Relations pluie-débit sur la Saône à Lyon

Abstract

The river Saone, a main tributary of the Rhone with a catchment area of some 30,000 km2 , has a major effect on electricity production of water-power stations below Lyon. For plant management purposes (especially in deciding on suitahle maintenance schedules), maximum discharge-forecasting efficiency is desirable ; hence the importance of the direct relationship between and future discharge. In the rainfall/runoff relationship, rainfall can be reduced to rainfall and the latter can be converted to runoff. Reduction is effected by means of the discharge coefficient, which depends on fhe three following parameters : 1. State of vegetation (seasonal parameter Tj) 2. State of soil (absorption parameter Qo) 3. Recent humidification (previous precipitation index IPA) K = 1 - Tj (1 - IPA/G) /.B c Q0!Q0 + C (1 - IPA/G) Variables Tj' Qo and IPA are normal seasonal temperature on the considered day at Besancon, initial discharge of the river Saone through Lyon and average in the catchment on the previous day respectively. Parameters B, C and G were first adjusted by trial and error from a sample of 92 Saone flood discharges measured at Lyon between 1961 and 1974 and sufficiently individualized over fhe considered period (B = 25°, C = 200 m3/s, G = 50 mm). The transformation is a linear operator (unit hydrograph) relating effective Pj to discharge on subsequent days (J to J + K). Coefficients Ai are obtained from the following equation by multiple correlation from fhe 92-flood sample : Q (J+K) = n Σ i = 1 AiP (J - i + K + 1) + DK Q0 The transfer and reduction function parameters were then corrected after simulating operation of the model on the sample whilst allowing for the flood-flattening effect of the river Saone's flood plain at discharges exceeding 1 000 m3/s. Finally, the model was tested continuously over the 1975 to 1977 period, which had not been used for its adjustment. The numerical results were compared with those a simple persistence model would have provided. The residual standard deviation of the model (g) compared to that of a persistence model is a maximum for fhree and four-day forecasts, then falling off fairly fast if cannot be forecast. Term J J + 1 J + 2 J + 3 J + 4 g 10 % 28 % 40 % 42 % 30 % Difficulties were experienced owing to interaction of the reduction and transfer functions. Any method whereby the parameters of both functions could be optimized at the same time, therefore, would considerably facilitate adjustment of fhe model, which is already suitable for medium-term operational decision-making in its present state.