Measuring flow in non-ideal conditions for short-term projects: Uncertainties associated with the use of stage-discharge rating curves

• We calculated uncertainties on flow rates in small streams.
• Flow rates were calculated using simulated rating curves derived from reference data.
• Reference data were obtained from stage and velocity measurements obtained in flumes.
• Rating curve hysteresis is the first factor in the overall uncertainty.
• Second factor is the inappropriateness of one mathematical function to describe the relation between stage and flow.

SummaryThe vast majority of hydrological stations are set up such that discharge can be estimated from the sole measurement of water height or stage above a local datum. Hydraulics laws show that in the right conditions there may be a unique and stable relationship between stage and discharge, which can be described by a rating curve. For short-term projects where there may be little choice for station location and time to construct a detailed rating curve, conditions for the use of rating curves may be less than ideal, potentially yielding high uncertainties on hydrologic measurements. This article evaluates uncertainties induced on instantaneous flow rates and cumulative annual flow volumes by the use of one-segmented rating curves in small streams. Uncertainty distributions were obtained by simulating rating curves calculated from random sampling of reference flow and stage data obtained with Doppler flowmeters. Factors tested included the number of manual gauged points, the type of rating curve (power vs polynomial), the use or not of the observed stage-of-zero flow, the spread of gauged points along flow range, and the measurement errors during gauging. Results could vary widely depending on the scenarios tested and sometimes yielded very high uncertainties. The best scenario yielded significant uncertainties on annual cumulative flow volume included between −13% and +14% for the low gradient streams and between −5% and +7% for the higher gradient streams, and for 22 manual gauged points per year. Our results show that, even in the best scenario, very significant uncertainty can result from using one-segmented rating curve in non-ideal situations in the field.