Optimising biodiversity assessments by volunteers: The application of occupancy modelling to large-scale amphibian surveys
Introduction
The use of volunteers in biodiversity monitoring schemes overcomes constraints caused by a lack of funding for professional researchers and increases public perception and involvement (Schmeller et al., 2009). In the USA at least 200 volunteer-based survey programmes exist (Cohn, 2008), whilst in the UK the Royal Society for the Protection of Birds mobilised in excess of half a million volunteers for its Big Garden Birdwatch in 2010 alone (http://www.rspb.org.uk/birdwatch/results/ accessed 10th May 2010). In the developing world long term biodiversity monitoring schemes may only be sustainable if volunteers are used (Holck, 2008). On the other hand, there are significant concerns over the reliability of data collected by volunteers (e.g. Engell and Voshell, 2002, Lotz and Allen, 2007, Schmeller et al., 2009). Even when simple presence–absence surveys are carried out, there may be a tendency to miss small populations, resulting in false absences. This problem may be compounded by the misidentification of some species (Genet and Sargent, 2003, Fitzpatrick et al., 2009).
Renewed interest in the global status of amphibians has precipitated coordinated volunteer surveys in a range of countries. However, detection probabilities for amphibian species are highly variable and often low (MacKenzie et al., 2002, MacKenzie et al., 2003). Failing to allow for variation in detectability generates unreliable data, especially with respect to false negatives (MacKenzie et al., 2002, Schmidt, 2003). Many volunteer amphibian survey schemes reported so far have relied on a single detection method, usually vocalisation by breeding males (Genet and Sargent, 2003, de Solla et al., 2005, Pellet and Schmidt, 2005, Pierce and Gutzwiller, 2007). Surprisingly little attention has been paid to the efficiency and variation between the methods used by volunteers for detecting amphibians, and how these fare in different landscapes and between different species.
In Britain, a National Amphibian and Reptile Recording Scheme (NARRS) was instituted in 2007 with a view to assessing future status changes of the herpetofauna of the UK, including the five widespread native amphibian species, common frog (Rana temporaria), common toad (Bufo bufo), smooth newt (Lissotriton vulgaris), palmate newt (L. helveticus) and great crested newt (Triturus cristatus). NARRS primarily uses trained volunteer surveyors who carry out presence–absence surveys using a standard protocol. In this study, we applied occupancy modelling to the current NARRS survey protocol. Occupancy modelling is based on the patterns of detection and non-detection of species at a range of sites over multiple visits, and estimates both site occupancy and detection probability (Mackenzie et al., 2002). There have been previous attempts to use this methodology to determine the survey effort required for the reliable inference of absence, rather than non-detection, of amphibian species (e.g. Pellet and Schmidt, 2005), but they have not yet been applied to a large-scale national survey programme using multiple detection methods. We used this approach to determine: (1) whether the range of methods recommended to volunteers can reliably detect presence and absence; (2) the number of surveys required to minimise the risk of false absences; and (3) covariates of detectability that may assist volunteers with the timing of surveys.
Section snippets
Study areas and field work
Two contrasting landscapes in Britain were chosen for comparison. A lowland area in east Kent, (Lat. 51°20′N, Long. 1°05′E) was compared with an upland area of central Wales approximately centred at Lat. 52°06′N; Long. 3°20′W. Land uses in the Kent study area were primarily woodland, arable and fruit growing, with extensive urban areas. In contrast, the study area in Wales had lower levels of human occupation with grazing predominant in the lower areas and open moorland with rough grazing on
Differences between study areas
Differences between the two study areas were confirmed in five variables (elevation, pond area, pond density, shade and macrophyte cover) by means of a MANOVA (Wilks lambda = 0.155, F5,42 = 45.7, P < 0.001). Comparison of individual variables showed that the ponds sampled in Wales were at higher altitude (means ± SD: Kent 72 ± 12 m, Wales 274 ± 81 m, F1,46 = 138.6, P < 0.001), larger (Kent 143 ± 154 m2, Wales 1185 ± 1199 m2, F1,46 = 17.1, P < 0.001), less-shady (Kent 50 ± 27%, Wales 15 ± 19% shading, F1,46 = 26.5, P < 0.001), had
Volunteer expertise
In addition to the problem of detection probabilities varying between species (de Solla et al., 2005), misidentification by volunteers can vary between 4% and 18%, with some species consistently confused (Genet and Sargent, 2003, Lotz and Allen, 2007). We do not believe that misidentification is a major issue for amphibian surveys by trained volunteers in Britain, as adults of the five widespread species are relatively easy to identify in the field. The most difficult to distinguish are palmate
Acknowledgements
We thank the Esmée Fairburn Foundation for financial support and the many students, and members of the Brecknock and Radnorshire Amphibian and Reptile Group, who helped with fieldwork. The adjustments to the collection of terrestrial habitat data were first suggested by Lee Brady. Additionally we thank Rachel McCrea, Gurutzeta Guillera-Arroita and Jose Lahoz-Monfort for their help and patience with the statistical analysis, and also Darryl MacKenzie who made a number of helpful suggestions at
References (30)
- et al.
Effect of sampling effort and species detectability on volunteer based anuran monitoring programs
Biological Conservation
(2005) - et al.
Monitoring distributions using call surveys: estimating site occupancy, detection probabilities and inferring absence
Biological Conservation
(2005) - et al.
Avoiding pitfalls when using information-theoretic methods
Journal of Wildlife Management
(2002) - et al.
Amphibians and Reptiles: A Natural History of the British Herpetofauna
(2000) - et al.
Model Selection and Multi-Model Inference
(2002) Citizen science: can volunteers do real research?
Bioscience
(2008)- et al.
Volunteer biological monitoring: can it accurately assess the ecological condition of streams?
American Entomologist
(2002) Great Crested Newt Mitigation Guidelines
(2001)- et al.
Observer bias and the detection of low-density populations
Ecological Applications
(2009) - et al.
Evaluation of methods and data quality from a volunteer-based amphibian call survey
Wildlife Society Bulletin
(2003)
How many clumps are there in a mass of frog spawn?
The British Herpetological Society Bulletin
A simple funnel trap for studying newt populations and an evaluation of trap behaviour in smooth and palmate newts, Triturus vulgaris and Triturus helveticus
Herpetological Journal
Participatory forest monitoring: an assessment of the accuracy of simple cost-effective methods
Biodiversity and Conservation
Inferring the absence of a species – a case study of snakes
Journal of Wildlife Management
Cited by (67)
Dynamic occupancy modeling of a cryptic ground-dwelling pheasant species in the Upper Blue Nile Basin in Ethiopia
2023, Global Ecology and ConservationObservers are a key source of detection heterogeneity and biased occupancy estimates in species monitoring
2023, Biological ConservationApplying economic and ecological criteria to design cost-effective monitoring for elusive species
2020, Ecological IndicatorsCitation Excerpt :Many elusive species, like mammals, require more sophisticated or combined methods (i.e., eDNA, Dorazio and Erickson, 2018) to obtain presence data (which increases the effort and costs to ensure the detection and occupancy). Moreover, even species known to occupy a certain patch could go undetected, generally because the surveys do not target periods when detectability is likely to be high (Sewell et al., 2010) or the monitoring schedule is not the most appropriate (i.e., short-term monitoring and irregular censuses among years). Finally, although we deemed landscape as constant in this study, climate and land cover changes could also drive spatiotemporal occupancy patterns of species (Kalle et al., 2018), which could be an important issue when estimating the costs of long-term monitoring projects and obtaining precise and accurate occupancy estimates.
Amphibian diversity in farmlands: Combined influences of breeding-site and landscape attributes in western France
2019, Agriculture, Ecosystems and EnvironmentGeodiversity inclusiveness in biodiversity assessment
2023, Progress in Physical Geography