Pull and plot water level time series at gauge locations for original NC9 mesh

[ththelen]

Acronyms: CB = Carolina Beach YB = Yacht Basin

[ththelen]

CB Nodes used for WL timeseries interpolation circled in blue. Not convinced that these are the 3 closest nodes to the sensor (circled in red), but they are in the right general area. However, the pullT code probably works well enough for the time being as long as the station coordinates are bumped to a location that is conducive for pulling water levels

Originally posted by @ththelen in https://github.com/NCSU-CHAZ/Sunny-Day-Flooding/issues/2#issuecomment-1158099320

[ththelen]

Water level comparisons between modeled and measured values are pasted below. Once I bumped the modeled locations to consistently inundated areas, I got peak water levels that were generally within 1 foot of the measured values. It is interesting to me that our model did a better job of predicting peak water levels at Wilmington and CB than it did on the open ocean at Wrightsville beach. The CB model levels seem to temporally lag the measured values compared to the other two locations; this lag seems like something we could potentially resolve by refining model bathymetry around CB.

However, I think it is encouraging that all results generally line up temporally and consistently show that we are slightly underpredicting peak water levels. I will be interested to hear what @caseydietrich (after he gets back from vacation) and @anardek think about these results.

[ththelen]

Good news and bad news on the ADCIRC wind data comparison @anardek and @caseydietrich .

Bad news: neither of the NOAA gauges had wind data for the time period we modeled. I had been hoping to use Wrightsville Beach wind data because there are no inland obstructions (e.g., buildings, dune line) between winds on the open ocean and this gauge. Alas.

Good news: modeled wind speed and direction match quite well with data from the CB FIMAN gauge. Our modeled wind speeds (top plot) from the period of highest wind velocity (June 5 - 8) are about 10 mph higher than what was measured at the Yacht Basin gauge. I suspect this is due to shielding from structures and dunes at our measurement location, whereas our modeled velocities probably reflect what would expect from a sensor located on the ocean shore. Our modeled wind directions (bottom plot: bearing relative to North - both 0 and 360 degrees reflect winds out of the North) are generally accurate to the data from the YB sensor.

[anardek]

So is the wind data from ADCIRC the wind applied at the boundary? Or from the closest node to the FIMAN gauge?

On Mon, Jun 20, 2022 at 3:24 PM Thomas Thelen @.***> wrote:

Good news and bad news on the ADCIRC wind data comparison @anardek https://github.com/anardek and @caseydietrich https://github.com/caseydietrich .

Bad news: neither of the NOAA gauges had wind data for the time period we modeled. I had been hoping to use Wrightsville Beach wind data because there are no inland obstructions (e.g., buildings, dune line) between winds on the open ocean and this gauge. Alas.

Good news: modeled wind speed and direction match quite well with data from the CB FIMAN gauge. Our modeled wind speeds (top plot) from the period of highest wind velocity (June 5 - 8) are about 10 mph higher than what was measured at the Yacht Basin gauge. I suspect this is due to shielding from structures and dunes at our measurement location, whereas our modeled velocities probably reflect what would expect from a sensor located on the ocean shore. Our modeled wind directions (bottom plot: bearing relative to North) are generally accurate to the data from the YB sensor.

[image: Image] https://user-images.githubusercontent.com/89156971/174665660-c8ff7fcf-bbb5-454c-b865-b723ae182709.png

[image: Image] https://user-images.githubusercontent.com/89156971/174665704-0edb3699-105f-4c05-9877-57880b337b2e.png

— Reply to this email directly, view it on GitHub https://github.com/NCSU-CHAZ/Sunny-Day-Flooding/issues/5#issuecomment-1160774955, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANXYL53RIPPVNTURVDBOSEDVQDAQFANCNFSM5ZDI2K2Q . You are receiving this because you were mentioned.Message ID: @.***>

[ththelen]

@anardek the wind data in the plots I sent is a time series of wind velocities interpolated from three points in the center of the YB. To answer your question, this result is most analogous to using the closest node to the FIMAN gauge. For consistency, I used the same three points I used to pull the water level time series. These 3 points were "bumped" off the exact location of the gauge and into the middle of the yacht basin to be at a location that was inundated for the duration of the ADCIRC times series.

Additionally, as I understand it, atmospheric forcings in ADCIRC are applied across the ADCIRC mesh, not just at the boundary conditions. @caseydietrich can correct me if I am wrong on this point.

[caseydietrich]

Thomas, these plots are great, thanks for making them.

For the water levels, I agree that these results are promising. To my eye, it looks like the ADCIRC water levels are biased low, which is expected, given that the model is not representing the longer-timescale oceanographic and atmospheric forcings. If we can quantify the bias, then we could add it into the model as an initial condition for the water levels. To start, can you please compute the mean water levels for the observations and ADCIRC during a time period with low winds, e.g. Nov 9-15? I'm hoping we get similar differences at all stations. I can also share the ADCIRC results from Oct, which would give a longer time series for comparison.

For the wind speeds, it is concerning that the NAM winds are too high by about 10 m/s during the event. This will likely push the model into overpredictions on the water levels, especially after we correct for the long-term bias. The NAM winds are given as full marine conditions, but then ADCIRC corrects them for overland roughness. So those effects are represented here, and the winds are still too strong. Not sure what to do about this. It would be nice to see a few more model-data comparisons. If you expand your search area, then are there other NOAA or FIMAN stations with wind observations during this timeframe? We could also look for NDBC buoys?

[ththelen]

TT note: use even number of tidal periods for WL averaging

[ththelen]

Hi team, I have answers for the two questions posed by @caseydietrich in his previous post.

1) What are the mean water levels for the observations and ADCIRC during a time period with low winds, e.g. Nov 9-15 The measured water levels seem to be about +0.7 feet higher than the modeled water levels Nov 9-15. This is pretty consistent (within +/- 0.1 ft) across all three sensors. I also note that the modeled tidal cycles lag the measured cycles by about 1 to 2 hours depending on the station

2) Since neither Wrightsville or Wilmington NOAA gauges have wind data from Nov 1-15, are there other NOAA or FIMAN stations with wind observations during this timeframe? Recall that the CB FIMAN wind speeds differed from modeled winds by almost 20 mph during peak winds. Yes, the Beaufort NOAA gauge did have wind data Nov 1-15. A comparison of this data with the modeled data pulled from near Beaufort is shown below. The correlation between measured and modeled is much better at the Beaufort site compared to the results I presented for the CB FIMAN wind data. Note that wind speeds are presented in m/s in these plots while the CB data was presented in mph. Since the modeled wind speeds are pretty similar between Beaufort and CB once you do the unit conversion, and the modeled data matches the Beaufort station quite well, I suspect the error comes from the CB FIMAN gauge data or my conversion of this data. I will have to do some more digging to get to the bottom of this issue, but it is encouraging to see that wind speeds match well at the Beaufort gauge.

[anardek]

@ththelen - can you share time series comparisons (plots) for the data you present in tabular form in #1 above?

For #2, I wouldn't be surprised if the FIMAN wind gauge data is bad (given that the data are not QC'ed, and the gauge itself is not as robust as the NOAA gauges). Is there any information on the gauge? Maybe measurement limitations (max wind speed the sensor can resolve? is what is reported an average? etc.). But even if the wind data from FIMAN are bad, the lag in the tides is concerning. @caseydietrich - any ideas on what might be causing the lag? Maybe the tide isn't propagating through the inlet properly (and instead is coming up the river?). @ththelen -- a movie of the tide propagation in planform might shed some light on this.

[ththelen]

Hi @anardek, the time series for the tabular data are the same 3 time series shown in this comment https://github.com/NCSU-CHAZ/Sunny-Day-Flooding/issues/5#issuecomment-1159199978

I could have been more precise with how I named some of my values in the table. Avg start WL could be more accurately described as Water level at the start of the averaging period, in this case the morning of Nov 9, and similarly for Avg end WL around the evening of Nov 14.

To get more information on the wind gauge I would have to reach out to the FIMAN folks. Their data sharing website is light on technical specs. We may get more insight into the tidal propagation question once we visualize results from the next ADCIRC run with new bathymetry that Casey has going right now.

[anardek]

Understood -- thanks!

[caseydietrich]

@ththelen Thanks for making these comparisons. The wind plots are especially encouraging, given the relative coarseness of the NAM forcing.

I will start a simulation with an initial condition to raise the water levels by 0.7 ft (0.21 m). I will use the original NC9 mesh, just so we can compare the effects of these two changes (higher initial water levels, deeper channels) independently. I should have these results in a few weeks.