To make the diverse organizations access the massive distributed heterogeneous spatial data easily, we build a Spatial Data Grid in peer-to-peer way, which composed of several spatial datacenters distributed in different agencies. Given a query, which often a range query in spatial application, the query parser will decompose the global query into several sub-queries and disseminate them to the peers who take responsibility for storing data of that area. In order to accelerate the query dissemination process, we utilize distributed index to locate object more quickly and accurately. We propose an adaptive indexing mechanism, Fully Distributed R-Tree Index, which is appropriate for efficient range query dissemination. It composes of two layers, on top of which is the spatial range of each peer. The under layer stores part of the R-Tree of each peer's neighbors, which can be adjusted according to the capacity of peer. There is no centralized control on top of the entire system, and it could adaptive to the environment change.