RE: Cyclonic Model weird result - Home - Delft3D
RE: Cyclonic Model weird result
Hi. I'm using delft3d Flow to simulate a storm surge event on Manila Bay, Philippines. The tidal simulation has good results but when I added the wind component of the storm, the result goes bad. Attached is a picture of the results using QUICKPLOT. The blue line is the tidal simulation (astronomic forcing only, no wind) and the red line line is the cyclonic model with a spiderweb file used for wind. There is a dip in water level that should have occurred. Please help.
A few things to check like: your horizontal viscosity is very low for a model of this scale, think about maybe 50 rather than 1. In the wind file, the wind is in knots, should it be in m/s??? The area covered by the wind file is quite large compared to your quite small model area so perhaps there is not enough resolution / detail of the wind? Is the wind direction convention correct (i.e. blowing to or blowing from)?
Perhaps someone with more experience of this kind of study might come along and have some thoughts as to whether your model area is too small to reproduce what you are trying to model?
Not sure if there is anything wrong with the model. If you look at the figure that I attached, you'll see that Manila experienced strong offshore winds at the moment (July 16th 00h00) you see the dip in water levels. These offshore winds are causing wind set-down in Manila Bay. The model does probably overestimate the set down somewhat, as the effects of land roughness (which typically reduce offshore blowing winds) in the Holland wind model of Delft Dashboard are not taken into account. Your model inputs look fine to me (don't worry about units of wind speed (indeed knots, as Phil pointed out) in the cyc file, as Delft Dashboard converts these to m/s when creating the spiderweb file). The one thing I would change are the settings for the wind drag coefficient Cd. It is now commonly excepted that the wind drag coefficient increases up to 25-30 m/s and then starts to decrease again. I have used the following setting in the mdf file in many modelling studies:
Wstres= 1.0000000e-03 0.0000000e+00 2.800000e-03 2.5000000e+01 1.5000000e-03 5.0000000e+01
Thank you so much for the insightful response, sir Marteen. However, the model would then diverge significantly from the actual data I'm comparing it with, as to assess its validity. Attached is the picture (and data file) of the actual measured data as compared with the forecasted tidal level generated by Delft dashboard at IHO tide stations. While there is a dip in water level, it isn't as significant as the simulation's. I'm not sure if the effect of surrounding topography is enough to explain such large discrepancy.
Your positive surge is pretty well reproduced, so the model isn't far off. In this context, for the negative surge, Maarten's suggestion of the lee effect of topography is sensible. Once you've dealt with that, along with changing wind drag in line with his suggestions and increasing hev in line with mine, then getting the positive surge profile spot on becomes simply a part of the normal calibration process.
Your model resolution is very low by the way; seemingly not important in this case for modelling levels, but if you want flow speeds it will need to increase significantly. Good luck. :-)
Thank you, sir. Will implement these changes. On the resolution, this is actually just my macro-grid. I will still nest a finer micro-grid (upto 500x500m resolution at the coastline) focused on my study area. Hopefully, my research gets approved soon. Really appreciated the feedback, quick and on point <3 <3
When you do come to refine your grid, make sure that orthogonality and smoothness are good, and also check your timestep in Quickin by using the Courant Number function. For your present coarse grid, for example, you can probably have a much larger timestep than the 0.5 minutes you're using. Although the model probably runs very fast anyway, as you have very few cells! :-) Good luck. Phil.
Just noticed that you're using spatially-varying bed roughness with Manning's n values of about 0.1 inside the bay. That is extremely high! It should be more in the order of 0.024. When I run with a spatially constant value of 0.024, the peak surge level goes up to ~1.2 which is more in line with the observations. The set-down just before the peak surge also increase a bit, which of course is not good. Unfortunately, it is not straightforward to fix this issue with current versions of Delft3D and Dashboard. They simply do not take into account the land effect on the wind field...
Sorry, I'm back again. My delft3d seems to be encountering an error during nesting. It can create the observation file for the macro-grid properly. Attached is the picture of the micro-grid within the macro-grid. When I run the nesting tool (1), it only creates one observation point very far from the micro-grid.
The example given at the manual (attached) were also overlapping. At this point, editing the macro-grid just seems too much work as everything else would have to follow. Also, I managed to capture the nesting command line during processing and here's what it reads (see 2nd attached screenshot).
I believe the forcing in the bnd file needs to be set to time series (T) instead of astronomic (A) for the first step of the nesting to work.
West Z T 1 17 1 2 5.0000000e+002
South Z T 2 1 37 1 5.0000000e+002
East Z T 38 2 38 26 5.0000000e+002
The overall and detailed grid are indeed usually overlapping.