Results 3.1. Identification of FA All three data sets (see Appendices S1–S3 in Supporting information) demonstrated highly significant side × leaf interactions, confirming the existence of FA in our samples and our ability to identify FA using repeated measurements of the given accuracy (Table 1). The left and right sides in a subsample of 150 leaves did not differ in the measured characters, indicating an overall absence of directional asymmetry; however, low accuracy measurements of 1331 leaves suggested that the width of the left half of the mountain birch leaf is, on average, 2% larger than the width of the right half (Table 1). Variance components from the mixed model ANOVA indicated that measurement error (ME5) accounted for 46% of the variation in FA when calculated from low accuracy measurements of leaf width. This value was 5.4% of the variation in FA when calculated from high accuracy measurements of leaf width, and only 0.3% when calculated from high accuracy measurements of leaf area. Consistently, ME1 for low accuracy measurements of leaf width was about a half of the absolute difference in values of leaf width from the left and right leaf halves (1.09 and 2.13 mm, respectively), whereas for high accuracy measurements ME1 did not exceed 11% of the absolute difference in values of the measured character between the left and right leaf halves (width: 0.20 and 1.85 mm; area: 5.4 and 88.1 mm2, respectively). The values of FA based on low and high accuracy measurements of the width of the same 150 leaves were positively correlated to each other (r = 0.41, n = 150, P < 0.0001), whereas only a weak correlation was found between values of FA based on high accuracy measurements of the area and width of the same leaves (r = 0.17, n = 150, P = 0.03).

2.4. Data analysis We conducted a mixed-model ANOVA for each of three data sets (low-accuracy width measurements of 1331 leaves; high-accuracy width measurements of 150 leaves; high-accuracy area measurements of 150 leaves) to test for presence of directional asymmetry and FA relative to measurement error (as described by Palmer and Strobeck, 2003; Merilä and Björklund, 1995). In this analysis, the leaf side was considered a fixed factor and the individual leaf a random factor (procedure MIXED; SAS Institute, 2009). We evaluated reproducibility of measurements by calculating the index ME5 = [MSi−MSm]/[MSi+ (n−1) × MSm],where MSi and MSm are the interaction and error MS from a sides × individuals ANOVA (Palmer and Strobeck, 2003). This index expresses FA variation as a proportion of the total variation between leaf sides, which includes variation due to both FA and measurement error. Second, we calculated ME1 = |M1 − M2|/N, i.e. the average difference between two subsequent measurements of one side of a leaf, M1 and M2 (Palmer and Strobeck, 2003). This index may be compared directly to an absolute difference in values of the measured character from the left and right leaf halves. The FA values were calculated as follows: FA = 2 × abs(L−R)/(L +R), where L and R are the respective values of the measured character from the left and right leaf halves. This index was widely applied in earlier studies of plant leaf FA (e.g. Kozlov et al., 1996; Kryazheva et al., 1996; Wilsey et al., 1998; Ivanov et al., 2015), and its use is justified by significant positive correlation between the absolute difference in the measured character between the left and right leaf halves and leaf size, whereas size-corrected FA values were independent of leaf size (data not shown). These values were square-root transformed to meet a normality assumption, averaged between two subsequent measurements and between two leaves collected from the same seedling, and analysed (separately for each data set) with a linear mixed model (procedure GLIMMIX). We considered treatment (control, heavy metals or drought), pollution at the site of seed origin (high or low), and their interaction as fixed effects, whereas site of seed origin, mother tree (nested within a site) and growth chamber were treated as random intercept effects. We facilitated accurate F tests of the fixed effects by adjusting the standard errors and denominator degrees of freedom by the latest version of the method described by Kenward and Roger (2009). The power of the analyses was calculated using http:// powerandsamplesize.com/Calculators/Compare-2-Means/2-SampleEquality for small, moderate and large increases in leaf FA (15, 30 and 60% of the average value in the control group, respectively). We also calculated the sample sizes necessary to achieve the recommended 80% power of the analyses (Cohen, 1988; Jennions and Møller, 2003).

3.2. Treatment effects on FA None of the three data sets showed significant increases in FA in response to either heavy metal or drought stress (Fig. 1a, c, e; Table 2). This result was independent of the pollution load at the site of seed origin and was consistent for FA values calculated from either the low or high accuracy measurements of leaf width, or from measurements of the width and area of leaf halves (Table 2). None of the random factors explained the variation in FA (data not shown). 3.3. Power of the analyses The analysis based on low accuracy width measurements of 1331 leaves from 665 seedlings, representing the progenies of 79 mother trees, had sufficient power (85%) to detect minor (15%) differences in FA between the treatments. Analyses of the smaller data sets (150 leaves measured with high accuracy, representing the progenies of 41 mother trees) had moderate power to detect 30% differences, but very high power to detect 60% differences in FA between the treatments (Table 3). Under the observed variability in FA, samples from 70 to 400 285

Ecological Indicators 84 (2018) 283–289

V. Zverev et al.

Fig. 1. Effects of experimental treatments on leaf FA (back-transformed values; estimated marginal means + SE) in mountain birch Betula pubescens subsp. czerepanovii. (a), (c), (e) Effects of treatment (C, control; D, drought; HM, heavy metals) averaged across all seedlings. (b), (d), (f) Effects of heavy metals on seedlings from clean (CLN) and polluted (POL) sites. FA values were calculated from (a), (b) low accuracy measurements of the width of leaf halves; (c), (d) high accuracy measurements of the width of leaf halves; (e), (f) high accuracy measurements of the area of leaf halves. Sample sizes (numbers of seedlings) are shown within the bars. All pairwise differences between means are not significant (P > 0.05); for the results of statistical analyses consult Table 2.

Table 2 Effects of treatment (TRT; control, heavy metals or drought) and pollution at the site of seed origin (PSO; low or high) on FA of mountain birch leaves calculated from different data (mixed model ANOVA, tests of fixed effects; the standard errors and denominator degrees of freedom adjusted by the latest version of the method described by Kenward and Roger, 2009). Source of variation

TRT PSO TRT × PSO

Width, low accuracy data

Width, high accuracy data

Area, high accuracy data

F

df

P

F

df

P

F

df

P

1.35 0.76 1.60

2,585 1,13 2,585

0.26 0.40 0.20

1.11 0.33 1.79

2,134 1,68 2,117

0.33 0.57 0.17

1.50 1.45 0.73

2,124 1,17 2,89

0.23 0.25 0.49

286

Ecological Indicators 84 (2018) 283–289

V. Zverev et al.

Table 3 Power of the analyses and minimum sample sizes required to achieve a power of 0.80. Variable

Power of the analysis (for α = 0.05)a

Sample size required to achieve a power of 0.80

a b

Measured character and accuracy of measurements

Width, low accuracy Width, high accuracy Area, high accuracy Width, low accuracy Width, high accuracy Area, high accuracy

Baseline for calculationsb

Relative difference between means (%)

Mean

S.D.

15

30

60

0.067 0.070 0.065 0.067 0.070 0.065

0.021 0.050 0.049 0.021 0.050 0.049

0.85 0.15 0.15 70 392 377

0.99 0.48 0.46 29 89 95

0.99 0.97 0.95 8 23 25

Calculated for actual sample sizes. Calculated for variation among progenies of individual mother trees under control conditions.

4.2. Power of the analyses

genetically different plant individuals per treatment would be needed to detect minor (15%) differences in FA between treatments or to accept the null hypothesis on the absence of these differences with high statistical confidence (Table 3).

The question of how much confidence to place in a negative conclusion is answered by statistical power analysis. Power (1–β), which is the probability of rejecting a false null hypothesis, increases with increases in the sample size, the probability of type I error (α) and the difference between the values of the response variable expected under alternative and null hypothesis, and it decreases with an increase in the variance of the study population (Cohen, 1988; Jennions and Møller, 2003). Thus, the estimation of the statistical power of the analysis requires that we specify the size of the effect we would like to detect. Earlier studies (Hodar, 2002; Llorens et al., 2002; Kozlov and Niemelä, 2003; Fair and Breshears, 2005; Souza et al., 2004) reported variable effects of drought on plant FA, from a 29% decrease to a 50% increase relative to controls; the average magnitude of the effect calculated from these studies was positive, but relatively minor (15%). At the same time, heavy metal stress was reported to result in substantial increases in plant FA, from 8% to 414% (Kozlov et al., 1996; Mal et al., 2002; Chudzinska et al., 2014; Kolbas et al., 2014; Ivanov et al., 2015; Smith, 2016), giving on average a two-fold increase relative to the controls. Thus, our analyses based on low accuracy measurements had sufficient power to detect the average effects of abiotic stressors on leaf FA. The general recommendation is that the experimental design should assure a statistical power of 0.80 (Cohen, 1988) or even 0.95 (Peterman, 1990). Under these recommendations, the probabilities of making a type II error, i.e. of accepting a false null hypothesis (β), are 0.20 and 0.05, respectively. However, the majority of published ecological studies have much lower power (Jennions and Møller, 2003), and this situation shows only rather slow improvement, if any at all (Smith et al., 2011). When compared with the studies reviewed by Jennions and Møller (2003), our analyses have a better-than-average statistical power. We therefore conclude that we may have overlooked the effects of heavy metals and/or drought on leaf FA in mountain birch in our experiment only if these effects were rather small, i.e. less than 15% of the FA value observed in control plants. However, patterns showed by the different data sets (Fig. 1) slightly differ from each other, thus leaving the question whether this discrepancy could have emerged due to methodological reasons.

4. Discussion 4.1. Abiotic stress and leaf FA in mountain birch Grime (1979) defined stress as the external constraints limiting dry matter production by plants. Our experimental plants demonstrated a substantial (23%) decrease in height in response to drought stress and significant decreases in leaf length (measured from the same leaves as FA) in response to both drought and application of heavy metals (13 and 6%, respectively: Eränen et al., 2009). Furthermore, chlorophyll fluorescence (measured in the same leaves) indicated adverse effects of these treatments on photosynthesis (Eränen et al., 2009). Thus, both treatments were clearly stressful for mountain birch seedlings in general and for leaves collected for measurements of FA in particular. Nevertheless, we still did not detect any treatment effect on FA. Seedlings from industrially polluted sites showed weaker changes in performance in response to heavy metal treatment than did seedlings from unpolluted sites, thereby demonstrating a heavy metal tolerance (Eränen et al., 2009). At the same time, these tolerant seedlings appeared more sensitive to drought than were the seedlings from unpolluted sites (Eränen et al., 2009). Despite this variation, the seedlings from the polluted and unpolluted sites did not differ in their responses of leaf FA to heavy metal and drought stress. Thus, neither of our predictions concerning the effects of abiotic stress on leaf FA in mountain birch were met. Our findings add to the current literature, which indicates that the FA of individuals experiencing unfavourable conditions does not always exceed the FA of individuals from benign environments. For example, Graham et al. (2010) listed 17 studies in which the expected increase in FA was not observed in response to the impacts of different environmental stressors, even those that clearly suppressed growth and increased the mortality of the study organisms. Only a half of the 34 papers on plant FA published in English from 2010 to 2014 and listed in the ISI Web of Science reported complete support for the predictions regarding the relationships between FA and different environmental factors or characteristics of the study plants. Three more papers found partial support for the tested hypotheses, and the remaining 14 papers reported only ‘negative’ results (Kozlov, 2017). Thus, at the current level of knowledge, an absence of any effects of drought and heavy metal stress on the FA of plant leaves is not particularly astonishing. However, the question remains whether we can accept the null hypothesis on the absence of these effects with a sufficient level of statistical confidence.

4.3. Research methodology We previously identified the low accuracy of FA measurements as the key reason for the observed low reproducibility of the results (Kozlov, 2015), and we suggested that a partial explanation for the discrepancies among the outcomes of FA-related studies might be the use of inconsistent and unstandardised methodology (Kozlov et al., 2017). However, in the current study, we found a weak but significant positive correlation between the values of FA obtained using the outdated methodology (that yielded low accuracy data) and using modern methodology (that yielded high accuracy data). Thus, the absence of 287

Ecological Indicators 84 (2018) 283–289

V. Zverev et al.

object question the use of FA in environmental ecology and management. Finally, the measurements of FA are far more laborious when compared to measurements of other traits used to reveal effects of environmental stressors on plants. First, leaves (or other plant parts) selected for FA measurements should be collected, labelled in a way that assures blinded measurements, scanned and preserved. Second, a minimum of four measurements (i.e. of the left and right sides of an object, conducted twice, preferably by different observers) should be conducted to obtain a single value for the FA, whereas all other morphological characters (such as leaf length, area or weight; shoot length or weight; plant height; and many others) are routinely measured only once. Third, measurements of FA often require additional operations, e.g. identification of the midpoint between the leaf tip and leaf base for measurements of leaf width, or separation of a leaf image into right and leaf halves for measurements of leaf area; all these operations should also be conducted twice to control for measurement error. Keeping in mind the large sample sizes required to achieve sufficient statistical power of the analyses aimed at detection of minor effects, and considering the inconsistent responses of FA to environmental stressors, this amount of work is not rewarded by the value of the information obtained for applied ecological research.

any effects of drought and heavy metals on leaf FA in our experiment, as we found long ago but did not publish due to scepticism of the reviewers, was not caused by the low accuracy of the measurements. Interestingly, the standard deviation is smaller for low accuracy than for high accuracy measurements (Table 3), so that the power to detect differences in FA between treatment and control samples is higher for low accuracy measurements. Thus, low accuracy measurements, which are less labor-consuming than high-accuracy measurements, may appear better suited for monitoring studies. At the time of sampling, our seedlings generally had 10−30 leaves. We planned to use these seedlings for further experiments, so we limited our collection to only two leaves per plant. This naturally gives a low accuracy estimate of plant-specific FA; but several earlier studies that measured FA in one or two leaves per plant (Llorens et al., 2002; Handy et al., 2004; Souza et al., 2004) still reported significant effects. Furthermore, the same set of leaves yielded significant effects of both treatment and seedling origin on leaf length and chlorophyll fluorescence (Eränen et al., 2009), giving no reason to assume that measurements of a larger number of leaves per plant would change our conclusions regarding treatment effect on leaf FA. We have recently demonstrated that studies of FA are prone to confirmation bias: the results obtained from the same set of leaf images differed significantly between the observers who were told that the samples originated from either ‘stressful’ or ‘benign’ environments (Kozlov and Zvereva, 2015). A recent survey of ecological, evolutionary and behavioural publications found that only 13% of the studies that could have been influenced by observer bias explicitly indicated that experiments were blinded (Kardish et al., 2017). To the best of our knowledge, earlier methodological papers, with rare exceptions (Kozlov et al., 2009), did not require that measurements of FA be conducted blindly. Intriguingly, a publication that reported a significant increase in FA of mountain birch near several polluters (Kozlov et al., 1996) was based on non-blinded measurements, whereas two later studies based on blinded measurements (Valkama and Kozlov, 2001; Kozlov et al., 2009) did not confirm this pattern. We suggest that some of the published data on increases in FA in pessimal conditions may have their origin in confirmation bias rather than in any actual effect of stress on the FA of study organisms. We therefore recommend that future studies of FA always be based on blinded measurements. One of the first studies of plant FA (Kryazheva et al., 1996) explored the effects of pollution on the FA of five characters of birch leaves and the authors concluded that FA values calculated individually from these five characters were only weakly correlated with each other. Subsequently, the FA responses to stressors have only rarely been compared between different characters of the same plants, and 71% of the papers published in English from 2010 to 2014 were based on measurements of a single plant trait (Kozlov, 2017), mostly leaf width or leaf area. This gives special importance to our finding that FA values calculated from width and area measurements of the same leaves were only weakly correlated and therefore showed different (albeit not statistically different) responses to our experimental treatments (Fig. 1c–f). Similarly, Llorens et al. (2002) found that only 25% of the variation in FA calculated from measurements of leaf area was explained by a variation in FA calculated from measurements of leaf width. Thus, the conclusions of the study could vary depending on the choice of the measured character (Llorens et al., 2002; Ivanov et al., 2002), and the researchers could arbitrarily select between these results. For example, Llorens et al. (2002) concluded that leaf FA is a more sensitive indicator of physiological stress than are leaf size or gas exchange measurements, although only FA calculated from leaf area measurements showed an expected pattern and only those data were illustrated in the publication. By contrast, the FA calculated from leaf width measurements did not change with warming and drought, and those data were not reported by Llorens et al. (2002). As long as no theoretical basis exists for selecting the character for measurements of FA, this inconsistency in responses of the FA calculated from different characters of the same

5. Conclusions Several years ago, Bjorksten et al. (2000, p. 165) suggested that researchers ‘abandon the search for a general link between FA and environmental stress’. Combined with the conclusions by the same authors on the ‘poor use of research money’ for ‘measuring different stresses and traits in different organisms’ (op. cit., p. 165), this opinion unequivocally suggests a blacklisting of FA studies. We do agree with this conclusion with respect to applied research, and we strongly recommend that the practical use of FA as an indicator of environmental stress be limited to study systems for which the existence of cause-andeffect relationship between the stressing impact and the changes in FA is confirmed by controlled, blinded experiments. At the same time, we appreciate that FA remains an attractive target for basic ecological and evolutionary research, because the detected inconsistency in outcomes of individual studies suggests that regularities in FA responses to environmental variation remain to be discovered. Acknowledgments We thank J. Nilsen and L. Lund for arranging the experiment, T. Klemola for statistical advices, and two anonymous reviewers for inspiring comments to an earlier draft of the manuscript. Funding was provided by the Academy of Finland (project numbers 201991, 211734 and 276671). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ecolind.2017.08.058. References Allenbach, D.M., 2011. Fluctuating asymmetry and exogenous stress in fishes: a review. Rev. Fish Biol. Fish. 21, 355–376. Beasley, D.E., Bonisoli-Alquati, A., Mousseau, T.A., 2013. The use of fluctuating asymmetry as a measure of environmentally induced developmental instability: a metaanalysis. Ecol. Indic. 30, 218–226. Berteaux, D., Diner, B., Dreyfus, C., Eble, M., Lessard, I., Klvana, I., 2007. Heavy browsing by a mammalian herbivore does not affect fluctuating asymmetry of its food plants. Ecoscience 14, 188–194. Bjorksten, T.A., Fowler, K., Pomiankowski, A., 2000. What does sexual trait FA tell us about stress? Trends Ecol. Evol. 15, 163–166. Black-Samuelsson, S., Andersson, S., 2003. The effect of nutrient stress on developmental instability in leaves of Acer platanoides (Aceraceae) and Betula pendula (Betulaceae).

288

Ecological Indicators 84 (2018) 283–289

V. Zverev et al.

czerepanovii, near the copper-nickel smelter: natural selection or phenotypic acclimation? Chemosphere 59, 189–197. Kozlov, M.V., 2015. How reproducible are the measurements of leaf fluctuating asymmetry? PeerJ 3, e1027. Kozlov, M.V., 2017. Plant studies on fluctuating asymmetry in Russia: mythology and methodology. Russ. J. Ecol. 48, 3–10. Kryazheva, N.G., Chistyakova, E.K., Zakharov, V.M., 1996. Analysis of development stability of Betula pendula under conditions of chemical pollution. Russ. J. Ecol. 27, 422–424. Lajus, D.L., Graham, J.H., Katolikova, M.V., Yurtseva, A.O., 2009. Fluctuating asymmetry and random phenotypic variation: history, achievements, problems, perspectives. Vestnik of Saint-Petersburg University, Ser. 3 Biology 0 (3), 98–110. Llorens, L., Penuelas, J., Emmett, B., 2002. Developmental instability and gas exchange responses of a heathland shrub to experimental drought and warming. Int. J. Plant Sci. 163, 959–967. Møller, A.P., Swaddle, J.P., 1997. Asymmetry, Developmental Stability, and Evolution. Oxford University Press, Oxford. Mal, T.K., Uveges, J.L., Turk, K.W., 2002. Fluctuating asymmetry as an ecological indicator of heavy metal stress in Lythrum salicaria. Ecol. Indic. 1, 189–195. Merilä, J., Björklund, M., 1995. Fluctuating asymmetry and measurement error. Syst. Biol. 44, 97–101. Nishizaki, M.T., Barron, S., Carew, E., 2015. Thermal stress increases fluctuating asymmetry in marine mussels: environmental variation and developmental instability. Ecosphere 6 (art. 85). Palmer, A.R., Strobeck, C., 2003. Fluctuating asymmetry analysis revisited. In: Polak, M. (Ed.), Developmental Instability: Causes and Consequences. Oxford University Press, Oxford, pp. 279–319. Palmer, A.R., 1996. Waltzing with asymmetry. Bioscience 46, 518–532. Parsons, P.A., 1992. Fluctuating asymmetry—a biological monitor of environmental and genomic stress. Heredity 68, 361–364. Peterman, R.M., 1990. Statistical power analysis can improve fisheries research and management. Can. J. Fish. Aquat.Sci. 47, 2–15. Rasmuson, M., 2002. Fluctuating asymmetry—indicator of what? Hereditas 136, 177–183. Ruotsalainen, A.L., Markkola, A.M., Kozlov, M.V., 2009. Mycorrhizal colonisation of mountain birch (Betula pubescens ssp. czerepanovii) along three environmental gradients: does life in harsh environments alter plant-fungal relationships? Environ. Monit. Assess. 148, 215–232. SAS Institute, 2009. SAS/Stat. User’s Guide, Version 9.2. SAS Institute, Cary, N.C. Sandner, T.M., Matthies, D., 2017. Fluctuating asymmetry of leaves is a poor indicator of environmental stress and genetic stress by inbreeding in Silene vulgaris. Ecol. Indic. 79, 247–253. Savelieva, N.A., Belova, E.E., Pervova, N.E., Kolontsov, A.A., 2017. Assessment of the effect of lead and cadmium ions on fluctuating leaf asymmetry in common bean (Phaseolus vulgaris L.). Vestnik of Moscow State University, Ser. 17. Soil Sci. 0 (4), 50–52. Smith, D.R., Hardy, I.C.W., Gammell, M.P., 2011. Power rangers: no improvement in the statistical power of analyses published in animal behaviour. Anim. Behav. 81, 347–352. Smith, M.C., 2016. Heavy metal contamination increases fluctuating asymmetry in Rhus glabra L. (Anacardiaceae). Southwest. Nat. 61, 156–159. Souza, G.M., Aidar, S.T., Oliveira, R.F., 2004. Developmental stability and network connectance in Phaseolus vulgaris L. genotypes under water deficit. Israel J. Plant Sci. 52, 205–212. Valkama, J., Kozlov, M.V., 2001. Impact of climatic factors on the developmental stability of mountain birch growing in a contaminated area. J. Appl. Ecol. 38, 665–673. Wiegolaski, F.E., 2005. Plant Ecology, Herbivory, and Human Impact in Nordic Mountain Birch Forests. Springer, Berlin. Wilsey, B.J., Haukioja, E., Koricheva, J., Sulkinoja, M., 1998. Leaf fluctuating asymmetry increases with hybridization and elevation in tree-line birches. Ecology 79, 2092–2099. Zakharov, V.M., 1990. Analysis of fluctuating asymmetry as a method of biomonitoring at the population level. In: Krivolutsky, D. (Ed.), Bioindications of Chemical and Radioactive Pollution. Mir Publishers, Moscow, pp. 187–198. Zverev, V.E., 2009. Mortality and recruitment of mountain birch (Betula pubescens ssp. czerepanovii) in the impact zone of a copper-nickel smelter in the period of significant reduction of emissions: the results of 15-year monitoring. Russ. J. Ecol. 40, 254–260.

Am. J. Bot. 90, 1107–1112. Chudzinska, E., Pawlaczyk, E.M., Celinski, K., Diatta, J., 2014. Response of Scots pine (Pinus sylvestris L.) to stress induced by different types of pollutants—testing the fluctuating asymmetry. Water Environ. J. 28, 533–539. Clarke, G.M., 1992. Fluctuating asymmetry: a technique for measuring developmental stress of genetic and environmental origin. Acta Zool. Fenn. 191, 31–35. Clarke, G.M., 1998. Developmental stability and fitness: the evidence is not quite so clear. Am. Nat. 152, 762–766. Cohen, J., 1988. Statistical Power Analysis for the Behavioral Sciences, 2nd ed. Erlbaum Associates, Hillsdale. Dale, V.H., Beyeler, S.C., 2001. Challenges in the development and use of ecological indicators. Ecol. Indic. 1, 3–10. Diaz-Gil, C., Palmer, M., Catalan, I., Alos, J., Fuiman, L.A., Garcia, E., Gil, M.D., Grau, A., Kang, A., Maneja, R.H., Mohan, J.A., Morro, B., Schaffer, J.J., Buttay, L., Riera-Batle, I., Tolosca, B., Morales-Nin, B., 2015. Otolith fluctuating asymmetry: a misconception of its biological relevance? ICES J. Mar. Sci. 72, 2079–2089. Eränen, J.K., Nilsen, J., Zverev, V.E., Kozlov, M.V., 2009. Mountain birch under multiple stressors—heavy metal-resistant populations co-resistant to biotic stress but maladapted to abiotic stress. J. Evol. Biol. 22, 840–851. Fair, J.M., Breshears, D.D., 2005. Drought stress and fluctuating asymmetry in Quercus undulata leaves: confounding effects of absolute and relative amounts of stress? J. Arid Environ. 62, 235–249. Freeman, D.C., Graham, J.H., Emlen, J.M., 1993. Developmental stability in plants—symmetries, stress and epigenesis. Genetica 89, 97–119. Graham, J.H., Freeman, D.C., Emlen, J.M., 1993. Antisymmetry, directional asymmetry, and dynamic morphogenesis. Genetica 89, 121–137. Graham, J.H., Raz, S., Hel-Or, H., Nevo, E., 2010. Fluctuating asymmetry: methods, theory, and applications. Symmetry 2, 466–540. Grime, J.P., 1979. Plant Strategies and Vegetation Processes. Wiley, New York. Handy, S.M., McBreen, K., Cruzan, M.B., 2004. Patterns of fitness and fluctuating asymmetry across a broad hybrid zone. Int. J. Plant Sci. 165, 973–981. Hodar, J.A., 2002. Leaf fluctuating asymmetry of Holm oak in response to drought under contrasting climatic conditions. J. Arid Environ. 52, 233–243. Ivanov, V.P., Ivanov, Y.V., Marchenko, S.I., Kuznetsov, V.V., 2015. Application of fluctuating asymmetry indexes of silver birch leaves for diagnostics of plant communities under technogenic pollution. Russ. J. Plant Physiol. 62, 340–348. Jennions, M.D., Møller, A.P., 2003. A survey of the statistical power of research in behavioral ecology and animal behavior. Behav. Ecol. 14, 438–445. Kardish, M.R., Mueller, U.G., Amador-Vargas, S., Dietrich, E.I., Ma, R., Barrett, B., Fang, C.-C., 2017. Blind trust in unblinded observation in ecology, evolution, and behavior. Front. Ecol. Evol. 3 (art. 51). Kenward, M.G., Roger, J.H., 2009. An improved approximation to the precision of fixed effects from restricted maximum likelihood. Comp. Stat. Data Anal. 53, 2583–2595. Knierim, U., Van Dongen, S., Forkman, B., Tuyttens, F.A.M., Spinka, M., Campo, J.L., Weissengruber, G.E., 2007. Fluctuating asymmetry as an animal welfare indicator — a review of methodology and validity. Physiol. Behav. 92, 398–421. Kolbas, A., Marchand, L., Herzig, R., Nehnevajova, E., Mench, M., 2014. Phenotypic seedling responses of a metal-tolerant mutant line of sunflower growing on a Cucontaminated soil series: potential uses for biomonitoring of Cu exposure and phytoremediation. Plant Soil 376, 377–397. Kozlov, M.V., Niemelä, P., 2003. Drought is more stressful for northern populations of Scots pine than low summer temperatures. Silva Fenn. 37, 175–180. Kozlov, M.V., Zvereva, E.L., 2015. Confirmation bias in studies of fluctuating asymmetry. Ecol. Indic. 57, 293–297. Kozlov, M.V., Wilsey, B.J., Koricheva, J., Haukioja, E., 1996. Fluctuating asymmetry of birch leaves increases under pollution impact. J. Appl. Ecol. 33, 1489–1495. Kozlov, M.V., Zvereva, E.L., Zverev, V., 2009. Impacts of Point Polluters on Terrestrial Biota: Comparative Analysis of 18 Contaminated Areas. Springer, Dordrecht. Kozlov, M.V., Cornelissen, T., Gavrikov, D.E., Kunavin, M.A., Lama, A.D., Milligan, J.R., Zverev, V., Zvereva, E.L., 2017. Reproducibility of fluctuating asymmetry measurements in plants: sources of variation and implications for study design. Ecol. Indic. 73, 733–740. Kozlov, M.V., 2003. Are fast growing birch leaves more asymmetrical? Oikos 101, 654–658. Kozlov, M.V., 2004. Leaf fall in white birch (Betula pubescens) is independent of leaf asymmetry. Can. J. Bot. 82, 910–913. Kozlov, M.V., 2005. Pollution resistance of mountain birch, Betula pubescens subsp.

289