Can buffer zones have detrimental effects on the conservation of Spanish imperial eagles?: a response from Gonález et al.

Abstract

We recently published the first study that empirically quantifies the reaction of the threatened Spanish imperial eagles Aquila adalberti towards a variety of human activities around the nest, and an evaluation of its influence on breeding success, with the aim of recommending buffer zones in order to enhance the Spanish imperial eagle’s reproductive success (Gonzalez et al., 2006). A reply of Ferrer et al. (2007) agrees with the interest and utility of our study, although they argued that we ‘underestimated that eagles can habituate to non-lethal human activities’, suggesting that ‘it seems counter-productive to impose restrictions on landowners and may even be detrimental if the presence of eagles diminished owners’ income’. Ferrer et al. speculate that a change in selective pressure (based on reinforcement of nest defence behaviour with increasing intruder frequency, Ferrer, Garcia & Cadenas, 1990) linked with the long reproductive life of the species would facilitate a decrease in the distance at which eagles would flush. This is not necessarily so. It could facilitate a decrease in the distance of attacks towards the intruders (as they state), but not necessarily a decrease in the distance from the intruders at which eagles would jump from the nest. Additionally, it is not clear whether the change in selective pressure indeed has the consequences predicted by Ferrer et al. (1990). For example, no further increase in aggressive behaviour of Spanish imperial eagles towards nest visitors has been observed, as Ferrer et al. (1990) expected. The small sample size in which their study was based may imply that those results were more individualdependent than applicable to the whole of the species. These authors argue that it would be more helpful to get the eagles used to human intrusions, assuming that if human presence were to be excluded from the territories, eagles would become increasingly sensitive to human presence. However, even if our results showed that some habituation does exist, this habituation was not enough to prevent a decrease in reproductive success, and so we disagree with the fact that it would be better to get eagles used to humans (the risk that it could create a decrease in productivity before the predicted ‘total’ habituation is not worth taking, in our opinion). Secondly, these authors minimize the importance of the effect of human activities on the eagles’ hatching rate because, according to them, the relation is only ‘marginally significant’. According to their re-analyses, the probability of the correlation between hatching rate and number of days with intrusions is P=0.052 with r= 0.627 (being rounded off in Gonzalez et al. (2006) as r= 0.63, Po0.05, n=10). The discrepancy may arise because of different ways of estimating the significance of r in different statistical software. Following the Sokal & Rohlf (1981) tables, with n=10, P is significant (0.05) when r40.576. Furthermore, Siegel & Castellan (1988) indicate that when testing for the significance of Spearman’s correlations, a special technique has to be used for a small sample size. According to that ouvrage, with n=10, P is significant (0.05) when r40.564. Thus, we reassert that there exists a significant trend for the hatching success to decrease with increasing intrusions in the eagle’s territories. And, as they are right to point out, there is also a significant (r= 0.677, P=0.02) and negative relationship between the hatching rate and intrusions per day, which also reaffirms our results. Ferrer et al. also argue that the significance disappears when excluding one of the territories, and so they conclude that the relationship is not important enough to support a proposal of a buffer around the nest. As Ferrer et al. know, it is difficult enough to have large sample sizes with these species (in terms of number of territories monitored). Even if significance decreased when excluding one of the nests, that could simply mean that the remaining sample size would not be large enough to detect significant results at 0.05 (a b test would be needed). Nevertheless, we have re-analysed the correlation eliminating the M-18 territory as suggested, and contrary to their arguments, the correlation is still significant (Po0.05) between the hatching rate and both variables (r= 0.6129 and r= 0.6129). Finally, Ferrer et al. (2007) argue that an increase in fecundity is less important than a decrease in survival (which