- Locations where the species is known to occur and the model predicts that it will occur;
- Locations where the species has not been recorded and the model predicts that it will not be found;
- Locations where the species is known to occur but it is not predicted by the model; and
- Locations where the species has not been found but the model predicts its occurrence.
Wednesday, 9 March 2016
In homage to the recording 'patch'
This last weekend I was asked which type of recording I favoured? A patch or wide-ranging recording? Bearing in mind a recording scheme is often judged by geographic coverage, it may come as a surprise to some that I thoroughly endorsed the 'patch' approach. Why might this be? Well, the big advantage of the 'patch' and regular recording, is that one starts to build up a year-on-year picture of a the local environment, with sufficient data to start to pick up trends. That is what we have lacked and what is needed to help to build the contextual picture.
Before the days of cheap transport, field naturalists tended to have a 'patch' and took an interest in many aspects of the natural history of their chosen site. Unfortunately, data collection systems were confined to card indexes and notebooks; many of which have probably been lost. The lack of systems to capture data mean that on the whole we have only a sparse record of what occurred prior to the 1960s. For invertebrates, museum collections are a great source of information, but inevitably they tend to hold examples of rarer species: there is not the space to retain all specimens of commoner species, recorders also limited the extent of their collections. Doubtless there were those who placed a lot of emphasis on seeing rarer species, but the 'patch' was an important part of their biological recording activity.
Cheap transport has allowed many naturalists to travel much further afield in pursuit of their chosen interest. In Entomology it is commonplace for field meetings to be organised at the richest sites or places where rarities might be found. I was no different to others in the early 1980s when one would turn up at a given locality for X or Y and find that others had arrived with the same intentions. Our collections were very similar and this took mothing into the realms of stamp collecting; hardly an ideal foundation for biological recording . I suppose that in a way we were the forerunners of today's moth twitchers who visit other moth traps to see the poor little soul that has flown off course, ended up in a moth trap and then in a pot in a fridge. What does this contribute to our knowledge of species' biology and conservation status? Precious little!
Thank goodness there was the Rothampstead Insect Survey, which established a network of locations where all moths were recorded. Although greatly trimmed in its coverage, this dataset is perhaps THE most important for invertebrate conservation. Its findings convey a sad story of decline in Britain's moth fauna. Efforts by the National Moth Recording Scheme (NMRS) to assemble the data from individual moth traps compliment the Rothampstead survey, but will always lack the benefits of the consistent approach that the Rothampstead survey has established. Some NMRS records will be from mobile traps that are run in more interesting places, but many will be people's gardens - the equivalent of the 'patch'.
Data for butterfly transects are also pretty robust and it is possible to say quite a lot about what has happened to them over the past 40 years. Lots of organisations and individuals run butterfly transects and these provide a hugely important resource that helps to chart the fortunes of Britain's butterflies.
Long-term monitoring of other taxa fares less well. There are comparatively fewer enthusiasts, and although there are recording schemes, the volumes of data for more challenging taxa are relatively sparse. This is arguably a serious gap in our knowledge because we cannot automatically extrapolate from the fortunes of moths and butterflies. The Lepidoptera occupy a relatively narrow niche: the vast majority are phytophages, whereas other Orders occupy much wider ranges of niches. Guilds of saprophages, mycophages, filter-feeders, parasites and predators are represented across the insect World beyond the Lepidoptera, and although we do have data they are nowhere near as robust. Nevertheless, it is still possible to detect trends; many of which mirror those for the Lepidoptera - perhaps 40-50% declining, 30-40% stable and 10-20% increasing.
For hoverflies we now have a run of about 35 years data in which more than 20,000 records have been submitted each year. This dataset is hugely important because hoverflies occupy a very broad range of niches (apart from parasitism) and can be used to convey important stories about the state of the wider countryside. The dataset is one of the bigger invertebrate datasets and it is possible to extract trends from the data for some but not all species (most notably lacking are the rarest species). The results are not pretty, paralleling those for many other taxa with about 40% of species declining significantly. Unfortunately, we don't have the long runs of data for individual sites like those of the butterfly or moth monitoring schemes.
How do we make something of the data?
There are various analytical techniques that can be used to examine the degree to which data are fully representative of a given area. Several work in similar ways, comparing the presence of given species against a basket of commoner species to determine the intensity of recording, and thus the likelihood that a species' absence is a function of recorder effort rather than being an actual absence. The system develops a 'matrix of confusion' which is somewhat Rumsfeldian in its outputs:
The 'matrix of confusion is used to test the reliability of predictive models and is highly dependent upon the availability of records for both commoner and rarer species. A measure of recorder effort can be determined from the proportion of commoner species that might be expected to occur that have actually been recorded in a given square and in surrounding squares.
Where species X is present in a given square and say only 40% of the commoner species that are likely to be constants in full species lists are present, this represents a very positive result. If it is absent, there remains the possibility that it will occur, assuming the location has the appropriate environmental parameters. Conversely if species X is absent from a square where all 50 of the commonest species have been recorded, it may be inferred that the level of recording has probably been sufficient to generate a reasonably comprehensive picture of the fauna of that square.
If only rarer species are recorded, the analytical outcome is likely to be a little distorted. Conversely, if species X is challenging to identify, it is highly likely to be under-represented in data that is dominated by records of easily identified species. The more comprehensive the dataset, the better the outcome.
Making something of trends on ad-hoc data is always going to be more difficult than an analysis of data collected using a constant system of survey such as butterfly transects, water traps, pitfall traps or a Malaise trap. But, regular repeat recording from a given 'patch' can result in remarkably valuable data. For example, Alan Stubbs has undertaken regular monitoring of his Peterborough garden. In the 25 years or so that he has done this, it has shown that overall species richness has dropped quite substantially. Stuart Ball uses Alan's data and the data generated by Jenny Owen's Malaise trap in Leicester as the only really detailed data that help to provide a picture of environmental change for hoverflies. Jenny stopped running her trap some while ago (and died recently), so that data run has come to an end. Alan is still monitoring his garden, but of course it is just one dataset. What we need is more people who are willing to regularly record from their garden or a favoured patch.
In the last year, several members of the UK Hoverflies Facebook page have been very active, regularly recording from their 'patch'. This is a welcome boost to the data and if it continues it will start to form the foundations for a pretty decent long-term monitoring project. It parallels the general concept of the garden hoverfly monitoring scheme that we have been trying to develop. Obviously, some people will drop out in time, but others may be recruited. The development of a big dataset from several constant effort sites should provide a really good foundation for the future.
So, for those who like to have a 'patch' do please make the effort to regularly record the hovers. Those who favour more wide-ranging activities - yes please, and don't forget the full species list. The critical issue is the full list of what is seen and can be reliably identified. Hopefully we can develop a system like Birdtrack, which is a really neat way of using 'ad hoc' data to demonstrate aspects of species' biology and to investigate trends at a national or regional level. And, if you do want to twitch special places, do please make sure the records reach the scheme - it is amazing how few records we see for New Forest specialities such as Caliprobola speciosa!