The plots show 10 MHz ground constants at eleven locations in Great Britain from a paper by R. L. Smith-Rose. At each location the soil was sampled at several depths to 10′ and then analyzed in a lab. ESU ∕ 108 = 11.1 mS/m.
Rudy Severns, N6LF, used a 12″ ground probe and NanoVNA-H4 to measure ground constants in a newly dug trench in Oregon bottomland. He inserted the probe horizontally in the trench wall. The soil was loam down to the bottom at 6′ where it became gravel. He inserted the probe vertically at 6′ and at the surface near the trench.
It's tempting to calculate effective ground constants by averaging the values at each depth weighted by antenna-induced current, which decays exponentially at rates the ground constants determine. But this simple model does not account for power reflected at soil layers due to ground impedance differences. Reflections and re-reflections alter the current. It's like cascading transmission lines of different characteristic impedance. Now consider an example with two soil layers. Reduce the ground constants of the upper layer until they become those of air. This increases antenna height by the depth of the first layer. But no set of effective ground constants can do this. This extreme case is on a continuum. Evidently multiple ground layers require a more complex ground model.
The effects may not be simple to calculate, but it is reasonable to expect that subsurface soil that differs from surface soil and has significant current may affect an antenna in ways not predicted by surface ground constants.
For total reflection at a subsurface soil layer one skin depth deep, reflected current at the surface will be 14% of incident current. For realistic partial reflection, it will be less. This suggests that nearly all subsurface soil effects occur within one skin depth. For the British measurements, surface soil skin depth is 2.5′ to 7.7′. For the N6LF data, it is 5.5′. For most of the soils, ground constants vary tens of percent within one skin depth of the surface. A few vary even more. Subsurface soil effects are most likely for poor soil at low frequencies where skin depth is greatest. Use this program to calculate skin depth.
This article explains how to determine ground constants with a low dipole. The method excites relevant ground. However, it is restricted to a single band, needs a large, flat, open area free of interfering wires, structures, vegetation, and ground conductors, requires meticulous construction, measurement, and modeling, and must be left in place or rebuilt to measure at different seasons. Despite these limitations, the method may be useful if you suspect your ground is inhomogeneous. Use bare copper wire. Knot dacron line over it near an end to avoid an unmodeled wire loop and insulator, or run dacron line through the tiniest possible loop. Measure and model the capacitance at the center insulator beyond that of straight wire (exclude connector). Attach the VNA directly using a male-male adapter, or add a wire to one side of the dipole model to represent the cable and VNA. Record measurements without touching the VNA. Recalculate the Sommerfeld-Norton parameters whenever you change ground constants. Since so many things can go wrong, do a sanity check against ground probe measurements. To overcome the single-band limitation, use this program to extrapolate ground constants to other bands.
Smith-Rose, R. L., "Electrical Measurements on Soil with Alternating Currents," Proc. IEE, Vol. 75, pp. 221-237, 1934. I skipped the two locations with indefinite depths and the location with no surface measurement.