High-precision river temperature models are of great significance for the comprehensive understanding of spatiotemporal characteristics of river temperature and ecosystem restoration. The data-driven air2stream model, known for ensuring prediction accuracy while avoiding computational complexity, has become a commonly used model in river temperature simulation. However, the original air2stream model does not consider the lag effect of river temperature, primarily caused by factors such as water's thermal inertia and hydrological conditions. This leads to the decreased practical accuracy of the model when flow data is unavailable. To address this issue, this paper employed Pearson correlation coefficient between air temperature and water temperature to determine time lag parameters, and constructed a new air2stream model with time lag. Furthermore, the new model's effectiveness and stability were validated by multi-year observed data from two monitoring stations in the middle and lower reaches of the Yangtze River. The results indicate that the new model offers higher accuracy and stability without additional observation data. Compared with the original model, the Root Mean Square Errors of the new model at the two monitoring stations were reduced by approximately 4.29% and 5.85% respectively. The new model is characterized by high precision and low hydrological data requirements, and can provide a basis for water environmental impact assessment and ecological protection in the middle and lower reaches of the Yangtze River.