INTERPRETIVE NOTES

Peak Flow Defined

The peak flow forecast represents the maximum mean daily flow at a point during the April through July period (the highest average flow for an entire day during the runoff season). It does not represent the instantaneous peak (the maximum flow at a single moment). In the case of smooth snowmelt regimes (hydrographs), it may be acceptable to approximate one with the other. In Arizona, the normal snowmelt period is from March to May. Occasionally, heavy rainfall events can produce higher peak flows than the snowmelt peak flows. For verification and calibration purposes, the maximum mean daily flow during the March through May period was used regardless of the runoff source. The Average Peak and Normal Time of Peak for a given gage are all derived from the period of record through 1990 whereas the Historic Peak is derived from the entire period of record, including the most recent years.

Forecast Probabilities

Peak flow forecasts are presented in terms of probabilities or, more specifically, exceedance probabilities. The so-called "most probable" forecast actually corresponds to a flow that is equally likely to be too high or too low, the 50% exceedance level (i.e., 50 chances out of 100 of being exceeded). The other exceedance probabilities associate the likelihood of exceeding other levels. In general, a close bunching of the exceedance forecasts indicates low variability and that the user can have a high degree of confidence in the forecast information. Conversely, a large spread in the exceedance forecasts indicates high variability.

Modelling Techniques

The peak flow forecasts that follow have been derived using a combination of (1) physically-based conceptual models and (2) statistical regression models. The conceptual model is the National Weather Service River Forecasting System (RFS) in the Extended Streamflow Prediction (ESP) mode. Since the conceptual model requires reservoir operation plans for up to five months into the future, ESP application is limited to basins where regulation is minimal (mostly in the headwater areas). The farther downstream a forecast point is, the more likely it is that a statistical regression was used between natural snowmelt runoff volume and the observed maximum mean daily flow to generate the forecast. Such an approach performs better when the correlation between regulated and unregulated flow is strong and is constant from year to year.