Regionalização de vazão mínima em bacias sem dados no semiárido da Bahia

Abstract

The lack or scarcity of data from hydrological monitoring is one of the major obstacles to the advancement of water resources management and planning. This information is extremely important for the semi-arid, a region that suffers from natural disasters such as extreme droughts, floods and flooding. In this sense, this study aimed to model hydrological variables for basins without data in the semi-arid of Bahia through the regionalization of flows with the Génie Rural model to 4 Paramètres Journalier (GR4J). For this, we used a data set from the Catchment Attributes and Meteorology for Large-sample Studies (CAMELS-BR), which included historical information from 1980 to 2018 in a set of 52 basins. In this study, k-Means and Ward methods were used to group the basins in regions with similar behavior. Within each homogeneous region, the values of the parameters of the model were correlated with physical and hydrogeological characteristics of the basins, in order to obtain the values of the parameters in basins without data, generating equations through linear regression for their determination. The results showed that the k-Means technique was satisfactory for obtaining homogeneous regions and some of the characteristics that influenced the heterogeneity between the clusters and homogeneity within the groups were: the area of the watershed, aridity, the average slope of the basin, the average duration of the maximum precipitation events (number of consecutive days 5 times the average daily precipitation), the average annual evapotranspiration, the longitude, the underground porosity of the watershed, the mean annual precipitation and the median depth of the water table. In general, the models present low efficiency in semi-arid regions, which was not different in this study. In the period analyzed, the GR4J obtained the Nash-Sutcliffe coefficient for the logarithm of flows (Nslog) higher than 70% in 13 (25%) stations in the calibration phase and 7 (13%) in the validation phase. In the process of regionalization of the parameters of the model the linear regression technique was not suitable for obtaining equations. The model was efficient to obtain the reference flow rates with 90% permanence in the historical series (Q90) and the minimum flow rate for seven days of duration with return time (RT) equal to 10 years (Q7,10), presenting a R² greater than 99%. When relating the Q90 obtained through simulated data and observed with the area of the basins, only one of the groups presented P-value lower than 0.05, showing the need to test other methods in order to support decisions regarding the planning and management of water resources in the semi-arid region of Bahia.