David Sewell’s research while affiliated with University of Kent and other places

Grass snakes (Natrix natrix) were monitored for nine years on a site in eastern England restored for an amphibian reintroduction. Counts of snakes increased between 2004 and 2012 from 1.25 to 3.83 snakes per survey visit. Grass snake counts were positively correlated with the number of common frog spawn clumps each year and peak counts of pool frogs. During surveys and incidental encounters 137 adult males, 161 adult females, 131 juveniles and 44 hatchlings were captured and individually photographically identified. Captures of hatchlings were erratic and recapture rates were low, so they were excluded from the analysis. Annualised capture data were analysed in the capture-recapture programme MARK, using the Cormack-Jolly-Seber model. The top ranked model gave an apparent annual survival rate of 0.66 (95% CI=0.543–0.755) and an individual detection rate of 0.17 (0.118–0.245). Population estimates based on this model ranged from 53 (95% CI=37–76) to 576 (95% CI=400–831) over the nine years of study. Grass snake population estimates were equivalent to densities of 4.8 to 52.4 individuals ha-1. Nevertheless, it is unlikely that these snakes were permanently resident within the study area, and annual survival may therefore be underestimated. A more plausible explanation for the large population estimates is that the snakes were temporarily resident within a patch of high quality habitat and moved through home ranges that included the study site.

Bridging the gap between conservation science and conservation practice is a widely acknowledged issue in applied ecology (Hulme 2011). Nowhere is the gap greater than in the area of data collection, analysis and interpretation. Population assessments for conservation are frequently based on traditional practices that use rules of thumb and quasi-quantitative methods. This means that important decisions that have far-reaching conservation, commercial and financial implications are often based on sketchy population assessments.This article is protected by copyright. All rights reserved.

Monthly variation in detection probability. (PDF)

Number of sites where species were encountered by month, year and sample size. (DOC)

Number of sites required to detect a decline. (TIF)

Biodiversity monitoring programs need to be designed so that population changes can be detected reliably. This can be problematical for species that are cryptic and have imperfect detection. We used occupancy modeling and power analysis to optimize the survey design for reptile monitoring programs in the UK. Surveys were carried out six times a year in 2009-2010 at multiple sites. Four out of the six species--grass snake, adder, common lizard, slow-worm -were encountered during every survey from March-September. The exceptions were the two rarest species--sand lizard and smooth snake--which were not encountered in July 2009 and March 2010 respectively. The most frequently encountered and most easily detected species was the slow-worm. For the four widespread reptile species in the UK, three to four survey visits that used a combination of directed transect walks and artificial cover objects resulted in 95% certainty that a species would be detected if present. Using artificial cover objects was an effective detection method for most species, considerably increased the detection rate of some, and reduced misidentifications. To achieve an 85% power to detect a decline in any of the four widespread species when the true decline is 15%, three surveys at a total of 886 sampling sites, or four surveys at a total of 688 sites would be required. The sampling effort needed reduces to 212 sites surveyed three times, or 167 sites surveyed four times, if the target is to detect a true decline of 30% with the same power. The results obtained can be used to refine reptile survey protocols in the UK and elsewhere. On a wider scale, the occupancy study design approach can be used to optimize survey effort and help set targets for conservation outcomes for regional or national biodiversity assessments.

Power simulation results. (DOCX)

Occupancy models from the 2010 45 sites dataset. (PDF)

Occupancy models from the 2009 29 sites dataset. (PDF)

Number of encounters by species, dataset and method of detection. (DOCX)