The use of time series methods in the analysis of hydrological data

Abstract

The account is based upon recursive time-series analysis

and its application to the study of river catchment behaviour in order to predict future events. The Lerderderg Representative Basin, in Victoria, was selected as a range of hydrological data was available for this catchment on magnetic tape from the Land Use Research Division of CSIRO. Additional information on soil morphology and fire history was obtained from other sources. The rainfall, runoff, evaporation and temperature records was analysed using the CAPTAIN package program and both short-term (hourly data) and long term (daily data) were considered. Since there were no available observations for soil moisture the non-linear soil moisture compensation algorithm of CAPTAIN was used. Transfer functions and steady state gain were calculated and impulse responses analysed. Short-term response was found to give a better explanation of the behaviour of the Lerderderg river system than that using long­ term response. For the long term analysis the model employing temperature effects was found to be more satisfactory than those based upon evaporation. The most successful model used dry bulb temperatures although as might be expected, very similar results were obtained for analysis based upon daily maximum temperature. Problems in using the technique are discussed and suggestions made for future lines of inquiry.