In-situ measurements of the scattering function are used to drive a channel simulator developed in the context of underwater acoustic telemetry. Two operation modes of the simulator are evaluated. A replay mode is accomplished by interpolation of measured impulse responses. A second, stochastic mode delivers multiple realizations of a given scattering function. The initial assumption of wide-sense stationary uncorrelated scattering is violated by strong phase correlations between taps. It is shown that time-varying Doppler shifts due to platform motion must be eliminated from measured scattering functions in order to provide the stochastic tap gains with the true Doppler spectrum of the channel. The simulator is validated through a comparison of acoustic data measured at sea, and emulated data, governed by the same scattering function. This comparison is based on scattering and coherence functions, multipath phase measurements, and application of a decision feedback equalizer. After the Doppler correction, the synthetic data are indistinguishable from the acoustic data in terms of delay-Doppler spread, temporal coherence, phase behavior, equalizer mean square error, and bit error ratio.