Improving spine-based age estimation in centrarchid fishes using otolith-derived training sets

Nonlethal age determination is a priority in fish biology contexts where sacrifice is restricted, but estimates derived from nonlethally attainable structures tend to suffer from poor precision and accuracy relative to necessarily lethal, otolith-derived estimates. Dorsal spines, while less accurate than otoliths, have shown higher precision compared to other alternative structures, making them a useful candidate for testing reader-based approaches to improved age estimation. Here, we test whether small, lake-specific training sets derived from paired otolith-spine images, derived from incidental mortalities can provide useful calibration to spine-based age estimates in four centrarchid populations from the northern Midwest, including two largemouth bass (Micropterus nigricans) and two bluegill (Lepomis macrochirus) populations. Five moderately experienced readers estimated ages from spine images before and after calibration with paired otolith-spine images. Calibration consistently resulted in increased estimated ages, particularly for older fish, and reduced variability among readers in both individual age estimates and predicted length-at-age trajectories. Specifically, we find that these structured training sets are useful in both reducing traditional underestimation biases and reducing reader-based variance effects at both the level of individual estimates and in growth modelling contexts. Therefore, they may offer a practical means of leveraging available incidental mortalities in the improvement of otherwise nonlethal aging programs. Accordingly, we suggest that where available, readers first be familiarized with lake-specific paired otolith-spine images before independently estimating spine ages.