There are two basic strategies for the definition of initial hydraulic heads for transient simulations:
[b]1. Interpolation from observations[/b]
If a large number of observations is available for the starting time of the model and the observations sufficiently cover the entire model domain (both horizontally and overlying aquifers), the initial conditions can be derived from the observed hydraulic heads by interpolation.
[b]2. Results of a steady-state simulation[/b]
In case that not enough observations are available for direct interpolation, initial hydraulic heads can be obtained from a steady-state simulation. As the hydraulic-head distribution in this case reflects average conditions, the model might need a starting period to adapt to actual conditions before producing reliable results.
For me it seems available data for the initial condition are rather sparse. Accordingly, the second option might be usefull. If you still want to use the first option you may try to regionalize the data by means of a parameter link.
The introductory tutorial, page 34 shows you one possible workflow.
https://www.mikepoweredbydhi.com/download/product-documentation#filters={44F569D8-FB28-4ED6-AE2E-192DC0A62123}|&c={9C84B9E1-0E5E-4D93-9355-DDA963D938CA}&o=asc&prefilter={79E8A24B-1EAC-4A6C-A678-999EE0D8865A}|{8BEDA1F4-1938-4FC7-AD3F-D2A6FFC8E7AA}|{9A70C578-9030-44CD-8B58-0A7D9369B9FB}|{6174231F-96A9-4DFD-B62E-595C173BDC1A}