Assessment of soil property spatial variation in an Amazon pasture: basis for selecting an agronomic experimental area

Our main objective in the present study was to assess the spatial variation of chemical and physical soil properties and then use this information to select an appropriate area to install a pasture rehabilitation experiment in the Amazon region, Brazil. A regular 25 m grid was used for collecting a total of 2955 soil samples (from 985 georeferenced soil pits) at 0 to10, 10 to 20 and 20 to 30 cm layers. Soil samples were analyzed for total carbon and nitrogen, δ13C and δ15N, pH in H2O, pH in KCl, clay, and sand contents. Conventional statistical methods and geostatistics were performed in order to analyze soil properties spatial dependence. Mean, standard deviation, skewness, and kurtosis for all measured variables were evaluated. All variograms generally were well structured with a relatively large nugget effect. Total C, total N, pH in H2O, pH in KCl, δ13C and δ15N semivariograms were best fitted by spherical models, while clay and sand contents were best fitted by exponential models. Two types of validation (“Jackknife” or cross-validation and external validation) were conducted, indicating a lack of bias for the used prediction models. Block kriging was used to interpolate the values at unmeasured locations, generating maps of spatial variation for each soil property. Those maps were processed using Geographic Information System and restrictive criteria were adopted in order to select the best area in which to install the pasture rehabilitation experiment. The study field was then divided into zones with similar homogeneity. The selected zone can now be subjected to different treatments once the natural initial conditions are well known, and could also be used as a baseline in carbon sequestration projects within the scope of the Kyoto Protocolʹs Clean Development Mechanism.