SOIL MAPPING QUALITY FOR SITE-SPECIFIC MANAGEMENT IN FRUIT FIELDS IN THE SEMIARID REGION OF BRAZIL

Abstract

Soil mapping is an important technique to apply site-specific management in fruit fields. Therefore, aims of this study were to evaluate soil sampling methods and the use of interpolator techniques in the quality of soil texture maps in a mango field; and to evaluate the efficiency of the use of independent data set in the validation of maps process. The experiment was in a 3.3 ha mango commercial orchard in the semiarid region of Brazil. The samples were collected following a regular grid containing 119 georeferenced spaced 14 x 20 m points, in addition to five micro-grids each containing six spaced 7 x 10 m points. To obtain an independent dataset for the maps validation, 40 random samples were performed. Soil texture was determined. Geostatistic and deterministic interpolation techniques were performed. Cross-validation method with internal and independent dataset was performed to validate the maps. All variables showed a strong spatial dependence. Kriging was slightly better than the deterministic interpolation techniques, showing root mean square errors of 2,40%; 2,84%; and 2,59% for clay, sand, and silt content, respectively. The use of micro-grid did not reduce the errors of the maps. The independent dataset showed efficient to validate soil texture maps since its values represent the real accuracy.