The study validates geo-statistical and conventional models for a soil fertility data set of the South Pacific agricultural lands in the coastal plains of Costa Rica. A metha-analysis was conducted and a best adjustment semivariogram employed to allow using Kriging interpolation. Statistical analysis included frequency distribution, means estimates, correlations and principal components analysis (CP). Values of selected variables were interpolated by ordinary Kriging following four consecutive validation types: field validation, cross validation, errors calculated for each interpolation of validations, subtraction of errors from original data observations, generation of a new field validation, and subsequent cross validation. Interpolations results were analyzed using error absolute average (PAE), error mean square (PCE), prediction affectivity (E) and determination (r2). The data set included soil available information of the counties Corredores, Golfito and Osa previously planted to oil palm, rice, forest and few other crops (named “cultivos varios”). Soils are mainly of alluvial origin in lower positions but with a little more development in the distal part of the hillsides, to include mainly Inceptisols, Ultisols, Entisols and a few Andisols. Variables investigated include pH, exchangeable acidity, Ca, Mg, K, P, Zn, Cu, Fe and Mn. Results shows high variation coefficients mainly for exchangeable acidity and availability of Mg, K, P, Zn, Cu Fe y Mn. Frequency analysis demonstrated abnormal distributions for all elements and tendencies between 25 and 75 percentiles however but normal for pH values. Mean values of variables by crops showed higher numbers under rice plantations pH (6.0), Ca (26.8 cmol (+) L-1), Mg (10.6 cmol (+) L-1) and Mn (34.2mg L-1). Values for oil palm plantations were significantly higher (α = 005) for exchangeable acidity (0.5 cmol (+) L-1), K (0.8 cmol (+) L-1), P (13.1mg L-1), Zn (2.8mg L-1) y Fe (99.8mg L-1). Spearman correlation analysis found proportional relationships between Ca, Mg, and K and inverse proportional correlation between pH y exchangeable acidity and soil available K and P. Other crops (“cultivos varios”) showed highly variable intermediate values. CP analysis explained 60.8% of nutrients variability in the study area with a relation between forest and other crops among CP1 and CP2 dimensions, probably related to the behavior of K and P under oil palm pH in rice fields. A relationship was also found among dimensions CP2 and CP3 for Fe and Cu under oil palm and pH under other crops. It was determined that soil acidity, pH and availability of Mg and K were strongly related to the nutritional management practices of the different crops and that of Ca particularly to the soil genesis on calcareous materials. Nutrients interpolation validation determined that PAE, PCE, E and r2 values improved prediction after removing interpolation errors. Corss validation after substration of interpolation errors showed the best interpolation adjustment as compared to field validation and both validations, and better estimation not subtracting nutrients distribution errors in the alluvial coastal plains of the South Pacific Costa Rica. It is concluded that the information provided by the maps build after interpolations well represent the special variability of the evaluated variables. This confirm that the tools employed is functional to develop relevant works in the diagnosis of nutrient problems in soils or soil fertility conditions for a region and other s similar with spatially referenced soil data.
Read full abstract