A.P.M. conducted 731 standard point counts over the 3 years, late May–early June 2006–2008, within forested areas around Chernobyl, each count lasting 5
min during which all birds seen or heard, bumble-bees, butterflies, spider webs and dragonflies (dragonflies were only censused in 2008) were recorded (Møller 1983
; Bibby et al. 2005
). Points were located at 75
m intervals. Grasshoppers were censused during early September 2007 at 374 points.
We conducted four line transects at 17 sites during early July 2008 within and just outside the southern border of the exclusion zone. At each site, we pre-selected plots with high and low radiation (high radiation: 0.60
(s.e.=1.29), low radiation: 0.38
(1.29)). At each plot, A.P.M. conducted two line transects along roads in opposite directions, each of 50
m. Walking speed was slow, each transect lasting 10
min. Half an hour later, the same line transects were conducted once more without reference to the first census results.
Invertebrate abundance estimates can be affected by habitat (agricultural habitats with grassland or shrub, deciduous forest or coniferous forest estimated to the nearest 10% of ground coverage within 50
m from each of the observation points), maximum height of trees estimated to the nearest 5
m, soil type (loam/clay or sand), presence or absence of open water within 50
m from each point, cloud cover at the start of each point count (to the nearest eighth), temperature (range 12–28°C) and wind force (Beaufort, range 0–4). For each point count, we recorded time of day (to the nearest minute). Because the activity of many invertebrates is curvilinearly related to time of day, with high levels of activity in the middle of the day (e.g. Barnes et al. 2001
), we also included time squared in the analyses.
Once having finished a 5
min census, we measured α
radiations two to three times at ground level using a hand-held dosimeter (Model: Inspector, SE International, Inc., Summertown, TN, USA). Cross-validation against data from Shestopalov (1996)
revealed a strong positive relationship (linear regression on log–log transformed data: F
=1546.49, d.f.=1,252, r2
<0.0001, slope (s.e.)=1.28 (0.10)), suggesting reliability of our estimates.
Abundance of invertebrates and radiation level were log transformed, while coverage with farmland and deciduous forest was square-root arcsine transformed. We developed best-fit models to assess the relationship between invertebrate abundance (dependent variable) and radiation, after the inclusion of potentially confounding environmental variables, as implemented in JMP (SAS Institute Inc. 2000
). Model selection was based on Akaike's information criterion (AIC), using the criterion of delta AIC<2.00 for exclusion of variables (Burnham & Anderson 2002
). We analysed line transect data by using abundance as the response variable and site, level of radiation and their interaction as predictor variables. Frequency distributions of invertebrate counts were skewed, with disproportionately many zeros. However, we obtained similar results with Kendall rank-order correlation and partial rank-order correlation (controlling for confounding variables in table S2 in the electronic supplementary material).