Status of priority species: distribution

Figure 1: Change in distribution of 506 UK priority species, 1970 to 2024

Figure 2: Long-term and short-term changes in 506 UK priority species’ distribution trends, 1970 to 2024

Figure 3: Change in distribution of 506 priority species, by taxonomic group, 1970 to 2024

Relevance

Priorities for species and habitat conservation are set at a country level through country biodiversity or environment strategies. Each country has an identified list of priority species, which are of high conservation concern due, for example, to restricted range or population declines. The indicator therefore includes a substantial number of species that, by definition, are becoming less widespread.

Measures of distribution are less sensitive to change than measures of abundance (see priority species abundance indicator). Nonetheless, if a threatened species that has been declining starts to recover, its distribution should stabilise, and may start to increase. If the proportion of species in the indicator that are stable or increasing grows, the indicator will start to decline less steeply. If the proportion declines, it will fall more steeply. Success can therefore be judged by reference to trends in both indicators, as well as other information on other priority species for which there are insufficient data for inclusion in the indicator.

International/domestic reporting

The suite of UK Biodiversity Indicators has been revised and updated to bring it in line with the Kunming-Montreal Global Biodiversity Framework (GBF) of the Convention on Biological Diversity (CBD). Some UKBIs will be used for the forthcoming UK national reports to CBD.

Web links for further information

• Aquatic Heteroptera Recording Scheme: Home page
• British Arachnological Society Spider Recording Scheme: Home page
• Bees, Wasps and Ants Recording Society: Identification guides to download
• British Bryological Society: Home page
• British Dragonfly Society: Recording Dragonflies and Damselflies in the British Isles
• British Lichen Society: Home page
• British Myriapod and Isopod Group: Centipede and Millipede recording schemes
• Bruchidae & Chrysomelidae Recording Scheme: Relevant BRC webpage
• Butterfly Conservation: Butterflies and Moths
• Conchological Society of Great Britain and Ireland: Home page
• Dipterists Forum: Cranefly,Empididae & Dolichopodidae, Fungus gnat Recording Schemes
• Gelechiid Recording Scheme: Home page
• Ground Beetle Recording Scheme: Home page
• Hoverfly Recording Scheme: Home page
• Hypogean Crustacea Recording Scheme: Home page
• Lacewings and allies: Recording schemes
• National Moth Recording Scheme: Home page
• Orthoptera Recording Scheme: Home page
• Riverfly Recording Schemes: Ephemeroptera, Plecoptera and Trichoptera: Home page
• Soldierflies and Allies Recording Scheme: Home page
• Staphylinidae Recording Scheme: Relevant BRC webpage
• Terrestrial Heteroptera Recording Scheme - Shield bugs and allied species: Home page
• UK Biodiversity Partnership: UK Biodiversity Action Plans
• UK Centre for Ecology & Hydrology – Biological Records Centre: Recording Schemes
• UK beetle recording: Weevil and Bark Beetle Recording Schemes
• Natural England: S41 List of priority species in England
• NatureScot Scottish Biodiversity List
• Wales Biodiversity Partnership Section 7 priority species in Wales
• Northern Ireland Environment Agency Northern Ireland Priority Species List

Background

The measure is a composite indicator of 506 species from 23 taxonomic groups (seven of the 30 groups originally modelled did not contain any species with sufficient data to be included in the final analysis), see Species List for a detailed breakdown of the species and groups in the indicator. The priority species identified in each of the four UK countries were highlighted as being of conservation concern for a variety of reasons, including their scarcity, their iconic nature or a rapid decline in their population. They are not representative of wider species in general. They do however include a range of taxonomic groups and will respond to the range of environmental pressures that biodiversity policy aims to address, including land use change, climate change, invasive species and pollution. The short-term assessment of change can be used to assess the impact of recent conservation efforts and policy aimed at halting and reversing species declines. However, natural fluctuations (particularly in invertebrate populations) and short-term response to weather may have a strong influence on the short-term assessment.

Regardless of advances in statistical techniques and the increase in the number of biological records collected, there are likely to be species on the priority lists for which little monitoring or occurrence data is available. Reasons for this include rarity, difficulty of detection, or those for which monitoring methods are unreliable or unavailable. In order for the indicator to be representative of priority species, a method of assessing the changing status of these remaining data-poor species would need to be considered.

The indicator represents annual change in the geometric mean estimated occupancy across the constituent species. The index is set to a value of 100 in the start year (the baseline), so that changes subsequent to this represent proportional change in occupancy; if on average species’ trends doubled, the indicator would rise to 200, if they halved it would fall to a value of 50.

Data sources and species-specific time series

Biological records data are used to produce an indicator of change based on annual estimates of the proportion of occupied sites (“occupancy”) for a set of priority species in the UK. Biological records are observations of species in a known place in space and time. These data are curated by the Biological Records Centre. They include data from the following recording schemes: Aquatic Heteroptera Recording Scheme, Bees, Wasps and Ants Recording Society, British Arachnological Society Spider Recording Scheme, British Bryological Society, British Isles Neuropterida Recording Scheme, Centipede Recording Scheme, British Myriapod and Isopod Group, Millipede Recording Scheme, Bruchidae & Chrysomelidae Recording Scheme, Conchological Society of Great Britain and Ireland, Cranefly Recording Scheme, British Dragonfly Society, Empididae & Dolichopodidae Recording Scheme, Fungus Gnat Recording Scheme, Gelechiid Recording Scheme, Ground Beetle Recording Scheme, Hoverfly Recording Scheme, National Moth Recording Scheme, Orthoptera Recording Scheme, Riverfly Recording Schemes: Ephemeroptera, Plecoptera and Trichoptera, Soldierflies and Allies Recording Scheme, Staphylinidae Recording Scheme, Terrestrial Heteroptera Recording Scheme - Shield bugs and allied species and the Weevil and Bark Beetle Recording Scheme + Scolytidae.

Most records are made by volunteer recorders and whilst these data may be collected following a specific protocol, the majority of records in these datasets are opportunistic. The intensity of recording varies in both space and time (Isaac et al., 2014a and b), which is a challenge for estimating robust quantitative trends. Fortunately, a range of methods now exist for producing such trends using unstructured biological records data (e.g., Szabo et al., 2010; Hill, 2012; Isaac et al., 2014a and b). We follow previous versions of these distribution indicators by using Bayesian occupancy models to analyse these species occurrence datasets (Isaac et al., 2014a and b). The occupancy approach attempts to model the data collection process and produces annual estimates for each species of the proportion of occupied sites (van Strien et al., 2013).

By using occupancy models to analyse occurrence records, greater taxonomic coverage was achieved for the 2015 version of the indicator (Eaton et al. 2015). Further improvements to the occupancy modelling framework (Outhwaite et al. 2018) enabled the inclusion of more taxonomic groups (for example, lichens, craneflies and weevils) in the 2020 indicator and more species from groups included in previous indicators. Since 2021, new rules-of-thumb for assessing the data available for each species (Pocock et al. 2019) have been applied.

Annual estimates of occupancy from between 1970 and 2024 were used for the multi-species indicator. However, some datasets finished at different years within this time period. Since the 2021 update of the indicator the Biological Records Centre has received substantial data updates for Ants, Aquatic bugs, Bees, Carabids, Longhorn beetles, Dragonflies, Empid & Dolichopodid Flies, Ephemeroptera, Fungus gnats, Hoverflies, Ladybirds, Leaf and seed beetles, Orthoptera, Plant bugs, Rove beetles, Shield bugs, Soldier beetles, Soldier flies, Trichoptera, Wasps and Weevils (Table 1).

The rules-of-thumb approach was used to exclude species occupancy models that were likely to be highly uncertain (Pocock et al. 2019). These rules of thumb include dropping rarely recorded species (less than 1 record in every 100 visits) if there were fewer than 3.1 records across the 10% of the best recorded years. More frequently recorded species were also excluded if there were fewer than 6.7 records across the 10% of the best recorded years (Pocock et al. 2019). These exclusion criteria are based on classification trees, selected to balance the rates at which species are excluded when not meeting precision thresholds and included when meeting the precision thresholds. These model quality tests were unavailable for the moth dataset, here species with fewer than 50 records across the UK (Outhwaite et al. 2019, Powney et al. 2019) were excluded. Previously, the rules-of-thumb were not available for the bryophytes, instead the 50 record threshold was used. In this version of the indicator, we now use the rules-of-thumb criteria to exclude bryophyte species, meaning 75 extra bryophyte species that would have failed the 50 record threshold are now included.

Table 1: Summary of species’ time-series included in the Priority Species Bayesian measure

Notes about Table 1

• MSI refers to the multi-species indicator.

Methodology

The composite indicator was produced using a novel hierarchical modelling method for calculating multi-species indicators developed by UKCEH (Freeman et al. 2020), which offers some potential advantages over the geometric mean method used to produce the 2020 indicator. It can be applied to multiple data types, improving the comparability between metrics derived from occupancy and abundance data and can account for the uncertainty associated with the underlying species-specific time series as well as uncertainty in the indicator arising from the subset of species that are included. Imputing missing values is informed by between-year changes in species for which data is available, assuming shared environmental responses. Additionally, a smoothing process is used to reduce the impact of between-year fluctuations, such as those caused by variation in weather, making underlying trends easier to detect. The smoothing parameter (number of knots) was set to the number of years divided by three following Fewster et al. (2000).

The Bayesian occupancy approach enables an estimation of species occurrence even though the data used in this indicator were collected without a standardised survey design (van Strien et al., 2013; Isaac et al., 2014a and b). For each species, records were extracted at the one kilometre grid square scale with records on different days being treated separately, and an annual time-series of the proportion of sites occupied was calculated. Each species-specific time-series was scaled so the first value in 1970 was set to 100. The annual index was estimated as the arithmetic mean of the scaled species-specific occupancy estimates. Each species was given equal weighting within the indicator. Uncertainty in the species-specific annual occupancy estimates is represented by the 95% credible intervals.

A hierarchical, state-space model based on species annual growth rates was used to estimate the multispecies indicator (Freeman et al. 2020). The hierarchical structure allows estimation of uncertainty from different sources including that associated with the underlying species-specific index values and from having an incomplete set of trends for priority species with which to estimate the indicator. A smoothing process reduces the impact of between-year fluctuations, such as those caused by variation in weather, making underlying trends easier to detect.

Assessment of change

Species were grouped into one of 5 categories based on both their short-term (over the most recent 5 years of data) and long-term (all years) mean annual change in occupancy (Table 2).

Asymmetric percentage change thresholds are used to define these classes as they refer to proportional change, where a doubling of a species index (an increase of 100%) is counterbalanced by a halving (a decrease of 50%).

The threshold values for each category were based on those of the wild bird indicator; whether an individual species is increasing or decreasing has been decided by its rate of annual change over the time period (long or short) of interest. If the rate of annual change would lead to an occupancy increase or decrease of between 25 per cent and 49 per cent over 25 years, the species is said to have shown a ‘weak increase’ or a ‘weak decline’ respectively. If the rate of annual change would lead to a population increase or decrease of 50 per cent or more over 25 years, the species is said to have shown a ‘strong increase’ or a ‘strong decline’ respectively. These thresholds are used in the Birds of Conservation Concern (PDF 1.6MB) status assessment for birds in the UK.

Table 2: Thresholds used to define individual species’ trends

Species list

The species list was derived from the biodiversity lists of the 4 UK countries. A link to the list for each country is provided below:

• England – Section 41 Species
• Northern Ireland – Northern Ireland Priority Species List
• Scotland – Scottish Biodiversity List
• Wales - Section 7 Priority species

Table 3: The biodiversity lists of the 4 countries of the UK

Table 4: Taxonomic breakdown of combined 4 country biodiversity list

The species lists are unchanged from those used in the previous version of this indicator. The criteria for inclusion in each of the 4 biodiversity lists are derived from those used to identify the UK Biodiversity Action Plan (UK BAP) priority species list, most recently in 2007, but there has been some divergence in approaches, see Table 3. For example, the Scottish biodiversity list and the Northern Ireland priority species list both have criteria based on rarity alone, whereas the UK BAP criteria did not consider rarity; rare species were only listed if they were considered threatened or declining.

For the purposes of this indicator, an inclusive approach has been taken, whereby a species only has to be included in one of the country lists to be included on the combined list. The Scottish Biodiversity list has a final criterion based on the importance of species to people, however, species listed as a result of this criterion were not considered here. The taxonomic composition of the combined Four Country List (FCL) is shown in Table 4.

Some countries have included a small number of taxa below the species level (that is, sub-species) on their biodiversity lists. Such infra-specific taxa were only retained on the combined 4 country biodiversity list if the associated species was not included. For example, a sub-species of the grass rivulet moth (Perizoma albulata) is included on the Scottish biodiversity list, but it is a full species on the Northern Ireland priority species list, thus on the combined list only the full species was retained.

For a full species list, please see the published datafile.

Development work

The method of fitting the species distribution models is currently implemented in sparta, a Bayesian R package (Lindskou et al., 2024). As the number and size of datasets used in these models grow, it has become increasingly unviable to run, even on High Performance Computing platforms. This particularly affects butterfly and moth datasets (15,900,392 and 28,081,910 records, respectively), but may soon become a problem for other groups such as dragonflies, bees and hoverflies. The UK Centre for Ecology and Hydrology (UKCEH) have explored a range of different options to address this problem, including splitting the model runs up into several chunks and trialling different software, such as NIMBLE (de Valpine et al., 2017). Ultimately, these methods were either slower than the original software or were too unstable for these purposes.

To address the critical challenge of generating indicators for the largest taxon groups, UKCEH investigated using the R package, occti, developed by Dennis et al., (2017). The occti method fits occupancy models using Generalised Linear Models (GLMs) separately by season, and offers a frequentist alternative to the Bayesian framework used in sparta (Dennis et al., 2017). It had been shown in previous research to produce occupancy estimates comparable to those from sparta, while offering faster computation times (Dennis et al., 2017). Completion times of occti model fitting for the larger datasets averaged around three hours - substantially faster than the sparta method.

Comparisons between the outputs from occti and sparta show that occupancy trends were broadly similar across the most common (frequently recorded) species. For rarer, less well-recorded reported species, trends diverged more markedly. A common theme (but not exclusively true for all species) was of much larger uncertainty in the occti model outputs compared to that of sparta. Across taxonomic groups, there were also differences. For dragonflies, sparta estimated a steeper increase in the occupancy index, while for hoverflies, sparta suggested a steeper decline than occti. Notably, occti indices had substantially wider 95% credible intervals. This is likely more realistic given the underlying uncertainty in these occupancy model outputs, and the intrinsic variability of the natural dynamics of these groups and in the raw data.

UKCEH also compared the overall composite indicators derived from the Freeman et al., (2021) method, applied to the 15 most and 15 least reported species (that had at least 50 records across all years). The results showed that the indicators for the most well-recorded dragonfly species were broadly similar between occti and sparta. However, all other indicators showed a high level of difference between occti and sparta, and that this was particularly noticeable for the least well-recorded species subsets. For the least reported species there were noticeably larger confidence intervals in the trends derived from the occti method.

Figure 4: A comparison of index scores derived from sparta (left) and occti (right). The first two rows show indices for the 15 most reported species in dragonflies (first row) and hoverflies (second row), while the third and fourth rows show indices for the 15 least reported species in dragonflies and hoverflies. All panels include 95% credible intervals.

Caveats

The Risk-Of-Bias In Temporal Trends (ROBITT) assessment framework (Boyd et al. 2022), was designed to identify and communicate potential sources of bias given the spatial, environmental, and taxonomic scope of an analysis. These risk-of-bias assessments use a range of metrics and figures to highlight patterns in the sample data, helping to judge whether the model outputs are sensible and reliable given the model assumptions and the underlying data.

Here, light-touch risk-of-bias assessments for all the updated insect datasets and the associated occupancy models in this indicator reveal a number of risks of bias. Notably, there are strong spatiotemporal and taxonomic patterns in sampled occurrence of species across the insect groups. These patterns vary across taxonomic groups, with some groups showing overall increases in recorder effort with the broad spatial pattern of recording remaining consistent, while others show large shifts in the spatial pattern of recording over time. For many groups, the data are strongly clustered in space, often with increased recording effort in southern England. There is widespread variation in the temporal pattern, many groups show a temporal increase in the number of 1 x 1 km records. However, this pattern is not always consistent, some groups have bursts of recording activity with distinct core periods of recording, this is potentially a result of targeted effort prior to the publication of a taxonomic group atlas.

Many groups have a substantial increase in available records in the most recent years. While the occupancy modelling approach used here is designed to control for variation in detectability, it is unclear how the model outputs are impacted by these large shifts in recording effort. For example, any change in the pattern in recording effort that correlates with a potential driver of species occupancy change (i.e. habitat destruction), is likely to cause bias in the underlying species trend outputs. It is likely that apparent changes in species trends may in part reflect variation in recorder activity rather than underlying species change.

As discussed above, occupancy models are useful for detecting species change, where they are designed to handle variation in detectability, such as that associated with an increase in recorder effort. While useful, these models are data hungry, requiring repeat visits to grid cells in the same year to estimate detectability. While no clear threshold for the number of repeat visits is available, a minimum of >4 repeat visits on average has been previously suggested. Here, the average number of repeat visits to a site across all groups was approximately 2.4 with the mean number of repeat visits to sites generally increased with time. As with other aspects of the data, the mean number of repeat visits varied across taxonomic groups ranging from 1.4 for the aquatic bugs and longhorn beetles to a high of 11.9 for the moths. The lack of repeat visits is therefore likely inhibiting the ability of the occupancy model to account for detection bias in the analyses, in turn influencing the ability to detect genuine change.

The performance of the Freeman et al. (2021) method is currently being explored in relation to the species abundance indicator. Many of the development plans for the abundance indicator are also relevant to the use of the Freeman et al. (2021) method in this indicator. The results and recommendations of that work will be considered in future versions of this indicator.

Development plan

We are keen to hear feedback from users of these statistics, please send your feedback to: biodiversity@defra.gov.uk.

Development plans over the next few years:

• We will update the remaining groups that were unable to be updated in this publication, such as moths.
• We will incorporate any developments that arise from the development plan of the species abundance indicators.
• We will further explore differences between the outputs of the occti and sparta models to decide on an approach that suits all taxonomic groups and has a viable computation time.