Interpolation of soil fertility data with Kriging and its validation

The goal of this study was to make and validate interpolated maps of 6 soil fertility variables. The maps were made from the results of Ca, pH, soil acidity, K, P and saturation of soil acidity of 138 soil samples that were taken from 1011 ha at Atirro, Costa Rica. The data were interpolated through...

Descripción completa

Autores Principales: Henríquez Henríquez, Carlos, Méndez, Juan Carlos, Masís, Ramón
Formato: Artículo
Idioma: Español
Publicado: 2015
Materias:
Acceso en línea: http://revistas.ucr.ac.cr/index.php/agrocost/article/view/12763
http://hdl.handle.net/10669/13974
Sumario: The goal of this study was to make and validate interpolated maps of 6 soil fertility variables. The maps were made from the results of Ca, pH, soil acidity, K, P and saturation of soil acidity of 138 soil samples that were taken from 1011 ha at Atirro, Costa Rica. The data were interpolated through ordinary Kriging. The validation was carried out using “field validation” and “cross validation” methods. Correlation coefficient (r) was estimated for both techniques between real and prediction values, and the prediction efficiency (E) as well. Other validation criteria used were the percent success by overlapping between real and estimated values, according to the uncertainty of soil analysis and to the success rate of overlap according to agronomic category. The r values using field validation varied from 0.09 to 0.87; and for cross validation were from 0.52 to 0.84. The variables Ca and pH had the highest prediction efficiency in both validation methods. The overlap criterium due to the uncertainty of analysis was 27 to 93% success, while the overlapping range that was due to agronomic category had 47 to 93% of success. In both cases, pH had the better values of success. It was concluded that the interpolated maps at a regional scale are a useful tool for to carrying out a good prediction on soil fertility properties, although it is important to perform a verification process in order to confirm these approximations, because this could change according to the type of variables.