Here's where we start to get to the nitty gritty of the Woking Pilot project. How many ponds were actually sampled to inform the distribution modelling? Was it enough to support the conclusions made in the consultation document?
The document published by Woking BC is intended to be read and understood by local residents and is therefore not scientific in its nature. And I certainly don't have a problem with that. Creating documents that are accessible to as many interested parties as possible should be encouraged. An attempt has been made to include Annexes that provide more detail and these include useful summary points. The problem with summaries though is that they can be misleading. The conclusions drawn may not be justified by available evidence. Without access to all data it may simply not be possible to undertake a detailed review. I won't let that stop me trying though.
Data on gcn distribution in Woking appears to have been collated from information held by Surrey Amphibian and Reptile Group, Surrey Wildlife Trust and directly sourced from an eDNA and HSI study funded by the pilot project. Let's dig a little and pull out some numbers.
Total number of ponds in Woking - not specified
Number of ponds assessed for HSI = 100 (pg. 20)
Number of ponds assessed for eDNA = (50 stated on pg. 20; 48 stated on pg. 33)
Number of ponds assessed for both HSI & eDNA = 36 (pg. 33)
Number of ponds with positive eDNA results = 5 (pg. 33)
Number of additional ponds with verified gcn records = 8 (pg. 33)
These are very (very!) small sample sizes. We don't even know what proportion of the available ponds they represent. I'm also confused as to why HSI assessments were not taken for all of the ponds sampled for eDNA (only 36 out of 50/48)?
There's also no confirmation that the HSI and eDNA data was collected following all appropriate guidelines. NE have quite rigorous requirements for eDNA sample collection. Were these guidelines actually followed?
The author(s) have attempted to correlate the results of the 36 ponds assessed for both HSI and eDNA with landscape level habitat variables derived from Land Cover Map 2000 (pg. 33):
"The results showed a positive correlation with the extent of urban areas within 250m (or a negative correlation with distance from urban areas) and a negative correlation with the extent of arable land. This suggests that urban areas may provide a greater diversity of habitats required by GCNs than arable areas. They also showed a weak negative correlation with the extent of woodland and conversely with distance to heathland."
That's quite a bold statement for a study based on just 36 ponds, only 5 of which were found to be occupied by gcn (assuming these 5 were all in the group for which HSI data was collected).
My own work investigating the relationship between gcn and landscape habitat variables does not support these conclusions. In Kent, gcn are negatively associated with urban areas and appear somewhat neutral with respect to woodland. The Kent dataset is based on a sample of several hundreds of occupied ponds across the whole county. Since the Woking project did not assess ponds across the whole of Surrey, the relationship between gcn presence and landscape habitat variables is highly biased since it is significantly influenced by the extent of each habitat category present in the borough. Looking at the Woking BC map, it appears that much of the borough is dominated by an urban landscape.
Results have been used to identify three metapopulations (Fig. 1, above) and these are described as follows:
Littleneck - 'a small number of ponds' .. 'largely in good condition'
Hook Heath - 'only one or two ponds'
Westfield - 'a larger number of ponds' ... 'in poorer condition'
These descriptions really are a bit woolly! Are these occupied ponds or all ponds? What does 'a small number of ponds' mean? Three ponds? Four ponds? At Hook Heath there were 'one or two ponds'. Which was it? Surely the authors know?! What does 'larger number of ponds' mean? Five or twenty five? Come on this wouldn't be acceptable for an undergraduate project, let alone a project attempting to define the conservation of gcn in a whole borough for the next 25 years!
In Woking, the metapopulaton with the highest number of occupied ponds can be inferred by reviewing the zones identified in Fig. 2 (pg. 23 of the report, shown above). Westfield, located to the south, appears to have the most occupied ponds and also happens to be the most rural. You'll also notice that few (if any) of the populations occur wholly within urban areas.
Although the apparent quality of individual gcn occupied ponds may well be better in locations situated closer to urban areas (especially when subject to positive management work), such ponds can be at greater risk from direct disturbance, introduction of fish and increased fragmentation of surrounding terrestrial habitat.
The interaction between gcn pond occupancy, pond quality and urbanisation has not been fully investigated within the report and such work is important if we are to more fully understand the Future Prospects of gcn in Woking. For example, if we assume a similar number of ponds in each of the three metapopulations, why was occupancy so low in two of the areas? Could it have been that the overall number of suitable ponds is lower in more urban areas? The HSI assessment does not appear to consider the quality of unoccupied ponds in each of the three metapopulations.
In fact I'll go a step further and propose a different interpretation of the data than that presented in the consultation document. Look again at Fig. 2 on pg. 23. The apparent relationship between gcn occupied ponds and urban areas has nothing to do with the alleged greater range of habitats that these areas may provide. Newts aren't colonising ponds on the edge of urban areas from the surrounding countryside. Urban areas in Woking are expanding and encroaching on existing newt habitat. Previously occupied ponds located within urban areas have been lost. Once urban areas expand to envelop the few gcn ponds that remain they will also be lost. If anything, the study presented here is an indictment of development and could be used to argue that in order to achieve a Favourable Conservation Status for gcn, there should be no more development projects in Woking! I'm fairly sure that's not what the council intended when they published the report for all to read. But I digress...
The results for Maxent modelling reveal the importance of larger sample sizes.
"Validating the results against the eDNA results, i.e. the 5 positive records used for training with the 43 negative results suggests that the model is providing a good fit to the training data and absence data GCN results."
Remember, the pilot project was using just 5 positive records to inform a distribution model for the whole borough. Is it really a surprise that using the same training data that was used to develop the model produced a good fit? What happens when independent data is tested?
"The SARG records contain eight validated records with 1m grid references within the Woking boundary. Testing these against the Maxent model trained with the eDNA positive records produces a poor fit."
Which contradicts an earlier statement made in one of the summaries (pg. 20):
"There is a strong coincidence between the predictions from the modelling and the field records"
The authors do appear to have subsequently incorporated the SARG records into a Maxent model:
"Training and testing the model using the available positive field records and the negative eDNA records shows an improved fit to the training data combined with the absence data."
It's quite revealing that the author(s) do not claim there to be a good fit using the SARG records, only an improved fit. An improvement on poor may well still be poor. Either way, it's something of a stretch to suggest a strong coincidence.
In conclusion, I believe that the sample sizes used in this pilot project are simply too small. A far more rigorous survey programme is required. The fact that there are (pg. 20):
"sparse, sporadic records elsewhere in the Borough."
Suggests to me that even in a limited areas such as Woking, sampling effort hasn't been sufficiently high to reliably identify all possible metapopulations. In a blog article such as this, I'm loathe to start referencing too much published research. However, there is one conclusion from a 2008 study on the performance of distribution models that I will quote:
"No algorithm predicted consistently well with small sample size (n < 30) and this should encourage highly conservative use of predictions based on small sample size and restrict their use to exploratory modelling."
M. S. Wisz, R. J. Hijmans, J. Li, A. T. Peterson, C. H. Graham, A. Guisan and NCEAS Predicting Species Distributions Working Group (2008) Effects of sample size on the performance of species distribution models. Diversity and Distributions, (Diversity Distrib.) (2008) 14, 763–773
