Visualize results of simulation run on original NC9 and NC9rev with water levels offset by 0.5 ft (0.15 m)

[ththelen]

Motivation for using 0.5 ft instead of 0.7 ft. Deriving offset from NOAA verified - predicted is more defensible (does not hide model errors) compared to NOAA verified - modeled

Comparisons of NOAA predicted vs verified (observed) water levels are done. For the month of October 2021, NOAA verified water levels were an average of 0.51 ft higher than predicted values.

I was also curious how the NOAA predicted versus verified stacked up against the ADCIRC modeled vs NOAA verified over the Nov 9-14 timeframe that we used to come up with the 0.7 ft=0.21 m WL offset. That comparison is shown in the tables below.

I still have questions about why we would use the predicted NOAA water levels to determine the ADCIRC offset instead of the modeled results. I understand that the NOAA 'Actual' water levels are influenced long term effects that both ADCIRC and the NOAA 'Predicted' water levels do not capture. However, 'Predicted' water levels do not capture what I will call medium-term effects, like wind setup for example, that would be captured by modeled water levels. To me, it makes sense to parse out the long-term effects that ADCIRC cannot simulate by comparing it to model results that include medium-term effects, not just precited tidal harmonics. It seems to me like we would want to address biases in the model through the WL offset by comparing modeled results to actual results.

Originally posted by @ththelen in https://github.com/NCSU-CHAZ/Sunny-Day-Flooding/issues/9#issuecomment-1208502743

[caseydietrich]

@ththelen These simulations are finished and posted on the cluster here:

/share/jcdietri/For/Thomas/220819-NC9+0.15m-Nov2021

/share/jcdietri/For/Thomas/220820-NC9rev+0.15m-Nov2021

[ththelen]

Hi @caseydietrich and @anardek, I finished calculations and visualization for the latest set of ADCIRC model results, and the results are very encouraging! In the attached PowerPoint you will find plots comparing the CB NC9rev+0.15m with FIMAN gauge data and results from all of our other runs. I have also included results comparing Wilmington and Wrightsville NC9rev+0.15m time series with the original NC9 results at these locations. Finally, I have calculated the RMSE between model results and NOAA/FIMAN gauge data at all locations to quantify the improvement we have seen in our model results. Takeaways from the plots and RMSE results are below.

• Adding the +0.15m offset and refining bathymetries around Carolina Beach both reduced RMSE compared to FIMAN gauge data.
• Carolina Beach RMSE compared to FIMAN is lower than Wilmington and Wrightsville RMSE compared to NOAA. If this is a reflection of the mesh improvements that we have made around CB, this is noteworthy give the complexity of the hydraulic connections around CB compared to the open ocean gauge at Wrightsville and Cape Fear gauge at Wilmington. Note that the RMSE of the original NC9 mesh at CB was about 0.2 ft higher than at Wilmington and Wrightsville.
• I think we are close to having our ADCIRC mesh substantially complete. The only lever I can think of tweaking at this point is refining the stair stepped Snows Cut mesh bathy adjacent to the Cape Fear.

20220825-NC9(rev)+0.15m.pptx

[ththelen]

Look at inlet resolution, and how it is reflected in ADCIRC model