Friday, 4 September 2015

A rationale for caution in photographic identification

There are several schools of thought about the use of photography as a means of biological recording:

1. Those who believe that a record can only be believed if it is supported by a preserved specimen identified by a competent taxonomist.

2. Those who believe that photography can replace traditional methods (identifications produced from preserved specimens, following keys based on museum specimens).

3. Those who believe that photography can be used to improve levels of recording, within definable limits.

There is no 'right' or 'wrong' answer, just a variety of views that reflect a combination of philosophical approaches and practical experience.

On balance it is difficult to see how the strictly taxonomic approach (Group 1) can be followed to ensure coverage of even a small country such as the UK. Professional taxonomists are rare and nobody is going to fund an increase in their numbers; if anything the emphasis is on reducing numbers still further. In the World of Syrphidae in the UK, there is nobody that I can think of that is now paid solely to work on hoverflies, although several University pollinator teams do employ people to identify specimens that are often checked by volunteer taxonomic specialists (I have done a lot of this sort of thing over the years). It is important to stress, however, that hoverflies are unusual: in the UK at least, they include significant numbers that are highly characteristic and that can be identified from one or more angles if the photograph is sharp, of good resolution and without colour bias. The same cannot be said for many taxa where microscopic characters and dissection are critical to a firm identification of possibly already tiny animals.

Conversely, those who think that photography is the answer will find that their views are not supported by experience. Some erect new characters based soley on photographs that are untested against firmly identified specimens (i.e. checked against the original type material or deseignated lectotypes or paratypes). In these cases they will find that they disagree with specialists used to examining specimens. A few move towards a half-way house where they use both photography and take specimens. On several occasions people who have done this have commented how much they subsequently realised that many species could not be done from photographs (and that once they used a microscope they found a much wider range of species).

There is then the issue of a growing demand for data by Government and its agencies. The ongoing theme has been 'how to increase volumes of data', and 'how to increase the numbers of people actively involved in biological recording'. That means that there have to be compromises. One compromise is to accept that the Group 1 (taxonomic approach) cannot be relied upon because it is too costly and socially exclusive. A second compromise is to establish mechanisms where existing Group 1 recorders put their effort into assisting group 2 recorders (with a corresponding reduction in the numbers of full datasets generated). This approach relies upon a cohort of recorders who are prepared to engage and to provide assistance within definable limits. It is the approach that has been embraced by the Hoverfly Recording Scheme.

There are also various groupings of biological recorders:

1. Those for whom an encounter with a particular animal or plant is strictly a personal issue with no real interest in what such records might be used for in a broader societal context. It may include the development of a personal list or even a competitive list amongst peer groups (some 'pan listers'). In these cases, the recorder sets their own confidence limits for the reliability of an identification.

2. Those for whom pleasure is gained from seeing a particular animal or plant, who want to be sure that the identification is as reliable as possible. This may include people who subsequently submit data to local or national recording schemes, but the principal driver is a personal apprreciation of the natural World.

3. People with a firm interest in a particular taxonomic group, who include the production of valid records as part of their fundamental interests.

4. Specialists whose interest lies in the development of datasets that can be used to establish a better understanding of the ecology of a particular group of plants or animals. At the simplest end this may be to improve understanding of biogeography, but there are many other applications for such data, including trend analysis that may be influential in wildlife policy promoted by NGOs or by various tiers of Government.

5. Taxonomists whose primary interest lies in refining the catalogue of plant and animal life. Taxonomic interests may not exclude a parallel interest in species' ecology, but there is usually a strong emphasis on refining identifications to the finest level of resolution.

Each view has its merits and disadvantages but it is perhaps worth teasing out the critical issues from the viewpoint of one person running a national recording scheme.

I have spent the majority of my life working in one (several) of the bodies responsible for delivering UK wildlife policy and governance. This has included a substantial time as a policy-adviser and as the front face of engagement with promoters of major development projects. I still bear some of the scars and am acutely aware of the ways in which data can be undermined to lessen the confidence in the science underpinning conservation policy and delivery. One need only reflect on the ways in which climate change sceptics represent (or perhaps misrepresent) data in order to weaken confidence in scientific predictions. The same situation obtains for outputs generated from biological recording. If there are weaknesses in the data then these can be magnified to undermine the veracity of any political messages that can be derived from trends.

At the moment, the trend data indicate that there have been dramatic declines in the abundance and richness of wildlife in some parts of Britain, and that there have been important shifts in species distribution. For me, these trends largely correspond to what I have seen in terms of hoverfly abundance over the past 30 years. I may be recalling past times with rose-tinted spectacles, but I simply cannot believe that I am remembering a mirage when I now find great stands of hogweed (mainly in SE Engalnd) with barely a hoverfly in attendance! Something has definitely happened and it is really a question of trying to work out the factors responsible for this change. If it has happened to hoverflies then the chances are it has also happened to organisms further up the food chain such as birds. The numbers of birds on red or amber on alert lists tell the story very clearly, but to understand what is happening we need to understand what is happening lower down the food chain – so the fate of predaceous, phytophagous and saprophagous invertebrates is of considerable importance.

A significant part of my interest in running the Hoverfly Recording Scheme lies in developing data that are robust enough to provide a reliable picture of what is happeneing to hoverflies (and by extension other wildlife). The robustness of the data has a very important bearing on the veracity of the messages imparted. If it can be shown that there has been rigour in the data collection process then the outputs of trend analysis are much more likely to be believed. If, on the other hand, it is clear that unreliable characters have been relied upon and determinations can be readily challenged, then the outputs will be weakened.

These are some of the problems that all recording schemes grapple with. One  wants to improve coverage and it is always good to see an increase in incoming records and interest in the taxonomic group. The question is whether this can be achieved without significantly reducing the robustness of the dataset. Setting practical limits on what can and cannot be identified is part of that process. Developing new analysis tools is a further component. Arguably, there is a developing need to be able to split datasets into those elements derived from photographs and those from traditional field work and retention of specimens for microscopic identification. Both have the potential to tell an important story.

For me, the important point about photographs is that they largely involve those species that are most visible to relative novices and can be used to generate much bigger datasets. Yes, there may be identication problems for a few, but the volume of data means that outputs are reliable within defined limits - possible at the level of particular guilds, possibly at generic level or perhaps to species in some cases. The big issue is to define the levels of robustness and to establish the analytical processes needed to achieve this. Within this dataset there are potentially useful indicators and it is really a matter of defining what these are and how best to generate useable data.

It is also clear that interactive forums that encourage photographic recording can substantially increase the numbers of major contribtors to a recording scheme. In the case of hoverflies, over the period from 1976 to 2009/10 a pool of 21 people supplied nearly 50% of the data (out of an average of about 22,000 records per year). In the following 6 years, that proportion has changed - several recorders have reduced their activity or have ceased to be active at all (a very aging group who have often been contributing for 35+ years).

Meanwhile, a significant group of photographic recorders has evolved. So far, just over 13,500 photographic records (to species) have been generated in 2015. Around 40% of these were provided by 24 people and 55% by around 70 people. This new paradigm suggests that there is considerable scope for growth in the dataset and in the constituency of contributors. Unlike other components of the dataset, we do know that these data have been supported by photographic voucher specimens and as such this starts to form a very robust sub-set of the database. It has its limitations, but so too do those data submitted by people who retain specimens.

The supplementary issue is how to establish a long-term philosophical approach that is robust enough to move with the times and yet retain the confidence of those who use the outputs (NGOs and policy makers). In line with this, there is the issue of succession management - it is all very well having a recording scheme, but what happens when the scheme organiser decides that they have run their course? I am firmly of the view that one should hand over the reins in good time to allow new blood to take a lead.

My intention is to start to pull away over the coming two to three years and to generate a new team to take over those parts of the scheme that I lead on. That means recruiting new people to extract data, and possibly additional technical identification skills. I have my eyes open for potential recruits!


  1. A nice exposition of the groups and issues involved. One idea that could help is to be able to add confidence level to records. None of the software I've used or the schemes, surveys and organisations to which I submit data, allow this.

    I've added a status code of U-Uncertain to my Wildlife Recorder databases (these don't cover hoverflies, but apply to birds, dragonflies, mammals, butterflies and moths, dragonflies, reptiles/amphibia and plants). Even such a binary flag is useful - but only to me if the destinations of the data don't use it.

    So in many cases, I make up my own rules. I'll illustrate with birds, which I have recorded since the 1970s. Don't tell the BTO, but when you conduct a Breeding Bird Survey, sometime you hear only a very brief call, or get only a glimpse of a bird. I generally record it on the survey if I am 80% sure or more, and the species is found on the survey site. (eg The flap of a wood pigeon's wings on take off from a tree is pretty distinctive, but if not seen it could just possibly be something else). If I am less sure or it is unusual on the site, I simply keep it in my own records, usually with an appropriate comment as they will reach local and national records.. But that's just me - every recorder must make some such judgements some of the time.

    The only two times I reported birds regarded as locally rare, my reports were rejected although in both cases I was absolutely sure as they were very distinctive species, and in one case had corroboration in a nearby county later that day. The regional rarities committees didn't approve my descriptions, or maybe trust me, so those records are basically lost. If they had included them as 'Likely but unproven' they could be telling us something.

    As Roger says, the 'strict' recorders will omit anything uncertain altogether. But it can be useful indication of something changing or previously unobserved. If systems, over time, could be adjusted to allow the degree of confidence - or just Certain or Uncertain - to be recorded, particular analyses can choose whether to include less than 100% or not. In reality, many inexperienced recorders in a particular group will record sightings, which therefore appear 'certain, many things which are incorrect - it would be better to allow openness.

    A similar idea is to allow more recording at family level e.g. Syrphus sp. This is not helpful for individual species - of course - but can give better indications of overall numbers and distribution of such groups.

    There are many challenges in non-professional biological recording, which I regard as being still in its infancy in many ways; this is just one.

  2. A very thoughtful and considered post. I would consider myself probably in category two of the biological recorders (observing for pleasure but wanting the names to be accurate), but also with elements of category three (once I am satisfied that I have accurately identified something then I want to contribute that record). I also have a local area that I put more effort into actively recording than I do if simply out in a new area.

    What I mainly wanted to say was how useful I have found the Introduction to British Hoverflies WildGuide. In particular the categorisation of species into identifiable in the field/require a closer look or specific features to be observed/requires microscopic examination have been very helpful. It has allowed me to have more confidence in identifying distinctive species and prevented me from assigning trickier ones to species level with false confidence. Also by researching key features at home, I have been able to take photographs from the right angle or of the essential areas to split species pairs.

    If observers (and scheme administrators) have a clear idea of what can and cannot be identified solely from photographs, then that will go some way to helping with the integrity of the data. I hope that the approach used in the WildGuide can be adopted by authors of guides to other taxa in place of the disheartening phrase "many similar species"!