Zenodo (https://zenodo.org) is an open-access repository operated by CERN (European Organization for Nuclear Research), which provides researchers with an easy and stable platform to archive and publish their data and other output, such as software tools, manuals and project reports. In the context of the ICEDIG (Innovation and Consolidation for Large scale Digitisation of Natural Heritage) project, Zenodo was investigated for its usability as a platform where digitized images of collection specimens could be archived and published. In a production digitization pipeline, we foresee the automated archiving of daily image production. If Zenodo could be used for this purpose, such a process would also immediately mean that data and images are published FAIR-ly (Findable, Accessible, Interoperable and Reusable) within hours of their creation.<br> To evaluate performance of the system, we first used a test dataset of 1800 herbarium specimen images, which was uploaded using Zenodo's API (Application Programming Interface) (Dillen et al. 2019). This dataset includes lossless TIFF images, label-segmented overlays and JSON-LD (JavaScript Object Notation for Linked Data) metadata using DwC (Darwin Core) terminology, constituting over 208 gigabytes of data. In addition, for all individual digital specimens the data about the specimen (in DwC) as well as metadata about its deposition on Zenodo (in Zenodo's internal data model) were available in multiple machine-readable formats. All data in DwC were provided as linked data with their DwC identifiers (e.g. http://rs.tdwg.org/dwc/terms/basisOfRecord). All individual specimens received minted DOIs (Digital Object Identifiers). A second upload of 280,000 herbarium JPEG images from a single institution (ca. 1 terabyte of data) with limited metadata (but using the same approach) was launched as well.<br> In this presentation, the workflow for proper usage of the API will be described as well as some performance metrics, flexibilities and functionalities of the platform. Some issues and potential developments to tackle them will be discussed. Currently, the rate of ingestion into Zenodo seems only fast enough for small scale digitization pipelines. However, a modest improvement in transfer rate would make this a realistic proposition for large volume usage.
