Lee Brady's Recording Blog

A Bumper Year for Toads?

Over the past week or so I have observed several juvenile common toad at survey sites across Kent. In discussing some of these observations with Steve Songhurst, the Vinters Valley LNR warden, Steve commented as follows:

"As regards the young toads they have been seen in large numbers all over the reserve this year by many people. It must have been a good breeding year for toads. Perhaps you may get other reports from around the county showing a similar position."

Whilst pondering his comments, I started to receive some interesting records from the KRAG online recording form. The first report was sent in by Mr. Christopher Goodbun who works at a garage near Broad Oak Nature Reserve in Canterbury. Mr. Goodbun reported seeing 40 to 60 juvenile toads:

"All the toads are youngsters, with the rain we have had thay are over the car park and in the showroom, each morning I collect them, and let them go back in the wooded river bank from where they came from." 

The second report came in a day or so later from Mr. Jason Steel who works at a Royal Mail Delivery Office. Mr. Steel reported that:

"Our work place has ben invaded with toads this week! In total we've had 76 toads come into our building this week with 35 of them being found today. Some were hiding and others were found crawling across the floor. They have been captured daily and released back on Crayford Marshes where I'm assuming they have come from. "

 

Mr. Steel sent me this and other fantastic photos of the toads that he found and has very kindly given me permission to post these on the blog.

 

Three independent reports of large numbers of juvenile toads doesn't necessarily indicate a trend. Last week's wet weather (after a very dry summer in Kent) may have resulted in a spike in activity. I would therefore be very interested to know whether there been similar reports of large numbers of juvenile toad from other areas. Whether the reports indicate a significant trend or not, the observations are fascinating and very useful records for KRAG's Getting Toads out of a Hole Project.    

Occupied Ponds v Breeding Ponds

Amphibians can be found in a wide variety of different aquatic habitats. Species that are known to breed in specific types of pond may still be found in other, perhaps very different, waterbodies. All waterbodies within which an amphibian has been recorded can be described as occupied. I have used occupied as the defining term for the recent series of blog entries that examine the impacts of fish, newts and pond density on common frogs. Some of the relationships revealed in the graphs don't appear quite as strong as we would perhaps expect. Whilst others appear a little surprising. For example, the number of fish ponds that are occupied by great crested newt in areas of high pond density appears to be higher than what our understanding of gcn ecology suggests. However, it is important to realise that just because a pond is occupied does not necessarily mean that the species of interest actually breeds there. 

Defining what constitutes a breeding pond can actually be quite tricky. It is a pond in which amphibians breed (obviously), but how do we define breeding. 

If breeding is defined only by the act of successful reproduction, we would need to follow eggs and tadpoles through their development and demonstrate that metamorphosis has actually occurred. Some may even argue that we also need to show that those metamorphs successfully survive to adulthood and return to the pond to breed once again. 

A pond that usually desiccates before tadpoles can metamorphose would not therefore be defined as a breeding pond, even though amphibians have obviously displayed breeding behaviour (i.e.. courtship behaviour, laying eggs etc). Yet in very wet years, the same pond may retain water long enough to allow tadpoles to successfully metamorphose. Metapopulation theory describes the process by which species may breed successfully in source populations whilst also displaying non-successful breeding in sink populations. Sources and sinks may even switch back and forth over time. 

When examining how habitat factors may affect species distribution, it is important to understand that some habitats such as ponds are often quite dynamic and their relative suitability may change over time. Think of a normally highly ephemeral pond in a very wet year, or a fish pond that fully desiccates during a drought. 

An occupied pond may not necessarily be a successful breeding pond this year, but it does represent a potential breeding pond for future years. 

Fish Ponds and Great Crested Newts

Although the dataset includes only a relatively small number of thoroughly surveyed fish ponds in areas of high pond density, I decided to run an analysis to investigate whether such ponds provide frogs with a refuge from great crested newts. The data may only reveal a trend, but I was hopeful that it could be enough to spark interest in fresh data collection. And maybe it will, but not necessarily in the way that I first thought... 

The following graph illustrates the interaction between pond density and fish. Data represents presence of common frog in fish ponds (yellow bars) and fish free ponds (red bars). 

There see, there does appear to be an interaction between fish and pond density! Hold on though, the interaction doesn't go the way I proposed it would... Frogs appear to do worse in fish ponds in areas of high pond density! What on earth is going on? Too little data? Biased survey results? Perhaps not...

The following graph illustrates the same information, but this time for great crested newt. 

 

As revealed previously, great crested newt pond occupancy increases (quite significantly!) as pond density increases. Indeed, in areas of high pond density, our data reveals that gcn occupy as many as 95% of the available fish free ponds! However, a similar trend also holds true for fish ponds. In area of low pond density, the presence of fish appears to have a very significant impact on newts. However, as pond density increases, newts start to occupy an increasingly greater number of fish ponds. 

Pity the poor frog then. Whilst breeding in fish ponds may provide a refuge from newts in areas of low pond density, the same isn't necessarily true in areas of high pond density. Newts that breed in other fish free ponds are perhaps more likely to enter fish ponds to breed - or perhaps just to feed on the frogs! 

Don't even get me started on those lucky frog tadpoles that manage to beat the odds and metamorphose. Did I also mention that grass snake distribution also appears to be positively correlated with pond density... ?

[update 13/10/09: couldn't resist just one more graph!]

Although we are fast approaching the point where there are too few surveyed ponds to show statistically significant relationships, I thought it would be interesting to illustrate the interaction between fish and gcn on common frog pond occupancy. Data represents presence of common frog in ponds that support gcn (yellow bars) and ponds with no gcn (red bars). 

What this graph reveals is that in ponds not occupied by great crested newts, fish appear to have little impact on common frog. There is even a small trend for frogs to occupy a greater number of fish ponds, but we will need more samples to determine whether this is significant. In ponds that are occupied by great crested newts, the additional pressure of fish seems to further reduce the number of ponds occupied by frogs. Note that the number of ponds found to support fish, gcn and frogs is very low and more survey work is required to determine whether the relationship is significant. 

Frogs, Newts and Fish

Yesterday's look at the impact of pond density on common frog pond occupancy proposed that: 

If newt predation is proportionally higher than fish predation, frogs may be more likely to occupy fish ponds in high density areas. 

We therefore need to examine the impact of both fish and great crested newts on frogs. First up is a graph to illustrate the impact of fish.

Quite revealing! It is apparent that the presence of fish has relatively little impact on pond occupancy by frogs. Whilst frog populations may be lower in fish ponds (this would need further research), it appears that frogs can at least tolerate the presence of fish. However, available data suggests that when fish are absent, there is actually a lower likelihood that a pond will be occupied by frogs. Could these fish free ponds already be occupied by great crested newts? What impact do newts themselves have on frogs? 

 

This graph is fairly clear. Ponds occupied by great crested newt are much less likely to be occupied by frogs. 

We already know that gcn are more likely to occur in areas of high pond density. What I now need to do is collect more data from fish ponds in areas of high pond density to determine whether they do indeed provide frogs a refuge from great crested newts. 

More Ponds, Less Frogs?

In a previous post, I considered the relationship between pond density and occupancy of ponds by great crested newts. There is a strong positive relationship. The higher the pond density the more likely the pond is to be occupied by gcn. If only life were so simple with the other amphibian species! 

Smooth newts appear to have a relatively high pond occupancy regardless of pond density, while frogs, toads and palmate newts are apparently more likely to occupy ponds in areas of low pond density. Think about that for a moment, the more ponds in the landscape, the less likely any one pond will be occupied by a frog. 

Here's a graph to illustrate the relationship between occupancy and pond density for common frog:

 

And here's one for common toad:

 

So what might be going on here? 

Should we not assume that since all of our native amphibians require waterbodies in which to breed, occupancy should increase with pond density for all species? 

Common frog tadpoles are very palatable to aquatic predators such as fish, invertebrates and that bad boy in the pond, the great crested newt. I propose that in areas of high pond density, the suitability of a pond for breeding frogs may be significantly reduced by the presence of newts. 

But surely the same constraint does not apply to common toad? Toad tadpoles have skin toxins that make them unpalatable to many aquatic predators. Toads also tend to breed in larger and deeper waterbodies than newts. Perhaps areas with high pond density typically include relatively few of the larger ponds? After thinking about this for a while I had to scurry off and look at the data to see if I could reveal the trend:

It takes a bit of squinting, but if you look at the graph you should be able to see that in areas of both moderate and high pond density pond size follows a normal distribution. That is, there are a few smaller ponds, a greater number of medium ponds and a few larger ponds. 

However, the distribution for low density areas is highly skewed towards the smallest ponds. My theory that larger ponds tend to occur in areas of low pond density appears to be completely wrong! 

Perhaps ponds in low density areas are more permanent and therefore more likely to support fish? 

Another big no to that idea then. Preliminary analysis of available data suggests that there is only a weak relationship between presence of fish and pond density and the relationship indicates more fish in high pond density areas (p does not add up to 1 btw, since ponds where the presence of fish could not be determined are not included in the graph). 

Determined not to give up on this, I wondered if survey effort was a confounding factor. I therefore redrew the graph illustrating the effects of fish - but this time I only included ponds that were surveyed with enough effort to reliably determine presence of common toad (ie. multiple visits rather than one off quickie assessments).  

Bugger. A small but potentially very important difference is revealed between the two graphs. 

In areas of low pond density, ponds supporting fish were proportionally more likely to be surveyed when compared to areas with high pond densities. I rather suspect that newt ponds were being cherry picked by surveyors. Since many of these ponds were surveyed as part of my own consultancy work, this probably also applies to my own data as well as that supplied by KRAG volunteers. In areas of low pond density, surveyors (including me) were probably happy just to have a pond to survey - regardless of whether it supported fish or not! If we assume that toads are more likely to occur in fish ponds, common toad occupancy should actually be (slightly) higher in areas of high pond density. 

[Edit: I should add that we do attempt to avoid this bias when undertaking survey work. The trend revealed in the above graph is only small and may not be statistically significant. However, it does suggest that we may need to be even more careful when collecting data in the future.]

Could the same issue affect the frog data? I don't believe so. Yet it is possible that there is a difference in the relative predation risk between fish and great crested newt. If newt predation is proportionally higher than fish predation, frogs may be more likely to occupy fish ponds in high density areas. That though, is an analysis for another day... 

Appeal to Save Historic Wildlife Reserve

A major appeal has been launched in a bid to raise £100,000 to protect a wildlife reserve.

The Kent Wildlife Trust this week called on the county to help fund a major restoration programme to ensure the continued success of Sevenoaks Wildlife Reserve – a site of special scientific interest – and a pioneering site in the history of wildlife conservation.

The trust’s reserves officer, Dave Hutton, said: “More than 50 years ago, this remarkable site made history when many local volunteers started to make it the first of its kind in Britain – a gravel and sand quarry transformed into a site for nature conservation.

“Today, it is hard to imagine that where there are now five beautiful lakes, reed beds, ponds and woodland, and home to more than 2,000 recorded species – was once a sterile industrial landscape.

“Since the trust took over the management of the site in 2005, continued progress has been made to improve habitats and facilities. 

“We now need to conduct a further programme of major restoration work, to build on conservation success already achieved.”

via www.kentnews.co.uk

I wonder how the trust will fund this one...

There are records of great crested newt, common toad, viviparous lizard, slow-worm and grass snake for the reserve.

[Testing the new 'Blog It' function that's available in TypePad btw] 

Kent Wildlife Conference 2009

The theme of this year's Kent Field Club organised conference is: 

Improving connectivity across our landscape: linking habitats and populations

Date: 17th October 2009

10.00 - 16.00

The Old Sessions Lecture Theatre, Canterbury Christ Church University College, Kent

The aim of this year’s conference is to look at how the interlinking of habitats and populations across the landscape influences our local wildlife. First of all, the flora and fauna need to be mapped on a site by site basis so we know where all the species are and the results made centrally available. This relies largely on the expertise and enthusiasm of the recording community. Further quantitative or semi-quantitative studies then provide us with additional information on where the largest, most viable populations occur and this is especially useful when developing plans for the conservation of rare or endangered species. The presence of suitable habitat on its own may not be sufficient for species persistence long term and close linkage (i.e. connectivity) to other high quality habitats is seen as increasingly important at a landscape scale in maintaining viable populations of many high profile species. Planning initiatives at local district, county and national levels have been implemented in an attempt to ameliorate the effects of habitat fragmentation and deterioration and several examples of these initiatives will be discussed during the conference. Counter-intuitively, motorways which normally exacerbate habitat fragmentation may act as corridors for selected groups of wildlife. 

Programme

9:30 - 10:00 Arrival and coffee

10:00 Welcome and Introduction. Richard Moyse (Kent Field Club)

Session 1

10:05 Wildlife and planning in Dover: the Green Infrastructure Approach. Nick Delaney (Dover District Council)

10:25 Mapping Great Crested Newt metapopulations across Kent. Lee Brady (Kent Reptile & Amphibian Group)

10:45 Linking marine intertidal communities. Ian Tittley (Natural History Museum, London)

11:05 - 11:30 Coffee and poster session

Session 2

11:30 Chiropteran Countryside commuters. Frank Greenaway (formerly, Natural History Museum, London)

11:50 Landscape planning and Habitat networks in the Canterbury district. Richard Andrews (Jacobs UK plc)

12:10 The Hitchhiker’s Guide to Kentish roadsides: botanical connectivity. Geoff Kitchener (Kent Field Club)

12:30 Connectivity across a woodland landscape: lessons from the management of Blean Woods and the Heath Fritillary butterfly.  Nick Bourn (Butterfly Conservation)

12:50 – 2:10 Lunch and poster session

2:10 Project and record requests. (All)

Session 3

2:30 The role of ponds in maintaining freshwater habitats at the landscape scale. Jeremy Biggs (Pond Conservation)

2:50 The role of Agri-environment schemes in restoring habitats and corridors. Dan Tuson (Natural England).

3:10 Reconnecting wildlife in Kent. Sue Young (Kent Wildlife Trust)

3:30 - 3:45 Discussion 

3:45 Tea and Close

The cost for the day will be £14.00 per person including all light refreshments and a buffet lunch (vegetarian dishes included). Please contact John Badmin, Kent Field Club, Coppice Place, Perry Wood, Selling, nr Faversham, Kent ME13 9RP (jbadmin@btinternet.com) for booking details and other information.

Predicting Species' Distribution Ranges: Kent Pond Data

In two previous articles (predicting range and survey effort) I have attempted to describe how I am starting to use recording data to more accurately model species' ranges. Ultimately, this will help us to better define 'Favourable Conservation Status'.  

Predictions of species' range need to consider several elements. Conceptually, the most important of these would appear to be 'recorded distribution'. If a species has been recorded at a specific location, it is reasonable to conclude that the location is situated within the species' range. The advent of the 'ecological consultancy' era, where animals found to occupy development sites are captured and relocated to sites that may be located many km away, may result in some suspect locations. Rest assured, suspect sites can be identified within the database and flagged as occurring 'outside of expected range'!

Although a species' range can be predicted using recorded distribution data, does that mean we expect the species to occur in all locations within its range? Indeed, does it mean we expect the species to only occur within its known range and not in other, perhaps unsurveyed locations elsewhere? Clearly, available habitat is an important factor. A site may be located well within the core range of a species, but if the site does not include any suitable habitat should we still expect the species to occur? 

Consider the photograph below that illustrates habitat in a site that is located within the range of adder and viviparous lizard. Should we expect these species to occur, even though the site consists of a highly managed, short grassland sward? 

 

Building habitat data into our range models is an important next step. 

Some time ago, I attempted to predict the number of great crested newt ponds that occur in Kent using the Kent habitat dataset. At that time, KCC had calculated that there were 41,000 ponds in Kent. However, we knew this was an overestimate due to limitations in the way pond data was identified. Phil Williams in the local Natural England team has filtered the pond data using an algorithm that removes linear features such as waterways and 'ghost polygons' around large features such as lakes. The filtered data includes 17,823 waterbodies that have been classified as ponds. Laura Wood at DICE very kindly analysed this data and extracted 10 figure grid references. This data has been imported into the KRAG database. 

For species such as great crested newt, I can now use pond density to refine range predictions that were previously based solely on recorded distribution. 

Great Crested Newts and Pond Density

In their now widely used method, Rob Oldham and colleagues found that pond density was an important factor in the development of a habitat suitability index. 

Reading from the graph, pond density ceases to be a limiting factor once it reaches 4 ponds per sq. km. This is the same as ~12 ponds located within 1 km of the target pond (12/pi = 3.8). I have therefore used 12 as a baseline figure from which to categorise low, moderate and high pond density:

Low  =     - 75% =     3 or less ponds located within 1 km

Moderate  =     - 75% to + 75%    =    4 to 20 ponds located within 1 km 

High =    + 75% =    21 or more ponds located within 1 km

Survey work undertaken by Calumma Ecological Services and KRAG volunteers at several hundred ponds in Southeast England has revealed a very close correlation between pond occupancy and pond density [edit: the relationship for species other than gcn is quite interesting and will be discussed in a future post]. Of the 252 surveyed ponds that were surveyed with sufficient effort to reliably determine the likely presence of great crested newt the relationship looks like this:  

 

As pond density increases, the rate of occupancy also increases. Note that other factors such as presence of fish, pond desiccation etc are also important in determining occupancy and these are considered in the habitat suitability index. 

Refining Great Crested Newt Range Predictions Using Pond Density

The KRAG database can now calculate the number of ponds that occur within any specified distance of a location. The risk assessment tool can therefore count the number of ponds that are located within 1 km of a search area and modify the likelihood of presence depending on whether the pond count is low or high:

High Pond Count 

Low Risk  --> Medium Risk

Medium Risk    --> High Risk

To ensure that likelihood of presence is not overstated in areas outside of the species' known range, the pond count is only applied for locations that are situated within the maximum expected range. 

Moderate Pond Count 

No change.

Low Pond Count 

Medium Risk  --> Low Risk

High Risk     --> Medium Risk

To ensure that likelihood of presence is not understated when a search is centred on an occupied pond that just happens to be located within an area of low pond density, the database also calculates the distance to the closest known species record. If the distance is within 250 m, low pond counts are ignored. 

A future blog post will consider how habitat data can be used to refine range maps (the process is a little different than the risk assessment method and will include habitat other than just ponds). 

Determining Survey Effort

This one's a little tricky. The risk assessment tool previously discussed is a useful way of predicting when a species is likely to be present at a particular site. But what happens when there is very limited survey effort in the area around the site of interest? How reliable is the risk assessment under those circumstances? If the risk assessment concludes that a species is unlikely to be present, could it simply be that nobody has actually looked in that area and no data is available?  

Clearly we need some estimation of survey effort so that we can put the results of the risk assessment in context. Now, here's the tricky bit. How do we define survey effort? 

I've thought long and hard about this and decided that one possible approach is to calculate relative survey effort. That is, how does survey effort around the site of interest compare to survey effort in the recording area as a whole. 

*** I define recording area as the total area covered by vice counties of interest. In Kent, the recording area is the combined areas covered by West Kent (vice county 16) and East Kent (Vice County 15). ***

One really important issue that needs addressing is the value of null observations. A site may be surveyed quite thoroughly (using appropriate methods at appropriate times under appropriate weather conditions etc), yet no observations of the target species are made. Such results are very important and must be considered when assessing survey effort. Null results should also be submitted to KRAG by recorders - please!

For any given search request (9 km block centred on the grid reference of interest), the database determines the number of 100 m squares that are represented by either a species record or a 'no observation' record and represents this as the number of surveyed squares per 1 km. The database then compares this value to the recording area average. Search requests in areas where the number of surveyed squares per km is about the same as that generated for the whole recording area are considered average. The score increases or decreases depending on how much the two values deviate from each other. The analysis is also undertaken separately for amphibians and reptiles to control for differences in effort between the two groups (and the different survey methods generally deployed). Of course, as more records are added to the system the county average changes (slowly). The assessment of survey effort is therefore dynamic and uses the most up to date information that is available at the time of the search request.  

In some recording areas (e.g. Lancashire), survey effort may be very low across the whole area. In these areas, relative survey effort in the area of interest may be high, yet effort may actually be quite low when compared to other recording areas (e.g. Kent). I'm looking at the database to see if this is likely to be a problem. At the moment I don't believe that it is, but I need to do further testing. 

Predicting Species Distribution Ranges

I've been hard at work on the recording database to help me better understand species distributions across Kent. Whilst simple dot distribution maps such as that for adder below are useful, they only hint at the true extent of a species' range. 

What we need is a map that illustrates not only a species' recorded distribution (as shown in the dots above), but also a species' predicted distribution. 

How do we predict a species' distribution? 

One approach is to consider the Pythagorean distance between actual observations and use this to predict the maximum distance between confirmed observations that is likely to be made. The technique borrows from the nearest neighbour analysis that is often used to determine whether observations are random or clumped. A guide to nearest neighbour is available here. 

Calculating Pythagorean distances relies on a simple mathematical formulae that relates the difference between eastings and northings between two points. 

I learnt a few things during development of these database routines. Rather significantly I discovered that computers are pretty naff at calculating square roots. That is, it takes a computer much longer to calculate a square root than a square (y1 X y1). Calculating Pythagorean distances relies on using the square root. For one off calculations that's fine, but those computing seconds start to add up into minutes and even hours when the database is analysing thousands of records. Remember, to find the nearest neighbour the database must check each observation against every other observation of that species. A nifty workaround is to only look at just the squares of the distances - and apply the square root to just the lowest value. Geeky stuff I suppose, but it does result in a faster database. If you want to read more, a couple of really useful summaries are available here and here. 

Rather than relying on just the maximum distance between observations, I have attempted to create range categories. These categories include the following:

core range - nearest neighbour distance within which 75% of observations occur

predicted range - nearest neighbour distance within which 95% of observations occur 

maximum expected range - distance from most isolated observation to nearest neighbour

The resulting analysis creates polygons than can be plotted on a map. I'm using Google Earth to illustrate range. 

Darker circles represent core range, lighter circles represent predicted range. If we add maximum expected range to the map, the result looks like this:

The maximum expected range of a species relies on calculating distance to outliers. For species with isolated records (particularly where this is a result of under recording), these outliers can result in a significantly exaggerated range and should be used with caution.  

A by-product of this process is that I have now also created a risk assessment tool for the database (similar to the one developed for Surrey ARG by Steve Langham). 

Using a simple interface, a user can type in a six figure grid reference and the database will reveal the likely risk of presence of each species at that location. The result for Canterbury Cathedral (TR 151 579) looks like this.

What the tool is telling us is that the cathedral lies within the core range of most of Kent's species! For sites with no waterbodies where there is a high risk of amphibians being present, terrestrial habitat within the site should be considered as potentially offering places for foraging, shelter and dispersal. This tool has now been built into the general database search enquiry that is regularly undertaken by KRAG on behalf of consultants and the like. I'm also working on making the tool available online to KRAG's conservation partners. The idea is that conservation based organisations can visit the KRAG website and log-in to the database. Users can then determine for themselves whether a site is likely to support different species. I believe that such an online tool would be tremendously useful for those organisations that undertake habitat management work as well as organisations (local authorities and the like) that review planning applications. 

Of course, predicting range from nearest neighbour relies on a good number of verified records that represent a reasonable distribution for the species. Lack of recording effort in some areas may give an under estimate of the likely presence of a species.  

Conversely, the likely presence of a species may be overstated when the site in question (although very close to an existing confirmed record) does not display any suitable habitat for the species in question. 

We therefore need to consider the extent of suitable habitat for each species across the county. Steve Langham has now accomplished this for Surrey and has been posting updates on his progress with the analysis on the RAUK forum.  

Thanks to Steve, I now have access to habitat data for Kent. In a future post, I will explain how this information will be used to estimate the extent of suitable habitat for Kent's herpetofauna.