Many applications for global climate change research heavily rely on analyzing geospatial data collaboratively and always accompanying with complex computational algorithms. It is desirable to study a flexible and efficient approach for easy interoperation among heterogeneous datasets and computation models, thus providing quick responses for decision making. Early research on services-based mechanisms can ensure interoperability of geo-data resources and interoperability of geo-processing sources, but cannot ensure interoperability of both geo-data resources and geo-processing sources simultaneously due to their different characteristics on service deployment and management. An important issue is to find a way to keep diversity of data sources and maintain advantages of various geo-processing algorithms whereas data resources and processing sources can be chained together at a certain level of granularity. To address this need, we propose Data Service Entity Node (DSEN) to take the spatial data sources and geo-processing function as isomorphic service in a uniform manner based on service clustering method. We have applied our on-demand data service model to China Spatial Information Grid (SIG) test-bed and report a case study of flood monitoring application. The result shows our advantages and powerful capabilities by running two algorithms and process models for water extraction in an efficient way. The proposed on-demand approach can be used to build reusable spatial data service and respond quickly, and dynamically support different requirements of targeted applications.