A data assimilation approach is demonstrated whereby seismic inversion is both automated and enhanced using a comprehensive numerical sedimentary basin simulator to study the physics and chemistry of sedimentary basin processes in response to geothermal gradient in much greater detail than previously attempted. The approach not only reduces costs by integrating the basin analysis and seismic inversion activities to understand the sedimentary basin evolution with respect to geodynamic parameters-but the technique also has the potential for serving as a geoinfomatics platform for understanding various physical and chemical processes operating at different scales within a sedimentary basin. Tectonic history has a first-order effect on the physical and chemical processes that govern the evolution of sedimentary basins. We demonstrate how such tectonic parameters may be estimated by minimizing the difference between observed seismic reflection data and synthetic ones constructed from the output of a reaction, transport, mechanical (RTM) basin model. We demonstrate the method by reconstructing the geothermal gradient. As thermal history strongly affects the rate of RTM processes operating in a sedimentary basin, variations in geothermal gradient history alter the present-day fluid pressure, effective stress, porosity, fracture statistics and hydrocarbon distribution. All these properties, in turn, affect the mechanical wave velocity and sediment density profiles for a sedimentary basin. The present-day state of the sedimentary basin is imaged by reflection seismology data to a high degree of resolution, but it does not give any indication of the processes that contributed to the evolution of the basin or causes for heterogeneities within the basin that are being imaged. Using texture and fluid properties predicted by our Basin RTM simulator, we generate synthetic seismograms. Linear correlation using power spectra as an error measure and an efficient quadratic optimization technique are found to be most effective in determining the optimal value of the tectonic parameters. Preliminary 1-D studies indicate that one can determine the geothermal gradient even in the presence of observation and numerical uncertainties. The algorithm succeeds even when the synthetic data has detailed information only in a limited depth interval and has a different dominant frequency in the synthetic and observed seismograms. The methodology presented here even works when the basin input data contains only 75 per cent of the stratigraphic layering information compared with the actual basin in a limited depth interval.