The Lake Kinneret hydrologic budget includes two unknown terms: evaporation losses Qe, and inflow from hot saline springs which represent the groundwater component, Qs. This last term affects also the salt budget as well as the energy budget of the lake, which is, in turn, the basis for evaporation estimates. A new approach consisting of the simultaneous solution of the water, heat, and salt balance equation is presented to evaluate Qe and Qs. This approach gives a better estimation of evaporation, since it accounts for the net energy advected to the lake. Neglecting this term led to overestimation of evaporation during the dry winter of 1991 and to underestimation during the whole year of 1992, which was particularly rainy. The maximal monthly error occurred for both years in February: ¿55% for 1991, and -29% for 1992. The application of the approach to the period of 1988--1991 shows that the annual hydrograph of the saline springs system is characterized by two distinct periods of high discharge during each year. The first period, in winter, seems to be related to the structure and timing of the rainfall season. The second occurs generally at the beginning of the summer, after the rainfall season has ended. The overall trend for the period is declining peaks and annual discharges, which is consistent with the hydrologic impact of a 3-year drought since 1989. The measured, as well as the estimated, components of the water balance equation are subject to uncertainty.

The formulation of the balance equation in terms of a state-space system and the application of the Kalman filter (KF) allows quantifying the effect of this uncertainty on the different components. Therefore for given noise statistics and a priori data, it is possible to determine the optimal linear estimates of the estimated components and to compare them to the values resulting from the simultaneous solution approach. The method was applied on monthly water budgets for a period of 60 months. The effect on evaporation varies with a range of ¿5%. The estimate of runoff from the ungauged part of the direct watershed was generally increased by the KF, especially during the rainy winter of 1988. The effect on this estimate varies within a range of ¿12%. The component that was affected the most is the estimated discharge from the saline springs system (¿15%). The KF increases by 15% the winter peak of 1988 and decreases all the discharges estimates for the dry years, especially those of 1990. Therefore the stochastic approach strengthens the agreement between the trends in the estimated values and in the climatic conditions during the period studied. ¿ American Geophysical Union 1993