[1pt] PR: BARC updates to cap xsec area and allow user to choose input bankfull geometry

[RyanSpies-NOAA]

These revisions aim to improve the technique and structure used for the bathymetry estimation. There are also some minor enhancements to the rating curve comparison workflow. This feature branch addresses the following github tickets:

• Test and optimize bathy estimation code variables [3pt] Test and optimize bathy estimation code variables #342
• Run bathy simulation with NWM-routing bankfull geometry input data [5pt] Run bathy simulation with NWM-routing bankfull geometry input data #343
• Bathy estimation solution: perform sierra solution (SRC vs USGS rating curve) for subset of HUCS [8pt] Bathy estimation solution: perform sierra solution (SRC vs USGS rating curve) for subset of HUCS #313
• Produce an aggregate csv with all of the USGS and FIM datum information for all eval sites [3pt] Produce an aggregate csv with all of the USGS and FIM datum information for all eval sites #401
• Rating curve comparison plots: plot FIM SRC points on top of the USGS rating curve points [1pt] Rating curve comparison plots: plot FIM SRC points on top of the USGS rating curve points #402

Changes

params_calibrated.env & params_template.env

• Added new env variable to control which file is used for the bankfull geometry input to bathy estimation workflow
• The default input bankfull geometry is currently set to the Wieczorek et al. (2018) dataset --> uses the Bieger et al. (2015) regional regression equations

bathy_rc_adjust.py

• Modified the bathymetry cross section area calculation to cap the additional area value so that it cannot exceed the bankfull cross section area value for each stream segment (bankfull value obtained from regression equation dataset)

run_by_unit.sh

• Using new env variable to define the input bankfull geometry data file

rating_curve_comparison.py

• Modified plot output to always put the FIM rating curve on top of the USGS rating curve (avoids USGS points covering FIM)
• Created a new aggregate csv file (aggregates for all hucs) for all of the usgs_elev_table.csv files (one per huc)

Screenshots

Example of a new rating curve comparison plot with the FIM synthetic rating curve now plotted on top of USGS rating curve:

Testing the NWM route link bankfull geometry

This feature branch was also used to evaluate the FIM Bathymetry Adjusted Rating Curve (BARC) performance using the estimated bankfull geometry dataset derived for the NWM route link dataset (#343). The goal of this evaluation was to determine if one input bankfull geometry dataset clearly performed better than the other. Results were analyzed for three different FIM versions:
FIM 3_0_16_1: the most recent FIM dev version at the time of evaluation
FIM NWMGeom: the input channel geometry database was developed using the CONUS derived regression equations from Blackburn-Lynch et al. (2017).
FIM Bieger: the input channel bankfull geometry dataset was developed by Wieczorek et al. (2018) using the Bieger et al. 2015 regional regression equations. This FIM version is the same as FIM 3_0_16_1 except that BARC now includes a bathymetry cross section area calculation limit to prevent values from exceeding the bankfull cross section regression value (see changes above).

Alpha Evaluation Summary

The area based CSI metrics were used to compare the mainstem FIM performance between FIM 3_0_16_1, FIM NWMGeom, and FIM Bieger. Aggregated area based CSI scores indicate that inundation outputs (with BARC) using the NWMGeom produce a slight regression for the action and minor thresholds but also produce a small improvement for the moderate and major thresholds:

Site specific results (median CSI score for all sites) hint at more noticeable regression for action and minor thresholds for the FIM NWMGeom compared to the FIM Bieger:

Sierra Evaluation Summary

The rating curve comparison evaluation compares the FIM synthetic rating curve vs. USGS rating curve for the same FIM versions outlined in the alpha evaluation above. Looking at the mean absolute elevation difference (units=feet) for a range of flow thresholds (1.5yr, 5yr, 10yr, action, minor, moderate, major) can provide and indication of synthetic rating curve performance between the different FIM simulations.

The following plot shows the histogram of "Mean Absolute Elev Differences" at 361 USGS gage sites. The Mean Absolute Elev Difference was calculated for the two FIM versions (FIM NWMGeom and FIM Bieger) and the resulting delta (FIM Bieger minus FIM NWMGeom) indicate which FIM version performed better. Negative bins (orange) in the plot below show the number of instances where FIM Bieger performed better than FIM NWMGeom and the positive bins (blue) indicate the inverse.

Conclusions

Based on these findings, the FIM Dev team decided to continue using the FIM Bieger version as the default input in the FIM workflow. Further development in the BARC methodology will likely evolve over time, and the FIM performance using NWMGeom (or other version) could potentially exceed the current FIM Bieger performance. The capability now exists to test the different channel bankfull geometry datasets seamlessly.

[BradfordBates-NOAA]

@RyanSpies-NOAA Please update the Changelog for this branch

[RyanSpies-NOAA]

Change log draft message:

v3.0.17.0 - 2021-06-04 - PR #393

BARC updates to cap the bathy calculated xsec area in bathy_rc_adjust.py and allow user to choose input bankfull geometry

Changes

• Added new env variable to control which input file is used for the bankfull geometry input to bathy estimation workflow.
• Modified the bathymetry cross section area calculation to cap the additional area value so that it cannot exceed the bankfull cross section area value for each stream segment (bankfull value obtained from regression equation dataset).
• Modified the rating_curve_comparison.py plot output to always put the FIM rating curve on top of the USGS rating curve (avoids USGS points covering FIM).
• Created a new aggregate csv file (aggregates for all hucs) for all of the usgs_elev_table.csv files (one per huc).
• Evaluate the FIM Bathymetry Adjusted Rating Curve (BARC) tool performance using the estimated bankfull geometry dataset derived for the NWM route link dataset.

[BradfordBates-NOAA]

@RyanSpies-NOAA Do you recommend I run a test on the VM to confirm successful runs by a different user? If so, please paste the desired usage for me to try. If not, I will go ahead and approve as this feature can be activated or deactivated easily, which is great.

[RyanSpies-NOAA]

@BradfordBates-NOAA probably not essential but feel free to run a quick test with the following (random HUC over Omaha, NE):
fim_run.sh -u 10230006 -e MS -c /foss_fim/config/params_calibrated.env -n ryan_test_wb/