This thesis comprises an empirical case study of scientific data production in oceanography and a philosophical analysis of the relations between newly created scientific data and the natural world. Based on qualitative interviews with researchers, I reconstruct research practices that lead to the ongoing production of digital data related to long-term developments of plankton biodiversity in the oceans. My analysis is centred on four themes: materiality, scientific representing with data, methodological continuity, and the contribution of non-scientists to epistemic processes. These are critically assessed against the background of today’s data-intensive sciences and increased automation and remoteness in oceanographic practices. Sciences of the world’s oceans have by and large been disregarded in philosophical scholarship thus far. My thesis opens this field for philosophical analysis and reveals various conditions and constraints of data practices that are largely uncontrollable by ocean scientists. I argue that the creation of useful scientific data depends on the implementation and preservation of material, methodological, and social continuities. These allow scientists to repeatedly transform visually perceived characteristics of research samples into meaningful scientific data stored in a digital database. In my case study, data are not collected but result from active intervention and subsequent manipulation and processing of newly created material objects. My discussion of scientific representing with data suggests that scientists do not extract or read any intrinsic representational relation between data and a target, but make data gradually more computable and compatible with already existing representations of natural systems. My arguments shed light on the epistemological significance of materiality, on limiting factors of scientific agency, and on an inevitable balance between changing conditions of concrete research settings and long-term consistency of data practices.

There is a pressing need for creative new data analysis methods whichcan sift through scientific simulation data and produce meaningfulresults. The types of analyses and the amount of data handled by currentmethods are still quite restricted, and new methods could providescientists with a large productivity boost. New methods could be simpleto develop in big data processing systems such as Apache Spark, which isdesigned to process many input files in parallel while treating themlogically as one large dataset. This distributed model, combined withthe large number of analysis libraries created for the platform, makesSpark ideal for processing simulation output.Unfortunately, the filesystem becomes a major bottleneck in any workflowthat uses Spark in such a fashion. Faster transports are notintrinsically supported by Spark, and its interface almost denies thepossibility of maintainable third-party extensions. By leveraging thesemantics of Scala and Spark's recent scheduler upgrades, we forceco-location of Spark executors with simulation processes and enable fastlocal inter-process communication through shared memory. This provides apath for bulk data transfer into the Java Virtual Machine, removing thecurrent Spark ingestion bottleneck.Besides showing that our system makes this transfer feasible, we alsodemonstrate a proof-of-concept system integrating traditional HPC codeswith bleeding-edge analytics libraries. This provides scientists withguidance on how to apply our libraries to gain a new and powerful toolfor developing new analysis techniques in large scientific simulationpipelines.