Key points are not available for this paper at this time.
Reproducibility is a cornerstone of the scientific method, yet it remains elusive in many domains of contemporary research, including the geosciences. In this Perspective, we examine reproducibility in the context of computational workflows for geospatial analyses. Building on established frameworks, we disentangle key dimensions — methodological, results, and inferential reproducibility — and identify critical barriers, including irreproducibility arising from obscurity, obfuscation, and uncontrollable conditions. We argue that enhancing reproducibility in geoscientific research requires both cultural transformation and practical, domain-specific interventions. Our focus lies on methodological and computational reproducibility, with particular attention to challenges posed by spatial data structures, diverse data sources and infrastructures, and the integration of statistical and machine-learning methods. We outline actionable guidance across the research workflow, including data governance, analysis design and documentation, code development, and long-term accessibility. Emphasis is placed on the use of open-source software, script-based automation, version control, and the adoption of FAIR principles to support findability, accessibility, interoperability, and reusability of data and code. Our aim is to provide a structured synthesis that supports reproducible and transparent geospatial research. By implementing even incremental improvements, researchers can strengthen the robustness, transparency, and reuse potential of their scientific contributions.
Schlögl et al. (Sat,) studied this question.