Spectral and cross-spectral analysis of three hydrological systems

The mean hydrological and hydrochemical behaviour is analysed from observed time series on three agricultural catchments, located in different regions of France, named respectively Kervidy, Melarchez and Mercube. The time ser ies are considered as input data or output data. The input data for hydrolo gy and chemistry are respectively rainfall and nitrate leaching, and the ou tput data are represented by stream discharge and nitrate concentration in the stream. The method is based on the spectral analysis of the data. First the spectral density function of each time series is calculated. It appear s that nitrate concentrations measured at the outlet of the three catchment s exhibit a strong and unique one year periodicity. This periodicity is due to the hydrological regime and the time distribution of the nitrate availa bility in the soil. Second, we performed a cross-spectral analysis between the input and output data. The catchment is then viewed as a linear and sta tionary system that converts an input to an output through a transfer funct ion. The transfer function characterizes the behaviour of the catchment and is related to the real physical and chemical processes. For each catchment , the observed transfer functions for water and nitrate are determined in t he spectral domain from the data. By comparing the observed frequency trans fer function with the spectral signature of the theoretical model, we ident ify for each catchment: (i) the major processes that govern water and nitra te transfer and (ii) the characteristic time scale of these processes. The spectral and cross-spectral method appear to be valuable in identifying the main transfer processes operating in the different catchments. (C) 2000 El sevier Science Ltd. All rights reserved.