The Controlled Shear Decorrelation Experiment (CSDX) linear plasma device provides a unique platform for investigating the underlying physics of self-regulating drift-wave turbulence/zonal flow dynamics. A 3D drift-reduced nonlocal cold ion fluid model which evolves density, vorticity, and electron temperature fluctuations, with proper sheath boundary conditions, is used to simulate dynamics of the turbulence in CSDX. The simulations show density gradient-driven drift-waves are the dominant instability in CSDX. However, the choice of insulating or conducting end plate boundary conditions affects the linear growth rates and energy balance of the system due to the absence or addition of Kelvin-Helmholtz modes generated by the sheath-driven equilibrium E×B shear and sheath-driven temperature gradient instability.

The physics study is then followed by an extensive quantitative validation study. For the direct comparison of nonlinear simulation results to experiment, a synthetic Langmuir probe diagnostic is used to generate a set of synthetic observables which are in turn used to construct the validation metrics. A significant improvement of model-experiment agreement relative to the previous 2D simulations is observed. An essential component of this improved agreement is found to be the effect of electron temperature fluctuations on floating potential measurements, which introduces clear amplitude and phase shifts relative to the plasma potential fluctuations in synthetically measured quantities. Moreover, systematic simulation scans show that the self-generated E×B zonal flows profile is very sensitive to the steepening of density equilibrium profile. To minimize the number of simulations required for uncertainty quantification, a more advanced seed sampling methods for efficient sampling of the input parameter space, and a rapidly converging non-intrusive Probabilistic Collocation method to map out a probabilistic model response is implemented. This approach is shown to yield significantly better constrained uncertainty estimates than conventional uniform sampling methods for practical numbers of nonlinear simulations.