Methods for recording butterfly transects
Fixed-route transects (Pollard Walks):
These transects form the core of the UKBMS with over 1,200 different transects walked in the first 30 years of the scheme (1976-2005). New transect sites are added annually with well over a thousand now recorded annually. Named after Ernie Pollard who helped design and pilot the methodology with colleagues, these transect walks provide the highest quality data within the UKBMS as they are walked weekly during a 26-week period between 1st April and 29th September each year. This allows calculation of robust measurements of changes in butterfly populations for over 50 species and production of site level trends when sufficient years of data have been collected. Additionally, it allows the generation of flight periods from which data from reduced effort surveys, or transect sites with incomplete annual walks, can be compared and subsequently utilised when on their own they cannot. However, this methodology is labour intensive and not the most suitable methodology for monitoring a small number of species that are either difficult to detect on transects, or inhabit remote habitats where weekly visits are not feasible. Thus other methods are also used in conjunction. More details of fixed-route transects follows, whilst detailed instructions for setting up and recording a transect are available here.
The methodology and development of transect monitoring for butterflies has been reviewed in detail elsewhere (Pollard and Yates, 1993). The predominant type of monitoring employed is the ‘Pollard Walk’. In brief, this is a fixed-route walk (transect) which is established at a site on which butterflies are recorded along the route on a regular (weekly) basis under reasonable weather conditions for a number of years. Transect routes are chosen to sample evenly the habitat types and management activity on sites. Care is taken in choosing a transect route as it must then remain fixed to enable butterfly sightings to be compared from year to year. Transects are typically about 2-4km long, taking between 45 minutes and two hours to walk, and are divided into sections corresponding to different habitat or management units.
Butterflies are recorded in a fixed width band (typically 5m wide) along the transect each week from the beginning of April until the end of September yielding, ideally, 26 counts per year. Transect walks are undertaken between 10.45am and 3.45pm and only when weather conditions are suitable for butterfly activity. Weather conditions have a considerable effect on the numbers of butterflies seen. To ensure that the counts are standardised as much as possible:-
- Transects are not walked when the temperature is below 13°C (in northern upland areas this may be reduced to 11°C)
- Between 13-17°C, a transect may be walked providing there is at least 60% sun
- Above 17°C, a transect may be walked in any conditions, providing it is not actually raining
- When wind speeds are above 5 on the Beaufort scale, transects should not be walked
Due to the vagaries of the British and Irish weather, it is rare in practice to achieve a full set of 26 weekly counts. However, a small number of missing values can be estimated using other counts during the season.
Reduced effort surveys:
Single species (as opposed to normal 'all species') transects have been increasingly established in recent years. Whilst such transects must follow the standard methodology and must record populations at least once a week throughout the flight period, the focus on a single (or small number of) species reduces both the time required to walk each transect and, more significantly, the number of weekly counts. With many demands on the time of site management staff and volunteer recorders, this reduced method has enabled population monitoring of particular threatened butterflies to be undertaken when otherwise it would not have been possible. By regularly recording a fixed route in standardised conditions, the number of butterflies seen on a transect can be compared from year to year.
For a number of habitat specialist species (especially the fritillaries) other ‘reduced effort’ methods are also used to monitor annual abundance at the site level, especially in more remote parts of the UK. These include adult timed counts, larval web counts (Marsh and Glanville Fritillaries) and egg counts (hairstreaks and Large Blue). For these methods, systematic recording is made on single days in suitable weather (when UKBMS recording criteria are met), with the counts converted to a site index that accounts for both the size of the colony and the time in the season when the count was made.
Wider Countryside Butterfly Survey (WCBS)
The WCBS was established in 2009 to improve data on the national population status of widespread butterflies across the countryside as a whole. This is important given that most site-based monitoring is biased towards good quality semi-natural habitat relatively rich in butterflies. The WCBS is based on the British Trust for Ornithology (BTO)’s Breeding Bird Survey (the ‘BBS’), counting along two parallel 1-km long transects subdivided into 10 sections, located within randomly selected 1-km squares throughout the UK. Butterflies are counted using the Pollard Walk method described above. However, sites are chosen randomly based on 1-km squares throughout the UK. Unlike for the standard BMS transects, only 2-4 visits are required for WCBS squares with a minimum of 2 visits in July/August and other visits, particularly in earlier in the year for spring species such as Orange-tip and the first broods of bi/multi-voltine species, encouraged.
Difficult species: a number of species may be difficult to determine from closely related species, especially in flight. However, it is not possible to accurately assess population trends of individual species from aggregate data and so it is important to try and record to species level at all times. Small and Essex Skipper, where both found together, are particularly difficult and require a good look at the underside of the tips of the antennae. This can be done using binoculars or by slowly approaching alighted individuals. However, given that these species can be abundant at certain times of the year in certain habitats, this can be extremely time consuming and we recommend that recorders try to determine the identity of a sample of 10-20 individuals (depending on how abundant they are) across the transect with which to calculate a proportion of the two species. This proportion can then be used to determine the approximate number of each species from a total of the aggregate of both species recorded. Data should then be entered for the two species separately and not for the species aggregate. Ideally this should be done at section level, as the two species may occupy slightly different habitats, and on a weekly basis as the two species also have slightly different flight periods as a result of differences in larval ecology. Other species groups that cause similar problems include the common whites (Large, Small and Green-veined) and Dark-Green and High Brown Fritillaries on the few sites where these two species are found together. Recorders may wish to adopt a similar technique for these, but the habitat associations are even more markedly different between these species so performing such measurements at section level is more important. Note that a licence is required to capture High brown fritillary and the use of nets may be prohibited in some areas – contact BC for details. Identification tips for such species can be found under the ‘Similar’ Species’ section for each species on the UK Butterflies website (http://www.ukbutterflies.co.uk/index.php).
Detailed instructions for setting up and recording a transect are available on our
There has been a great development in data capture techniques in recent years in the UKBMS. An online recording system is now in place and will phase out the Transect Walker programme developed previously, specifically for capture of transect data. The online system also allows data entry for other survey types, including the Wider Countryside Butterfly Survey (WCBS) and non-transect surveys, and enables recorders to visualise theirs and others’ data much more easily and as the season progresses rather than only at the end. The UK is split into distinct regions, for each of which there is a regional coordinator responsible for collecting and checking the data from all monitoring sites in that region. Data from regional coordinators is then sent to Butterfly Conservation where it is compiled into one dataset. The data is then sent to CEH where it is appended to the main UKBMS database in Oracle and undergoes a series of automated and manual validation processes.
Method of analysis:
UK and national Collated Indices of abundance...
Statistical models are continuously developing and the UKBMS analysis currently adopts a combination of approaches dependent on the amount and type of data available. For habitat specialist and regular migrant butterflies where there are fewer data, a greater reliance on reduced effort surveys and/or where annual numbers fluctuate greatly, analysis is broken into two distinct processes. Firstly, weekly counts from transects are summed to generate site annual abundance indices. For sites with missing weekly counts, a statistical model (a Generalised Additive Model, ‘GAM’) is used to impute the missing values and to calculate a ‘site index’ (Rothery & Roy 2001). Counts from reduced effort schemes are adjusted according to the time of season and the size of the colony to produce site indices for these data. Site indices are an annual estimate of the abundance of a butterfly species at site. A site index is not an absolute measure of the size of a population, but has been shown to relate closely to other, more intensive, measures of population size such as mark, release, recapture (MRR) methods (Pollard et al. 1995) The site index can be thought of as a relative measure of the actual population size, being a more or less constant proportion of the number of butterflies actually present. The proportion seen is likely to vary according to species; some butterfly species, such as Marbled White are conspicuous, whereas others such as Dingy Skipper are much less easy to see.
Although a relative measure, site indices can be combined to derive regional and national collated indices and be used to estimate trends over time. However, this collation is not a straightforward calculation because not all transect sites in the UKBMS dataset have been recorded each year; some transect sites have operated for twenty years or more but the great majority have not and some have only been recorded for a few years. A statistical model is therefore needed to produce a regional or national index of how butterfly populations have changed each year.
Next, the site indices are used to calculate UK and national collated indices – a single index for each species over a defined range. In common with most butterfly and bird monitoring schemes in Europe (ter Braak et al. 1994), a log-linear Poisson regression model is used. In this approach, the expected count at a particular site in a given year is assumed to be a product of a site and a year effect. Put more simply, the model attempts to take account of the fact that some years are generally better than others for numbers of a particular butterfly species (the year effect), e.g. if weather is generally favourable. Similarly, the model accounts for some sites supporting higher numbers of a particular species than other locations (the site effect), e.g. if habitat conditions are highly suitable. In this way, for years where a transect site has not been recorded, the model imputes an estimated site index that allows for the general conditions of the year in question and the how favourable the site is. The national collated index is then calculated as the mean (on a log scale) of the imputed and recorded site indices for each year.
In 2013 we implemented a new analysis method for wider countryside species, to make better use of available transect data, and incorporating WCBS data in order to compile more representative national and UK indices. Briefly, the new method (Dennis et al. 2012) uses a single two-stage model. Firstly, all butterfly counts in a season from both traditional UKBMS transects and wider countryside squares are used to estimate the seasonal pattern of butterfly counts for that year using a GAM. This stage relies heavily on the traditional UKBMS transect data with good coverage throughout the season. A second stage of the model is then applied to the full set of annual counts, accounting for where the counts occur within the flight season, to then calculate annual population indices using a GLM accounting for site and year as in the old analytical method.
Collated indices have been calculated for butterfly species that have been recorded from five or more sites per year. For most species, this allows a plot of the collated index to be provided from 1976 to the current year, showing how the overall population abundance of each species has changed over this time.
Long-term and decadal trends are calculated for each species at UK and country level where sufficient data is available, using linear regression models on the national collated indices. Biodiversity indicators use multi-species (composite) indices of abundance for different groups of butterflies e.g wider countryside and habitat specialist species, and butterflies in different habitats e.g farmland and woodland. Composite indices are calculated following methods developed for UK birds, derived by calculating the geometric mean index across each species assemblage. Trends and confidence intervals in these indicators are then assessed by structural time-series analysis using the program Trendspotter. These indicators are updated and published annually and can be viewed at: http://jncc.defra.gov.uk/page-1824
The raw count data collected on standard UKBMS transects is also used to calculate phenological metrics for each butterfly species (e.g. first appearance, length of flight period). A linear regression analysis is then used to determine long-term and decadal trends in the timing and duration of butterfly flight periods for each species.
Method of analysis: Colonisation
and extinction of populations...
Colonisation is defined as 'absence' from a site
followed by 'presence';
Extinction is the reverse with 'presence' followed by 'absence'.
The identification of population extinction and foundation can rarely be certain
however. In particular, the absence of a species from a site can not be proven by
the absence of records and the presence of records does not prove the existence
of a breeding population, as butterflies may fly through areas in which they do
not breed. In spite of these difficulties, definitions have been adopted to identify
extinction and colonisations of populations from butterfly monitoring data. Full
details of the method are given in Pollard and Yates (1992).
We assume that there was a breeding population at a site if a species was seen in
four successive flight-periods and assume that there was no breeding population
if it was not seen in four successive flight-periods. Population extinction was
therefore assumed if 'presence' was followed by 'absence' at some later period;
population foundation or colonisation was assumed when the converse occurred. Thus,
a run of data of eight flight periods (note: 4 years for a bivoltine species; 8
years for a univoltine species) was required for either extinction or foundation
to be identified.
Method of analysis:
Trends on monitored sites (increasing, decreasing, stable)...
Trends are calculated using indices of abundance at individual
The regression slope of log index on years was used as the measure of trend.
The significance of the trends was also assessed to classify trends as:
- Increasing where the regression slope is positive and p < 0.05;
- Declining where the regression slope is negative and p < 0.05;
- Stable where p >= 0.05, e.g. not significant.
A minimum of eight index values was adopted for the inclusion of the data set for
a site in the calculation of trends and tests of significance.
Years with zero index values were omitted from the individual site data for the
calculation of trends.
Dennis, E.B., Freeman, S.N., Brereton, T. & Roy, D.B. (2013) Indexing butterfly abundance whilst accounting for missing counts and variability in seasonal pattern. Methods in Ecology and Evolution, 4, 637-645
Pannekoek, J. & van Strien, A. (1996) TRIM (TRends & Indices
for Monitoring data) Statistics Netherlands, Voorburg, The Netherlands.
Pollard, E. & Yates, T.J. (1992) The extinction and foundation
of local butterfly populations in relation to population variability and other factors.
Ecological Entomology, 17, 249-254.
Pollard, E., Moss, D., & Yates, T.J. (1995) Population trends
of common British butterflies at monitored sites. Journal of Applied Ecology
, 32 , 9-16.
Rothery, P. & Roy, D.B. (2001) Application of generalized additive
models to butterfly transect count data. Journal of Applied Statistics
, 28 , 897-909.
ter Braak, C.J.F., van Strien, A.J., Meijer, R., & Verstrael, T.J. (1994).
Analysis of monitoring data with many missing values: which method? In Bird Numbers
1992: Distribution, monitoring and ecological aspects. (eds W. Hagemeijer
& T. Verstrael), pp. 663-673. SOVON, Beek-Ubbergen, Netherlands.
The UK Butterfly Monitoring Scheme is organized and funded by Butterfly Conservation (BC), the Centre for Ecology and Hydrology (CEH), the British Trust for Ornithology (BTO), and the Joint Nature Conservation Committee (JNCC). The UKBMS is indebted to all volunteers who contribute data to the scheme.