This directory saves example configs for spatiotemporal super resolving GAN models for wind, solar, and temperature climate change data. All generator model configs should start with "gen_*" and should have two "_4x" tags (4 for example) that represents the spatial and temporal enhancements that the generator is designed for respectively and one "_3f" tag that represents the number of output features.
For example "gen_2x_24x_2f.json" is a model that would enhance a 4km daily spatiotemporal field to 2km hourly with 2 output features.
Unique model designs are utilized for each unique variable set. For example, when doing spatial super resolution of wind fields, a custom model with mid-network topography injection via a "Sup3rConcat" layer is used. For wind temporal super resolution, a 24x enhancement is used to go from daily to hourly, but for solar an 8x enhancement is used to go from 3 days to 24 hours of the middle day. Also, note that the "_trh_" model tag stands for temperature and relative humidity.
These configs are only examples and are not guaranteed to be the models used in producing actual production datasets. For the final model architectures, see the global file attributes associated with sup3r output h5 files which should contain all model meta data.