16 January 2025 in Water Resources Research
Physics‐informed neural networks (PINNs) are increasingly being used in various scientific disciplines. However, dealing with non‐stationary physical processes remains a significant challenge in such models, whereas fluid motions are typically non‐stationary. In this study, a PINN‐based method was designed and optimized to solve non‐stationary fluid dynamics with shallow water equations in a polar coordinate system (PINN‐SWEP). It was developed and validated with a classic circular basin case that is well‐documented in scientific literature. In the validation case, the wind‐induced water surface fluctuations are less than 1 cm, posing challenges in modeling. However, our PINN‐SWEP model can accurately simulate such tiny water surface fluctuations and resolve complex fluid motions based on limited and sparse data. A boundary discontinuity problem associated with the use of a polar coordinate system is further discussed and improved, thereby enhancing the applicability of PINN in water research. The methodology can provide an alternative solution for numerical or analytical solutions with high accuracy.
5 November 2020 in Journal of Advances in Modeling Earth Systems
Particle filters are non‐Gaussian filters, which means that the assumption that the error distribution of the ensemble should be Gaussian is unnecessary. Like the ensemble Kalman filter, particle filters are based on the Monte Carlo approximation to represent the distribution of model states. It requires a substantial number of particles to approximate the probability density function of states in high‐dimensional models, which is prohibitive for real applications. In order to overcome problems with high dimensionality, localization was applied in an Ensemble‐type data assimilation system. This study combines the localization in LETKF (Local Ensemble Transformation Kalman Filter) with particle filters and proposes a new local particle filter with the model state space correction using Gamma test theory for high‐dimensional models. A series of tests with various parameter settings, including different the numbers of particles, observation intervals, localization scale, inflation factors, and observation operators, were used to evaluate the performance of this new method using a Lorenz model with 40 variables. Besides, the proposed filter was applied in the Lorenz model with 1,000 variables to evaluate its performance in the model with higher dimensions. The results show that this approach can deal with the issue of dimensionality, which otherwise leads to the collapse of the particle filters in high‐dimensional systems. The local particle filter is stable and has considerable potential for complex higher‐dimensional models.