Data on numbers of toads and frogs moved by volunteers during spring migrations were collected at the end of the migration season which typically runs between March and April. Standard forms were completed by the patrol leader every night of the season (typically for 25–30 nights every year) collated into a season total, and sent off to be centralised. We used the seasonal total of toads and frogs moving towards the pond rather than the daily counts in the analysis. Volunteers collected and counted all the individual toads by repeatedly surveying a stretch of the road during each survey night. However, at some sites in the UK and most sites in Switzerland, temporary plastic fences were used in order to prevent toads from accessing the road, and the amphibians were then collected along the fences [22]. Given the timing, selected to coincide with the breeding season, the vast majority of the individuals recorded was represented by adults. Toads and other amphibians were collected in plastic buckets, taken to the other side of the road and released, often near the breeding area.

We used data from volunteer groups collected from 1975 to 2014 in the UK and from 1973 to 2012 in Switzerland (see Supporting Information), and we only included sites with at least 5 years of data so that enough years were included to remove the bias effects of natural population variations [20, 31, 32] and to establish a trend. Given the small number of populations between 1975 and 1984 for the UK, we used 1985–2014 data only. For the UK, we included a regional component by splitting the data into 5 regions following a similar previous analysis [15]. The populations were structured by county and separated into Northern, Western, Central, Eastern and South-Eastern. Regions are defined as follows: North (Scotland, Northumberland, Cumbria, Durham, Yorkshire, Lancashire, Cheshire and Derbyshire); West (Wales, Shropshire, Staffordshire, Herefordshire, Worcestershire, Gloucestershire, Wiltshire, Dorset, Somerset, Devon, Cornwall); Central (Nottinghamshire, Leicestershire, Northamptonshire, Rutland, Cambridgeshire, Warwickshire, Oxfordshire, Buckinghamshire, Bedfordshire); East (Lincolnshire, Norfolk, Suffolk, Essex, Hertfordshire); South-East (Berkshire, London, Hampshire, Surrey, Kent, Sussex, Isle of Wight). For Switzerland, we also included data on the abundance of the common frog (Rana temporaria) for comparison.

After removing sites with less than 5 years of data, a total of 153 sites were included in the analysis for the UK and 141 for common toads and 135 for common frogs for Switzerland. The average length of time series for common toads in the UK was 14.1 years. In Switzerland, time series lengths for common toads and common frogs were 14.3 years and 14.0 years, respectively.

The data were collected by a large number of volunteers in many different places and years using similar but not identical methods. It is thus important to account for variation in sampling effort between sites and years [33]. Estimating detection probabilities and adjusting counts is a commonly used method to adjust for variation in sampling effort [34]. Therefore, we used the open N-mixture model for spatially replicated repeated count data [35] because this model is highly suitable for our data [36]. This model can be used to estimate population trends while accounting for variation in sampling effort and detectability because it assumes that the number of individuals is generally underestimated. The Dail-Madsen model was successfully employed for the analysis of similar amphibian population data [37–39].

Population trend is commonly estimated using a regression of abundance against calendar year [40–42]. We used population growth rate as our estimate of trend because we believe that it is a biologically more meaningful metric [43–45]. The model is [35, 36] Ni,t∼Poisson(λNi,t-1) (eq 1) where N is population size, i and t are indices for site and year, respectively, λ is population growth rate, and N_i,t is a Poisson random variable which allows for demographic stochasticity. Given this definition, the trend quantifies a proportional change in abundance (i.e., the population multiplication rate) rather than an absolute change. The biological trend model is related to the counts y_i,t using yi,t∼Binomial(p,Ni,t) (eq 2) where p is detection probability [35, 36]. We fitted separate models for the Swiss und UK data and estimated parameters for both countries in independent analyses. For the purpose of our analysis, we estimated a mean λ per decade. Within the UK, we estimated a region- and decade-specific λ in a second analysis. Mean λ reduces the effects of random variation among years and sites. Detection probability p was allowed to vary among decades and regions (the latter only in the second analysis).

Model fitting followed the procedures outlined in [46]. We modified code available from the electronic supplement of Hostetler and Chandler [36] and fitted the models in the Bayesian software JAGS [47] using R and the R package jagsUI [48, 49]. To fit the models, we used a Poisson distribution to describe initial abundance N_i,1. Because time series started in different years, we wrote the model in JAGS in such a way that N_i,1 was replaced with N_{i,first t} where first was the first year of the time series i. We specified diffuse uniform priors for all parameters (p, λ and N_i,first) to be as uninformative as possible. For each model, we ran three Markov chains with 10000 iterations each, discarded the first 2000 iterations as burn-in and thinned the remainder by one in ten. Convergence was assessed using the Brooks–Gelman–Rubin R^ statistic [46].

Citizen science projects have been gaining attention due to their potential for collecting large amounts of data over wide spatial and temporal scales as well as a means of engaging the public in conservation efforts. Importantly however, the primary motivation for “toad patrols” was the conservation action of protecting amphibians from road mortality rather than surveys or data collection, making this dataset relatively unique as volunteers collect data measuring not just the number of animals present but also their own success in moving them to safety. Expanding such schemes that provide both highly valuable monitoring data and a measurable conservation action that attracts and engages people could be a successful model for other citizen science wildlife projects [50]. Cooke [19] demonstrated that the decline in numbers of toads recorded killed on roads at three unpatrolled sites in the UK between 1974 and 2010 correlated strongly with numbers of adults reaching the ponds, making the numbers of individuals at road crossing points a suitable measure of population abundance. In our analysis, we used statistical models to account for imperfect detection and variation in sampling effort. Detection probabilities were sometimes low but this may be an effect of the structure of the model. If repeated counts within years were available, we might have been able to model detectability in a better way [36]. Allowing for environmental stochasticity in the models [36] might lead to higher estimates of detection probabilities because the model may be better able to disentangle variation in detection and natural population fluctuations.

The selection of our sites was not random but rather volunteers chose them based on observing amphibian mortality on specific road sections, and as such, it is unclear how representative they are for toad populations at a national scale in each country. However, declines of common toads (based on presence/absence data) in Switzerland were also independently found in a survey where sites were selected randomly [17]. There are two reasons why we believe that the studied populations are representative nationally and important. First, roads are known to have strong negative effects on amphibian populations [51]. Both the UK and Switzerland have some of the densest and most intensely used road networks in the world when compared to the total surface area of the country [52]. For example, Switzerland has a road density of 2.7 km/km² [22] and considerably higher in the lowlands. Similar to other Western or Central European countries, given the high habitat fragmentation degree in both the UK and Switzerland, there are probably few areas other than mountains where toad populations migrating 1 km or more towards the breeding area would not have to cross at least one road. Thus, it seems likely that the majority of UK and Swiss toad populations are affected to some extent by roads. Second, even though the sample is not random, this dataset represents hundreds of populations that benefit from conservation action, and one might thus expect that they fare better than other populations with no such conservation action.

An annual trend of 0.96 implies that a population will decline by more than 30% in less than ten years. Given this magnitude of decline, a population would qualify as “vulnerable” under IUCN Red List criterion A2 (“Population reduction observed, estimated, inferred, or suspected in the past where the causes of reduction may not have ceased OR may not be understood OR may not be reversible.”). The observed declines in our dataset were close to this figure or even surpassed it at both regional level and national level (common toads in the UK between 1995 and 2004). The fact that toad abundance significantly declined across both countries since the 1980s even at sites where there was an obvious and long term conservation action in place, (i.e. moving individual toads across the roads by volunteers) is troubling. Our results appear to confirm that such actions are not sufficient to prevent long-term declines, as previously observed for 33 toad populations in Italy [16]. Reasons for the decline of the common toad might include the fact that volunteers in “Toads on Roads” projects only target adult migration towards the wetland but not the unpredictable adult return migration or the movement of juveniles dispersing from the breeding area in late summer. Thus, the postmetamorphic juvenile stage, whose survival is crucial for amphibian population dynamics [53–55], is not protected from road mortality [22]. Road traffic and car numbers have increased substantially in both countries since 1980, almost doubling in the UK within that period to 35 million vehicles registered in 2013 [56], potentially making it increasingly difficult for juvenile toads moving over the road to escape car traffic. Thus, large and exceptional amphibian populations should be protected from road impacts and habitat loss by adequate mitigation such as the construction of underpasses rather than solely through such volunteer actions. However, questions remain about the effectiveness of toad tunnels for juveniles [22].

Additionally, road mortality represents only one potential cause of decline, and other factors such as climate change and increasingly mild winters [18], agricultural intensification and widespread habitat loss or degradation, as well as emergent diseases [57], remain unresolved and ongoing and may affect toads both in their aquatic and terrestrial habitats. The fact that toad declines appeared most pronounced in South-East England, the most densely populated region in the UK, is important and requires further investigation. Similarly, the steep decline in Switzerland since the 1980s might be linked with both road traffic increases, changing climate, and widespread habitat degradation as already observed in declines of other widespread species. A potential solution to halt declines would be to urgently incorporate widespread and common species such as toads into wider conservation and planning policy initiatives and target large-scale improvements of the general landscape using networks of agri-environment scheme options and landscape-scale connectivity projects.

It is likely that in the absence of volunteer efforts toad declines would have been more marked and more populations would have become extinct given the unsustainable levels of adult mortality, as road mortality can drive toad populations to extinction [58]. This was the case at 3 sites monitored in Cambridgeshire, UK where no animals were moved by volunteers and instead effort concentrated on annual recording of toad mortality and traffic levels on the road from 1974 until all three populations collapsed by 2010 [19]. Local extinction sites are almost certainly under-represented in our datasets as volunteers may become disheartened following prolonged declines and give up patrolling the site once the numbers are low.

Data collected partly from the same period seems to indicate that for the other widespread and abundant amphibian species, the common frog, populations appear to have stabilised in both Switzerland and the UK ([15, 17], this study), although our analysis indicates recent declines in Switzerland. The reasons for this disparity in trends are unknown. Carrier and Beebee [15] suggested that common frogs are better adapted to breed in shallow and small bodies of water, including garden ponds, which seem to have provided important refuge for the species in the face of lost habitats in farmland areas. It is, however, also possible that common frogs are less impacted at the population level by road mortality either for juveniles only or for both juveniles and adults. Understanding why one common generalist species declines while the other appears to be stable would be an important focus for future research.

Despite some recent increase in interest in the value of common species and the importance of their declines [4], conservation resources, including monitoring as well as government policy at a national level, remain almost entirely focused on rare species which are facing regional or global extinctions in the near future. Both common birds and common butterfly species, including former pest species, are declining in Europe while effective conservation programs for habitat specialists mean that at least some of the rarer species are increasing in abundance [7,9]. However, commonness itself is rare, and even moderate declines in a small number of common species can inherently mean the loss of vast numbers of individuals and biomass over large areas with consequences across the trophic chain, habitat or ecosystem function [2, 10]. Common toads for example, are important invertebrate predators and also a regular food source for a wide range of species in Europe, including mammals, reptiles and birds (for a review on ecosystem functions of amphibians see Hocking and Babbitt [29].