The last few months have been extremely busy at Proteus, but here's a quick overview of some of the things we've been up to since our last update.
Redesigning kangaroo surveys
In November 2017, when we first relaunched our website, we mentioned an analysis we were performing for Parks Victoria (PV), Australia, estimating kangaroo abundance from distance sampling surveys in some national parks in north-western Victoria. In June we worked with PV staff to redesign the distance sampling surveys to ensure better coverage of the parks with their survey effort. This involved clearly determining their areas of interest, defining survey strata based on barriers in the landscape that could not be easily crossed by a person on foot (e.g., fences, large creeks or streams), then use a probabilistic sampling scheme to select transect lines that would be surveyed. An additional consideration was that areas of the national parks get flooded periodically, including during the period of the kangaroo surveys. We therefore defined alternative transect lines that could be searched in years when some of the primary transect lines can't be searched due to flood waters.
Gecko monitoring - goodbye to trend detection
Over the last few years we've helped Auckland Council to analyse data they've collected from gecko monitoring programmes, and provided advice on the design of such programmes. Like many monitoring programmes around the world, Auckland Council were wanting to design their monitoring programme around the concept of being able to detect a 'significant' trend in a specified timeframe. We advised against that because a) they didn't have the budget to do the required level of monitoring for the arboreal gecko species of interest, and more importantly, b) focusing on trend detection misses the much more salient issue of identifying what the management goal is for the population. As an alternative, we suggested taking an idea from industrial statistics and use a control chart to assess how the population is doing over time. A control chart involves plotting a relevant metric of interest that is calculated at each survey period to indicate how that metric is changing over time. Control limits are defined to indicate when the process is deemed to be 'in control', and when the metric crosses such a limit, the process is 'out of control' so managers would be expected to take suitable actions to get the system back in control. While a control chart-based approach may not have a major influence on the design of the monitoring, it does force managers to think about their goals for the population and possible management actions if the population is heading away from the goal. Control charts have seen limited use in ecological applications, but it is growing.
Scottish sojourn
In the last week of June, Darryl travelled to Glasgow, Scotland (the image on this post in the Glasgow City Chambers), to give an introductory-level occupancy modelling course to a group of researchers, mainly from the UK and Europe, but also included one from Costa Rica and one from Mexico! They were a great group and Oliver Hooker from PR statistics was a fantastic host for the course. Glasgow was also a great host, turning on some fantastic weather for us, with the temperature hitting 32 degrees! That's centigrade not Fahrenheit by the way. Following a short visit to Inverness, where Darryl shot a video on interpreting confidence interval overlap, he then headed to St Andrews to attend the 2018 International Statistical Ecology Conference (ISEC). It was a great conference and while he was there presented a graphical method for assessing the fit of occupancy models. The conference finished with an entertaining tribute to Steve Buckland, of distance sampling fame, who is about to shift to semi-retirement after an incredibly productive career.