In some cases, a record of historical flow or stage can provide all the information needed for the decision making. For example, suppose that the 0.01 Annual Exceedance Probability (AEP) stage at a floodplain location is required for regulating floodplain activities. If a long continuous record of measured stage is available, fitting a statistical distribution to the record (following procedures described in EM 1110-2-1415 - USACE, 1993) and using this fitted distribution to find the stage will provide the information required for the decision making.
Historical records are not often available or are not appropriate for the decision making. The record length may be too short for reliable statistical analysis, the gage may be at a location other than the location of interest, or the data of interest may be something that cannot be measured.
For example, to compute Expected Annual Damage (EAD) with which to compare proposed flood-damage measures in a watershed, runoff peaks are required. But until the measures are implemented and floods occur, no record of peaks can be available. Implementing the measures and waiting to see what impact the changes will actually have is unacceptable, as the benefits of the measures must be determined before decisions can be taken to expend funds to implement the measures.
Similarly, a record of inflow is needed to determine appropriate reservoir releases should a tropical storm alter its course and move over the contributing watershed. But until the rain actually falls and runs off, no record of such inflow will be available. Waiting to observe the inflow is not acceptable, because actions must be taken beforehand to protect the public and property.
In these cases, flow, stage, velocity, and timing must be predicted to provide the required information. This can be achieved with a mathematical model of watershed and channel behavior – a set of equations that relate something unknown and of interest (the model's output) to something known (the model's input). In hydrologic engineering studies, the known input is precipitation or upstream flow and the unknown output is stage, flow, and velocity at a point of interest in the watershed.