We analyze data and meta-data modeling challenges to provide curated collections that can be easy to explore. Data exploration can use provenance tools to give insight into the conditions in which data are collected. We are concerned with data curation and exploration in seismic geophysics. We believe that the tasks involved in graph exploration depend highly on the knowledge domain. The discussion about possible solutions is driven by the hypothesis that graphs can be well-adapted data models to represent, explore and analyze seismic data. Given that data curation is done by human agents, it is essential to provide automatic tools to add provenance to seismic data.