hydropml Goto Github PK

dataset4hydropml

dataset4landslidenets

https://doi.org/10.5281/zenodo.10294997

disasternets

https://arxiv.org/abs/2306.09815

floodcast

Large-scale flood modeling and forecasting with FloodCast (https://doi.org/10.1016/j.watres.2024.122162)

floodcastbench

hydropml.github.io

Offitial website for HydroPML

hydropml_flood

The operational rainfall-runoff-inundation forecasting system based on HydroPML

hydropml_hydrodynamic

Hydrodynamic models are mathematical models that attempt to replicate fluid motion and typically require solving computationally. These models simulate water movement by solving equations formulated by applying laws of physics. Hydrodynamic models can realize the simulation of hydrological process in seconds, hours to days time scale.

hydropml_hydrodynamicbench

Large-scale data sets and benchmarks for hydrodynamic modeling based on physics-aware machine leaning

hydropml_rainfall_runoff

Rainfall-runoff Forecast Meets Physics-aware Machine Learning

landslidecast

HydroPML for landslide dynamic process modeling and forecast (https://doi.org/10.1029/2023EA003417)

paml

Physics-aware ML (PaML) aims to take the best from both physics-based modeling and state-of-the-art ML models to better solve scientific problems (https://arxiv.org/abs/2310.05227)

paml_pahl

Physics-aware Hybrid Learning (PaHL) directly combines pure physics-based models, such as numerical methods, climate, land, hydrology and earth system models, with ML models. According to the hybrid way, hybrid learning can be divided into serial way, parallel way, and complex way.

paml_pdgml

Physical Data-guided Machine Learning (PDgML) is a supervised DL model that statistically learns the known or unknown physics of a desired phenomenon by extracting features or attributes from raw training data.

paml_peml

Physics-embedded Machine Learning (PeML) is achevied by embedding physics in the model frameworks or modules.

paml_piml

PiML is a widely used approaches to incorporate physical constraints, which can be trained from additional information obtained by enforcing the physical laws (for example, designing loss functions (regularization))