Wafer Resources Management 4: 79-95, 1990. o 1990 Kluwer Academic Publishers. Printed in the Netherlands. 79 Extreme Dry Weather Intervals of the Growing Seasonin BaCka, Yugoslavia MILICA BERI6, EMIR ZELENHASIk, and BOJAN SRDJEVIe University of Novi Sad, Faculty of Agriculture, Institute for Water Resources Development and Management, Veljka Vlahovi6a 2. 21000 Novi Sad, Yugoslavia (Received: 23 July, 1989; revised 28 February, 1990) Abstract. A method of describing and analyzing the stochastic process of droughts, which are defined here as the upper extremes of intervals of no rainfall, is recommended. All important components of extreme dry weather intervals such as their duration, time of occurrence, their total number in a given time interval [0, t], the longest drought duration in a given time interval [0, t], and time T(t) of occurrence of the longest drought are taken into consideration. Application of the method is performed using the records of nine meteorological stations in BaEka, Yugoslavia and a good agreement is found between the theoretical and empirical distribution functions for all analyzed components of the process. On the basis of the performed computations, a set of maps showing the contours of extreme dry weather intervals, having return periods from 2, 5... up to 100 years, related to the growing season (1 April - 30 September), for the region of BaEka in Yugoslavia is obtained. If the period of exploitation of an irrigation system is 60 years, then it could be expected to appear as 20 dry weather intervals having 3 year return periods. The duration of dry weather intervals is given on the set of maps. The obtained results give a prognosis of an average state of droughts during long time intervals (60, 100,200... years). Key words. Drought, rainless period, stochastic process of extremes. 1. Introduction A method of analyzing extreme dry weather intervals in the growing season at a given meteorological station and over a given region is presented in this paper. Extreme dry weather intervals are also synonymously called extreme rainless periods, or droughts, in this paper. The method attempts to develop a general stochastic model of extreme rainless intervals and uses the data on rainless periods above a given reference value, Y,. All important components of the process, such as drought duration, time of occurrence, number of droughts in a given time interval [O,t], the longest drought in a given time interval [OJ], and its time of occurrence, are taken into consideration. Droughts are defined here as the upper extremes of dry weather intervals and are treated as a random number of random variables in an interval of time [O,t]. The method is based on the assumption that droughts are independent, identically distributed random variables and that their occurrence is subject to the Poisson probability law. This is confirmed for each of nine analyzed meteorological stations of the region considered in this paper. An application of the method is shown for