Abstract: The groundwater environment during tunnel construction is highly dynamic, making it challenging to fully capture the uncertainties of tunnel construction risks under complex multidimensional geological conditions using a one-dimensional cloud model. To address this, a dynamic evaluation model for water inrush and mud burst risks in tunnels is proposed, combining a two-dimensional cloud model with fuzzy mathematics theory and probabilistic algorithms. The model evaluates risks from two dimensions: severity of consequences and likelihood of occurrence. A fuzzy mathematics normal distribution membership function is used to quantify and transform the traditional risk matrix. Deviation degree is introduced to replace membership degree, further improving the two-dimensional cloud model. Multi-source data acquisition and expert system are used to generate cloud droplets, whose distribution reflects risk uncertainty and quantifies risk probabilities. This approach enables the determination of risk levels and early warning states. Engineering application results demonstrate that this method dynamically updates risk environment information as tunnel construction progresses and provides real-time assessments of water and mud inrush risks.