Aplicaςao da anaálise estatistica espacial em levantamentos pedolólicos semidetalhados

Abstract

Soil data from Guaíra, north of São Paulo State, Brazil, were generated from semidetailed pedologic survey by Agronomic Institute of Campinas (IAC, SP). Textural and chemical analysis data were submitted to geostatistic methodology and trend surface analysis. These methods were tested to distinguish soil's chemical and physic characteristics and potencialities. The intention was to add spatial and quantitative approach to conventional methods used in pedologic surveys, that means field prospecting, fotointerpretation and classic non-spatial statistics. Geostatistic results have shown that kriging is not recommended as interpolation method for these data, despite all the advantages of the method. All the semivariograms, except for carbon variable, revealed excessive or pure nugget effect (Figures 6-7). These results reveal spatial variation of data in a smaller scale than the sampling one. Estimating by kriging would be an incorrect option as there is no covariance among data for the sampled intervals in this scale. Better results were obtained with trend surface analysis (Figures 8-14) and such results represented additional information about Guaira's soils. These methods are simple to use and represent a good general view of data spatial distribution, revealing anomalies, helping in systematic survey's stage and distinguishing among different types of soils in a detailed scale. Residual maps for textural variables revealed oriented bodies (aprox. NW-SE) with sand levels above the mean (Figures 9-10). Simple interpolation for values were estimated by inverse distance to a power method producing thematic maps (Figures 15-19). Therefore it is recommended that spatial statistic analysis should be used in semidetailed pedologic surveys as an aditional tool, with emphasis to trend surface analysis and elaboration of residual and thematic maps. Care with the coordinate information on colecting data is an obvious but really important providence. Geostatistic should be preliminary tested to verify spatial continuity of data and, if results indicate excessive nugget elect, this method should be abandoned as an interpolator.