marinedb: An open-source, flexible and scalable Python workflow for curating, standardizing, and integrating marine biodiversity data

Abstract. The digital era and the emergence of data-aggregation initiatives have greatly expanded access to biodiversity occurrence records, but accessibility alone does not ensure usability for ecological modeling. Marine occurrence data must be cleaned and reconciled across spatial, temporal, and taxonomic dimensions before they can be reliably linked to environmental predictors and used to model species distributions. Here, we introduce marinedb, an open-source, modular, and scalable Python package for curating marine occurrence data. The package provides a suite of tools that can be freely combined into user-defined workflows tailored to study objectives and data characteristics. We applied marinedb to three major biodiversity resources: the Global Biodiversity Information Facility (GBIF), the Ocean Biogeographic Information System (OBIS), and the Jellyfish Database Initiative (JeDI). Together, these resources span hundreds of millions of occurrence records located at sea and more than one hundred thousand species. First, we used the framework to assess data quality non-destructively across the three sources. The analysis revealed substantial and source-dependent issues, including records located on land, missing depth information, heterogeneous, incomplete, or low-precision temporal information, records lacking species-level information or referring to non-marine taxa, weak or unspecified evidential support, absence records, and strong taxonomic imbalance. Second, we applied a tailored dataset-production workflow to produce a ready-to-use global marine occurrence dataset for large-scale modeling of present-day species distributions, which comprises 158,910,536 records representing 139,844 species. Together, the reusable Python tools and the curated dataset derived from GBIF, OBIS, and JeDI lower the technical barrier for multi-species and deep-learning approaches, both of which require extensive and diverse occurrence records, while preserving flexibility for future use-case-specific curation.