Issue |
La Houille Blanche
Number 7-8, Novembre 1981
|
|
---|---|---|
Page(s) | 605 - 608 | |
DOI | https://doi.org/10.1051/lhb/1981059 | |
Published online | 01 December 2009 |
Analyse statistique des périodes de sécheresse
Statistical analysis of dry spells
Université Technique de Budapest, Hongrie
Abstract
Average rainfall in Hungary is between 6 and 800 mm according to the area. Its distribution over time is quite capricious. The Great Plain of Hungary is subject to drought effects. The statistical parameters for dry spells and their formation during the summers have been examined. The analysis has been based on the most reliable observations of rainfall conditions in the Great Plain where the Szeged station is equipped with a shade recorder, with a series of measurements spanning the period 1926 to 1968. The periods of lightest rainfall have been taken as drought periods. Thus periods of three days or more close to h = 0 where determined, in which the daily rainfall takes the following maximum: values 1 mm, 5 mm and 10 mm. The lack of precision of short dry spells (up to 3 days) has thus been counteracted. Such dry periods have no effect on crops. The duration of a dry spell is counted in tenths of a day. The number of elements in the series of data varies from 84 to 123. The empirical statistical functions for series of observed data (Fig. 1) have been determined. The statistical parameters in question are the average (F) and the standard deviation (Ot). The variations of these parameters over time are shown represented (Fig. 2). These curves give a better portrayal to which the statistical functions relate of the differences between the different months. The statistical sample of long dry periods (K) can be approached, according to the result, by the geometric distribution: the full average of the sample can be calculated if part of the function is known, and it is assumed that the full and the truncated function are also geometrically distributed. The approximation of the distribution of dry spells by fitting a geometrical distribution to the data is satisfactory, with a 5 percent confidence interval and using the Smirnov Kolmogorov test.
© Société Hydrotechnique de France, 1981