Continuum discussion about internal_waves_barkley_canyon HOT 3 CLOSED

• Use of Eq. 19 and proper E_c/E_{GM} ratios.
• Check Klymak (2006) Hawaii paper for M2-dissipation relation.
• Factor of 2 in D (depth-mean) between spring and neap.
• They found D ~ E^(1.0 +/- 0.5).
• Refer to as /eps ~ E_{M_2}^(b +/- db).
• Want a range on power law fits (b_axis = 0.64 +/- 0.03).
• Four plots, with each constituent as x-axis.
• Summary plot (as above) for strongest (M2) with the dots 'coloured' to show relative strength of next biggest contributor (with colourbar). e.g. sub-diurnal at Slope, NI at Axis.

Plot showing each constituent vs dissipation (both sites, all years):

• There is little obvious correlation between dissipation and constituents, other than the semidiurnal.
• Even semidiurnal is iffy, at Slope.
• Based on scatter plots and seasonal correlations, the subdiurnal (Slope) and near-inertial (Axis) are the most likely 'secondary' contributors.

Plot showing semidiurnal vs dissipation at each site (all years) with markers coloured based on the strength of the secondary constituent:

• The relationship between dissipation and the secondary constituents (coloured markers) is obviously different, temporally, from the semidiurnal influence (i.e. high values of secondary constituent band-power do not match high values of semidiurnal band-power).
• However, higher values of secondary constituent band-power do trend towards higher values of dissipation, suggesting some weak correlation that is independent of the semidiurnal influence.

from internal_waves_barkley_canyon.

• Refocus write-up, based on above. Remember continuum fit potentially not a good estimate of dissipation.
• Can cut correlation plots (still mention results) and just show scatter.
• Update outline based on this Issue (and archived).
• Do a percent-good for each fit (difference of fit value vs dissipation value, red vs blue).
• Added information to plots.
• Try a few other p0 estimates to confirm best fit.
• Other estimates (widely varying), as well as not providing estimates, all result in the same fits.
• Bootstrap for better variance estimates on exponents, using scipy.stats.bootstrap method.
• 4x2 plot (left Slope, right Axis) for each constituent, with the interesting ones coloured for power of their co-contributor.
• Multi-factor analysis scipy.optimize minimisation for ax^b + cy^d
• Also used bootstrap method for multi-variate curve_fit.
• Slope: (5.9e-7)(M2)^(0.83+/-0.17) + (7.5e-8 )(subK1)^(0.59+/-0.13)
• Axis: (3.4e-5)(M2)^(1.47+/-0.48) + (1.8e-8)(NI)^(0.24+/-0.06)
• At Slope, the multi-variate fit is a big improvement over just the M2 fit, though neither are perfect. Suggests a combination of forcing.
• At Axis, both fits are good, with each having different periods of brief inaccuracy, suggesting the components may act independently to drive dissipation.

from internal_waves_barkley_canyon.

Archived for reference.

from internal_waves_barkley_canyon.