Fig 1.
a) The number of BNM records as a function of time and b) the number of sites (1 km squares) with records each year.
Fig 2.
AIC comparison using varying covariates for detection probability.
The solid black line compares having the seasonal covariate and list length as two covariates, rather than with just list length. The dashed blue line compares having the seasonal covariate and list length as two covariates, rather than with just the seasonal covariate.
Fig 3.
Index Comparison for models B (black, circles) and C (blue, triangles).
The 95% confidence bands follow directly from the MCMC of the Bayesian analysis and from an approximate bootstrap approach in the classical case. In model C, detection probability has two covariates, the seasonal covariate and list length, and the data are restricted in date.
Fig 4.
Trend comparison for models B and C.
Trend comparison for 2005-2014 from model C and the trend from [15]—the State of UK Butterflies 2015 (SOBUK). Species codes are defined in Table A in S1 File. The 95% confidence bars for model C and result from the approximate bootstrap approach, while for model B we use the results of [15].
Table 1.
Simulation check of the Bayesian (B) and classical (C) models.
Table 2.
Estimated covariate coefficients for occupancy for Large Skipper.
Fig 5.
Illustrative regional occupancy indices for Large Skipper.
Confidence intervals are suppressed for clarity. The indices are presented across two plots for clarity, and are loosely grouped for northern and southern regions.
Fig 6.
Illustrative spatial distribution maps.
Maps are shown for Large Skipper, Small White and Silver-washed Fritillary in 2014: a) observations b) estimated occupancy probability c) estimated standard error.