Large-scale ecological surveillance data were analysed to determine the locations of apparent eutrophication effects across common British vegetation types between 1990 and 1998. Plant species composition was recorded from a total of 9514 fixed plots located in a stratified, random sample of 501 1 km squares across Britain. Changes in plant species composition along a gradient of substrate fertility were inferred from statistical tests of change in mean Ellenberg fertility value calculated for each fixed plot. Plots were grouped by eight vegetation types, five landscape features (hedges, road verges, watercourse banks, small biotopes, larger units in fields and unenclosed land) and six environmental zones. Tests of change in mean Ellenberg value were carried out on all combinations of these strata. Decision tree modelling was used to identify groups of test outcomes sharing the same direction of change and where each group was defined by a minimum number of strata. Post hoc power analysis was used to select statistically non-significant test outcomes that could be used to infer stability in the sampled sub-population.
Out of a total of 142 tests of change in fertility value, 67% were increases, 8% were reductions and 25% indicated stability. The best overall predictor of increases in fertility value, and therefore of shifts in favour of plants suited to higher substrate fertility, was vegetation type. Irrespective of landscape feature and environmental zone, increased means were associated with infertile grasslands, moorland, upland woodlands and heath/bog. Already highly fertile grasslands and woodland assemblages in lowland Britain remained largely stable. The small number of decreasing test outcomes were associated with arable land in Scotland, Wales and western England. These patterns of change are hypothesized to reflect pervasive land-use drivers combined with the inherent responsiveness of the vegetation.