Researchers calibrate urban flood models using machine learning latent variables and adjoint equations

Researchers have developed a new method for calibrating urban underlying surface parameters, which is essential for accurate urban flood simulations. This approach frames the calibration as an optimization problem within a Bayesian framework, incorporating latent variables to account for uncertainties. To enhance efficiency, the method utilizes the adjoint equation of the surrogate model for gradient information and employs parameter sharing and localization techniques to reduce computational complexity. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces novel techniques for parameter calibration in complex simulations, potentially improving accuracy in environmental modeling.

RANK_REASON This is a research paper detailing a new method for parameter calibration in urban flood simulations. [lever_c_demoted from research: ic=1 ai=0.4]

Read on arXiv cs.LG →

• arXiv
• Bayesian framework
• Manning's coefficient

• paper
• other