LEITE, E. P. F.; http://lattes.cnpq.br/6392892671835124; LEITE, Eugênio Pacelli Fernandes.
Resumo:
The work was conducted on the Rio Cabelo hydrological micro basin, located in
João Pessoa, State of Paraíba, Brazil, aiming to establish hydrologic parameters of the
basin and characterize physic-hidrically the existing soils and their spatial variability. All
the procedures used in the study were executed through out free computational systems,
the R Statistical and the SIG GRASS system. During the elaboration of the Land
Numerical Method (MNT) the Spline Line interpolation methods, with regularized tension,
gave altitude values closer, than those obtained when used the universal kriging, and also
had a better performance according with the AIC and Quadratic Error Mean criterion for
model selection. The basin had an area of 9,786 Km2 and the flux lines a predominant
dendritic estructure. When constructing the empirical semivariograms, during the
geoestatístic analyses of the soil parameters, using boxcox transformations of the data,
with and without removal of linear trend, it was observed that with exception of the 60-
90cm particle density data, the trend removal permitted to predict values very much closer
than the obtained in the field. To estimate the parameters of the covariance models
(spherical, exponential, gaussian, matern, powered-exponential, cauchy) were utilized the
Ordinary Least Squares (OLS), Weight Least Squares (WLS) and Maximum Likelihood
(ML) methods. This last one, presented resulted more reliable than the OLS and WLS
methods, when considered the AIC and Maximum Likelihood logarithm criterion for the
models selection. The space dependence range of the models, were established for
prediction of values not sampled for the textural separates clay, silt, sand, for soil density,
particle density and for the hydraulic conductivity.