To address the problem that the non-euclidean characteristic of spatio-temporal data restricts the efficiency of data clustering and query pruning in restricted network, we proposed a quotient space oriented non-euclidean spatial-temporal indexing model based on the domain of water resource data. Through the modeling and analysis of the network of river basin, in conjunction with the non-euclidean characteristic of restricted spatial-temporal data, we introduced the multi-granularity of quotient space to model the spatial-temporal data, utilized the query algorithms of spatial network to convert non-euclidean query into query in Euclidean space, constructed the framework of index management model, and then verified the model. Experimental result showed that this work provides a feasible solution to model the index of non-euclidean data.