problems transforming NAVD88 to MHW near Westport WA

[rjleveque]

Thanks team for producing these tools, they should be very useful in our tsunami modeling work!

I just tried transformez 0.5.1 to transform one of the recent CUDEM tiles, which covers Westport WA:
ncei19_n47x00_w124x25_2025v2.tif
from NAVD88 to MHW.

I compared the results to a MHW version of the same DEM that was earlier created by NCEI for NTHMP modeling,
sowa_mhw_19_n47x00_w124x25_2025v1.nc
from https://coast.noaa.gov/htdata/raster2/elevation/NCEI_ninth_Topobathy_2014_8483/wash_outercoast/index.html

I found that the two agree right around the shoreline, but they deviate inland, with the result of transformez transitioning from the correct MHW elevations back to NAVD88 not far inland, see the figures below.

I understand that datums like MHW do not make sense far inland, but in this case the transition happens at less than 2 meters elevation in parts of this region. Since a Cascadia tsunami can have runup of more than 10 m, this makes the result unusable for tsunami modeling.

I initially tried doing this in a Jupyter notebook, rendered version here. I also tried it at the command line and got the same result (command and output included at the bottom of the notebook).

Please let me know if I'm using this tool incorrectly.

The figures show the 2-meter contour around Westport (correct MHW in blue, transformez result in red) and the elevations along one transect, where there should be a constant offset of about 2.1 m between NAVD88 and MHW at all points.

[matth-love]

Hi @rjleveque Thank you so much for this detailed report!

The figures and the Jupyter notebook made it very easy to see exactly what is happening here. You are using the tool perfectly, but you have run into a default safety setting in transformez that is specifically at odds with tsunami modeling.

What is happening:
By default, transformez applies a decay_pixels value of 100. This is designed for general terrestrial mapping to prevent tidal shifts (like MHW) from being incorrectly applied miles inland to mountains. Because your DEM is high resolution (1/9 arc-second, or ~3 meters per pixel), that 100-pixel decay forces the MHW shift to smoothly transition back to NAVD88 just ~300 meters inland. As you noted, for a Cascadia event with 10m+ runup, 300 meters inland is nowhere near far enough!

The Solution:
We built a specific override for hydrodynamic modeling. If you set the decay pixels to 0, transformez disables the inland decay and applies a continuous, infinite extrapolation of the coastal MHW shift as far inland as the DEM goes. This guarantees that constant ~2.1m offset you need for your transects.

For your CLI command, just add: --decay-pixels 0

For your Jupyter Notebook (api.transform_raster), just pass: decay_pixels=0

Give that a try on the Westport tile and let us know if it aligns as expected with the NTHMP reference DEM. We will also make sure to highlight this specific setting for hydrodynamic modelers in the documentation.

Thanks again for using the tool and for the great report!

[rjleveque]

Thanks @matth-love, that's great to know. I tried it out and it does a much better job, although there are still some minor differences from the MHW version in the 2m contours. I'm traveling at the moment but will work with it some more before long, and will let you know if run into other major issues.