Sensor networks employed by scientific applications often need to support localized collaboration of sensor nodes to perform in-network data processing. This includes new quantitative synthesis and hypothesis testing in near real time, as data streaming from distributed instruments, to transform raw data into high level domain-dependent information. This paper investigates in-network data processing mechanisms with dynamic data requirements in resource constrained heterogeneous sensor networks. Particularly, we explore how the temporal and spatial correlation of sensor measurements can be used to trade off between the complexity of coordination among sensor clusters and the savings that result from having fewer sensors involved in in-network processing, while maintaining an acceptable error threshold. Experimental results show that the proposed in-network mechanisms can facilitate the efficient usage of resources and satisfy data requirement in the presence of dynamics and uncertainty.