Connected environments are infrastructures providing valuable data and services for applications in various domains, e.g., energy, transportation, environmental domains, to cite a few. Indeed, the resources of the connected environment and their sensed observations should be available upon requests by authorised users or machines from any location at any time. To this end, efficient indexing and discovery mechanisms are required to swiftly locate and retrieve relevant resources/observations that fulfil the query requirements. In this respect, we introduce here a new scalable indexing and discovery scheme in large scale connected environments. One of the main thrusts of the proposed scheme is its fully decentralised architecture that enables to obtain a higher scalability. In addition, a formal model is provided for describing the connected environment resources, their observations as well as their sensing zones. Interestingly enough, our scheme relies on geospatial indexing techniques allowing to reduce the search space and then to swiftly locate the target resource/observations. Furthermore, a resource ranking mechanism is provided to consider the environment constraints and the resources capabilities. The performed experiments show the effectiveness and the efficiency of the introduced scheme compared to a baseline approach from the literature.