Super-resolution and Uncertainty Estimation from Sparse Sensors of Dynamical Physical Systems

This is supporting code for the article

Collins, A. M., Rivera-Casillas, P., Dutta, S., Cecil, O. M., Trautz, A. C., & Farthing, M. W. (2023). 
Super-resolution and uncertainty estimation from sparse sensors of dynamical physical systems. Fontiers in 
Water,5, 1137110. https://doi.org/10.3389/frwa.2023.1137110

Email: adam.m.collins@erdc.dren.mil for any questions/feedback.

Super-resolution framework

Getting Started

• Download the relevant datasets from the following links and save them in respective subdirectories under the data directory.
  • Link for the mean flow dataset. Individual cases to be saved as data/kepsilon, data/komega etc.
  • Link for the NOAA Optimal SST dataset. To be saved in data/noaa/.
  • Link for the global soil moisture dataset. To be saved in data/sm/.
• The first run for each example generates relevant tessellated fields and saves them in subdirectories under data/npy/.

Dependencies

• Python 3.x
• Tensorflow TF 2.x. Install either the CPU or the GPU version depending on available resources.
• A list of all the dependencies are provided in the requirements file.

Executing program

• Training and evaluation for the mean flow dataset can be performed using examples/Mean_Flow/mf.py. The script is set up to parse various arguments during the execution call which typically looks like python mf.py 'unet-kepsilon-PHLL_case_0p5 PHLL_case_0p8 PHLL_case_1p0 PHLL_case_1p2 BUMP_h20 BUMP_h26 BUMP_h31 BUMP_h38 CNDV_12600-Ux Uy-PHLL_case_1p5 BUMP_h42 CNDV_20580-Ux Uy-train'
• Training and evaluation for the NOAA SST dataset can be performed using examples/Sea_Surface_Temperature/sst.py.
• Training and evaluation for the soil moisture dataset can be performed using examples/Soil_Moisture/soilmoisture.py.