Diffusion Model-based Precipitation Downscaling
We generate satellite-based ensemble precipitation using a residual diffusion model.
Streamflow Prediction in Ungauged Regions with Uncertainty Quantification
Satellite-based ensemble streamflow prediction in ungauged regions is realized by leveraging satellite observations, hydrologic modeling, and uncertainty quantification methods.
Uncertainty Quantification for Data-Driving Hydrologic Modeling
Satellite precipitation, neural network parameter, training sample limitation and calibration target uncertainties are quantified to solve the over-confident issue in date-learning based ensemble streamflow prediction.
Uncertainty Quantification for Physics-based Hydrologic Modeling
Satellite precipitation uncertainty quantification greatly improves the reliability and accuracy of ensemble streamflow predictions. Hydrologic model parameter uncertainty is quantified simultaneously.
Satellite-based Precipitation Error Modeling
We created a first-ever novel near-Realtime quasi-global satellite-only ensemble precipitation dataset STREAM-Sat. It fills in the regions with no ground-based precipitation measurements.
