engleski

Spatial variation of soil nutrients on sandy-loam soil

The spatial variability of plant available phosphorus, plant available potassium, soil pH and soil organic matter content in central Croatia was investigated using geostatistical tools and geographical information system to create nutrient maps and provide useful information for the application of inputs that will also be used for the design of an adequate soil sampling scheme. In a regular grid (50 m
50 m), 330 samples were collected on sandy loam Stagnic Luvisol. Soil available phosphorus and plant available potassium showed relatively high spatial heterogeneity, ranging from 105 mg kg-1 to 310 mg kg-1, and from 115 mg kg-1 to 462 mg kg-1, respectively. Content of soil organic matter and pH had lower variability ranging from 1.26% to 2.66% and from 3.75 to 7.13, respectively. Investigated soil properties did not follow normal distribution. Logarithm and Box–Cox transformation were applied to achieve normality. Directional exponential model for soil available phosphorus, potassium and pH and spherical model for soil organic matter was used to describe spatial autocorrelation. Fourteen different interpolation models for mapping soil properties were tested to compare the prediction accuracy. All models gave similar root mean square error values. Available phosphorus, potassium and pH evaluated by radial basis function models (CRS, IMTQ and CRS, respectively) provide a more realistic picture of the structures of analyzed spatial variables in contrast to kriging and inverse distance weighting models. For soil organic matter datasets the most favorable model was LP1. According to the best model soil nutrient maps were created to provide guidance for site- specific fertilization and liming. Soil fertility maps showed sufficient concentrations of soil available phosphorus and available potassium. Acidity map showed that the largest part of the investigated area is very acid and acid. For future management it is necessary to provide more liming materials while fertilization rate should be lower.

geostatistics; nutrient maps; interpolation models; spatial variability

nije evidentirano