Biologia 69/5: 691695, 2014Section ZoologyDOI: 10.2478/s11756-014-0356-x
Validation of scale-reading estimates of age and growthin a brown trout Salmo trutta population
Libor Zvorka1,2, Ondej Slavk3 & Pavel Hork3
1Water Research Institute Tom Garrigue Masaryk, Department of applied ecology, Podbabska 30, CZ-16000 Prague 6,Czech Republic; e-mail: libor email@example.comCharles University in Prague, Faculty of Science, Institute for Environmental Studies, Albertov 6, CZ-12843 Prague, CzechRepublic3Department of Zoology and Fisheries, Faculty of Agrobiology, Food and Natural Resources, Czech University of LifeSciences Prague, CZ-16521 Praha 6, Czech Republic
Abstract: In this study, we validated a scale-reading method estimating age and growth in brown trout Salmo trutta in wild,landlocked, stream-dwelling populations from mountain headwaters in the Elbe catchment area of the Czech Republic. Thevalues estimated from scale reading were compared with measured values, collected using a mark-recapture program overeight consecutive years. The age-corrected absolute percentage error was 10.71%, primarily because the ages of the oldestindividuals according to scale reading were underestimated, and the ages of juvenile individuals were slightly overestimated.The back-calculated length was slightly underestimated (the mean error was 4.60 mm), but it was not signicantly dierentfrom the real measured length. This study suggests that in cold mountain headwaters, scale reading is a suciently accuratemethod for age and growth estimation in juvenile brown trout; however, the results for adult individuals must be takenwith caution.
Key words: ageing; growth back-calculation; mark-recapture; sh
Fishery scientists and managers measure individualgrowth and age structure to describe sh population dy-namics and evaluate sh stock management (Leonardos2001; Yule et al. 2008). The age and growth estimatesfrom calcied bony structures are commonly used formany ectothermic species, such as sh (Das 1994), am-phibians (Liao 2011) or reptiles (Castanet 1994). Scales,otoliths, spines, vertebrae, n rays and skull bones aretypically used for age and growth estimates in sh (Helf-man et al. 2009).
Estimating age and back-calculating growththrough scale reading is based on annuli counts alongthe anterior-posterior axis of a scale (Lee 1920; Pear-son 1928) and it is the oldest, but still commonly usedmethod (Alvord 1954; Maceina et al. 2007). The pri-mary advantage of scale reading is that sampling is non-lethal for sh and has no inuence on viability, whereasremoving bony structures such as otoliths requires sac-ricing the sh (Hubert et al. 1987). Therefore, scalereading is especially useful if the target population isendangered or protected because it facilitates harmlessevaluation of many individuals in a population. How-ever, lower precision in age and growth estimates fromscales compared with other calcied structures has beenreported, for example in van der Meulen et al. (2013).
Scales grow by accretion at their margins and formannuli during a regular period of annual slow growth(Helfman et al. 2009), which is determined by low watertemperatures during winter seasons in temperate Eu-ropean climates (Elliott 1989). Variability in individualgrowth across seasons is caused by many endogenousand exogenous stress factors, such as disease, injury,food and nutrient unavailability, maturation, reproduc-tive behaviour and temperature (DeVries & Frie 1996;Helfman et al. 2009; Zhi-Hua et al. 2010). Such growthvariability can be associated with irregularity in annuliformation on scales, which can lead to errors in age andgrowth estimates (Beamish & McFarlane 1983). Ageand growth estimates in sh populations based on scalereading are likely less precise for populations with longand indistinct growing seasons (Hoxmeier et al. 2001).On the other hand, slow growth may produce crowdedannuli, which decrease the precision of scale reading,and as a result, diminish the validity of this information(Power 1978). Thus, knowledge of scale reading preci-sion is necessary to validate such information (Beamish& McFarlane 1983; Hork et al. 2010; Lopez Cazorla &Sidorkewicj 2011).
The aim of this study was to validate age andgrowth estimates from scale reading in a landlockedpopulation of brown trout Salmo trutta L., 1758, whichis a widespread and ecologically important species of
c2014 Institute of Zoology, Slovak Academy of Sciences
692 L. Zvorka et al.
European salmonids (Kottelat 1997). The estimatesfrom scale reading were compared with results obtainedfrom measuring values through mark-recapture obser-vations in mountain headwaters in Czech Republic.
Material and methods
In total, 6,240 of brown trout individuals were caught inthe headwaters of the Otava River (491 N, 1329 E) andthe Tepl Vltava River (4858 N, 1339 E) in the Elbecatchment area, the Czech Republic. Both sampled streamshad similar hydrological conditions and species structure.For a detailed description of the study site and sh pop-ulation, see Slavk et al. (2012) and Zvorka et al. (2013).The sh were sampled twice a year (May and October) fromautumn 2005 to autumn 2012 and were caught with a back-pack electro-shing device (EFKO, Germany). Every spec-imen was anesthetized (2-phenoxiethanol), measured (stan-dard length to the nearest mm), weighed (to the nearest g),and individually tagged using Visible Implant Alphanumerictags (Northwest Marine Technology, Shaw Island, Washing-ton) or passive integrated transponder (PIT) tags (Trovan,UK). From 1,754 randomly selected individuals, 510 scaleswere sampled. The scales were removed from the left bodyside, one row above the lateral line in the intersection ofdorsal and ventral n (DeVries & Frie 1996).
The scales were placed between two microscope slides,magnied (67.5 times), and photographed with a cameraconnected to a stereoscopic microscope (Arsenal, www.arsenal.cz, Czech Republic). The scales from 7% of thesh sampled were discarded because they were distorted orscarred. The scale readings were conducted without priorknowledge of the sh size. The age was determined usinga standard notation following Pearson (1928), which in-volved counting winter annuli. The scale size and annualincrements were measured along the anterior-posterior linefrom the scale centre to its margin (Martinson et al. 2000)using picture analysis software (UTHSCSA ImageTool3.0., http://compdent.uthscsa.edu/dig/itdesc.html, USA).Growth was back-calculated using the Fraser-Lee equation(Lee 1920), which is dened as follows:
Lt =StSc(Lc c) + c (1)
where Lt is the length at age t, Lc is the total length, St isthe radius of a scale annulus at age t, Sc is the scale radius,and c is the empirical constant. The value of constant cwas 10, estimated based on the brown trout larval lengthin mm (Baru & Oliva 1995; Ojanguren & Brana 2003). Inthe case of this study, the calculation of the constant c asthe intercept from the length-scale relationship regressionwas inappropriate, as it yielded unrealistic values for theback-calculated length.
The scale reading validation was based on 53 recap-tured individuals, for whom both initial and recapture scalesamples were available. The recaptured individuals were ex-posed in the stream for seven months in average (rangingfrom 4 to 25 months). Seventeen individuals were exposedto the winter conditions, while 36 individuals were exposedin stream from spring to autumn between the capture. Theindividuals age at recapture was compared with the stan-dard age, which was calculated as the initial capture ageplus the time between captures. We used the earlier esti-mate as a reference for the standard age, because it isgenerally accepted that age estimates in younger sh are
more precise (DeVries & Frie 1996). Two coecients wereused to express the error in aging. The standard percentageerror, which is the ratio between the number of errors in ag-ing and the sample size, and the absolute percentage error(PE) relative to standard age (equation 2) were used. Theabsolute percentage error measures the error size in relationto the sh age (Riart et al. 2006), as follows:
PE = 1001N
where N is the sample size, Xi is the standard age of theith sh and Yi is the estimated age of the ith sh. Theprecision of the initial capture age estimates was evaluatedby two independent skilled readers. The error of estimationwas calculated using the following coecient of variation byChang (1982):
CV = 1001N
(Xij Xj)2R 1
where N is the sample size, R is the number of times eachsh age was determined, Xij is the ith age determination forthe jth sh, and Xj is the average age calculated for the jth
sh.The back-calculated length accuracy was assessed us-
ing a Virtual Mark procedure, in which the structure radiusat the initial capture was modelled on the recaptured sample(Zymonas & McMahon 2009) as follows:
SVM = AnR(
]SVM + c (4b)
where SVM is the scale radius at the virtual mark, SCI is thescale radius at the initial capture, SCR is the scale radiusat recapture, AnR is the nth annulus radius at recapture,AnI is the nth annulus radius at the initial capture, LVM isthe back-calculated sh length at the virtual mark, LR isthe sh length at recapture, and c is the empirical constant.The accuracy of the back-calculation was validated only inindividuals with correctly estimated age in both samples(initial capture and recapture). The dierences between theback-calculated and measured lengths were evaluated usinga paired t-test.
The estimated age of captured sh ranged from 1+ to8+. The age of the recaptured sh was estimated withan overall 20.80% standard error and a 10.71% abso-lute age-corrected percentage error (n = 53). The lowerstandard error and the lower absolute percentage er-ror in younger age groups were found (Table 1). Thescale reading accuracy was aected neither by winterconditions, nor the exposition time (Table 2). The vari-ation coecient of the age estimates in two independentreaders was 7.46 (n = 53). The accordance of reader A
Scale-reading estimates of age and growth in brown trout 693
Table 1. Error rate (%) in age estimation according to the time of exposition; 95% condence interval is given for standard errorpercentage only.
ErrorNumber of winters Number(range of exposition of sh Number Overestimated Underestimated Standard error Condence Age correctedtime in days) percentage interval (95%) error percentage
Zero (124161) 36 8 4 4 22.2% 8.036.5 10.7%One (211365) 14 1 0 1 7.1% 0.022.6 2.4%More (571720) 3 2 0 2 66.7% 0.0100.0 25.8%
Total 53 11 4 7 20.8% 9.532.0 10.7%
Table 2. Error rate (%) in age estimation according to age classes; 95% condence interval is given for standard error percentage only.
ErrorAge Number Standard error Condence Age corrected
of sh Number Overestimated Underestimated percentage interval (95%) error percentage
1+ 12 1 1 0 8.3% 0.026.7 8.3%2+ 30 5 5 0 16.7% 2.530.8 10%3+ 7 2 1 1 28.6% 16.673.7 9.5%4+ 2 1 0 1 50% 0.0100 12.5%5+ 1 1 0 1 100% 40%6+ 0 7+ 0 8+ 1 1 0 1 100% 37.5%
Total 53 11 7 4 20.8% 9.532.0 10.7%
Table 3. Age frequency table summarizing pairwise comparison of age estimates from two readers. Data indicate the number of shindividuals.
Reader BAge 1+ 2+ 3+ 4+ 5+ 6+ 7+ 8+ Total
1+ 10 2 122+ 2 21 7 303+ 2 4 1 7
Reader A 4+ 2 0 25+ 1 0 16+ 07+ 08+ 1 0 1
and reader B decreased with increasing estimated ageof sh, being highest in the age classes 1+ and 2+ (Ta-ble 3).
The size of the individuals tested ranged from73 mm to 276 mm at the initial capture, and 115 mm to290 mm at the recapture. The measured length did notdier from the back-calculated length for the VirtualMark (paired t-test: t = 0.91, df = 76, P = 0.37). Themean for the dierences was 4.60 mm (see Fig. 1).
Our results showed higher errors in age and growth es-timates compared to studies using otoliths (Hubert etal. 1987; Schill et al. 2010); the age and length wereslightly underestimated, which is consistent with resultsfrom earlier studies focused on scale-reading validation(Kruse et al. 1997; Riart et al. 2006; Heidarsson et
al. 2006; Zymonas & McMahon 2009). We found a ten-dency to overestimate the age of juvenile individuals;however, the results indicated generally bigger errors inadult individuals. This assumption was also supportedby the nding that the accordance of two independentreaders decreased with increasing age of individuals.We suggest that the energetic requirements for breed-ing together with the harsh environmental conditionsin mountain headwaters decrease the growth of ma-ture trouts (Bohlin et al. 2001) and may lead to annulicrowding. Consequently, the crowded annuli increasedthe risk of age underestimation in mature individuals(Alvord 1954; Power 1978). The time of exposition instream and number of winters between captures, whichcould increase the risk of erroneous scale reading, didnot aect the precision of scale reading.
To summarize our results, the age and growth es-timates based on scale reading were found to be a su-
694 L. Zvorka et al.
Fig. 1. Boxplot of error distribution in length back-calculation. Errors are expressed as a dierence between the real measured length andthe length modelled based on the Virtual Mark procedure (equations 4a and 4b). Outlying values are marked with circles accompaniedwith their values in mm.
ciently precise method for juvenile individuals until ma-turity, which included the sh in age classes 1+ and 2+for the observed streams (Baru & Oliva, 1995). How-ever, scale-reading does not provide reliable estimatesof age and growth for adult individuals in brown troutdwellings in cold mountain streams. These ndingsseem to be of importance for eld biologists and man-agers focused on populations of free living salmonids incold waters.
The authors thank the technical sta for their assistanceduring the experimental period. We would also like to thankPetra Hork for valuable assistance in scale reading andto Zuzana Harmkov and Alena Slavkov for the lan-guage correction. The study was nancially supported bythe Technology Agency of the Czech Republic within theTA01020875 project.
Alvord W. 1954. Validity of age determination from scales ofbrown trout, rainbow trout and brook trout. Trans. Am. Fish.Soc. 83: 91103.
Baru V. & Oliva O. 1995. Mihulovci a ryby (2). Fauna R a SR.1. Academia, Praha, 698 pp.
Beamish R.J. & McFarlane G.A. 1983. The forgotten requirementfor age validation in sheries biology. Trans. Am. Fish. Soc.112: 735743.
Bohlin T., Pettersson J. & Degerman E. 2001. Population den-sity of migratory and resident brown trout (Salmo trutta) inrelation to altitude: evidence for a migration cost. J. Anim.Ecol. 70: 112121. DOI: 10.1111/j.1365-2656.2001.00466.x
Castanet J. 1994. Age estimation and longevity in reptiles. Geron-tology 40: 174192.
Chang W.Y.B. 1982. A statistical method for evaluating the re-producibility of age...