I'm happy with this. I'm curious how bad the test value changes are...I would hope they're not significant.

Max absolute difference: 0.37513416
Max relative difference: 0.11791906
 x: array([ 2.062733  ,  0.96246913, 11.22554254, 12.83861839,  6.64413777,
        6.90054376])
 y: array([ 2.18346161,  0.86094706, 11.6006767 , 12.53639395,  6.89206916,
        6.69867051])

Probably a little bigger than I'd like (percentage-wise) but not really scary.

@dcamron Do you want to fix up the tests to get this in the release?

If there's no other community input and we feel good about these changes, I'm glad to.

Hello @dopplershift @dcamron! After reviewing #2092 and #2144 , I think using mean_pressure_weighted for bunkers storm motion calculation might be better. Because the get_layer function in weighted_continuous_average only return pressure profile to pres_prof, which is used in calculating [trapezoid(var_prof, x=pres_prof) / (pres_prof[-1] - pres_prof[0]) for var_prof in others]
Considering that even for isobaric layer data, the mass of air blocks between pressure layers is not consistent, using mean_pressure_weighted might be a better method for calculating the average wind speed of the entire layer. In the function corfidi_storm_motion for calculating MCS storm motions using the Corfidi method, the mean_pressure_weighted function is used to calculate the average wind.

@yfNJU Hi! I would point you to #1361, where we intentionally changed bunkers_storm_motion to use a non-pressure-weighted mean for the wind, as that's what is explicitly specified in the Bunkers (2000) paper. In fact, they arrived at the non-pressure weighted average as it empirically yielded the best mean absolute error (MAE).

EDIT: If you have an updated peer-reviewed publication indicating broad acceptance that using a pressure-weighted mean yields an improved technique, I'd love to see it and would be interested then in updating MetPy's implementation to reflect the latest science.

@yfNJU Hi! I would point you to #1361, where we intentionally changed bunkers_storm_motion to use a non-pressure-weighted mean for the wind, as that's what is explicitly specified in the Bunkers (2000) paper. In fact, they arrived at the non-pressure weighted average as it empirically yielded the best mean absolute error (MAE).

EDIT: If you have an updated peer-reviewed publication indicating broad acceptance that using a pressure-weighted mean yields an improved technique, I'd love to see it and would be interested then in updating MetPy's implementation to reflect the latest science.

Thank you very much for you reply! I reviewed #1361 and confirmed that the method mentioned therein aligns with Eq. (1) in Bunker [(2014)]. Additionally, I agree with #2092's consideration that Eq. (1) in Bunker (2014) appears to assume even sampling of levels in a vertical profile, whereas #2092's approach of using the average value between adjacent layers multiplied by the pressure difference as a weighting factor is more convincing.

Additionally, I recommend adding Bunker (2014). to the references of bunkers_storm_motion. This paper explicitly defines the non-pressure-weighted formulation. It is also worth noting that it introduces the "effective layer" method for wind averaging, which utilizes pressure-weighted vertical integration and has already been applied in the SHARPpy algorithm (See def bunkers_storm_motion in line 2407). Maybe we can consider developing a effective_layer_bunkers_storm_motion function. Thanks!