The large number of geo referenced data sets provided by Open Data portals, social media networks and created by volunteers within citizen science projects (Volunteered Geographical Information) is pushing analysts to define and develop novel frameworks for analysing these multisource heterogeneous data sets in order to derive new data sets that generate social value. For analysts, such an activity is becoming a common practice for studying, predicting and planning social dynamics. The convergence of various technologies related with data representation formats, database management and GIS (Geographical Information Systems) can enable analysts to perform such complex integration and transformation processes. JSON has become the de-facto standard for representing (possibly geo-referenced) data sets to share; NoSQL databases (and MongoDB in particular) are able to natively deal with collections of JSON objects; the GIS community has defined the GeoJSON standard, a JSON format for representing georeferenced information layers, and has extended GIS software to support it. However, all these technologies have been separately developed, consequently, there is actually a gap that shall be filled to easily manipulate GeoJSON objects by performing spatial operations. In this paper, we pursue the objective of defining both a unifying view of several NoSQL databases and a query language that is independent of specific database platforms to easily integrate and transform collections of GeoJSON objects. In the paper, we motivate the need for such a framework, named J-CO, able to execute novel high-level queries, written in the J-CO-QL language, for JSON objects and will show its possible use for generating open data sets by integrating various collections of geo-referenced JSON objects stored in different databases.