Validating Back‐Calculation Models Using Population Data

Abstract

AbstractOtolith back‐calculation, an estimation of size and growth at some previous age, is a valuable and widely used method of tracking the traits of surviving individuals. Although numerous back‐calculation models have been developed, the accuracy of model prediction is rarely assessed before using a model on wild populations. Using the Atlantic Silverside Menidia menidia, we tested the precision of four commonly used back‐calculation models (body proportional hypothesis, modified Fry, biological intercept, and time‐varying growth). We calculated the moments of the observed length frequency distribution (mean, variance, skewness, and kurtosis) and compared them with the moments of the back‐calculated length frequency distribution predicted by each model. We also compared how the most accurate back‐calculation model changed under various scenarios of size‐selective mortality and found that bias in back‐calculation could not be explained by size‐selective mortality. The differences between functional forms over a wide range of fish sizes were slight for the biological intercept and modified Fry models, whereas the time‐varying growth and body proportional hypothesis models differed considerably. We found that all models performed well for at least one measure of accuracy. However, the modified Fry model was the most accurate overall when back‐calculating to the observed population moments.