A comparison of different approaches to monitoring bird density on New Zealand sheep and beef farms
- Centre for the Study of Agriculture, Food and Environment, University of Otago, PO Box 56, Dunedin 9054, New Zealand
When designing large-scale bird monitoring schemes, financial constraints often require researchers to make trade-offs in the spatial resolution and precision of the density estimates by varying the number of sites monitored and the intensity of sampling effort per site. Here, I compare density estimates of four common farmland bird species (skylark, blackbird, song thrush, Australasian magpie) on South Island sheep & beef farms collected using a large-scale monitoring scheme with equivalent estimates from a more intensive survey, and discuss possible sources of imprecision and bias in each. Density estimates from the intensive survey were generally lower and more precise than the monitoring-scheme ones, and the surveys were susceptible to different types of bias. These effects were linked to combined differences in modelling methods and sampling effort distribution, and to observer-related issues. In long-term designs, estimates from data pooled over several annual surveys are likely to become accurate quickly, but an increase in monitoring effort per site may be required to increase the precision of individual-survey estimates. Species that pose challenges to observers may be hard to estimate accurately with the monitoring-scheme design explored here, and the use of pilot surveys is recommended. Raw count data from the same species were tested for their usefulness in the creation of reliable relative population indices. A sufficiently constant ratio of plot-count indices to absolute density estimates was found in skylarks and thrushes, while inclusion of a correction parameter to account for woody vegetation effects on detectability was necessary in blackbirds, and magpie estimates proved unreliable. Similar analyses are recommended for all monitored species when trend estimation using relative indices is intended.