Add notes about power law interpolation vs linear

[ppinchuk]

@grantbuster Did I miss anything in the main text that should be added?

I tried to place the note in a couple different places so it would be easier to find, but let me know if you think I missed a good place to put it.

[grantbuster]

@bnb32 for your awareness: the original WTK code does linear interpolation between WRF vertical levels. I was confused because the WTK report says power law interp and we had previously discussed the benefits of such an approach. I reached out to Michael Rossol who originally implemented WTK processing and he came back with this explanation why we want to use linear. This might change how we think about processing ERA data. Power law might be okay because the model might learn to produce negative wind shear, but linear interp might better encourage this.

[bnb32]

@bnb32 for your awareness: the original WTK code does linear interpolation between WRF vertical levels. I was confused because the WTK report says power law interp and we had previously discussed the benefits of such an approach. I reached out to Michael Rossol who originally implemented WTK processing and he came back with this explanation why we want to use linear. This might change how we think about processing ERA data. Power law might be okay because the model might learn to produce negative wind shear, but linear interp might better encourage this.

You can still get negative wind shear if you use a negative exponent though. I was using log interp which does allow for negative wind shear in a similar way, but I switched back to lin interp a since I didn't see much difference. Interesting that the wtk paper says power law interp though.

[grantbuster]

Notes from our discussion: difference between log/power law fit and log/power law interpolation between WRF levels. If we fit a power law exponent between two WRF levels we can have a continuous power law interpolation, but it might behave poorly with sharp gradients at local maximum windspeeds. Based on this, i think we can proceed with linear interpolation under the argument that it is guaranteed to be a well-behaved interpolation method under negative wind shear conditions.