How reliable is the monitoring of permanent vegetation plots? A test with multiple observers

Questions: A multiple plots design was developed for permanent vegetation plots. How reliable are the different methods used in this design and which changes can we measure ? Location: Alpine meadows (2430 m) in the Swiss Alps. Methods: Four inventories were obtained from 40m-plots: four subplots (0.4 m) with a list of species, two 10m-transects with the point method (50 points on each), one subplot (4 m) with a list of species and visual cover estimates as a percentage and the complete plot (40 m) with a list of species and visual estimates in classes. This design was tested by 5-7 experienced botanists in three plots. Results: Whatever the sampling size, only 45-63% of the species were seen by all the observers. However, the majority of the overlooked species had a cover < 0.1%. Pairs of observers overlooked 10-20% less species than single observers. The point method was the best method for cover estimate, but it took much longer than visual cover estimates, and 100 points allowed for the monitoring of only a very limited number of species. The visual estimate as a percentage was more precise than classes. Working in pairs did not improved the estimates, but one botanist that repeats the survey is more reliable than a succession of different observers. Conclusion: Lists of species are insufficient for monitoring. It is necessary to add cover estimates to allow for subsequent interpretations in spite of the overlooked species. The choice of the method depends on the available resources: the point method is long but gives precise data for a limited number of species , while visual estimates are quick and exhaustive but allow for recording only large cover changes Constant pairs of observers improve the reliability of the records.