Abstract
Factors influencing detection probability in line transect distance sampling were investigated to estimate the abundance of four common farmland birds on 12 sheep & beef farms in the South Island of New Zealand. Our primary aim was to evaluate the necessity of employing distance methods to correct for heterogeneity in detection probability. Detections of skylark Alauda arvensis, blackbird Turdus merula, song thrush Turdus philomelos, and Australian magpie Gymnorhina tibicen were recorded using ten 500-m unbounded line transects on each of 9-10 visits, and modelled using multiple covariate distance sampling methods. Covariates of detectability played a strong role in model fitting, but showed few consistent directional trends within species. Increased woody vegetation cover on farms greatly decreased detectability, while few seasonal or geographical effects were found. No detectability differences were found between farms using certified organic, integrated management or conventional farming systems, indicating that bird population dynamics might be compared between systems using simpler index counts. However, unless detectability parameters can be standardised to a high degree within a survey, we recommend the use of analysis methods that incorporate heterogeneity in detection probability.