This paper describes a service oriented architecture for mobile and web applications and the enablement of participatory observations of the environment. The architecture hosts generic microbial risk forecast models in bathing zones, which are trained by heterogeneous input data. Open observation data sources, specializing in water quality indicators and environmental processes are used for the construction of such applications. Nevertheless, the encountered integration of the open data sources was challenging due to the various incompatibilities found in data samples. These included gaps in the data with diverse temporal and spatial coverage as well as conflicting collection policies.