The increasing amount of publicly available data streams of environmental observation stations opens up new opportunities: do- main experts in the field of environmental observations are provided with extensive observations covering large areas with high density of environmental sensors, which could hardly ever be provided by a single organization. However, these opportunities come at the cost of new challenges regarding trustworthiness and comparability of such observations. In this paper, we address the challenges of semantic validation and enrichment of heterogeneous observation streams by exploiting collaboratively created and curated annotations. For this purpose, we introduce and discuss the Linked Stream Annotation Engine (LSane) to validate obser- vation messages from heterogeneous sensors. We enrich these observation messages with statements derived from annotations of provenance information of observations. We showcase an implementation of LSane with observation messages from public and private environmental observation stations which are mapped to explicit semantics, and validate and enrich the mapped messages based on annotations from a collaboration platform.