Synthetic spectra for assessing statistical fitting methods used to estimate ocean turbulence
Description
Many turbulence estimates require fitting expected model forms, typically nonlinear equations, to observations that have been mathematically converted into spectral observations. This database contains synthetic spectral observations used to evaluate several fitting methods, along with the estimated parameters. The impact of degrees of freedom (spectral averaging) was assessed using two datasets with variability generated using two different statistical distributions. The repository contains the following datasets:
- SyntheticSpectra.nc that contains:
- 3200 synthetic spectra with the variability drawn from lognormal distribution
- 3200 synthetic spectra with the variability drawn from chi2 distribution
- LorNormal.nc that contains fitting results for lognormal dataset
- Fitting 10 samples in each spectrum
- Fitting 100 samples in each spectrum
- Chi.nc that contains fitting results for chi_d^2 dataset
- Fitting 10 samples in each spectrum
- Fitting 100 samples in each spectrum
The parameters estimated include the sought turbulence (epsilon) quantity and the spectral slope (beta_1).
The testing was done as part of ATOMIX SCOR working group #160, with support from NSF grant #OCE-2140395 and contributions from national SCOR committees. The ATOMIX wiki has more information about the group's activities. A manuscript has been submitted relating to the above testing, and this repo will be updated once the citation is available.