Implications of design and data quality for the analysis of a nationwide biodiversity monitoring scheme

Biodiversity monitoring schemes are designed to infer trends in biodiversity over long time periods. The value of a biodiversity monitoring program depends largely on its data quality. High quality data allow to estimate temporal trends without bias and with high precision. Data quality largely depends on the initial design of the monitoring scheme, on properly conducted fieldwork, on various aspects of quality control mechanisms, and on the methods to analyse the data. In my thesis I show and discuss implications of design and data quality presenting five case studies using data from the Swiss Biodiversity Monitoring Scheme (BDM). The BDM is a long-term programme of the Swiss Federal Office for the Environment and was initiated in 2001 to monitor Switzerland’s biodiversity. The programme focuses on changes in species richness and surveys selected species groups in a systematic sampling grid all over the country.
Defined and constant sampling methods are needed to allow for unbiased and precise estimations of biodiversity trends. In Chapter I, we analysed inter-observer variation of double-sampled vegetation plots. We could show that both systematic (directed) methodological errors and random variance of species counts were small. We concluded that BDM methods are adequate for detecting biodiversity trends. In the meantime this conclusion has been widely confirmed with recent data from quality control. Chapter II focuses on detectability of species that provides the link between a raw species count and true species richness. Variation in detectability between species or habitats may considerably bias trend estimates in biological studies. We therefore asked if capture-recapture methods were suitable to analyse differences in species detectability of butterflies and looked for underlying factors that may cause variation in detectability. Because the methods available at that time were not allowing the analysis of butterfly surveys over the whole season we had to restrict it to three mid-season surveys. We found that average detectability per count was 0.61 and was influenced by observer, transect and region. Individual species during one count were detected with a mean probability of 0.50. Since the study has been published in 2007 statistical methods have been substantially developed and nowadays enable detailed analyses of butterfly communities.
In the study in Chapter III we demonstrated how data from the systematic BDM surveys could be used in combination with environmental variables. We tested different sets of variables for modelling plant species richness and produced species richness maps for Switzerland by predicting species richness for each kilometre square. We found that the final models performed similarly well. Average elevation was the best single variable for explaining plant species richness nationwide. Species richness maps typically showed belt-like patterns of highest richness at intermediate altitudes. We discussed different approaches for explaining such “mid-elevational peaks” of species richness.
In the frame of the BDM vascular plants, butterflies and birds are surveyed on the same sites during the same years. These simultaneous studies may be considered as a major advantage of the BDM compared to the monitoring programs in other countries. In the final two chapters we therefore inferred patterns between the species groups. Chapter IV is based on data of the first iteration of surveys. We looked at the changes that had happened in surveyed species communities of plants, birds and butterflies within the period of 5 years. As a response to climate warming we expected species to shift their distribution towards higher altitudes. We used the “Community Temperature Index” (CTI) to test for differences in reaction to climate change. As expected, in the lowlands birds and butterflies tracked climate warming with an average uphill shift of 42 and 38m respectively, while plants showed a shift of only 8m. At higher elevations there was no significant CTI change in plants and butterflies. In general our results supported the idea that reactions to climate change in alpine landscapes were lowest and alpine landscapes could be safer places because of their highly varied surfaces. In the study in Chapter V we examined to what extent distribution patterns of butterfly species are shaped by interactions with their individual host plants or, alternatively, by environmental factors. Our findings indicated that butterfly - host plant interactions were not relevant in benign environments. In contrast, at the cold distribution limits there was a strong coincidence between butterfly and plant ranges. We argued that this could be evidence for butterfly species being limited by the distribution of their host plants in harsh environments and discussed the implications of the findings under climate change conditions.
Finally I summarized the most important results and also included more recent experiences from other studies using BDM data and from unpublished analyses, e.g. from quality control. I concluded in discussing the strength and weaknesses of long-monitoring programmes and pointed out that they should be considered as a complementary data source and reference for experimentally orientated research.