In order to achieve the effective control of water resources systems, one must know the future behavior of the inputs to that particular system. Because of the uncertainties inherent in water resources processes, the prediction algorithm, to be constructed, should include stochastic elements, too. Moreover, the algorithm should be recursive to avoid cumbersome computations and to be able for real-time forecasting. 

In this paper we present a method which is applicable for both linear and nonlinear hydrologic systems having not completely time-invariant properties. The algorithms are based on the state space description of the processes involved and utilize the Kalman stochastic filtering technique. Due to the unknown nature of noise processes, the basic algorithms were changed to be adaptive. Using the algorithms the joint handling of water quantity and quality data becomes feasible.