Testing the power of an experiment to measure predator control and habitat complexity impacts on farmland bird abundance

In this study I assess the statistical power to detect a significantly greater increase in bird population size on treatment farms than on control farms given that there is a substantial treatment effect. Computer simulations of bird populations on New Zealand sheep/beef farms were used to generate significant changes in bird abundance from (a) controlling predation by introduced small mammals, (b) habitat structural complexity, and (c) an interaction of both. A simplified computer model of bird population dynamics was developed that predicted a birth pulse of 357% when predators were controlled and 110% if not, and a target of detecting the experimental elevation of bird abundance at a statistically significant level (P < 0.05) in 75% of all attempts was set. If at least four farm pairs (treatment vs non-treatment) are monitored, this is feasible for 15 of 23 species common on farmlands for which sampling error of abundance estimation was below ~40%. A second virtual experiment measured the power of tests of whether habitat complexity and predation in combination led to added increases in bird abundance. It showed that a 75% detection of elevated benefits of predation control in complex habitats could only be achieved if at least 48 farms were monitored, and then only for species for which abundance could be estimated with <10% error. Researchers are advised to invest in increased within-site monitoring to achieve a reasonable precision in bird abundance estimation before increasing the number of replicates.