We have assembled and tested, in real time, a space weather modeling system that starts at the Sun and extends to the Earth through a set of coupled, modular components. We describe recent efforts to improve the Hakamada-Akasofu-Fry (HAF) solar wind model that is presently used in our geomagnetic storm prediction system. We also present some results of these improvement efforts. In a related paper, Akasofu [2001> discusses the results of the first 2 decades using this system as a research tool and for space weather predictions. One key goal of our efforts is to provide quantitative forecasts of geoeffective solar wind conditions at the L1 satellite point and at Earth. Notably, we are addressing a key problem for space weather research: the prediction of the north-south component (Bz) of the interplanetary magnetic field. This parameter is important for the transfer of energy from the solar wind to the terrestrial environment that results in space weather impacts upon society. We describe internal improvements, the incorporation of timely and accurate boundary conditions based upon solar observations, and the prediction of solar wind speed, density, magnetic field, and dynamic pressure. HAF model predictions of shock arrival time at the L1 satellite location are compared with the prediction skill of the two operational shock propagation models: the interplanetary shock propagation model (ISPM) and the shock-time-of-arrival (STOA) model. We also show model simulations of shock propagation compared with interplanetary scintillation observations. Our modeling results provide a new appreciation of the importance of accurately characterizing event drivers and for the influences of the background heliospheric plasma on propagating interplanetary disturbances. ¿ 2001 American Geophysical Union