The European landscape is dominated by intensive agriculture which leads to widespread impact on the environment. The frequent use of agricultural pesticides is one of the major causes of an ongoing decline in flower-visiting insects (FVIs). The conservation of this ecologically diverse assemblage of mobile, flying insect species is required by international and European policy. To counteract the decrease in species numbers and their abundances, FVIs need to be protected from anthropogenic stressors. European pesticide risk assessment was devised to prevent unacceptable adverse consequences of pesticide use on FVIs. However, there is an ongoing discussion by scientists and policy-makers if the current risk assessment actually provides adequate protection for FVI species. The first main objective of this thesis was to investigate pesticide impact on FVI species. The scientific literature was reviewed to identify groups of FVIs, summarize their ecology, and determine their habitat. This was followed by a synthesis of studies about the exposure of FVIs in their habitat and subsequent effects. In addition, the acute sensitivity of one FVI group, bee species, to pesticides was studied in laboratory experiments. The second main objective was to evaluate the European risk assessment for possible deficits and propose improvements to the current framework. Regulatory documents were screened to assess the adequacy of the guidance in place in light of the scientific evidence. The suitability of the honey bee Apis mellifera as the currently only regulatory surrogate species for FVIs was discussed in detail. The available scientific data show that there are far more groups of FVIs than the usually mentioned bees and butterflies. FVIs include many groups of ecologically different species that live in the entire agricultural landscape. Their habitats in crops and adjacent semi-natural areas can be contaminated by pesticides through multiple pathways. Environmentally realistic exposure of these habitats can lead to severe effects on FVI population parameters. The laboratory studies of acute sensitivity in bee species showed that pesticide effects on FVIs can vary greatly between species and pesticides. The follow-up critical evaluation of the European FVI risk assessment revealed major shortcomings in exposure and effect assessment. The honey bee proved to be a sufficient surrogate for bee species in lower tier risk assessment. Additional test species may be chosen for higher tier risk assessment to account for ecological differences. This thesis shows that the ecology of FVIs should generally be considered to a greater extent to improve the regulatory process. Data-driven computational approaches could be used as alternative methods to incorporate ecological trait data in spatio-temporal scenarios. Many open questions need to be answered by further research to better understand FVI species and promote necessary changes to risk assessment. In general, other FVI groups than bees need to be investigated. Furthermore, comprehensive data on FVI groups and their ecology need to be collected. Contamination of FVI habitat needs to be linked to exposure of FVI individuals and ecologically complex effects on FVI populations should receive increased attention. In the long term, European FVI risk assessment would benefit from shifting its general principles towards more scientifically informed regulatory decisions. This would require a paradigm shift from arbitrary assumptions and unnecessarily complicated schemes to a substantiated holistic framework.

Since the invention of U-net architecture in 2015, convolutional networks based on its encoder-decoder approach significantly improved results in image analysis challenges. It has been proven that such architectures can also be successfully applied in different domains by winning numerous championships in recent years. Also, the transfer learning technique created an opportunity to push state-of-the-art benchmarks to a higher level. Using this approach is beneficial for the medical domain, as collecting datasets is generally a difficult and expensive process. In this thesis, we address the task of semantic segmentation with Deep Learning and make three main contributions and release experimental results that have practical value for medical imaging. First, we evaluate the performance of four neural network architectures on the dataset of the cervical spine MRI scans. Second, we use transfer learning from models trained on the Imagenet dataset and compare it to randomly initialized networks. Third, we evaluate models trained on the bias field corrected and raw MRI data. All code to reproduce results is publicly available online.

The distributed setting of RDF stores in the cloud poses many challenges. One such challenge is how the data placement on the compute nodes can be optimized to improve the query performance. To address this challenge, several evaluations in the literature have investigated the effects of existing data placement strategies on the query performance. A common drawback in theses evaluations is that it is unclear whether the observed behaviors were caused by the data placement strategies (if different RDF stores were evaluated as a whole) or reflect the behavior in distributed RDF stores (if cloud processing frameworks like Hadoop MapReduce are used for the evaluation). To overcome these limitations, this thesis develops a novel benchmarking methodology for data placement strategies that uses a data-placement-strategy-independent distributed RDF store to analyze the effect of the data placement strategies on query performance. With this evaluation methodology the frequently used data placement strategies have been evaluated. This evaluation challenged the commonly held belief that data placement strategies that emphasize local computation, such as minimal edge-cut cover, lead to faster query executions. The results indicate that queries with a high workload may be executed faster on hash-based data placement strategies than on, e.g., minimal edge-cut covers. The analysis of the additional measurements indicates that vertical parallelization (i.e., a well-distributed workload) may be more important than horizontal containment (i.e., minimal data transport) for efficient query processing. Moreover, to find a data placement strategy with a high vertical parallelization, the thesis tests the hypothesis that collocating small connected triple sets on the same compute node while balancing the amount of triples stored on the different compute nodes leads to a high vertical parallelization. Specifically, the thesis proposes two such data placement strategies. The first strategy called overpartitioned minimal edge-cut cover was found in the literature and the second strategy is the newly developed molecule hash cover. The evaluation revealed a balanced query workload and a high horizontal containment, which lead to a high vertical parallelization. As a result these strategies showed a better query performance than the frequently used data placement strategies.

Constituent parsing attempts to extract syntactic structure from a sentence. These parsing systems are helpful in many NLP applications such as grammar checking, question answering, and information extraction. This thesis work is about implementing a constituent parser for German language using neural networks. Over the past, recurrent neural networks have been used in building a parser and also many NLP applications. In this, self-attention neural network modules are used intensively to understand sentences effectively. With multilayered self-attention networks, constituent parsing achieves 93.68% F1 score. This is improved even further by using both character and word embeddings as a representation of the input. An F1 score of 94.10% was the best achieved by constituent parser using only the dataset provided. With the help of external datasets such as German Wikipedia, pre-trained ELMo models are used along with self-attention networks achieving 95.87% F1 score.

Water uptake, respiration and exudation are some of the biological functions fulfilled by plant roots. They drive plant growth and alter the biogeochemical parameters of soil in the vicinity of roots, the rhizosphere. As a result, soil processes such as water fluxes, carbon and nitrogen exchanges or microbial activity are enhanced in the rhizosphere in comparison to the bulk soil. In particularly, the exudation of mucilage as a gel-like substance by plant roots seems to be a strategy for plants to overcome drought stress by increasing soil water content and soil unsaturated hydraulic conductivity at negative water potentials. Although the variations of soil properties due to mucilage are increasingly understood, a comprehensive understanding of the mechanisms in the pore space leading to such variations is lacking. The aim of this work was to elucidate the gel properties of mucilage in the pore space, i.e. interparticulate mucilage, in order to link changes of the physico-chemical properties in the rhizosphere to mucilage. The fulfilment of this goal was confronted to the three following challenges: The lack of methods for in situ detection of mucilage in soil; The lack of knowledge concerning the properties of interparticulate mucilage; The unknown relationship between the composition and the properties of model substances and root mucilage produced by various species. These challenges are addressed in several chapters. In a first instance, a literature review picked information from various scientific fields about methods enabling the characterization of gels and gel phases in soil. The variation of soil properties resulting from biohydrogel swelling in soil was named the gel effect. The combined study of water entrapment of gels and gel phases in soil and soil structural properties in terms of mechanical stability or visual structures proved promising to disentangle the gel effect in soil. The acquired methodical knowledge was used in the next experiments to detect and characterize the properties of interparticulate gel. 1H NMR relaxometry allows the non-invasive measure of water mobility in porous media. A conceptual model based on the equations describing the relaxation of water protons in porous media was developed to integrate the several gel effects into the NMR parameters and quantify the influence of mucilage on proton relaxation. Rheometry was additionally used to assess mucilage viscosity and soil microstructural stability and ESEM images to visualize the network of interparticulate gel. Combination of the results enabled to identify three main interparticulate gel properties: The spider-web effect restricts the elongation of the polymer chains due to the grip of the polymer network to the surface of soil particles. The polymer network effect illustrates the organization of the polymer network in the pore space according to the environment. The microviscosity effect describes the increased viscosity of interparticulate gel in contrast to free gel. The impact of these properties on soil water mobility and microstructural stability were investigated. Consequences on soil hydraulic and soil mechanical properties found in the literature are further discussed. The influence of the chemical properties of polymers on gel formation mechanism and gel properties was also investigated. For this, model substances with various uronic acid content, degree of esterification and amount of calcium were tested and their amount of high molecular weight substances was measured. The substances investigated included pectic polysaccharides and chia seed mucilage as model polymers and wheat and maize root mucilage. Polygalacturonic acid and low-methoxy pectin proved as non-suitable model polymers for seed and root mucilage as ionic interactions with calcium control their properties. Mucilage properties rather seem to be governed by weak electrostatic interactions between the entangled polymer chains. The amount of high molecular weight material varies considerably depending on mucilage´s origin and seems to be a straight factor for mucilage’s gel effect in soil. Additionally to the chemical characterization of the high molecular weight compounds, determination of their molecular weight and of their conformation in several mucilages types is needed to draw composition-property profiles. The variations measured between the various mucilages also highlight the necessity to study how the specific properties of the various mucilages fulfill the needs of the plant from which they are exuded. Finally, the integration of molecular interactions in gel and interparticulate gel properties to explain the physical properties of the rhizosphere was discussed. This approach offers numerous perspectives to clarify for example how water content or hydraulic conductivity in the rhizosphere vary according to the properties of the exuded mucilage. The hypothesis that the gel effect is general for all soil-born exudates showing gel properties was considered. As a result, a classification of soil-born gel phases including roots, seeds, bacteria, hyphae and earthworm’s exuded gel-like material according to their common gel physico-chemical properties is recommended for future research. An outcome could be that the physico-chemical properties of such gels are linked with the extent of the gel effect, with their impact on soil properties and with the functions of the gels in soil.