Pathological levels of removals from study area

[Jason Waite]

We've been using SECR to successfully monitor a population of grey wolves in Alaska for the past several years using hair samples collected in hair snares to genetically identify unique individuals. In order to minimize bear interactions (both in terms of snare destruction and non-target sample deposition), our sampling session takes place between October and December, which unfortunately partially overlaps the local hunting and trapping season for wolves. Generally, most wolves are taken by trappers in the very tail-end of our sampling session, which has minimized the number of animals removed from our study area. Additionally, trappers had been required to have the hides sealed within 14 days of take and report a (general) location of take. Sealed hides are also sampled for DNA to identify the individual. Using this information, we have, up until now, been able to account for these removals in our models.

However, in part due to a growing frustration with our agency's population estimates (locals think it should be higher), there was a well-coordinated effort among the local trappers (and probably newly-recruited trappers) last season to greatly increase trapping effort. Concurrent changes in the trapping season and sealing requirements made it possible for an unprecedented level of wolves to be harvested over the course of our sampling session. Greater than 95% of our mean population estimate for the previous year was harvested. To make matters worse, sealing requirements changed such that taken animals only had to be reported within a certain amount of time after the close of the season. As such, we have only been able to assign a month of take to each harvested animal.  Additionally, we have some evidence that suggests our hair snare locations may have been targeted by trappers, as our field crew found traps and snares in the immediate vicinity of several of our hair boards. 

My questions are: 

1. Will it even be possible to use SECR to estimate a (pre-trapping) population size when such an extreme number of animals were taken from the study area, especially considering time of take is at the month resolution and reported location may be questionable and vague?

2. Is the potentially purposeful lethal take around our detectors something that I should be concerned about? My feeling is that it would affect the detection function around that particular detector in that surrounding the hair snare with lethal traps is akin to setting up a camera trap and surrounding it with a shoddy fence--some animals may make it through to the detector, but the chances are reduced. My first idea was to include a detector-specific covariate (targeted / not-targeted), but we cannot say for sure which of our detectors were affected.

Thanks in advance for any suggestions or comments on this.

[Murray Efford]

Wow, you do have problems. I don't have deep answers. First thoughts:

1. Does the 95% take make you doubt your previous estimates?

2. The old (Otis et al. 1978) advice for dealing with harvested (dead) animals in non-spatial models was to take them off the top and add them back in after the analysis. To do that spatially you need to know their spatial distribution (perhaps assume uniformity over some area). And it is not optimal because it doesn't use any information from the truncated histories.

3. The challenge is to describe/model the harvest process in space and time: if you can do that then I assume there is a smart way to include it in the SECR model, but this is a research question.

[Jason Waite]

No, I don't necessarily doubt our previous estimates. Due to the amount of time it takes to get the DNA results back, our estimates lag by about a year. This means the population gets managed based on the previous year's pre-harvest level and does not take into account the normally low harvest and the unknown amount of recruitment (we don't really know much about the recruitment rates for this population). So, 95% harvest of the previous year's estimate (2 years ago) may actually be a lower percentage of what was available last year when the huge harvest took place. There is an extensive road system on the island, and the most of the remaining areas not accessible by road can be reached by skiff. These are a group of trappers that are self-proclaimed wolf haters and some have literally devoted their winters to eliminating the wolves on the island. So, even if our estimates are dead-on, there's not a doubt in my mind that this group could accomplish a near extirpation of the population in a single season. Of course, we still need to wait for all of the DNA from the trapped animals to come back so we can tell if any of the wolves were turned in more than once or came from different parts of the state (both of which were rumors we heard several times). 

Now, that being said, it is certainly possible that the estimate has been biased low all this time, but we really don't have a good way of assessing that. However, the estimates we've been getting are consistent with historical trends, tend to be fairly stable, and the inter-year differences have all been well within the range of biological plausibility. In my mind, these estimates (for this particular project) should be viewed more of a stable, conservative trend by which to manage the population (had the regulations not been changed). 

Once the data is available, I may begin by examining just how much the timing of the removals affects the estimate by running multiple models with the removals entered at different times (i.e., shift them by a week each time, within the reported month of removal, with some sort of randomization as to which harvested animals get shifted where/when). We have *some* idea of the spatial distribution of the removals as the trappers do report a general area which we've been able to somewhat standardize. The spatial scale is a lot coarser, but it is probably better than nothing.

Thanks for the response.