The second set of \(n=\) 60 observations was obtained in the years of 2012 and 2013. Because there are many limitations to the access in the area (geographic barriers and prohibition by landowners), and because of the shortage of workforce, budget, equipments and time, soil sampling was performed using randomly selected representative transects. The definition of the population of representative transects was achieved through the elicitation of the knowledge of three researchers with field experience in the area. Each researcher delineated sixty transects (400 m), taking into account the factors that determine the spatial distribution of soil properties.
Instruí os especialistas a delinear os transectos ao longo dos maiores gradientes de variação dos componentes ambientais (relevo, geologia, vegetação e solos). Entretanto, esses gradientes deviam estar localizados onde o acesso é garantido, o que corresponde, em termos estatísticos, à uma grande probabilidade de serem visitados. Para isso, alertei os especialistas para a necessidade de evitar o delineamento dos transectos sobre estradas e outras construções antrópicas, uma vez que esses locais não são possíveis de ser visitados para a amostragem de solo. Em termos estatísticos, esses locais possuem probabilidade nula de ser visitados.
A atividade de delimitação dos transectos pelos especialistas foi desenvolvida utilizando o software Google Earth. Para isso, cada especialista recebeu um arquivo no formato KML (Keyhole Markup Language) contendo o limite da área de estudo. Tanto esse arquivo, como aqueles produzidos pelos especialistas contendo os transectos, foram armazenados e manipulados no interior de uma pasta do Dropbox. Depois de finalizada a atividade pelos três especialistas, analisei os transectos quanto a coerência com a solicitação feita no início da atividade. Nessa análise observei que alguns transectos foram delineados sobre lagos e residências, o que inviabiliza a sua utilização. Tais transectos foram substituídos por outros em local próximo. Depois de analisar todos os 180 transectos produzidos pelos especialistas, agrupei os mesmo em um único arquivo e os submeti à seleção utilizando um algoritmo do software QGIS.
Twelve out of \(m=\) 180 transects were randomly selected. Iterations were performed until there were no intersecting transects, and there was at least one transect located in each of the three major landforms of the area (plateau, slope and valley floor). Finally, \(n=\) 5 sampling points separated by equidistant intervals of 100 m were placed in each transect. Every data point was identified with a number in increasing order, following the order in which the observations were made (341-400).
The sampling points were identified in the field using a GNSS signal receiver with a horizontal precision of less than \(\sim\)8 m. A soil pit was opened within a radius of 2m from the sampling point. Soil sampling and description was carried out using the same procedure previously used for the first set of \(n=\) 340 observations. Georeferencing was performed using a Differential Global Positioning System (DGPS) with centimetric horizontal positional error.
The third set of \(n=\) 10 observations correspond to the A horizon of modal soil profiles, from which a single soil sample was collected and used for laboratory analysis. The resulting sampling depth varies from 12 to 30 cm, with a mean of 22.6 cm. This data was compiled from the studies of \cite{Pedron2005} and \cite{Miguel2010}. Every data point was identified with the same identification used in the studies from where they were compiled.