Density-dependent detectability in dynamic occupancy survey: a case study on a vulnerable beetle species in hollow trees
(Oral and Poster)
Fabien Laroche
, Heidi Paltto
, Thomas Ranius
SEE PEER REVIEW
Conservation of threaten species living in fragmented habitats crucially relies on evaluating their occupancy and their ability to colonize and persist in habitat patches. Fitting stochastic patch occupancy models (SPOMs) to occupancy data can help assessing these features. However it is critical to account for the limited detectability of target species in this type of analysis to avoid severe biases in estimation.
Detectability of a population in a habitat patch often tightly depends on the local density of individuals. This connection between density and detectability has rarely been used in SPOM analysis, even when abundance data are available. The two quantities are often considered independent and estimated separately. Here, we propose a framework for using density-dependent detectability in the analysis of a SPOM.
We illustrate our approach with the example of T. opacus, a beetle inhabiting hollows in old trees. We use a 6-year survey of adults abundances in an woodland pasture area harbouring a high density of old oaks, in Östergötland, south-east Sweden. T. opacus is classified as "vulnerable" on the Swedish red list.
We first modeled abundance data in occupied trees as a function of tree and environmental features. We used this model to predict density-dependent estimates of detectability in all the trees of our study site. Importantly, we could explore how the environmental features affect carrying capacity and detectability of trees, with the latter aspect being rarely explored in metapopulation studies. Secondly, we showed the good match between our density-dependent estimates of detectability and those obtained from repeated occupancy surveys (Pearson r = 0.54, p<2E-16). Thirdly, we found that omitting density-dependent detectability had little effect on metapopulation parameter estimation in our example (except lowering the goodness of fit to data), probably because we performed many repeated surveys in sites. Ultimately, accounting for density-dependent limited detectability did not eliminate unstructured colonization from our study system, which may indicate the contribution of the cryptic larval stage of T. opacus.
Our study thus shows that density-dependent framework allows for a simple integration of limited detectability into metapopulations studies, based on a more thorough use of abundance data than classic approaches. In particular, it may be applied even with no or a limited number of repeated surveys, although a statistical survey would be neceassry to assess this point. More generally our work also suggests that accounting for limited detectability is only the first step towards deriving reliable metapopulation model estimates for conservation planning.
INTRO: Authors propose a framework for using density-dependent detectability in stochastic patch occupancy models. Their framework allows integration of limited detectability into metapopulation studies. However, they found that omitting density-dependent detectability did not affect on metapopulation parameter estimation, or eliminate unstructure colonization from their study system. They think that the number of repeated surveys and the cryptic larval stage of their study organism are the reasons for the lack of effect of including the density-dependent detectability in thei model.
MERITS: Density-dependent detectability and how to include it into different models and also in monitoring and evaluation schemes should be interesting for all conservation scientists.
CRITIQUE: No.
DISCUSSION: Information of density-dependent detection is needed for interpreting results of monitoring schemes and evaluating conservation status of species.
Many times, repeated visits for survey sites are not possible due to lack of monitoring resources - perhaps authors could consider also such cases?
- - -
INTRO: The authors describe how abundance data can be used to account for detectability in occupancy modelling. The method is illustrated using multi-year field observations of the oak inhabiting beetle Tenebrium opacus. The results show that density dependent detectability estimates correlate with the estimates based on the traditional repeated surveys method.
MERITS: The aims and methods of the study are clear and justified. The abstract gives a good overview of the scope of the study and seems to summarize its major points. The data used seem well suited to demonstrate and test the proposed method.
CRITIQUE: How the density dependent detection is integrated into a SPOM remains unclear and also whether in the study detectability is only estimated in the model or also compared to control data with very high detectability. The possibility of a cryptic life-stage raises concerns about the detectability estimates for this particular species, but should not affect the general results and conclusions.
DISCUSSION: The results and study system are relevant for population monitoring and conservation planning and thus this is a suitable study for the ECCB conference. Understandably, an abstract does not provide enough space to elaborate on the technical details so that any the merits and weaknesses from the modelling point of view cannot be evaluated. It sounds very promising, but as the authors note, detectability is only one of the potential issues of practical applications of SPOMs.