Hi everyone!
I’m using the InVEST Coastal Vulnerability Model to run two different scenarios: one for the present and another for a future projection. I’d like to incorporate future-projected wind and wave data instead of relying solely on the default WaveWatch III (WW3) dataset. Although I wasn’t able to find projected wave data, I decided to at least include projected wind speed to make the future scenario more realistic.
To do this, I downloaded monthly near-surface wind speed data from the HadGEM3-GC31-MM (UK) model for the SSP5-8.5 scenario in the year 2070. I used monthly northward and eastward near-surface wind components from the same model, scenario, and year.
I’m now trying to figure out how to convert this data into a format compatible with the WW3 structure used in InVEST — specifically, how to distribute it into the 16 directional sectors.
Has anyone tried doing something similar? Any advice or guidance would be greatly appreciated!
Thank you in advance!
Hi @marianamendesm,
The InVEST User’s Guide provides instruction about this: Coastal Vulnerability Model — InVEST® documentation
Let me know if you have any specific questions!
Hello,
I’ve carefully reviewed the entire guide to ensure the file was calculated and formatted correctly. I used Python to automate the process and compute the REI_VX values. However, most of the results are unfortunately returning zero.
I’m not sure whether this indicates an error in my code or a misunderstanding on my part, but I assume this is not the expected outcome for a climate change scenario.
I’m attaching my script in case someone is able to take a look and provide some guidance. I would really appreciate any help.
Thank you in advance!
wind2.txt (4.6 KB)
Hi @marianamendesm,
Would you be able to share your data so I can take a look?
Thanks!
Hi @csimpson,
Here is the link to the files: wind - Google Drive
Thank you very much in advance!
I keep wondering whether I should be considering only the ocean points… Since the wind data comes from WaveWatch III, maybe including land points is artificially raising the threshold? Let me know what you think!
Hi @marianamendesm,
Yes, I believe you’re correct and the data you’re creating to mimic WaveWatchIII should only include points in the ocean. Please let me know if this solves your problem!
Hi @csimpson
It still hasn’t worked… Something I’ve been thinking about is that, since I’m using monthly data over a 10-year period, there might be a high heterogeneity in the values, which could affect the top 10% selection and cause the points to cluster in certain regions of the globe.
The user guide suggests considering the top 10% of all available values, which in my case would be 120 values (monthly data from 2070 to 2079) for each point in the shapefile.
However, do you think it would be reasonable — in order to avoid this heterogeneity and clustering — to first calculate the mean wind speed for each point over the entire time period, and then select the top 10% of those mean values for the calculations?
This might help avoid spatial clustering of the points (for example, ending up with points mostly concentrated off the east coast of Africa, where the highest wind speeds seem to occur).
I’d really appreciate hearing your thoughts on this.
Thank you in advance!
Hi @marianamendesm,
Sorry to hear you’re still having trouble! I believe I have identified your issue - are you currently calculating the top 10% of wind speeds globally across all grid points and times steps? If so, this will result in only the windiest parts of the globe contributing values to the REI calculation (and everything else being 0). Instead, the top 10% of wind speeds should actually be calculated from the timeseries for each location, as a measure of the local wind conditions. In other words, the 90th percentile threshold should be calculated and applied locally, not globally
The user guide suggests considering the top 10% of all available values , which in my case would be 120 values (monthly data from 2070 to 2079) for each point in the shapefile.
Exactly, so if you have 120 total values for each point/pixel, you’d take the highest 12 values for each point’s timeseries, assign each value to its associated directional sector, and then compute the average wind speed per directional sector to get REI_VX.
Please let me know if this helps!
Hi @csimpson ,
Thank you very much for your help! I was indeed calculating the top 10% from all available values globally… So just to confirm: the top 10% calculation is done individually for each point, right? I take the highest 10% of the time series for each point, and from these I calculate the REI_VX for that point.