Testing the reliability of assessment of land management impacts on Scottish upland vegetation

Abstract

Background: Surveys of land management impacts (e.g. of herbivores) are carried out in the Scottish uplands to meet conservation objectives. The reliability of such assessments has seldom been tested. Aims: To assess the inter-observer reliability of impact assessments. Methods: In the upper Spey valley, central Scottish Highlands, up to 17 observers were asked to assess in 5 m by 5 m plots and 1 ha plots (a) drying and peat loss in blanket bog, (b) browsing of dwarf-shrub heath, (c) grazing and trampling of wind-clipped heath, and (d) trampling and grazing in soligenous mire. Results: Averaged across indicators, complete disagreement between any two observers (high versus low impact) was found to be very unlikely (P < 0.06, and often < 0.001). On the other hand, complete agreement was moderately likely (between P = 0.53 and 0.75). Reliability varied considerably but was greater for large test plots than for small test plots, and for wind-clipped heath and dwarf-shrub heath. Conclusions: Reliability was similar to that experienced in previous studies of land cover mapping, but appeared to be substantially worse than for the estimation of plant cover. However, reliable discrimination between the extremes of high and low impact, as found in the present study, provides adequate precision for the majority of land management planning and monitoring in extensive, open range conditions.