Vízügyi Közlemények, 1978 (60. évfolyam)

3. füzet - Abonyi István-Zsuffa István: Folyamatos vízállás-előrejelzés jelentősebb mellékfolyó nélküli folyószakaszon

450 АЬопуг I.—Zsuffa I. the differences in the vegetation covering the flood bed have greatly impaired the reliability of the forecasts issued. The forecasts should therefore be based on the data relating to the most recent conditions. For this reason the series of aids com­piled for the Danube is revised at 10-day intervals and each study is founded on the data of the last 80 day period. Over the Danube secLion Írom Pozsony down to the mouth of the Dráva River series of relationships for forecasts of the afore­mentioned advance in time have been compiled for 12 gaging stations, using all information available. The large volume of information must be filtered appro­priately. Attempts have first been made to find relationships between the daily changes in stage by relating individually each station to all upstream stations, determining hereafter the cross-correlation functions for the changes in stage obtained by pairs of stations. The travel time at which the cross-correlation function assumes maxi­mum value is determined at the maximum of the r(t) function. Using this approach the number of variables at each station could be reduced to the number of gages upstream of the station considered. The set of data thus filtered is still too large, especially over the downstream reaches, to yield a reliable relationship based on the stage data of 80 days. In the second stage of filtration the partial correlation coefficients habe been examined. Any station for which a negative partial correla­tion coefficient has been found, was eliminated as superfluous and even confusing, and the data thereof have been neglected in the next study. This filtration accord­ing to the partial correlation coefficients has been organized for computer cycles, omitting a single factor only in each analysis. This method proved to reduce the number of stations adequately and proportionally. After the second stage of filtra­tion the absolute values of the stages at the remaining stations are also included, together with the squared changes in stage. The first forecasts were issued on the 1st August, 1975, producing since daily forecasts and revising the relationships at 10 day intervals. Retrospective analyses have shown the accuracy of 1 to 2 day forecasts to be acceptable, and to decrease significantly if the revision of data is made less frequently. Entirely misleading results are obtained if, instead of revision, the system is founded on a single relation­ship. Forecasts covering more than ,'i days were, found to be of lower accuracy but an autocorrelation study of the differences have indicated the possibility of refin­ing even these longer-term forecasts. The new solution separating low- and high stages is expected to improve the accuracy of these forecasts, too. * * * kontinuierliche Wasscrstandsvorlicrsagc iiir einen Flussabschnitt ohne bedeutendere Zubringer von Dipl.-Ing. István Abomji und Dr.-Ing. István Zsuffa Die Methode der kontinuierlichen, 1 — 6 Tage fristigen Vorhersage des Wasser­standsregimes der Donau wurde im Jahr 1953 entwickelt. Eine nachträgliche Ana­lyse der Vorhersagen weist daraul hin, dass deren Zuverlässigkeit während der Niedrigwasserperiode durch Sandbankbewegungen, während des Hochwassers aber durch den verschiedenen Pflanzenwuchs des Deichvorlandes erheblich beeinflusst werden kann. Es ist immer zweckmässig, die Vorhersagen auf Grund einer auf den neuesten Stand bezogenen Datenmenge zusammenzustellen. Demgemäss wird die für die Donau ausgearbeitete Behelfsreihe alle 10 Tage erneuert und bei jeder Unter­suchung werden die Daten der jeweils letzten 80 Tage herangezogen. Unter Ver­wendung aller zur Verfügung stehenden Informationen wurden solche Beziehungs­reihen für 1 — 6 Tage l'ristige Vorhersage für 12 Pegelstationen des Donauabschnittes Bratislava-Draumündung ermittelt. Die Verlasser suchten zuerst Beziehungen zwischen den täglichen Wasser­standsänderungen, indem sie jede einzelne Station mit sämtlichen flussaufwärts

Next

/
Thumbnails
Contents