We analyse the correlative relationships between various quantities derived
from 2-D inhomogeneous and time-dependent model atmospheres and between se
lected simulated line parameters to compare them with height-dependent corr
elations derived from spectral observations. We detect three photospheric r
egions: thermal convection, overshooting convection and a transition layer.
We also show that correlations found for the model data and those computed
within simulated spectral observations are a good testing tool for line fo
rmation depths. As an example, we examine two criteria, providing heights o
f line core formation in LTE, and conclude that the approach which defines
this quantity as geometrical height at line center optical depth tau (lambd
a0) = I is likely more suitable for diagnostic purposes than the method bas
ed on depression contribution function.