For earthquake prediction, we should strategically link several methods, exchange and share earthquake-related information. The hydrological method has been recently developed and not yet had a suitable theory which can adequately explain hydrological precursors to earthquakes. Therefore it is important to clarify the mechanism of earthquake-related hydrological changes and turn the hydrological information into the information which can be easily understood by seismologists. From this point of view, we propose four roles of hydrological methods for earthquake prediction; (1) providing crustal deformation data estimated from hydrological changes, which is usually useful for the short-term prediction, (2) providing the information related to displacements of ground surface caused by groundwater level changes, which will enable us to eliminate some non-tectonic deformations of ground surface and contribute to improving S/N ratio in the geodetic measurement such as GPS observations, which will be useful for the mid-term or long-term prediction, (3) providing the information related to temporal changes in permeability in and around active fault zones, which contributes much to understanding the earthquake-cycle in the active faults and the long-term prediction, and (4) providing the information of pore-pressure changes in the focal region. At present, only the first role can be applied to the practical short-term prediction. Therefore detecting more reliable hydrological precursors to earthquakes depends on how they are connected to preseismic crustal deformation. The ‘strain model’ attributes groundwater changes to volumetric strain changes, and is the only model that can quantitatively explain the precursors at present although it needs more improvement. For the improvement, comparing groundwater level changes with crustal deformation is effective.
