We develop a mean-field theory for Escherichia coli chemotaxis based on the coupled spatiotemporal dynamics of the cell population and the mean receptor methylation level field. This multiscale model connects the cells' population level motility behavior with the molecular level pathway dynamics. It reveals a simple scaling dependence of the chemotaxis velocity on the adaptation rate in exponential gradients. It explains the molecular origin of a maximum chemotaxis velocity. Simulations of our model in various spatiotemporal stimuli profiles show quantitative agreements with experiments. Moreover, it predicts a surprising reversal of chemotaxis group velocity in traveling wave environments. Our approach may be used to bridge molecular level pathway dynamics with cellular behavior in other biological systems.