The heat and particle transport in a magnetized fusion plasma is driven by turbulent processes. Collective effects, charged particles and electromagnetic fields, make turbulence in fusion plasmas more complex than in fluid. Such complexity together with the large number of parameters motivate to test the data-driven discovery approach. Database exploration has been used very successfully for parameters such as the energy confinement time, but this systematic approach has not been applied to data such as turbulence measurements.
A systematic study of the density fluctuations probed by reflectometry is undertaken. Because each fluctuation reflectometry measurement produce tens of thousands of values, data reduction must first be applied to get a handful of parameters to characterize each measurement. These parameters together with the cutoff position, the local and global parameters can then be stored in a database. First investigations point out strong differences between the fluctuation spectra in Lower Hybrid (LH) and Ion Cyclotron Resonance Heated (ICRH) discharges. The effective collisionality was identified as a key parameter to unify LH and ICRH observations. Change of the dominant turbulent instability with increasing effective collisionality is proposed as the underlying explanation for the spectrum evolution.
Reflectometry is a radar-like technique able to access the density fluctuations with high sensitivity and good spatial and temporal resolution. Fixed frequency plateaux are the standard method to probe density fluctuations along the radius. On Tore Supra and now WEST, the D-band (105-155 GHz) has probed from the low field side up to the inner side in the X mode polarization . Compared to Tore Supra, the accessibility on WEST is shifted toward the outer side. Several times per discharge, 20 to 40 frequency steps are used. Fourier transform is the first obvious step. To reduce further the data, the spectrum is modelled by a sum of components. Three components and a noise level are enough to fit the spectra. The two central components -the low frequency (LF) and direct reflection (DC) [2,3]- are fitted by a Gaussian. For the Broadband (BB) component, a fourth parameter such as the exponent of the generalized Gaussian function is needed to fit the BB (Gaussian to Lorentzian shape). Eleven parameters are then enough to describe the 1024 values of a frequency spectrum .
The amplitudes are normalized to the spectrum energy to remove the signal amplitude variation. Other narrow components are currently neglected, as their contribution to the spectrum is small, but the inclusion of quasi-coherent modes  is planned.
Broadband component collisionality and turbulence regime
Although the eye detects spectrum modifications with the probing frequency, or with the experimental settings, a lack of quantification prevented further investigation of this control-room impression. The decomposition method makes these studies easier and systematic.
At similar heating power, the contribution of the BB component is usually much higher in ICRH plasmas than in LH plasmas. The effective collisionality ν_eff was found to bring coherence. With increasing ν_eff, the normalized BB amplitude E_BB increases both in Ohmic and L-mode discharges as shown in figure 2. In contrast, the LF component, which is the main component at low ν_eff, vanishes at high ν_eff in heated discharges. Overlapping LH and ICRH data and mixed LH & ICRH discharges show that this trend is independent of the heating scheme.
These spectrum modifications could be caused by a change of the turbulent instability from TEM (Trapped Electron Modes) to ITG (Ion Temperature Gradient) . A change of the turbulent instability with increasing ν_eff is supported by a lower density peaking at high ν_eff. Gyrokinetic simulations  predict a broad density fluctuation spectrum when ITG are dominant. For the TEM case, the spectrum exhibits a narrow intense peak at low frequency. This peak is separated from the TEM induced BB component, which is less intense and narrower than in the ITG case. Such medialisation is coherent with the spectrum observations.
Parametrization of the reflectometry spectrum allows us to build a database with hundreds of thousands measurements in very different plasma conditions. The modifications of the reflectometry spectrum with the effective collisionality are interpreted as a consequence of a change of the main turbulent instability. This could open the way for an experimental control-room clue to identify the turbulent instability regime.
 R. Sabot, et al, Nuclear Fusion 46 (2006), S685–92
 V.A. Vershkov, et al, Nucl. Fusion 45 (2005), S203–S226
 A. Krämer-Flecken, 44 (2004) 1143–57
 Y. Sun, et al., 89 (2018), 073504
 H. Arnichand, et al, Plas. Phys. Cont. Fus. 58 (2016) 014037
 Y. Sun, et al, Phys. Plas. 26 (2019) 032307
|Country or International Organization||France|