Abstract Spatiotemporal patterns of large-scale spiking and field potentials of the rodent hippocampus encode spatial representations during maze run, immobility and sleep. Here, we showed that multi-site hippocampal field potential amplitude at ultra-high frequency band (FPA uhf ) provides not only a fast and reliable reconstruction of the rodent’s position in wake, but also a readout of replay content during sharp wave ripples. This FPA uhf feature may serve as robust real-time decoding strategy from large-scale (up to 100,000 electrodes) recordings in closed-loop experiments. Furthermore, we developed unsupervised learning approaches to extract low-dimensional spatiotemporal FPA uhf features during run and ripple periods, and to infer latent dynamical structures from lower-rank FPA uhf features. We also developed a novel optical flow-based method to identify propagating spatiotemporal LFP patterns from multi-site array recordings, which can be used for decoding application. Finally, we developed a prospective decoding strategy to predict animal’s future decision in goal-directed navigation.