Métodos de interpolação no mapeamento de atributos químicos do solo

Abstract

The use of geostatistical technique is important for the identification of spacial variability and mapping the soil and the plants attributes. The aim of this project was to characterize the spacial variability of chemical attributes on a dystrophic Oxisoland compare the efficiency of interpolation of data by ordinary kriging, kriging by blocks and the inverse square distance. The experimental area is located in Fazenda São Luiz, in Fátima do Sul-MS. The soil sample was made in interlaced points of a regular mesh of 11 lines and 11 columns, with 20 m spacing between each point, totaling 121 points of samples. There were determined the content of pH, Ca, Mg, K, H+Al, which were used for the determination of saturation by bases. The spacial variability of the soil attributes was described based on the parameters of the semivariogram obtained from geostatistc analyses. After the interpolation, there were made maps of the spacial distribution, which contains the liming necessity according to the different interpolators. The descriptive analyse of the chemical attributes revealed the high values of saturation by bases, there were no need for limestone application (NC). Moreover, in the points that there was a necessity, the quantity to be applied would be very low. The geostatistic techniques detected and described the spacial variability in most of the soil attributes. However, the adoption of smaller sample mesh to identify the spacial dependence of all the soil attributes is needed. The sample spacing used was appropriate to verify the spacial dependence. The attributes K, have shown strong spacial dependence. However, it must be noticed that the model adjusted have shown a low determination coefficient (R2 = 0,34), V% have shown weak spacial dependence, and the rest of them have shown moderate spacial dependence. There were not seen differences between the values estimated by the punctual kriging and the kriging by blocks, as the parameters of the fit of crossed validation were the same or have been very close. The estimator inverse square distance have shown the worse estimation, observed by the results of crossed validation.