The potential of co-location to mitigate sampling bias in Research Infrastructures

Research Infrastructures (RIs) are key to face the grand environmental challenges humanity encounters by providing standardized, large-scale data. However, the in-situ facilities of RIs frequently exhibit a sampling bias. A promising and cost-efficient approach to mitigate sampling bias is ”co-location”, i.e. the usage of in-situ facilities by more than one RI. We investigated the bias mitigation potential of co-location in Europe exemplarily for eLTER RI. Therefore, we added 50 candidate facilities from seven pan-European RIs and peer networks to the eLTER RI and investigated bias reduction for current environmental conditions and anticipated changes under RCP4.5 and RCP8.5. Additionally, we investigated which facilities contribute to bias reduction. It required 5, 10 and 25 of 832 candidate facilities, respectively, to reduce sampling bias for RCP8.5, RCP4.5 and current conditions. However, after the addition of 50 sites, 60 % (RCP4.5) to 80 % (current conditions) of the initial bias remained. Bias reduction was most conclusive in Eastern Europe and Fennoscandia, while bias in the Mediterranean was only removed for RCP4.5. The analysis recommended 197 candidate facilities for bias reduction, 156 in Fennoscandia, Poland, Lithuania and Spain. Thus, co-location alone can mitigate but not overcome bias in eLTER RI. The mitigation potential was limited by the small number of candidate facilities in the Iberian Peninsula. Therefore, for eLTER RI we recommend: (i) aim to co-locate the facilities recommended by this analysis, (ii) search for or establish further facilities in the Iberian Peninsula and (iii) mitigate bias in data analyses by using less biased or unbiased subsets. • First analysis of co-location on European scale. • Co-location alone can mitigate but not overcome bias within eLTER RI. • Facility recommendations driven by unifiying analyses of representativity and transferability. • Analysis provides clear overview which facilities are most important in mitigating bias.