GeoClaw has a parameter sealevel (see Specifying GeoClaw parameters in setrun.py) that can be used to specify the initialization of the fluid depth relative to the specified topography (see Topography data). Unless a different set of initial conditions is specified (see qinit data file parameters), the default is to initialize with zero velocity and depth h chosen so that h+B = sealevel at any point where B < sealevel, where B is the topography or bathymetry in the grid cell (as determined by interpolation from the specified topo files as described in Topography data).
It is important to know what vertical datum the topography is relative to. Coastal bathymetry developed for tsunami modeling (e.g. from NOAA NGDC inundataion relief) is often relative to Mean High Water (MHW), in which case setting sealevel = 0. corresponds to assuming the water level is initially at MHW. To adjust to use a different tide level, the value of sealevel must be set appropriately. The relation between MHW and other tide levels such as Mean Sea Level (MSL) can often be found from the NGDC webpages for a nearby tide gauge. For example, if you go to a station page such as Hilo Bay, you will see a Datums link on the left menu. (Be sure to switch from feet to meters!)
Note that the difference between MHW and MSL can vary greatly between different locations. Global bathymetry data such as the ETOPO1 data (available from GEODAS Grid Translator - Design-a-Grid) is generally relative to MSL. However, this data has a resolution of 1 arc-minute, more than 1.5 km, and is not suitable as coastal bathymetry, so this data will presumably only be used in grid cells away from the region where coastal bathymetry is available. Since the difference between MSL and MHW is at most a few meters, the use of different vertical datums for regions of vastly different resolution will generally have little effect.
If GeoClaw is used to compare inundation or tide gauge values to observations from past tsunamis, it may be important to know the tide stage when the largest tsunami waves arrived. Ideally it would be possible to model the actual rise and fall of the tides during the duration of the event, but this is not currently possible. Tidal currents may also have a significant effect on observed inundation patterns, but these are also ignored in GeoClaw since the water is assumed to be at rest before the tsunami arrives.
If GeoClaw is used for hazard assessment based on potential tsunami scenarios, then thought should be given to the appropriate value of sealevel to assume. The NGDC coastal bathymetry data is referenced to MHW because this is often the level assumed for tsunami hazard assessment, but higher tide levels such as Mean Higher High Water (MHHW) or the Astronomical High Tide (AHT) are sometimes used for worst-case scenarios.
See Some sources of tsunami data for some other sources of data.