The study examines the potential applications of artificial neural networks in predicting the time series of water levels and river flows. The data for computations comprised the time series of daily water levels and river flows at the gauging sections in Gdów and Stróża on the Raba river and in Marwin on the Krzyworzeka stream from the hydrological years 1979-1981 and the daily precipitation totals from 1978–1981 recorded at the meteorological station in Dobczyce. Two separate neural networks were created for the time series of water levels and river flows. It was decided to build perceptron networks differing in architecture. An analysis showed that the efficiency of a perceptron network in predicting water levels and river flows at the gauging section in Gdów on Raba is higher when the water levels and river flows at the Stróża gauging section on Raba and the Marwin gauging section on Krzyworzeka along with the daily precipitation totals in Dobczyce are assumed as input parameters than in the case in which the model takes no account of precipitation. Slightly better results of prediction were obtained for river flows than water levels, which may be due to somewhat different factors determining the former.

59-73