Context. A proper estimate of the chromospheric magnetic fields is thought to improve modelling of both active region and coronal mass ejection evolution. However, because the chromospheric field is not regularly obtained for sufficiently large fields of view, estimates thereof are commonly obtained through data-driven models or field extrapolations, based on photospheric boundary conditions alone and involving pre-processing that may reduce details and dynamic range in the magnetograms.

Aims. We investigate the similarity between the chromospheric magnetic field that is directly inferred from observations and the field obtained from a magnetohydrostatic (MHS) extrapolation based on a high-resolution photospheric magnetogram.

Methods. Based on Swedish 1-m Solar Telescope Fe I 6173 Å and Ca II 8542 Å observations of NOAA active region 12723, we employed the spatially regularised weak-field approximation (WFA) to derive the vector magnetic field in the chromosphere from Ca II, as well as non-local thermodynamic equilibrium (non-LTE) inversions of Fe I and Ca II to infer a model atmosphere for selected regions. Milne-Eddington inversions of Fe I serve as photospheric boundary conditions for the MHS model that delivers the three-dimensional field, gas pressure, and density self-consistently.

Results. For the line-of-sight component, the MHS chromospheric field generally agrees with the non-LTE inversions and WFA, but tends to be weaker by 16% on average than these when larger in magnitude than 300 G. The observationally inferred transverse component is systematically stronger, up to an order of magnitude in magnetically weaker regions, but the qualitative distribution with height is similar to the MHS results. For either field component, the MHS chromospheric field lacks the fine structure derived from the inversions. Furthermore, the MHS model does not recover the magnetic imprint from a set of high fibrils connecting the main polarities.

Conclusions. The MHS extrapolation and WFA provide a qualitatively similar chromospheric field, where the azimuth of the former is better aligned with Ca II 8542 Å fibrils than that of the WFA, especially outside strong-field concentrations. The amount of structure as well as the transverse field strengths are, however, underestimated by the MHS extrapolation. This underscores the importance of considering a chromospheric magnetic field constraint in data-driven modelling of active regions, particularly in the context of space weather predictions.