Yiyin Chang

1 Introduction Estimating species abundance presents significant challenges. Because it is impractical to count every individual, abundance estimates must be derived from sampling, which requires the use of statistical models (MacKenzie and Nichols 2004; Bonar et al. 2011; Bailey, Mackenzie, and Nichols 2014). Several widely used approaches exist, each with different data requirements and assumptions.

Assessing road bias We examined the effect of road bias on model prediction because the sighting records are clearly biased by observer accessibility (Figure 4). In our exploratory models, Euclidean distance to road was consistently selected as an important variable when included as a predictor. However, when projected, the model prediction with Euclidean distance to road shows very strong, unrealistic, habitat suitability tied to roads (Appendix ).