The recent technological advancements and emergence of the open data in environmental and life sciences are opening new research opportunities while creating new challenges around data management. They make available an unprecedented amount of data that can be exploited for studying complex phenomena. However, new challenges related to data management need to be addressed to ensure effective data sharing, discovery and reuse, especially when dealing with interdisciplinary research contexts. These issues are magnified in interdisciplinary context, by the fact that each discipline has its practices, e.g., specific formats and metadata standards. Moreover, the majority of current data management practices do not consider semantic heterogeneity existing among disciplines. For this reason, we introduce a flexible metadata model that describes the datasets of various disciplines using a common paradigm based on the observation concept. It provides a key vision for articulating the user point of view and underlying scientific domains. In this study, we therefore decide to mainly reuse the SOSA lightweight ontology (Sensor, Observation, Sample, and Actuator) to efficiently leverage others existing ontologies to improve datasets discovery and reuse coming from Earth and life observation. The main benefit of the proposed metadata model is that it extends the technical description, usually provided by existing metadata models, with the observation context description enabling the need of a user viewpoint. Moreover, following the FAIR principles, the metadata model specifies the semantics of its elements using ontologies and vocabularies, and reuses as much as possible ontological and terminological existing resources. We show the benefit and applicability of the model through a case study we identified as representative after interviewing researchers in environmental and life sciences.