Non-point source (NPS) pollution has become a major threat to watershed environment and security in recent decades. The occurrence and transport of NPS pollution include several dynamic processes which are characterized by randomness, complexity, uncertainty, etc. Existing monitoring network is not sufficient to accurately evaluate the contribution of NPS pollution and understand the underlying mechanisms. Efficient and sound assessment tools are needed to facilitate preventing and controlling NPS pollution. Nowadays many watershed models become increasingly popular to simulate, predict, and analyze NPS pollution processes and assist in decision makings for integrated watershed management. HSPF (Hydrological Simulation Program-FORTRAN) is one of the most comprehensive and flexible NPS models around the world. It has been developed and maintained mainly by the U.S. Environment Protection Agency and is highly recommended in flood-control planning and operations, water quality planning and management, conservation practices evaluation, etc. The HSPF model conceptualizes a watershed system into surface, soil, and stream segments to simulate the continuous, storm event, or steady-state behavior of both hydrologic and water quality processes in natural or manmade water systems. The model has proven its ability to be used with reasonable accuracy in diverse hydrologic regimes, which leads to its broad application worldwide. The research and practice of the HSPF model applied to watersheds in China have expanded rapidly in recent years, but the model is also facing challenges in terms of localization for model theory and parameters, refinement during the model setup process, and uncertainty embedded in the simulated results. Therefore, this paper focused on the research progress of the HSPF model in NPS pollution modeling and control by reviewing existing relative papers. The conceptualization principles of the HSPF model to simulate NPS pollution was first described. This was followed by extensive reviews of the modeling methods and application in changing environment, as well as the ideas and methods for parameter identification, uncertainty analysis, control measures evaluation and total emission control. The extended development of HSPF model under the modern environmental modeling situation was also summarized. PyHSPF and HSP² as two representatives were specifically discussed in their improvement in compiling environment, data management, dynamic representation of watershed properties, parallel computing, and Application Program Interface. The suggested developments in future studies include: localization model improvement for China's watershed characteristics, large-scale refinement modeling to cope with the fine regulatory services of the river chief system, coupling with big data statistics and artificial intelligence for crossfeed ensemble modeling.