2014: End of Project Summary

Note: throughout these pages you can right click on any image and download a full size copy in a new tab.

Objectives Met?

So how has the project progressed against its objectives?

I set out on the journey in January 2012 to:

1. Complete a systematic survey of the bat species presence in the Parkhurst Forest and surrounding area

2. Generate a species distribution model linking the bat presence records to the key habitat features

3. Demonstrate the value of maximum entropy modelling at a micro-scale 

In total, 936km was covered (232hrs sampling time) over the three seasons and a total of 13,081 presence records identified. The distribution of presence records across the species range is given below :

In total, 16 species of bat were identified, of which 10 were significant enough in number to generate species distribution models. Not unexpectedly, the main species present is the common pipistrelle, but it was surprising to see such a relatively large number of barbastelle bats present. If you refer to this section you will see the reasoning behind my estimate that that there are approximately 12 times more barbastelle bats present than would have been expected, which indicates that the forest is a particularly important habitat for this species. The next most significant species was the Serotine then the soprano pipistrelle, followed by the Noctule, Whiskered, Leislers, Natterers, Alcathoe and the Brown long-eared bat. The most unexpected find though was the presence of the Greater Horseshoe bat on the eastern edge of the forest, the first confirmed sighting of this species on the Isle of Wight for nearly ten years.

You can see the main activity periods in the graph below:

The general pattern of activity matches emergence norms indicating that all the species recorded are roosting within or very close to the forest, except for the NSL guild, which showed that a significant number of bats may well be commuting into the forest area rather than roosting since the early activity tail of the distribution is over 20mins from sunset, and the peak activity over an hour from sunset. With a flight speed of 40-50km/hr, this could indicate a commuting range for the noctule and leisler’s bats that is significantly outside the study area in some cases. The late arrival of the majority of the Serotine bats suggests a significant proportion of these presence recordings could also be from bats commuting to the forest area to feed.

The graph above shows the impact of the ambient temperature on the recorded activity. In most cases, the activity is peaking when the ambient temperature is around 15 degrees C. This temperature corresponds to the early to mid July period, which is when most juveniles will have fledged, and so is not that unexpected. It is interesting to note that the NSL guild bats tend to require a higher temperature before becoming active, which supports other observations that indicate a link between the ambient temperature and prey abundance.

The overall species richness picture of the area based on the 9 most prevalent species is shown below:

The first distribution indicates the widespread use of the forest and surrounding fields around the forest boundary, and it picks out the use of the hedgerows by a small number of species. In general, the northern half of the forest supports a wider range of species which is due to the more balanced plantation plan and the presence of the ancient woodland areas in its core. The areas to the south are predominantly coniferous/spruce plantation with little or no deciduous planting, except in those areas that support more than 5 species indicated above. The impact of the habitat background becomes more pronounced when you look at the second distribution which has been thresholded at a logistic probability of 0.4. This second distribution indicates the habitat "hot-spots" and these are the areas that support the widest range of bat species. These hot-spots lie in areas where there is a mix of both evergreen and deciduous planting, and areas where human interaction is more limited. One interesting feature is that the forest exterior edge region is particularly species rich on the sheltered eastern boundaries.

The individual, partially tuned models are available to view for each species here. These models are corrected for detection range by increasing the beta-multiplier depending on the "loudness" of the bat species call, but they are not pruned back to use only the most significant variables. I've done this so that the reader can see the relative importance of the different environment variables employed, whilst preventing the model from being over-fitted. Some of the deeper findings will be published in due course once written up.

So, I think I've achieved all my objectives set for this work... In addition, one "unwritten objective" of this work was to show that species distribution modelling is not a "black art" and that it is within reach of anybody with an interest in ecology and who has a good grasp of basic computing skills. All the key tools to undertake a project like this are available for free too! If you want to make a start with species distribution modelling, make sure you download my presentation and workshop notes for the BCT National Bat Conference 2013 from here.

Survey Data Summary

Click on the thumbnails below to see the summary of the data records. Note, the monthly record count has two bars, the blue bar is the raw data and the red bar shows the result after re-scaling the record data by the survey effort for each month.

Barbastelle Habitat Preference Shifts

One of the particular aspects of the work was to investigate whether the method would be sensitive enough to detect year on year habitat shifts. This is of particular importance at the Parkhurst Forest site due to the very high numbers of Barbastelles present and the fact that the Forest is a working forest with regular harvesting activities. The technique I used involved partitioning the presence data into years and running models for each year data-set, and then subtract the model generated using the full three year dataset. The results are given below for each year:

In these plots, red hue pixels indicate an increase (blue indicates a decrease) in the logistic probability. It is interesting to note that there is a really significant shift in habitat use in 2014, which corresponds to the season after a major thinning/harvest activity in centre/north of the forest and the total removal of all the trees from under the power lines near the eastern edge. The changes in the south western area are unexplained, the only change that has taken place is the closure of the Camp Hill prison  in 2013 and the subsequent reduction in ambient lighting in the area. Clearly there is some shift in the habitat preferences occurring, but more work needed to establish the root cause. One point that is apparent however is that the core areas of activity are always in the ancient woodland areas of the forest.

For comparison, the table below gives the year on year change:

These plots show the habitat shift in a more dramatic way as the full three year model tends to average out the habitat use. In the case of the Barbastelles present,  there is a very noticeable shift in habitat use between the seasons. There is a note of caution here relating to the time taken to build up habitat use models and their intended application. At one limit, one might assume that the model will converge as the number of presence samples is increased over time, and the final model would represent the least biased case. However, it is clear from these results that the Barbastelles are changing their habitat use year on year, which is a potential issue if you were undertaking a survey ahead of a major development where you might be surveying in just one season.

What next..?

I really hope you've found this record of my work useful, whether amateur or professional. The Parkhurst Forest programme of work is complete now so far as this website is concerned (though nay be a couple of publications with more detail in due course). My future plans are to look at macro level modelling and generating a HSM for the whole of the Isle of Wight... So watch out for the Isle of Wight Bat Project!

The Isle of Wight Bat Project is now live! Please visit https://sites.google.com/site/iowbatproject/ for more detail.