The precise regulation of gene expression is crucial for almost all biological processes.  In development, spatio-temporal patterns of gene expression are controlled by extensive regulatory networks, where the activity of transcription factors converge on cis-regulatory modules (CRMs) or enhancer elements.  The location and even combinatorial occupancy of CRMs can be experimentally measured using ChIP-seq at specific stages of development, at high-resolution.  A current major challenge however, is how to interpret these transcription factor's binding data in terms of the resulting spatio-temporal enhancer activity.  Using the integration of a machine learning approach with enhancers of known activity, we recently demonstrated that transcription factor occupancy alone is sufficient to predict enhancer spatio-temporal activity during development.  We have now complemented this by generating cell-type specific information on chromatin state within the context of a developing embryo using a new method that we developed called BiTS-ChIP.  The data reveals heterogeneous combinations of chromatin marks linked to active enhancers.  Using a Bayesian network, we show that chromatin state is sufficient to predict, not just the location, but activity state of regulatory elements, accurately distinguishing between enhancers in an active versus inactive state.  The model revealed that Pol II occupancy is highly predictive for the precise timing of enhancer activity and is tightly correlated with both the timing and location of transcription factor occupancy.  Taken together, this approach provides a systematic and high-resolution view of dynamic enhancer usage during development, and essential step toward deciphering developmental networks.