ºÝºÝߣ

Editor's Notes

1. Abundance and distribution of a species are key quantities for conservation and management strategies but remains challenging to assess in the field.
2. The main issue when we study natural population is that we can¡¯t have an exhaustive census of all the animals. We cannot count them all like we would do with sheeps. Most of the sampling methods give access to some portion of the population.
3. Capture-recapture methods are the most famous methods giving access to abundance estimates while correcting for imperfect detection. But we have seen in previous studies that when the population and the detection probability are low, CR models give quite rough abundance estimates.
4. Moreover, CR methods mean that individual data are required and as a consequence imply a great commitment of time and money.
5. Just to give you an example, the first camera trapping sampling design we deployed in France in 2011was quite expensive : 163 700 euros
6. In order to cope with that last issue, some monitoring programs are only based on the collection of presence/absence data such as indirect presence sign of the species (tracks, hairs, scats¡­.). That kind of data can be used to estimate the proportion of area occupied by a species (occupancy) by explicitly modelling detection probabilities.
7. The monitoring of the Eurasian lynx in France is a quite good example of that kind of issues.
8. In 2011 we started a camera trapping project in the Jura Mountains.
9. We sampled 33 sites
10. on which we settled 2 camera traps facing each other during 2 months
11. We managed to collect 39 photographs and identified only 9 individuals. When you work next to another PhD student that work on birds with 1 thousand ring data, you¡¯re quite frustrated ;)
12. So, when I started my 3year-PhD I had a dataset with 9 individuals¡­ and no more money to invest in the monitoring¡­
13. First, i cried for my mommy and pray for a miracle to have good quality data enough to have sound abundance estimates
14. But then, my spiritual guide, as known as my supervisor, told me
15. Indeed, in parallel, a network of volounteers collected 172 presence signs from 2009 to 2011 in the same area.
16. Up to now, data collected during intensive and extensive surveys were analyzed separately but as the lynx is a territorial species, we can assume the presence signs collected belonged to individuals that were also exposed to camera trapping and we can analyze all these data together.
17. So, we had presence/absence data informing on the occupancy on one hand and CR data informing on the abundance on the other hand. Since then, abundance and occupancy were considered as distinct variables but¡­
18. If we dig a little, we can easily see that the two state parameters are linked, and that¡¯s the key concept of this study.
19. When a site is known to be occupied by a species, there is at least one individual of this species on this site (that means N, as we usually call the abundance is >1).
20. That means we can have some information on N if we have information on the occupancy of the species, right ?
21. Thanks to occupancy models, we estimate psi, the probability that a site is occupied
22. and 1-psi the probability that a site is unoccupied.
23. We also take into account the imperfect detection by modelling the detection process. People from the network go repeatedly on the field and the repetition of presence signs in time enables us to estimate p, the detection probability.
24. The camera trapping data were used to estimate N and were analized using CR models. In the capture-recapture model we assume that N is a realization of a Poisson distribution so the probability to have k individuals was equal to¡­
25. so the probability to have k individuals was equal to¡­
26. And the probability to have no individual is equal to exponential minus lambda
27. As a consequence, the probability to have at least one individual is 1 minus the probability to have 0 individual 1-exp - lambda
28. In CR model we also model the detection process thanks to the repetition in time and among individuals that are detected by the camera traps..
29. The detection process enables to take into account individuals that are missed during the camera trapping period and correct for the sites that are truely occupied but detected as unoccupied in the occupancy model
30. So, when a site is occupied
31. That means the site contains at least one individual
32. And the probability to have at least one individual in CR models equals 1 ¨C exp ¨C lambda which is exactly what is called the probability of occurrence in the occupancy models
33. And the probability to have at least one individual in CR models equals 1 ¨C exp ¨C lambda which is exactly what is called the probability of occurrence in the occupancy models
34. So, we can use occupancy data
35. to capture information on abundance
36. and made the link between the two state variables explicit in the model. This is achieved by expressing lambda, the rate of the Poisson distribution, as a function of the probability of occurrence and implement that function in the combined model.
37. And of course, in the combined model we also took into account the detection process !
38. The results are quite clear, the abundance estimates coming from the combination of presence-absence and CR data are drastically improved ! When the sole CR model with heterogeneous detection probability was fitted to camera trapping data p was really low (0.11) and N was estimated between 9 and 35. Using our new approach abundance was estimated between 9 and 13.
39. Must be sure the closure assumption is valid so the period not during breeding season and geographically closed (territorial species)
40. Be careful with potential false positives in presence data. All the data must go through a validation step.
41. N was considered as a homogeneous poisson random variable and can lead to underestimating the variance of N
42. The best way to overcome this issue would be to resort to SECR surveys using a nonhomogeneous poisson process to model animal home range locations.
43. Abundance is the most informative state variable in conservation that¡¯s why most of the time we are looking for individual data. But that require huge financial resources and most of the monitoring programs are quite limited.
44. Presence-absence data are an added value and can arise from citizen science programs so affordable and quite useful in large scale surveys.
45. So, combining these data remains useful for monitoring programs with small budgets or targetting elusive species and for monitoring programs that combine surveys with 2 resolutions such as count data or individual data on a small scale and presence-absence data on a larger scale.