An effort to calibrate model parameters of watershed states

In the early days of hydrologic simulation, in the Stanford years, “pseudo calibration” was coined to describe attempts to calibrate model parameters within the range of influence of an initial watershed state. Pseudo calibration fails because the watershed state (like soil moisture or snow levels) is described by variables. Parameters like land slope and soil profile characteristics are constants.

Streamflows produced by a watershed model (and by watersheds!) are dependent on the initial watershed state as a model run begins. Watershed state is a combination of soil moistures, active groundwater storage, snowpacks, transient storage in upstream channels, and storage in lakes and reservoirs.

A current or historic streamflow observed at a point in time is entirely dependent on the watershed state. As forecasts for future streamflow are made, weather becomes more and more influential. With time, the continuously variable weather that the watershed experiences becomes more and more dominant, and the initial watershed state becomes less important. Finally, the streamflows created by a model for a given historic or future weather sequence become wholly dependent on the model parameters and the model structure independent of the initial watershed state.

Then and only then is reliable calibration of model parameters possible.

In most watersheds, the range of influence of the initial watershed state is six to twelve months, so simulated versus recorded comparisons are not given weight within the first year of a model run. In watersheds where huge aquifers or glaciers contribute to streamflows, the initial watershed state will be influential for many years.