Prediction of plant diversity response to land-use change on Scottish agricultural land

Surveys of 87 sites within nine locations on Scottish agricultural land, during 1995-1997, recorded 335 species of vascular plants and 95 bryophyte species, within 31 recognisable plant communities. Multivariate analysis placed the vegetation into five main vegetation types, with differing plant species richness (S: number of species per 100 m(2)). These were upland grassland (mean S = 37); mesotrophic semi-improved grassland (mean S = 25); mire/heath vegetation (mean S = 22); intensive grassland (mean S = 12); and arable land and weed communities, including set-aside (mean S = 9). Management intensity was a good predictor (R-2 = 0.729) of vegetation type, with low intensity being typical of upland grassland, and high intensity typical of arable vegetation. The data were used to develop and test a set of minimal linear models of plant diversity response to land-use changes, operating at field scale, under the agro-geo-climatic conditions of mainland Scotland. The models utilised predictor variables drawn from a suite of functional vegetation attributes, environmental data, and management information. All-plant-species S was well predicted at quadrat level (i.e. within field, excluding margins) by three models with high R-2 values (0.82-0.93) utilising management variables such as crop-type, livestock density, time since last cultivation; vegetation state variables such as bryophyte cover, sward height, litter cover; and environmental variables such as soil P content. Bryophyte S was well predicted by four models with high R-2 values (0.83-0.92). Bryophyte percentage cover was the most important predictor variable in these models. Vascular plant species S was less well predicted than either all-plant-species or bryophytes-only models. The models can be used to assess the implications of land-use change scenarios, and to aid management decisions to help to maintain plant diversity within the Scottish agricultural landscape. (C) 2002 Published by Elsevier Science B.V.