Using occupancy estimates to fine‐tune conservation concerns

Our study (Sberze et al., 2010) combined three personal contributions: a professor’s (M. C.-H.) experience in neotropical ornithology, a post-doc’s (G. F.) practice with occupancy estimation, and a graduate student’s (M. S.) interest in nocturnal birds. The work was developed in the Biological Dynamics of Forest Fragments Project (BDFFP), a 30-year landscape manipulation experiment focused on biological change in artificially isolated forest patches created by forest clearing in the early 1980s (Laurance et al., 2002). Circumstances beyond control of the BDFFP led to gradual abandonment of cleared land and subsequent growth of c. 30 km of secondary forest (SF) surrounded by old growth (OG) and embedding the experimental patches. Our occupancy study of nocturnal birds was at once an effort to take advantage of unplanned secondary growth (as accurately pointed out in Sekercioglu’s commentary) and a contribution to the ongoing debate about the conservation value of SF (Laurance & Wright, 2009). We are grateful to the editors of Animal Conservation for featuring our paper and to Ken Feeley, Cagan Sekercioglu and Daisy Dent for their kind and insightful commentaries. We aim this reply at clarifying some of our methodological choices and identifying two avenues for future progress. We believe that our study convincingly illustrates the potential conservation relevance of SF thanks to a set of simple methodological choices. Both abundance and occupancy estimates provide biologically meaningful quantifications of population state. Abundance estimates, however, require mark–recapture or counts of individuals at measurable distances from the observer (Williams, Nichols & Conroy, 2002). Our focus on occupancy was in part due to the difficulties of counting nocturnal birds mentioned by Sekercioglu (2010). At the same time, the occupancy approach provided a sound conceptual basis for comparing different types of land cover and allowed us to replicate sampling at a large and reasonably representative spatial extent. While attempting to cover as much ground as possible on foot, MS had to limit the time per sampling point. This temporal constraint reduced the probability of detecting species at each visit, increasing the importance of explicitly accounting for detection failure in the analyses, which we did. Also due to walking constraints, sampling points had to be relatively close to each other. We addressed this problem by quantifying the influence of occupancy in neighboring points upon occupancy at each focal point (spatial autocorrelation). Since we did not find strong evidence of spatial autocorrelation, the procedure gave us confidence that our sampling points are not pseudo-replicates. The spatial analysis was our strongest reason to use a Bayesian framework, although Feeley (2010) correctly suggests that we could have further explored the framework by using informative priors. We chose not to do this because we were interested in finding out whether our data revealed a priori expectations despite the use of priors that did not convey those expectations. Although most of our predictions were supported by the data, we were surprised by the general preference for SF over OG among owl species. We suspect that relatively high abundance of rodents in SF may attract owls, but we have no data to back up this explanation. We agree with Feeley (2010) that birds in SF may be in a demographic sink, that is a place where reproduction fails to keep pace with local mortality (Pulliam, 1988). This should inspire caution in the interpretation of results and motivate further investment in life-history study, a need that all commentaries have identified. Dent (2010) is particularly clear about this point, emphasizing how life-history information will help generalizations about what types of species are more or less vulnerable to landscape change. We believe that multi-taxon occupancy studies backed with detailed life-history information should be a major avenue of future progress (Gardner et al., 2009). The challenge resides in how much rigor will be required of life-history assessments. Life history studies are field-intensive and must be conducted one species at a time: more rigor will mean smaller sets of species and more limited contribution to generalization. Progress on this front requires a combination of smart data-mining, conceptual simplicity, and analytical sophistication. Ideally, this should be done without forgetting phylogenies, because species often share more than the life-history traits that we decide to measure (Harvey, 1996; Owens & Bennett, 2000). Relentless effort in life-history study should be complemented with further development of occupancy studies. This second avenue of progress presents analytical challenges but relatively little additional fieldwork. Recent work by Zipkin, DeWan & Royle (2009) illustrates how multi-taxon models of occupancy developed in a Bayesian framework make it possible to borrow information across species and thus produce inferences about species for which there are relatively few data. Perhaps even more important for the SF–OG comparisons is the infusion of a dynamic perspective in occupancy studies (MacKenzie et al., 2003). Our nocturnal bird study treats OG and SF as static environments for which we estimate average site-occupancy values. It would be desirable to estimate not only the occupancy at a