USGS-R / drb-inland-salinity-ml

Code repo for Delaware River Basin machine learning models that predict inland salinity.
Creative Commons Zero v1.0 Universal
3 stars 4 forks source link

Add maps of error metrics for different time periods #238

Closed jds485 closed 1 year ago

jds485 commented 1 year ago

This PR adds maps of RMSE by calendar month, season, and for each calendar year in the dataset. There is a new function that replaces the previous targets {commands} to make the code a little cleaner for these targets.

Currently, the plot does not have a label for what is plotted in it. That information is only within the file name. I think that's enough for our needs, just wanted to note it.

I also added reservoir locations to these maps as black circles. I had trouble getting those to display in the legend because the existing legend (RMSE) is for a continuous variable.

Example plot for year 2000 (one of the very high residual years): RMSE_log10_year_2000_full

Closes: #233

jds485 commented 1 year ago

Thanks, Lauren!

Overall, I'm also wondering to what extent the functions plot_nhdv2_attr and plot_maps are expected to serve multiple purposes within this pipeline?

I think plot_nhdv2_attr is used for all maps and I've been thinking of this as generic. plot_maps is a wrapper around that function for different ways to aggregate information to plot on the map, and I also view that as generic. I forgot about the site vizualizations in the synoptic_site_viz branch. I'm also okay with keeping that separate, but it is a similar function. I don't intend on merging that branch with the main repo.

Is this something you were hoping to add but just not sure how to implement?

No, I wasn't planning to add this information to the plots.

jds485 commented 1 year ago

I am going to wait to rebuild targets until merging #240 because of the merge conflict. I most recently ran #240 targets, and I think if I rebuild these targets I will need to rebuild in #240 as well