核融合装置の周辺プラズマ輸送モデルを実測値で較正・比較

Cross-field turbulent transport remains one of the largest uncertainties in edge plasma simulations and is commonly approximated through empirical transport coefficients. In this work, we calibrate and assess several cross-field transport models implemented in SOLPS-ITER using measurements from the TCV-X21 reference case. The considered models range from conventional constant-diffusivity descriptions to the self-consistent k-model, in which anomalous diffusivities evolve along with plasma conditions. Model parameters are estimated through gradient-based optimization by minimizing discrepancies between simulated and experimental upstream and divertor profiles in forward field configuration. The calibration results show that increasing the number of free parameters substantially improves agreement with the calibration dataset. However, these more flexible models exhibit poor predictive capability when applied to the reversed field configuration, indicating overfitting. In contrast, the simplest constant-diffusivity model provides the best overall predictive performance while requiring only a small number of calibrated parameters. The k-model achieves a calibration quality comparable to the constant-diffusivity model and reproduces the experimental profiles with similar accuracy, while simultaneously providing a physics-based description of the spatial variation of anomalous transport. Predictions for a density scan reveal differences that are not apparent near the calibration point. The k-model predicts increasing transport levels around the separatrix with increasing density, leading to broader upstream profiles and an earlier onset of divertor rollover compared to the constant-diffusivity model. The presented framework provides a systematic and efficient route for model calibration in SOLPS-ITER, and a set of calibrated k-model parameters to be employed in future studies.