Mots cles : gestion de reservoir, previsions, apports, risque, incertitude. Abstract: The behaviour of a hydric system depends on three factors : (i) the state of the installation, ( ii) the operating rules, and (iii) the inflows. While the first two factors are (in theory) known to the manager, the third can only be esti - mated by means of more or less precise forecasts. A significant part of the risk, to which is subjected the system at a given time, is induced by the uncertainty in the future inflows. The evaluation of this uncertainty is therefore a first step in the incorporation of risk into management. Its evaluation is then a stage preliminary to the integration of the risk in management. A method of construction of inflow scenarios starting from an arbitrary date t of the year is devel- oped in this paper. It uses a Markovian process formerly developed by the authors to model short-term uncertainty in stream flow. These scenarios, which are not equiprobable, are built to reproduce the statistical behaviour of the river or reservoir and have the shape of an event tree whose structure is defined by the user before application of the method. Two examples of application on two rivers located in Quebec, Canada, are presented.
[1]
Emanuel Parzen,et al.
Stochastic Processes
,
1962
.
[2]
Gwilym M. Jenkins,et al.
Time series analysis, forecasting and control
,
1972
.
[3]
O. Seidou,et al.
Modélisation de l'incertitude sur les séquences futures de débits en rivière
,
2002
.
[4]
N. Matalas.
Mathematical assessment of synthetic hydrology
,
1967
.
[5]
Jose D. Salas,et al.
Gamma‐Autoregressive Models for Stream‐Flow Simulation
,
1990
.
[6]
A. I. McLeod,et al.
Advances in Box-Jenkins modeling: 1. Model construction
,
1977
.
[7]
Jose D. Salas,et al.
Periodic Gamma Autoregressive Processes for Operational Hydrology
,
1986
.
[8]
Gideon Weiss,et al.
Shot noise models for the generation of synthetic streamflow data
,
1977
.