The spatial variability and spatial dependence structure of physical water soil attributes (total porosity - VTP, saturated hydraulic conductivity - k(0), drainable porosity DP and soil moisture at field capacity - theta cc) was evaluated for an Oxisol representative of Ribeirao Marcela watershed, in the Alto Rio Grande domain. Spherical and exponential semi-variograms were fitted by the weighted minimum square (WMS) and maximum likelihood (ML) methodologies. Undisturbed and disturbed physical water samples were collected from the 0-0.15 m layer for characterization, in grids of 240 x 240 m and 60 x 60 m, totaling 165 sampling points. VTP and theta cc presented low variability, with coefficients of variation (CV) of 7.54 % and 10.22 %, respectively. The mean DP variability was intermediate, with CV of 43.2 %, and was high for k(0) with a CV of 88.3 %. All attributes presented spatial dependence structure, which was strong for VTP, theta cc and k(0), with a dependence degree (DD), respectively, of 81.82, 75.31 and 82.61 %. The spatial dependence structure of DP was moderate with a DD of 54.66 %. The domains of VTP, qcc and DP were around 1.000 m and for k(0) it was about 77.46 m. The results of both methodologies were similar, so that both are recommendable for the fitting of semi- variograms to physical water soil attributes. Based on cross-validation, spherical semi-variograms can be indicated for VTP, qcc and k(0), with a Mean Square Error (MSE) close to 0.00161, 0.2795 and 0.00155, respectively, and the exponential model recommended for DP, with MSE of 0.004070.
