Bat Presence Estimates
One of the objectives of this work has been to establish the relative numbers of the different bat species within the Parkhurst Forest. Estimating actual numbers is extremely difficult given that visual sighting of the bats is very difficult once the sky darkens, and of course in areas where there is substantial foliage cover. The way I've approached the problem is to make few assumptions so I can use the relative call density (number of echolocation calls per second per km^2) as a proxy for the bat numbers, and then apply the principle of scaling to the intra-specific spatial abundance (see p31 Maximum Entropy and Ecology, John Harte, OUP). These assumptions are basically as follows:
The national bat population estimates for England (Harris et al) can be used to establish the relative abundance of each species
The national figures can be rescaled to a common area where all the represented species are found (based on the BCT bat atlas)
The re-scaled population can be used to generate a normalisation constant for the recorded bat calls
Not all bats have the same call period, so all calls are normalised with respect to longest call period
Not all bats have an equal probability of detection, so this needs to be factored in too (using my own measurements + some results from: Adams A.M., Jantzen M.K., Hamilton R.M., Fenton M.B., Do you hear what I hear? Implications of detector selection for acoustic monitoring of bats, Methods in Ecology and Evolution 2012, 3, 992–998)
A hypothesis can then be made for the number of bat calls/species based on this re-scaled and normalised data
A Chi Square test or similar statistical test can then be applied to see if this hypothesis is correct
What I found was somewhat unusual, some species where exceptionally close to the expectation value, but others were significantly different. This is shown in the graph below where I've taken the ratio of observed and expected call density.
In this example, the closer the ratio is to 1, the nearer the observed data is to the expectation value. Taking the brown graph bars first (which are uncorrected for the probability of detection) the first thing that strikes you is that the observed Barbastelle population is approximately 9 times this expection value. Given the known importance of the site for Barbastelles from the work of Ian Davidson-Watts, a significant population was expected, and these results reinforce the findings of this earliar capture/radio tracking work. In addition the leisler's and serotine bats are also well above their expectation values and the bechstein's and brown long-eared bats are seemingly very low against the expectation value.
However, if you refer to the blue graph bars, which are corrected for the probability of detection, you can see that the seemingly low detections of the bechstein's and brown long eared species are actually to be expected based on the detection performance of the Elekon BatLogger. In fact, the number of bechstein's detected is higher than the expectation value by a factor of over 16. Before anyone starts to criticise, the BatLogger performance, my measurements and independent published tests suggest that there are only two commercially available detectors that will give a reasonable chance of finding the quiet call species like bechstein's and brown long-eared bats by acoustic detection, and one of them is the BatLogger... The poor detection probability of most commercial detectors probably partially explains the lack of success of finding the quiet call species using acoustic detection methods.
Clearly, these results are strongly dependant on the accuracy of the original Harris et al population estimates and the accuracy of the species distributions, however the method does allow distributions from large scale surveys in different areas to be compared. Using the pipistrelles as a marker species, this model gives values of 1.3 for the Parkhurst Forest survey, and 1.4 for the Norfolk Bat survey by Newson, which is a very close set of results and provides some degree of confidence in the method.
I'll publish the detailed maths and some examples applied to other researchers' big bat surveys in due course.