Utilização de técnicas de classificação automática para definir bacias hidrográficas homogêneas em termos da pluviometria e fluviometria.

Abstract

The present thesis constitutes a study of the north-east region of Brasil, with the objective of defining the groupings of raingauge stations and flow measuring stations, that have similar characteristics. Techniques of automatic cIassifiction as applied to a set of variables were utilised herein. These variables were obtained for a combination of reference periods (being a year, semester or trimester) among the parameters of the station data, such as arithmatic mean, standard deviation, coefficient of variation skewness coefficient and the maximum value. The rainfall data for 400 raingauge stations between the years 1937 to 1973 were utilized in the study. Thirty (30) years of data, with a superposition of atleast 5 years between the stations, were utilized for raingauge stations. The data for flow measuring stations, numbering 97, consisted of reliable data over an eight-year period. The Methods of "Quick Cluster" and "K-Means" (which belong to the techniques of non-hierarquic classification) were applied to the set of precipitation variables and the Methods of," U/ard", Simple Linking, Complete Linking and Centroid (which pertain to hierarquical techniques) were applied to the set of flow variables. The applicability of each of these methods is discussed here-in.