To investigate the risk propagation process of the watershed water resources system, a risk propagation model was established based on a risk multi-link network, which includes the coordinated initiation stage, propagation stage, and dissipation stage. Risk propagation indicators, such as node risk sensitivity index, link carrying risk, network risk diffusion index and network structure vulnerability index, were proposed. Through these indicators, the comprehensive analysis of network risk propagation under typical hydrological years was conducted. And risk prevention strategies were constructed, such as betweenness control, degree control, node risk sensitivity index control, and random control. The results indicate that there is a positive correlation between the node risk sensitivity index and the node load increment. The risk carried by the link is negatively correlated with the number of transmissions, and positively correlated with the increase in load of the risk source. In general, the network risk diffusion index shows a first increasing and then decreasing trend, while the network structure vulnerability index shows a continuous decreasing trend, and both dynamically change with the attenuation coefficient and load retention coefficient. Under typical hydrological years, risk propagation indicators change significantly in extremely dry years and dry years. When slowing down the spread of risks, the prevention and control effects are ranked from good to poor as betweenness centrality prevention and control, degree value prevention and control, node risk sensitivity index prevention and control, random prevention and control, and no prevention and control strategies, which means prioritizing the regulation of the correlation elements with high betweenness centrality can reduce the spread rate of network risks. When reducing the number of risk nodes, the order of prevention and control effectiveness from good to poor is node risk sensitivity index prevention and control, random prevention and control, betweenness prevention and control, degree value prevention and control, and no prevention and control strategy. This indicates that prioritizing the control of associated elements with high node risk sensitivity index can reduce the vulnerability of network structure.