Multivariate technique for validating historical hydrometric data with redundant measurements

The aim of this research was to develop an automated methodology for valida ting chronological series of natural inflows to reservoirs. Theoretically, gauges located on the same reservoir should indicate the same reading. Howe ver, under the influence of meteorological and hydraulic factors, or simply because of failed measuring equipment, there may be large deviations betwe en the various measurements. Since the calculation of historical natural in flows is directly linked to the measurement of reservoir level by the water balance equation, there will be as many series of natural inflows as there are of reservoir levels. A multivariate filtering technique is used to val idate the historical natural inflow computed by each water level variation. The multi filter methodology has the advantage of balancing the water volu me of natural inflows to the reservoir when applied over a relatively long period of time. As a result, the validated flood peaks are not systematical ly overestimated or underestimated and the validated natural inflows are ne arly identical for all the gauges. The proposed technique has been incorpor ated into a software program called ValiDeb, which has been successfully te sted on-site on the Gatineau River in Quebec.