Banca de DEFESA: FRANCISCO IVO DOS SANTOS AGUIAR

Digital mapping of soil properties as a support in sustainable land use planning

Ordinary Kriging, Soil survey, Data mining, Environmental covariates, Geostatistics.

In recent decades, there has been a growing demand for information about the spatial distribution of soil properties and classes, so modern computational techniques as Digital Soil Mapping have been developed. The general objective of this study was to predict the spatial variability of the soil properties: sand, clay and organic carbon contents by ordinary kriging and to correlate the results with topographic and remote sensing covariates. The study area took place in the municipality of Bom Jardim, RJ. To compose the soil database, 209 soil profiles were used, in a total of 603 samples. The digital elevation model (DEM) was generated to represent the altimetry of the study area and, subsequently, the basic attributes of the terrain were extracted: slope, slope height, aspect, valley depth index, topographic humidity index and factor - LS. Seeking to achieve the objective of the present study, orbital images were obtained from the Multispectral Instrument (MSI) sensor, installed on board the Sentinel-2B satellite. The free software QGIS 3.10.4 was used for pre-processing the images, processing the NDVI (Normalized Difference Vegetation Index), SAVI (Soil Adjusted Vegetation Index), Clay Minerals Ratio (CM), Iron Oxide Ratio (IO ), and bands 1 to 12. The statistical procedures to analyze the data and implement the soil depth functions were performed in the open access program R. To plot the soil depth functions, the Algorithms for Quantitative Pedology (AQP) package was used. Regarding the basic statistics of the data set, to identify outliers, boxplots were used. To determine the normality of the data, the Kolmogorov-Smirnov test was applied, at a 5% probability level. After the statistical analysis, the geostatistical analysis of the data was carried out. The modeling of the experimental semivariogram followed the principles established by the intrinsic hypothesis. After modeling the semivariograms, the ordinary kriging (KO) technique was used to interpolate values in locations not sampled. It is concluded that the terrain covariates derived from the digital elevation model can represent the soil forming factors and can be used in the digital mapping of soils in the area, as well as to establish areas of agricultural suitability in mountainous areas, such as Bom Jardim - RJ. All soil properties showed spatial dependence structure and better fit the spherical model. The spatialization of soil properties in the study area proved to be very efficient in understanding the spatial variability with the use of Ordinary Kriging