search

USGS-R / river-dl

Deep learning model for predicting environmental variables on river systems
Creative Commons Zero v1.0 Universal
21 stars 14 forks source link

Minor edits to GW loss #191

Closed janetrbarclay closed 2 years ago

janetrbarclay commented 2 years ago

This PR makes 2 relatively minor updates to the GW loss code.

  1. It moves the list of additional reservoir-affected reaches (that aren't in the reach attributes file) out of the gw_utils.py and into the config file
  2. It standardizes treatment of temps < 1 C in the annual stats calculations. Previously they were dropped in the calculations of the targets (to avoid freeze / thaw dynamics) and included in the calculations within the loss function (because the fft doesn't play nicely with missing values). Now temps < 1 are set to 1 in both locations. Setting the temps < 1 to 1 resulted in stats that more closely matched the results when they were removed than did leaving them in or setting temps < 0 to 0.