This has been discussed a little in a previous thread. You may want to use the search option for "underground structures" and "seismic". However, for you IJR, I will repeat part of my response to that inquiry and hopefully address yours as well.
At rest lateral earth pressure will be larger than the rankine or coulomb active pressure however it is a "rule of thumb" treatment of the actual exercise. Of course, for dynamic considerations the inertia of the wall and soil mass must be included in the analysis.
Mononobe-Okabe (MO) is a psuedostatic method based largely on coulomb's active earth pressure theory. Therefore it is not technically accurate for non-yielding walls but is a good approximation of typical free-standing retaining walls. Another distinction here is the application of the equilvalent lateral force, which is is applied at mid-height of a rectangular shaped pressure diagram rather than at the third point of a triangular shaped pressure diagram.
For walls that are non-yeilding or as in the case of a box culvert or similar structure it can be assumed that coulomb's and rankine's active earth pressures cannot be developed and the wall pressures may be obtained from the elastic solution for the case of a uniform, constant, horizontal acceleration applied throughout the soil. Thus your coefficient is really a(h)/g as opposed to Ka. Where a(h) is the horizontal acceleration coefficient. And the 1/2 quantity which arises from triangular distribution is now 1.0 for rectangular distribution.
However, there are other concerns with these underground applications that must be discussed for a complete analysis. More focus needs to be placed on displacement based concerns rather than force based concerns as above. Of course, this goes beyond whether the wall yields but what really happens when one section of the box displaces out-of-phase with another section etc.
For a more complete discussion I refer you to Soil Dynamics by Shamsher Prakash.
