Spatial Network Design of the eLTER Research Infrastructure

Abstract

Research Infrastructures (RIs) are key to face the grand environmental challenges humanity encounters by providing standardized, accessible, large-scale data. Yet, observation locations in RIs tend to be biased. These biases cannot simply be judged by spatial distance. Rather, meaningful RI-specific indicators describing a status quo and future conditions should be the basis to uncover biases and inform an RI’s spatial network design. This thesis studies the spatial network design of the Integrated European Long-Term Ecosystem, critical zone and socio-ecological Research Infrastructure (eLTER RI). Therefore, eLTER RI’s target region was defined and a survey was conducted to describe eLTER RI’s network of in-situ facilities. As relevant indicators, six indicators collectively describing current ecosystem functioning were chosen and four indicators describing Global Change Pressure on ecosystems were developed. Based on these indicators, using a representativity analysis sampling bias was identified. Lastly, it was investigated whether in-situ facilities of other RIs can resolve sampling bias within eLTER RI. eLTER RI currently consists of 204 in-situ facilities, which underrepresent agricultural lands, dry bioclimates, areas with low economic density, areas projected to experience a strong increase and a strong decrease in precipitation, little change in biotemperature, both strong and hardly any land use change, and areas where water availability during a year becomes more uniform. Spatially, these conditions and therefore underrepresented regions cluster in the Iberian Peninsula, Fennoscandia and Eastern Europe. Facilities of other RIs can convincingly counteract bias in Eastern Europe, but not in the other regions. As long as biases within an RI persist, they must be addressed in data analyses, e.g. by removing data or explicitly mention biases uncovered. Only an unbiased in-situ facility network would allow confident upscaling of all data. The procedure to design the spatial network of an RI outlined in this thesis can function as a blueprint for continental or global scale RI development towards minimizing sampling bias.