Confrontation between modelled solar integrated observables and direct observations I. Radial velocities and convective blueshift

Stellar variability strongly impacts the search for low-mass exoplanets with radial velocity techniques. Two types of planet-free time series can be used to quantify this impact: models and direct solar observations after a subtraction of the Solar System planetary contribution. Comparing these approaches is necessary for simulations. Our objective is to validate the amplitude of the convective blueshift in plages used in our previous works, particularly in blind tests, with HARPS-N solar data. We applied our model to the structures observed at the time of observations and compared the radial velocity time series. To complete our diagnosis, we studied the observed radial velocities separately for each diffraction order derived from the individual cross-correlation functions, as well as our line-by-line radial velocities. We find that our previous model had been underestimating the amplitude of the convective blueshift inhibition by a factor of about 2. A direct estimation of the convective blueshift in the spectra explains the difference with previous estimations obtained with MDI/SOHO Dopplergrams, based on the properties of the Ni line. We identified several instrumental systematics: the presence of a 2 m/s peak-to-peak signal with a period of about 200 days in radial velocity and bisector, which could be due to periodic detector warm-ups, a systematic dependence of the long-term trend on wavelength possibly related to the variability of the continuum over time, and/or an offset in radial velocity after the interruption of several months in Oct. 2017. A large amplitude in the convective blueshift inhibition of (360 m/s) must be used when building synthetic times series for blind tests. The presence of instrumental systematics should also be taken into account when using sophisticated methods based on line properties to mitigate stellar activity when searching for very weak signals.