Proxies in Practice: Calibration and Validation of Multiple Indices of Animal Abundance

Abstract

Abstract The abundance of individuals in a population is a fundamental metric in basic and applied ecology, but sampling protocols yielding precise and unbiased estimates of abundance are often cost-prohibitive. Proxies of abundance are therefore common, but require calibration and validation. There are many ways to calibrate a proxy, and it is not obvious which will perform best. We used data from eight populations of Chinook salmon Oncorhynchus tshawytscha on the Oregon coast where multiple proxies of abundance were obtained contemporaneously with independent mark–recapture estimates. We combined multiple proxy values associated with a single level of abundance into a composite index and then calibrated the composite index to mark–recapture estimates using several different techniques. We tested our calibration methods using leave-one-out cross-validation and simulation. Our cross-validation analysis did not definitively identify a single best calibration technique for all populations, but we could identify consistently inferior approaches. The simulations suggested that incorporating the uncertainty associated with mark–recapture estimates into the calibration technique reduced precision and introduced bias when mark–recapture estimate uncertainty increased with point estimate values. Cross-validation techniques can be used to test multiple methods of calibrating multiple proxies to an estimate of abundance. Critical uncertainties with the application of calibrated proxies still exist, and cost-benefit analysis should be performed to help identify optimal monitoring designs.