van der Veen has explored and further developed new approaches for multispecies modelling in ecology. The project is to a large extent built around the development and application of a type of statistical models referred to as generalized linear latent variable models (GLLVM). These models are very flexible and can be used for modelling species distributions, and composition of ecological communities and how composition is influenced by variation in the environment. The analysis of community composition in ecology has traditionally been done using ordination techniques. However, the classic ordination techniques used in community ecology have some important theoretical and practical issues and limitations. In this PhD project, GLLVM is used as a framework to develop model-based ordination methods that can be an alternative to classic ordination techniques, and have several benefits compared to these. Model-based ordination using latent variable models is a more flexible and transparent framework for analysis of community composition compared to classic ordination. The model-based ordination techniques are, as the name indicates, based on statistical models with assumptions that can be checked and evaluated. The GLLVM framework makes it possible to include additional random effects in the model if necessary, and it is thus possible to create models that better represent the data. Another important benefit of using the GLLVM framework is that model building and model selection can be done with many of the same tools as in more standard regression models. The PhD thesis includes seven papers and manuscripts. In one of the published papers, a new method is proposed that relaxes the assumption of equal niche widths for species in model-based ordination. Model-based constrained ordination is developed as a new extension to existing model-based ordination methods. Model-based ordination is also used to build a statistical model that predicts niche overlap among species. The thesis provides examples of how the models and techniques developed in the project can be used in more applied ecological research. Finally, the thesis includes two manuscripts with more conceptual discussions of the way forwards for developing statistical tools for multispecies modelling in ecology. Wenng?s work focused on the connections between agricultural practices, climate, and freshwater environment. On a Nordic scale, Wenng contributed to two papers. The first discussed the potential impacts of a future Nordic bioeconomy on surface water quality, advising that a sustainable implementation would be best for the water resources, and recommending maintaining long time series of water quality and quantity data in the Nordic countries. The second examined 69 headwater catchments with different land use, concluding amongst others that catchments under forestry had larger nutrient losses than near-pristine catchments, although lower than in agricultural catchments. Next, the scale of the study was narrowed down to Norwegian conditions, first by examining seven small catchments at different locations of the country, and performing a multi-criteria analysis of, amongst others, trends in the length of the growing season and nitrogen losses. Nitrogen losses are receiving increased attention in Norway, especially since the Oslo Fjord has major ecological challenges. It was found that in catchments with cereal production, the increased growing season length corresponded to a reduction in nitrogen concentrations, but there was no such relationship in grassland catchments. Improved understanding of catchment processes was also the focus area of a fourth paper from this project, where two catchments in South-Eastern Norway were studied. Here, turbidity sensors were used as a proxy to understand sources and pathways of sediments, and hence, sediment associated substances such as phosphorus. By studying the way turbidity increased and decreased in storm events it was found that the turbidity peak came before the water discharge peak, which can imply that the sources of both sediments and phosphorus derive from locations close to the river course. This, again, increases our understanding on which environmental measures will have the best effect on water quality. The fifth paper concentrated on water discharge patterns. Seven small headwater catchments were studied, but this time by scrutinizing 22-26 years of meteorological and hydrological data. A significant increase was found in annual and seasonal air temperature. Annual changes in hydrology were more difficult to detect, while seasonal differences were more apparent, with increasing trends in water discharge primarily in winter and autumn. This confirms that the Norwegian winter might no longer be a hydrological inactive season and further demonstrated the hydrological differences between rain-fed and snow-fed regions, and highlighted the importance of studying local hydrological regimes.
Stipendiat-prosjektet til Wenng har gitt økt kunnskap om viktige tema knyttet til effekten på vannressurser av kombinasjonen bioøkonomi og klimaendringer,
Some of the statistical methods developed as part of the PhD project to van der Veen has been made available to the research community through the R package gllvm (https://cran.r-project.org/web/packages/gllvm/). Bert van der Veen has also been included as an author on this R package.