Researchers have developed an adaptive surrogate modeling method for problems with high-dimensional spatio-temporal outputs. This approach uses dimension reduction to map the output to a lower-dimensional space and then constructs a surrogate model in that space. The method also includes an adaptive sampling technique to improve the surrogate model's accuracy with minimal runs of the expensive physics-based model. The authors demonstrate their method using thermo-mechanical analysis of a gas turbine engine blade. --- Why it matters: This work is important for researchers and engineers working on complex systems with high-dimensional outputs, such as climate models or materials science simulations, who need to balance computational efficiency with accuracy. Source: https://arxiv.org/abs/2608.17250