Model Parameters from the thesis "Bounding Anomalous Transport: Prior Distributions for Tokamak Simulation Validation"

Data description — exported prior/model parameters The entries are prior/model parameters over plasma particle transport coefficients (diffusion coefficient $D$), derived from ASDEX Upgrade (AUG) data. Everything is expressed on a base-10 log scale $x = \log D$ and resolved per normalized poloidal flux radius $\rho_\text{pol}$. The underlying reference scales follow the physical ordering $D_\text{classical} \lesssim D_\text{gyro-Bohm} \lesssim D_\text{Bohm}$ (classical = collisional lower bound, Bohm $\propto B^{-1}$ = upper bound, gyro-Bohm $\propto B^{-2}$ = expected anomalous-transport regime). Each quantity is provided in three coordinate variants: - $D_*$ — raw physical estimates, SI units m²/s - $D_{\rho,*}$ — converted to the normalized flux coordinate via the metric factor $\langle |\nabla \rho_\text{pol}|^2 \rangle$ (flux-surface-average, now in flux-coordinate units, no longer m²/s) - $V' D_{\rho,*}$ — additionally flux-volume-scaled by $V' = \mathrm{d}V/\mathrm{d}\rho_\text{pol}$ 1. Hierarchical-prior models (eta-boost and mixture families) A hierarchical Bayesian prior over $\log D$ built from physical beliefs alone. It encodes the bounds $D_\text{classical} < D < D_\text{Bohm}$, a preference for the gyro-Bohm region, and a proximity decay whose exponential form follows from the maximum-entropy principle and whose scaling is fixed by translation/scale invariance. Columns per radius: $$(\rho_\text{pol}, L, U, a, b, \tau_{L,\min}, \tau_{U,\min}, \tau_{a,\min}, \tau_{b,\min})$$ - $L, U$ — lower/upper bounds of the broad support $[L, U]$ (deep-tail quantiles of the classical-electron and Bohm log-diffusion samples) - $a, b$ — edges of the preferred gyro-Bohm region $[a, b] \subset [L, U]$ (central bulk of the gyro-Bohm samples) - $\tau_{j,\min}$ — data-informed minimum softness scales for each boundary $j \in {L, U, a, b}$, set so a Student-t boundary window covers at least the empirical tail mass observed beyond that bound Two families are exported separately. Mixture enforces the gyro-Bohm mass preference ($\Pr(R)\ge\tfrac12$) additively — a two-component mix of an in-region and out-of-region prior. Eta-boost enforces it multiplicatively — tilting the base density in log-space by $\exp(\eta, g_R(x))$. The two agree in the least-informative limit. Both marginalize the remaining construction uncertainty (quantile levels, degrees of freedom, softness, decay scale) by Monte Carlo. Use: recommended for synthetic validation and general plausibility checking of transport/reconstruction codes — draw coefficient profiles from the prior as known ground truth and check the code recovers them across the plausible range. Modular by design, the physical beliefs are separate likelihood factors that can be added, removed, or replaced, with the resulting uncertainty propagated automatically. The same $\log D$ prior also applies to the heat diffusivities $\chi_e, \chi_i$ (both gyro-Bohm order) and maps to a pinch-velocity prior via the additive shift $-\log a$. 2. Regime distributional priors (Student-t mixtures) Empirical two-component Student-t mixtures fitted per radius to the AUG estimates, capturing the observed multimodal regime structure. Columns per radius: $$(\rho_\text{pol}, w, \nu_1, \nu_2, \mu_1, \mu_2, \sigma_1, \sigma_2)$$ - $w$ — weight between the two components - $\nu_1, \nu_2$ — degrees of freedom (tail heaviness) - $\mu_1, \mu_2$ — component locations (in $\log D$) - $\sigma_1, \sigma_2$ — component scales One entry per reference regime: Bohm, classical-electron, classical-ion, gyro-Bohm. Use: recommended specifically for regime-combination recovery testing — probing whether a code recovers correct behavior when the coefficients ($\chi_e, \chi_i, D, v_\text{pinch}$) simultaneously sit in different regimes (each mixture component represents a distinct empirical regime). For general validation the hierarchical prior is sufficient.