Thursday 14 January 2016

What happens to our photographs when we shuffle off this mortal coil?

Last year I posted a reflection on the need for an archive of photographs. As far as I know, there has been no progress on this issue and it is very difficult to see how best to generate a debate that leads to action.

Perhaps there is no real need for huge numbers of photos of common species; especially if they are not accompanied by details of where and when taken? But, photographs accompanied by a record of where and when they were taken start to become a very important biological archive. Some may be instantly identifiable but others will not. Do we really understand the true potential of photographic recording? I think not, apart from a few noteworthy exceptions - the data assembled by iRecord, iSpot and by some recording schemes that have committed to photo identification as part of their repertoire.

Amongst dipterists I detect two greatly differing camps. One is broadly receptive of the role that photography can play in generating both interest and data. Others cannot see the point. It is not uncommon to hear the comment that so little can be identified that it is not worth bothering. I agree that in general some families will be far too difficult to do from photos, but then there remains the taxonomic conundrum - should we be wedded to the idea that the only way animals (or plants) can be identified is from specimens? What I have learned from hoverflies is that many animals have subtle but important differences in life that one cannot seen in a preserved specimen (and vice-versa). Learning to explain why a photo is or is not a particular species/family is an immensely useful process as it helps the taxonomist understand how to convey information that might be sufficient to get people off the ground.

I therefore think there is an urgent need for the community of recording scheme organisers to develop a wider understanding of what can and cannot be done from photos. I have been trying to do so for hoverflies for the past six or seven years and am now pretty clear about its potential (for hovers). It is considerable but there are limitations that make traditional collection of specimens a continuing necessity. What I also start to realise is that we might be able to develop keys that help the photographic recording community improve their accuracy. I have some way to go before trying to put these ideas into action, but I do wish I had been armed with that information before we wrote the WILDGuide.

As with all taxonomic revision the important issue is to have access to lots of material - and that is what photographers are generating. These photos are the equivalent of voucher specimens but are possibly easier to house (at least in terms of space). Unlike museum specimens Anthrenus will not be a problem, but I expect long-term issues such as backward file compatibility and maybe file corruption may strike.


I am in no position to do much more than to champion the potential merits of more recording schemes working with photographic recorders to improve our knowledge of what really can and cannot be done (assuming the photos are sharp and of good resolution). I believe we need to be thinking in a far broader way about photographic recording. It is not just a question of getting more people involved in taking photographs. iSpot, iRecord and various Facebook pages have shown how to generate interest. The bigger issue is how to retain and grow that interest in conjunction with growing skills amongst both photographers. I know of recorders who would engage, but who get disillusioned because there is a lack of engagement - that is a huge pity because there is an untapped constituency of support to engage with.

In this respect we need a much wider spectrum of specialists to engage with photographers. We also need to look at photographs in the same way as many of us would for traditional specimen-based recording.  When I start to deal with an unfamiliar family I store up specimens until I can start to put pieces of the jigsaw together. The same holds for photographs; hence the need to develop archival systems that can be used by coming generations of taxonomists. Perhaps there would be a need to restrict storage to a group of invited photographers? Not all photos one sees are necessarily of a standard that might be useful in such studies, but I can think of the work of maybe 30 photographers that is quite exceptional and ought to be recognised by creating such an archive. I am also aware of several who wonder what will happen to their photos and how to deal with this legacy. Perhaps it is time for an initiative to take this aspect of biological recording and taxonomy into the 21st Century?


Sunday 10 January 2016

Misidentification - not a new phenomenon



Earlier today, I posted a response on the UK Hoverflies Facebook page concerning the factors that affect data reliability. For some unknown reason it disappeared. Rather than post again in that medium I thought it best to tease out the issues in a more stable location where it can be reached by a wider audience.

The issue raised was the degree to which data reaching the Hoverfly Recording Scheme were reliable. These last couple of days I have worked my way through the maps and phenology histograms of the revision of the latest hoverfly atlas. That reality check revealed:

·         61 species in which there were at least one, and often several records well outside what might be considered a reliable date range. Often records occurred far too early in the year or far too late in the year.
·         24 species where there was at least one data point that needed to be questioned for its location - often much further north than its known distribution, but in a few cases too southerly or easterly.

The breakdown of issues is listed below against genus:
Genus
Phenology
Distribution
Other
Notes
Cheilosia
17
5
1

Chrysogaster
1



Chrysotoxum
3
2


Criorhina
3



Dasysyrphus
2



Epistrophe
3



Eumerus
2
1


Helophilus
1
1
1

Heringia

1


Mallota
1


Possibly larval records
Melangyna
4



Melanogaster
1



Meligramma
1
1


Microdon
2
1

Possibly larval records
Myolepta
1


Possibly larval records
Neoascia
1



Orthonevra
2



Paragus

1


Parasyrphus
2
1


Parhelophilus
1



Pipizella

2


Platycheirus
5
4


Pocota
1



Portevinia
1



Rhingia

1


Scaeva

1


Trichopsomyia

1

Single record queried
Sericomyia
2



Tropidia
1


Possibly larval records
Volucella
1



Xylota
2
1



It is possible that some of the phenology problems relate to larval records - where the recorder has not noted the stage of the specimen recorded. This is something that we might be able to resolve by checking back with the authors. It will resolve some of the problems but by no means all of them.

A highly likely second reason is that instead of running specimens through full keys and checking against details of phenology and biology, a record has been made by 'matching' to a picture in a guide book. This problem is often obvious in datasets that I check manually - lots of records of Platycheirus immarginatus from inland sites is a bit of a give-away! Elsewhere, mistakes are possible by experienced people using keys. For example I have seen several specimens labelled as Melangyna quadrimaculata that turn out to be Leucozona laternaria.

The problem of picture matching ID is not something confined to recent years - I'm certain it has happened ever since the initial publication of Stubbs & Falk. It is just that as the data become more substantial the problems also start to stand out. We have resolved the problem of Platycheirus immarginatus by mapping records attributed strictly to males from known reliable recorders - it is an entirely coastal species and we suspect some misidentifications involve Platycheirus perpallidus in places such as the Spey Valley. I think we may also find that the distribution of Platycheirus nielseni will prove to be strictly northern when we map males of this species.

The question was posed: was it that modern data are much better and show up the older data (paraphrased).

This is very complicated - the short answer is no, but that is simplistic. Way back when recorders relied on Verrall or Coe's keys they had few illustrations, and possibly little access to reliable museum specimens of difficult taxa. I'm certain there are lots of dodgy records from 100 years ago and probably to more recent dates from this source. Some arise because of more recent species splits (e.g. some Sphaerophoria). Since the publication of Stubbs & Falk we have seen more people getting interested in recording but have also come across a fair amount of picture-matching. There is a classic mistake in Chinnery that makes it possible to identify this source as the guide book!

So access to a good key does not necessarily mean that records will be any more reliable. It is amazing how many datasets I see where species listed include montane species from coastal grasslands and coastal species from inland woods! I can recognise these but data in-putters at a local records centre will not - so data we get from LRCs can be pretty messy.

What has changed is that we now get to see an awful lot of photographic records and start to see the sorts of regular mistakes that are made when picture-based recording happens. The photographic data extracted in recent years are likely to be as reliable as any - not least because they get scrutinised by three specialists who will also correct each other's oversights (these things happen!).

What is becoming increasingly clear is that most hoverflies have quite distinct flight periods that will vary by a couple of weeks between north and south, and perhaps a few days east to west along the same latitude.  A definable flight envelope can be developed that takes this into account for different latitudes and records that fall outside these parameters  can be flagged for further investigation. It won't solve all problems, but it helps. These are some of the 'rules' that are being developed for 'Record Cleaner' and as time passes they will get refined. BUT we must not deceive ourselves that there will always be a small number of mistakes in the dataset. The bigger the dataset the less important the mistakes are.

Changes in geographic range are more problematic, but extreme outliers can often be picked up and checked - some prove to be valid and a surprising addition to the knowledge. Others prove to be erroneous. One must always retain an open mind as surprises do happen - e.g. the recent range expansion of Callicera rufa. Nevertheless, an odd location reported by a recorder of unknown expertise must be investigated and treated with caution. Dubious records are marked as 'requires verification' and do not show in the maps but remain on the database.

So to return to the original question: 'Could it be also that more volunteers are recording giving better data that perhaps wasn't there with less records'? I am inclined to say that there has probably been little change in the proportion of dodgy records over the last 100 years. The reasons for mistakes have almost certainly changed but one cannot escape the problem that people take short cuts and have to learn their craft. Those who never make mistakes are few and far between!

To my way of thinking the important point about developing sound biological recording is to encourage adoption of a set of straightforward rules:

            i.            When you start recording the chances are that what you encounter will be the commonest species - finding rarer species often involves specialised field craft. So when you start beware if your first record is a stonking rarity! It does happen but not that often.
           ii.            Check your records against distribution maps and phenology histograms - be self-critical and get records that lie outside these parameters double checked. If the guide says it is a Scottish species and you are in Wiltshire, maybe that is a pointer to a misidentification. If in doubt, leave it out!
         iii.            Learn to understand what can and cannot be done from picture association - and start to read the keys - knowing the keys will help you to understand why a particular animal is classified to a particular place in the taxonomic order.
         iv.            Develop field craft - try to understand the ecology of your target group.
          v.            When travelling to far-flung parts of the country, remember that there are possible additions to what you normally see and also species that you won't see.
         vi.            If using photography then keep a photo voucher - it can help to photograph specimens securely held from several angles.
       vii.            Consider retaining specimens of some more difficult groups - even if to pass them on to a specialist to identify - this is probably the best change you have of securing long and reliable lists for your favoured site.
      viii.            Keep comprehensive records - they will help recording schemes but may also help you to better understand your chosen subject.