Utilizing random forest analysis with otolith mass and total fish length to obtain rapid and objective estimates of fish age

Age estimates from otolith morphometrics represent a rapid and objective alternative to traditional ageing techniques though use has been limited to marine and short-lived freshwater species. We utilized random forest analysis with otolith mass, total length, and several temporal and spatial predictor variables to assess variable importance and accuracy of age estimates for age-0 through age-11 yellow perch (Perca flavescens) in southwestern Lake Michigan. Accuracy of age predictions decreased with increasing age as 95% of juvenile (age-0 through age-2) ages were predicted correctly compared with 55% for adults (age-3 through age-11). Precision of age estimates within 1 year of reader-assigned age were high for both juvenile and adult yellow perch at 100% and 86%, respectively. Otolith mass was the most important predictor variable; however, substantial overlap existed among adult ages. Random forest analysis utilizing otolith mass, total length, and other pertinent predictor variables represents an applicable tool to reduce subjectivity and resource expenditure while providing accurate age estimates for juvenile and shortlived fishes.