Using Virtual Simulation in a Geographic Information System to Optimize Abundance Survey Designs When Logistic and Biological Conditions are Constrained

Estimating wildlife abundance is central to many resource and ecosystem management problems. Early statistical work on wildlife abundance estimation focused on small-scale problems, but there is a growing need for such information at large spatial scales. An emerging area of research is the optimization of survey designs and methods for large-scale inference, given that agencies need to manage over large scales but operate within tight logistic and financial constraints. We used a geographic information system to explore how candidate regional-scale sample survey designs performed with regard to bias, field efficiency, and potential disturbance using a case study where biological and logistical constraints were severe (a regional-scale ground survey of Adélie penguins [Pygoscelis adeliae] in Antarctica). Some design options enabled gains of up to 50% in field efficiency and 80% in reduced disturbance without any bias or loss of precision. Biased abundance estimates were obtained when small sub-colonies were selected as sample units for convenience in counting. Probabilistic sampling using either plots or sub-colonies returned unbiased estimates. Improvements in field efficiency and reduction in disturbance were achieved in increments through a number of design features. Design decisions often resulted in opposing gains and costs in field efficiency for various survey activities. The optimal outcome of these opposing trends was not obvious without examining the breakdown of overall survey time by activity. Design requirements for optimizing criteria of bias, field efficiency, and disturbance were often opposing and competing. Identifying an optimal overall outcome for these competing criteria depends on their relative importance in the context of the management objectives, logistical constraints, and ethical values.