If you want "realistic" initial conditions, you can use the real.exe utility with reanalysis data to get starting states for the model. Most of the time for hydrologic models we run an extensive spinup before evaluating model outputs to let soil moisture, groundwater, etc. stabilize, so the initial conditions matter less. The create_wrfinput R (and now python - see
https://ral.ucar.edu/projects/wrf_hydro/pre-processing-tools) scripts will take the geogrid file, copy over the needed static variables (e.g., soil type, veg type), then put in global values for the initial conditions for states like snowpack and soil moisture. This is OK if you plan to run a spinup, but if you do not plan to run an extensive spinup then you might want to go through the extra effort of running real.exe.
Hope that helps.