Factors Influencing Precision of Age Estimation from Scales and Otoliths of Bluegills in Illinois Reservoirs

—We examined the effects of population-specific variation on age estimates from scales and otoliths of bluegills Lepomis macrochirus from Illinois reservoirs. We divided 12 reservoirs into four groups (north stunted, north quality, south stunted, and south quality) to examine the effects of latitude and fish size distribution on the precision of estimated ages. Ages of 40 bluegills from each reservoir (total N 5 480) were independently estimated by two readers. Otoliths provided more precise age estimates than scales. Population size structure and sex had no effect on precision of ages estimated from either structure; however, latitude, age, and maturity stage all affected precision of ages estimated from scales. Age also affected precision of estimates from otoliths. Percent agreement, coefficient of variation, and age bias plots all provided useful interpretations of the data. Our results demonstrate the importance of examining population-specific sources of variation with multiple statistical methods when comparing ages of fish from different populations. The assessment of ages within and among populations is a critical component of many fisheries investigations. Data on fish ages are necessary to examine growth, mortality, and reproductive characteristics (e.g., age of maturation) of a population. Although fish ages can be estimated through statistical analyses (e.g., length histograms, age– length keys), direct estimation from bony structures such as scales and otoliths is currently the most widely used method (DeVries and Frie 1996). Because biologists must often depend on multiple readers and multiple structures to evaluate fish ages in long-term data sets, the precision of age estimation has received much attention (e.g., Beamish and McFarlane 1987; Chilton and Stocker 1987; Hammers and Miranda 1991; Hoenig et al. 1995). Previous studies have primarily addressed precision among multiple readers or multiple aging structures with fish collected from either single or a few populations (Boxrucker 1986; Sharp and Bernard 1988; Kruse et al. 1993; Robillard and Marsden 1996). Although these investigations have indicated which structures or methods are most precise for various fish species, they have not considered the effects of important external sources of variation associated with estimating ages from multiple populations. Latitude has been suggested to influence the precision of ages estimated from scales (Kruse et al. 1993); however, these * Corresponding author: [email protected] Received March 10, 2000; accepted October 1, 2000 effects have not been directly assessed. Population size structure (e.g., stunting) may also influence the precision of estimated ages, given that annuli are more compressed in slowly growing populations. In addition, growth rates can decline greatly with the onset of maturation (Roff 1983; Jennings et al. 1997). The effects of sex and maturation schedules on age estimates often have been overlooked (but see Sharp and Bernard 1988). Because management activities frequently cover large geographic scales with multiple populations, a synthetic examination of the effects of biotic and abiotic factors on precision between readers and structures is necessary. There is currently considerable debate regarding appropriate statistical methods for examining precision among readers and structures. Percent agreement is considered a simple, intuitive statistic that provides valuable insight into differences in precision (Hoenig et al. 1995). Use of this statistic has been discouraged, however, because it is a relative measure of precision and does not allow for comparisons among species. The range of ages among species may be so different that percent agreement among multiple species becomes meaningless (Beamish and Fournier 1981; Chang 1982). Alternatives to percent agreement, such as average percent error (APE; Beamish and Fournier 1981) and coefficient of variation (CV; Chang 1982), have been suggested to compensate for this deficiency because being age-dependent, they allow for comparisons among species and populations. Construction of age bias plots also has been sug375 ESTIMATING BLUEGILL AGE FIGURE 1.—Location of bluegill study lakes used to compare the effects of latitude and population size structure located in northern and southern Illinois. Quality populations (Q) were those in which mature, male bluegills greater than 180 mm total length (TL) were abundant, whereas in stunted populations (S) the majority of mature male bluegills were less than 160 mm TL. gested for detecting both linear and nonlinear biases (Campana et al. 1995). Our goal was to expand the scope of previous studies and quantify the effects of potential sources of variation on the precision of age estimation. Specifically, we examined the effects of latitude, population size structure, sex, maturity level, and age on the precision of age estimation between readers and structures (scales and otoliths) of bluegills Lepomis macrochirus. We chose bluegills because use of both their scales and otoliths has been validated (Regier 1962; Schramm 1989), but no previous studies have examined precision of bony structures for this common sport fish species. Additionally, bluegills exhibit complex life history strategies that often result in differences among populations in characteristics such as growth rate, population size structure, longevity, and maturation schedule (Claussen 1991). This variability makes the bluegill particularly suited for examination of precision of age estimation in multiple populations. Additionally, we have compared statistical methods to determine which, if any, is most appropriate for identifying and quantifying differences in precision between both readers and structures.