ABSTRACT Soil reflectance spectroscopy has become an innovative method for soil property quantification supplying data for studies in soil fertility, soil classification, digital soil mapping, while reducing laboratory time and applying a clean technology. This paper describes the implementation of a Graphical User Interface (GUI) using R named AlradSpectra. It contains several tools to process spectroscopic data and generate models to predict soil properties. The GUI was developed to accomplish tasks such as perform a large range of spectral preprocessing techniques, implement several multivariate calibration methods, generate statistics assessment and graphical output, validate the models using independent dataset, and predict unknown variables using soil spectral data. AlradSpectra has four main modules: Import Data, Spectral Preprocessing, Modeling, and Prediction. The implementation of AlradSpectra is demonstrated by applying visible near-infrared reflectance spectroscopy for soil organic carbon (SOC) prediction. The data contains the value of SOC and Vis-NIR reflectance for 595 soil samples. The prediction statistic assessment of SOC was performed applying all spectral preprocessing and methods. The R 2 considering all models ranged from 0.54 to 0.80. In the partial least squares regression (PLSR) models, the performances were similar to multiple linear regression (MLR) and support vector machines (SVM). The lowest error in the SOC prediction was achieved by PLSR method with standard normal variate (SNV) preprocessing reaching an R 2 of 0.80, the smallest root mean square error (RMSE) of 0.47 %, and ratio of performance to inter-quartile distance (RPIQ) of 3.12. The capacity of performing multiple tasks, being free and open-source, easy to operate, and requiring no initial knowledge of R programming language are features that make AlradSpectra a useful tool to perform different modeling approaches and predict the desired soil variable.

ABSTRACT The success of soil prediction by VIS-NIR-SWIR spectroscopy has led to considerable investment in large soil spectral libraries. The aims of this study were 1) to develop a soil VIS-NIR-SWIR spectroscopy approach using legacy soil samples to improve spectral soil information in a regional scale; (2) to compare six spectral preprocessing techniques; and (3) to compare the performance of linear and non-linear multivariate models for prediction of sand and clay content. A total of 1,534 legacy soil samples, stored by Epagri, were collected from agricultural areas in 2009 on a regional scale, covering 260 municipalities of Santa Catarina. Six spectral preprocessing techniques were applied and compared with reflectance spectra (control treatment) in the development of sand and clay prediction models. Five multivariate regression models, Support Vector Machines, Gaussian Process Regression, Cubist, Random Forest, and Partial Least Square Regression were compared. The scatter-corrective preprocessing groups produced similar or better performance than spectral-derivatives. In addition, preprocessing spectra prior to regression analysis does not improve sand prediction, since reflectance spectra achieved the best performance using Cubist, SVM, and PLS models. In general, clay content presented better prediction accuracy than sand content. The best multivariate model to predict sand and clay content from soil VIS-NIR-SWIR spectra was Cubist. The best Cubist performance was achieved combined with reflectance spectra (R2 = 0.73; root mean square error = 10.60 %; ratio of the performance to the interquartile range = 2.36) and MSC (R2 = 0.83; root mean square error = 7.29 %; ratio of the performance to the interquartile range = 3.70) for sand and clay content, respectively. Considering the mean RMSE values of the validation set, the predictive ability of the multivariate models decreased in the following order: Cubist>PLS>RF>GPR>SVM for both properties. The predictive ability of VIS-NIR-SWIR reflectance spectroscopy achieved in this study for sand and clay content using legacy soil data and heterogeneous samples confirmed the potential of the spectroscopy approach.

ABSTRACT The study of the relationships between the yield potential of forest stands and the conditions offered for plant development is fundamental for the adequate management of the forest when aiming at sustainable high yields. However, these relations are not clear, especially in commercial forests, on rugged terrain where relationships between the landscape, soil, and plants are more complex. Considering this, we tested the hypothesis that the morphological aspects of the soil conditioned by topography are the main limiting factors for tree development. Our objective was to evaluate the edaphic and topographic influence on the dendrometric variation of Pinus taeda L. of a forest stand in a subtropical climate at high altitude. For that, Spearman’s correlation analysis and canonical correspondence were performed on two data datasets containing pedological, topographic, and dendrometric information of a commercial plantation of Pinus taeda, in the Campo Belo do Sul, Santa Catarina state, Brazil. Soil sampling and characterization was performed in two distinct designs. The first design was based on the morphological description of 11 soil profiles. The second was performed with intensive prospecting of the area, with 102 sampling locations determined through conditioned Latin hypercube sampling. For each sampling point, the height and diameter of the four nearest trees were measured and the terrain attributes were calculated from the digital elevation model. The solum depth, the thickness of the superficial horizon, elevation, and vertical distance to channel network were the main conditioning factors of the dendrometric variation, wherein taller trees were found in deeper soils, with a thicker surface horizon, in lower areas that are vertically closer to the drainage network. Our results showed that the selection of topographic and morphological variables has a significant effect on the tree height and should therefore be used to select homogeneous areas for the development of the species. In addition, we showed the importance of using an intensive sampling survey to understand dendrometric variation.

Resumo: O objetivo deste trabalho foi testar a extrapolação das relações solo-paisagem de uma área de referência (AR), por meio de mapeamento digital de solos (MDS), para uma carta topográfica (1:50.000), e comparar os resultados aos obtidos em estudos similares anteriormente desenvolvidos no Brasil. O trabalho consistiu no levantamento de solos, com técnicas convencionais de mapeamento de uma AR de 10 km2 (na escala 1:10.000), para mapear uma área fisiograficamente similar de 678 km2 (na escala 1:50.000), tendo-se utilizado o MDS. A técnica de árvore de decisão (AD) foi utilizada para a construção do modelo preditivo de extrapolação, com base nas classes de solos e em oito atributos de terreno da AR. A validação do MDS, com pontos de observação de campo, resultou em 66,1% de exatidão global e 0,36 de índice kappa. Os solos mais representativos da área foram preditos corretamente, enquanto solos menos representativos e de menor ocorrência na paisagem e, consequentemente, com amostragem reduzida, tiveram sua predição comprometida. A proporção da AR, igual a 1,5% da área total, é um fator limitante à formulação das relações solo-paisagem para representar precisamente a área mapeada por MDS.

Abstract: The objective of this work was to test the extrapolation of soil-landscape relationships in a reference area (RA) to a topographic map (scale 1:50,000), using digital soil mapping (DSM), and to compare these results to those obtained in similar studies previously conducted in Brazil. A soil survey in a 10 km2 RA, using conventional mapping techniques (scale 1:10,000), was made in order to map a 678 km2 physiographically similar area (scale 1:50,000) using DSM. The decision tree technique was employed to build a predictive extrapolation model based on soil classes and eight terrain attributes in the RA. The validation of DSM by application of field observation points resulted in a 66.1% global accuracy and in 0.36 kappa index. The most representative soils in the area were correctly predicted, whereas the less representative and less frequent soils in the landscape (and consequently with reduced sampling) had their prediction compromised. The RA proportion, which equals 1.5% of the total area, is a limiting factor in the formulation of soil-landscape relationships to precisely represent the mapped area by DSM.

As imagens com resolução espacial submétrica, como as adquiridas por Veículo Aéreo Não Tripulado (VANT), apresentam alta variabilidade espectral. A análise baseada em objetos possibilita a redução dessa variabilidade e a geração de atributos, aumentando a dimensionalidade do conjunto de dados. Funções de seleção de atributos relevantes e de reforço, disponíveis no algoritmo C5.0, e a análise baseada em objetos facilitam a classificação dessas imagens. Este trabalho teve como objetivos: (i) avaliar as classificações dos objetos em relação aos parâmetros de seleção de atributos (winnow), de reforço (trial) e do Número Mínimo de Amostras (NMA), (ii) determinar os atributos preditivos mais importantes e (iii) comparar a classificação por Árvore de Decisão com Máquina de Vetores Suporte. Para segmentar a imagem foi utilizado o método crescimento de regiões e para a classificação o algoritmo C5.0. Os valores dos parâmetros trial (10) e NMA (5) resultaram acurácias superiores a 0,8. Com esses parâmetros o valor do kappa foi superior a SVM. Ao habilitar o parâmetro winnow foi observado uma redução da dimensionalidade do conjunto de dados de aproximadamente 30%. Os dois atributos mais importantes na discriminação das classes foram a razão entre as bandas verde e azul e a média dos valores das elevações

Remotely sensed images acquired with submetrical spatial resolution, such as those captured by Unmanned Aerial Vehicle (UAVs), usually present a high spectral variability. An object based approach allows the generation of attributes, increasing the dimensionality of the original dataset. Selection functions for relevant features and booster, available in the C5.0 algorithm, and the objects based analysis facilitate the classification of such datasets. This study aimed at: (i) to evaluate object classification approach in relation to the parameters of attributes selection (winnow), booster (trial) and Minimum Sample Size (MSS), (ii) to determine the most predictive features, and (iii) to compare the classification performance of both Decision Tree and Support Vectors Machine. A region growing method was used in order to segment the image. C5.0 algorithm was used in the classification procedure. The values of the trial (10) and MMS (5) parameters resulted in accuracies higher than 0.8. The kappa value was greater that the SVM, with these parameters. A significant decrease in the data dimensionality up to 30% was observed when the winnow parameter was habilitated. The two most important features in the breakdown of classes were ‘ratio between the blue and green bands’ and ‘average of the elevation values’.

ABSTRACT: Relief characterization using a digital terrain model (DTM) is widely applied in erosion, soil and vegetation modeling. However, factors, such as acquisition technology and the spatial resolution of the digital model, affect modeling results. The aim of this study was to characterize noises in a DTM of the entire state of Santa Catarina recently made available through the state's Sustainable Economic Development Secretary and to evaluate different methods of interpolation and smoothing of the original 1 m resolution to a new digital model with a spatial resolution of 15 m. Using the SAGA GIS program, spurious data that appeared as peaks and sinks were removed from the digital model. Of five processing procedures, the following three were used for smoothing: a Gaussian filter, a Lee filter, and mesh denoising. The remaining two were for interpolation: nearest-neighbor, and ordinary kriging. Altimetric reference data were collected in the study area of 11,597 km2 with two dual-frequency GPS RTK receivers located in the Celso Ramos community in the municipality of Frei Rogério (SC). The root mean square error showed that the documented values of the aerial survey report were consistent with the findings of this study. However, the GPS RTK data showed a difference of 1.07 m compared with the original DTM with 1 m spatial resolution. There were pixels with peaks and sinks in the digital model of the Santa Catarina State.

ABSTRACT Terrain attributes (TAs) derived from digital elevation models (DEMs) are frequently used in digital soil mapping (DSM) as auxiliary covariates in the construction of prediction models. The DEMs and information extracted from it may be limited with regard to the spatial resolution and error magnitude, and can differ in the behavior of terrain features. The objective of this study was to evaluate the quality and limitations of free DEM data and to evaluate a topographic survey (TS) underlying the choice of a more appropriate model, for use in DSMs at a scale of 1:10,000. The study was conducted in an area of 937 ha in the watershed of Lajeado Giruá, in southern Brazil. The DEMs: DEM-TS, DEM-Topographic Map (TM), DEM-ASTER, DEM-SRTM, and DEM-TOPODATA were evaluated with regard to the precision elevation by statistical tests based on field reference points, the root mean square error (RMSE), identification of the number and size of spurious depressions, and the application of the Brazilian Cartographic Accuracy Standards Law (BCASL) to define the scale of each DEM. In addition, the TA derived from each DEM was compared with the TA from DEM-TS, considered to be terrain reality. The results showed that the elevation data of DEM-TS had the best quality (RMSE = 1.93 m), followed by DEM-SRTM (RMSE = 5.95 m), DEM-Topographic Map (RMSE = 8.28 m), DEM-TOPODATA (RMSE = 9.78 m) and DEM-ASTER (RMSE = 15.57 m). The DEM-TS was well-represented at a 1:10,000 scale, while the DEM-Topographic Map and DEM-SRTM fitted 1:50,000, the DEM-TOPODATA 1:50,000 and the DEM-ASTER a 1:100,000 scale. The results of DEM-SRTM and DEM-TOPODATA were closest to terrain reality (DEM-TS) and had the lowest number of spurious depressions and RMSE values for each evaluated attribute, but were inadequate for not fitting detailed scales compatible with small areas. The techniques for the acquisition of elevation data of each DEM and mainly the flat to gently undulating topography were factors that influenced the results. For a DSM at a scale of 1:10,000 in similar areas, the most appropriate model is DEM-TS.

ABSTRACT Diffuse reflectance spectroscopy (DRS) is a fast and cheap alternative for soil clay, but needs further investigation to assess the scope of application. The purpose of the study was to develop a linear regression model to predict clay content from DRS data, to classify the soils into three textural classes, similar to those defined by a regulation of the Brazilian Ministry of Agriculture, Livestock and Food Supply. The DRS data of 412 soil samples, from the 0.0-0.5 m layer, from different locations in the state of Rio Grande do Sul, Brazil, were measured at wavelengths of 350 to 2,500 nm in the laboratory. The fitting of the linear regression model developed to predict soil clay content from the DRS data was based on a R2 value of 0.74 and 0.75, with a RMSE of 7.82 and 8.51 % for the calibration and validation sets, respectively. Soil texture classification had an overall accuracy of 79.0 % (calibration) and 80.9 % (validation). The heterogeneity of soil samples affected the performance of the prediction models. Future studies should consider a previous classification of soil samples in different groups by soil type, parent material and/or sampling region.

Increasingly intensive use of soil and its disorganized occupancy result in its degradation. There is a lack of more detailed information on soils and on soil surveys. Soil mappings are key for land sustainable management because they provide basic information about this natural resource and its distribution on the landscape. The production of highly detailed soil class and attribute maps, will only be possible using geotechnologies in combination with field data. The objective of this review was to emphasize the importance of soil surveys within a new paradigm, defined as Digital Soil Mapping (DSM). DSM requires: i) the use of data from field sampling and analytical results from laboratories, including existing as well as newly collected data ii) the process of inference and of spacing includes a proposition of mathematical and statistical models based on soil observation, environmental covariates and other scorpan factors iii) results should be presented as soil spatial system format with continuous and/or discrete data, added to the uncertainty of this information, which could be updated whenever new observations become available. The application of DSM will make it possible for soil spatial information, attributes and classes to be generated and be readily available in digital format with its associated uncertainty. Among the challenges for further implementation of several strategies available in DSM, the training of future generations of soil scientists should be a priority.

O uso cada vez mais intensivo do solo e sua ocupação desordenada resultam em sua degradação. Existe uma grande carência de informações mais detalhadas de solos e seus levantamentos. Por apresentarem informações básicas sobre esse recurso natural e sua distribuição na paisagem, os mapeamentos são fundamentais para o manejo sustentável das terras. A produção de mapas de classes e atributos, com maior nível de detalhamento, só será possível com a utilização de geotecnologias combinadas com dados coletados no campo. Assim, o objetivo desta revisão foi ressaltar a importância dos levantamentos de solos dentro de um novo paradigma, definido como Mapeamento Digital de Solos (MDS). A prática de MDS requer que: i) os dados utilizados tenham sua origem em coletas de campo e resultados analíticos de laboratório, incluindo dados legados e novas coletas; ii) o processo de inferência e espacialização inclui a proposição de modelos matemáticos e estatísticos entre as observações do solo, as covariáveis ambientais e os demais fatores scorpan; e, iii) os resultados estejam na forma de um sistema espacial em solos, com dados contínuos e/ou discretos, agregados à incerteza dessas informações, podendo ser atualizados sempre que novas observações estiverem disponíveis. A aplicação do MDS irá permitir que informações espaciais em solos, de atributos e classes, sejam geradas e prontamente disponibilizadas em formatos digitais e com a incerteza associada. Entre os desafios para uma maior aplicação das várias estratégias disponíveis em MDS está a capacitação e formação das futuras gerações de pedólogos.

RESUMO O conhecimento sobre a cobertura da terra é fundamental como informação para o planejamento e o estudo dos efeitos da substituição de paisagens naturais por paisagens antropizadas. Este estudo objetivou analisar a dinâmica da cobertura da terra entre os anos de 1989 e 2011, na bacia hidrográfica do rio Marombas (SC), empregando o classificador árvore de decisão (AD). Foram utilizadas bandas espectrais do satélite Landsat 5, índices de vegetação e atributos de terreno extraídos do modelo digital de elevação. Esses dados foram utilizados como atributos de classificação da cobertura da terra, nos anos de 1989, 1991, 1993, 1997, 2001, 2004 e 2011. A qualidade do classificador AD foi avaliada por um conjunto de 500 pontos aleatórios e independentes, gerados para cada ano, o que permitiu calcular os parâmetros índice Kappa e exatidão global a partir das matrizes de confusão. O algoritmo AD obteve desempenho médio próximo a 83% para o índice Kappa e exatidão global média de 86%. Esses valores permitem considerar a classificação como excelente, o que permitiu uma associação segura entre a influência antrópica e a dinâmica da cobertura da terra na bacia hidrográfica estudada. Foi diagnosticado o aumento das atividades agrícolas e silvicultoras em detrimento das coberturas naturais, além de uma fragmentação dos corredores ecológicos da Floresta Ombrófila Mista, no intervalo analisado de 22 anos.

ABSTRACT Land-cover information is essential for planning and studying the effects of changes from natural to disturbed landscapes. This study aimed at studying the dynamics of land cover from 1989 to 2011 in the Marombas River basin using a Decision Tree (DT) algorithm. Landsat-5 spectral bands, vegetation indices, and terrain attributes from elevation models were used as attributes for classification in the years of 1989, 1991, 1993, 1997, 2001, 2004, and 2011. DT classifier quality was assessed by a set of 500 independent random points, generated for each year; which allowed calculating Kappa index parameters and overall accuracy from the confusion matrices. The DT algorithm achieved a mean Kappa index of about 83% and a mean global accuracy of 86%. Therefore, it can be stated an excellent classification, from which we can securely associate anthropogenic influence with land-cover dynamics in this basin. We also observed an increase of agricultural and silvicultural activities at the expense of more natural land covers. Adding to that, the results showed that rapid fragmentation has occurred in the natural mixed ombrophilous forest along this interval of 22 years.

Uma questão crítica no mapeamento digital de solos é a seleção do método de amostragem dos dados para treinamento do modelo preditivo. Uma abordagem emergente aplica a seleção de instâncias (observações) para reduzir o tamanho do conjunto de dados, selecionando amostras relevantes para obter um subconjunto representativo, o qual seja grande o suficiente para preservar as informações pertinentes, mas pequeno o suficiente para ser facilmente manipulado pelos algoritmos de aprendizagem. Embora existam sugestões para distribuir a amostragem de dados em função da proximidade de limites de unidades de mapeamento de solos (UM), ainda existem contradições entre as recomendações de pesquisa para localizar amostras mais perto ou mais distantes desses limites. Foi realizado um estudo para avaliar os métodos de seleção de instâncias com base na coleta de dados espacialmente explícita usando a localização em relação aos limites de mapa de solo como o principal critério. Realizou-se análise de árvore de decisão para a modelagem de mapeamento digital de classes de solo usando dois esquemas de amostragem diferentes: a) selecionando pontos de amostragem localizados fora das áreas marginais aos limites das UM e b) selecionando pontos de amostragem situados dentro das áreas marginais aos limites das UM. Os dados foram preparados para a geração de árvores de classificação para incluir somente dados pontuais localizados dentro ou fora de faixas com larguras de 60, 120, 240, 360, 480 e 600m ao redor dos limites de UM. Ambos os métodos de seleção de instâncias foram eficazes para reduzir o tamanho do conjunto de dados usado para calibração de árvores de classificação, mas não trouxeram vantagens para o mapeamento digital de classes de solos.

A critical issue in digital soil mapping (DSM) is the selection of data sampling method for model training. One emerging approach applies instance selection to reduce the size of the dataset by drawing only relevant samples in order to obtain a representative subset that is still large enough to preserve relevant information, but small enough to be easily handled by learning algorithms. Although there are suggestions to distribute data sampling as a function of the soil map unit (MU) boundaries location, there are still contradictions among research recommendations for locating samples either closer or more distant from soil MU boundaries. A study was conducted to evaluate instance selection methods based on spatially-explicit data collection using location in relation to soil MU boundaries as the main criterion. Decision tree analysis was performed for modeling digital soil class mapping using two different sampling schemes: a) selecting sampling points located outside buffers near soil MU boundaries, and b) selecting sampling points located within buffers near soil MU boundaries. Data was prepared for generating classification trees to include only data points located within or outside buffers with widths of 60, 120, 240, 360, 480, and 600m near MU boundaries. Instance selection methods using both spatial selection of methods was effective for reduced size of the dataset used for calibrating classification tree models, but failed to provide advantages to digital soil mapping because of potential reduction in the accuracy of classification tree models.

A espectroscopia de reflectância difusa (ERD) pode ser utilizada como alternativa para quantificação de atributos como granulometria e matéria orgânica do solo (MOS). Essa técnica pode ser opção para quantificar esses atributos em grande volume de amostras de solos, visto ser rápida, com menor custo e sem a geração de resíduos químicos. O objetivo deste estudo foi desenvolver modelos usando análise de regressão linear múltipla para predizer o teor de argila, areia, silte e MOS, utilizando dados de ERD em uma área de relevo e geologia complexa localizada na região central do Rio Grande do Sul. No estudo, foram utilizadas 303 amostras coletadas na profundidade de 0,00-0,20 m para determinar os teores de argila, areia, silte e MOS por meio da análise laboratorial e de reflectância espectral. O desempenho dos modelos de predição apresentaram bons resultados, com capacidade de explicação da variância de 77 e 72 % para areia e argila, respectivamente. Mesmo com a complexidade geológica e pedológica, os resultados evidenciaram que a técnica é promissora, sendo possível a aplicação dessa técnica para predição da granulometria e teor de MOS.

Diffuse reflectance spectroscopy (DRS) can be used as an alternative in identifying and quantifying some soil properties such as particle size and soil organic matter (SOM). This technique may be an alternative to quantifying those properties in a large volume of soil samples since it is faster and less costly and does not produce chemical residues. The aim of this study was to develop models using multiple linear regression analysis to predict the content of clay, sand, silt, and SOM using DRS data in an area of complex topography and geology located in the central region of Rio Grande do Sul, Brazil. In the study, 303 samples were collected at a depth of 0.00-0.20 m for determination of clay, sand, silt, and SOM by laboratory and spectral reflectance analysis. The predictive models produced high-quality results, explaining 77 and 72 % of the variance for sand and clay, respectively. The interpolation maps of the observed and predicted properties revealed that the spatial patterns are mainly associated with the topography and geology of the area. Even with the geological and pedological complexity, the results indicated that the technique is promising and it may be applied for prediction of particle size and SOM.

A cobertura da terra é uma informação espacial de extrema relevância para uma série de modelos, sendo utilizada para estimar a produção de sedimentos e para mensurar a potencialidade da paisagem em sequestrar carbono. A classificação da cobertura da terra pelo método de classificação supervisionado necessita de áreas de treino, já que essas áreas devem ser representativas para cada classe de cobertura da terra. Para o algoritmo de classificação por árvore de decisão (AD), a complexidade da AD resulta em diferentes valores de acurácias para os mapas temáticos. Desse modo, o objetivo deste estudo foi determinar a densidade mínima de amostras em um modelo por AD, a fim de discriminar as classes de cobertura da terra e avaliar o tamanho da AD gerada quanto ao seu número de folhas. Além disso, preocupou-se em identificar as classes da cobertura da terra de mais difícil classificação. Nesse contexto, foram utilizadas bandas da imagem do satélite RESOURCESAT-1 e índices espectrais. A densidade mínima de amostras variou entre 0,15 e 0,30% da área total para cada classe. Esse intervalo de amostragem possibilitou resultados melhores que 80% para o índice kappa. O menor agrupamento entre observações em uma mesma folha terminal foi de 45, e as classes mais difíceis de classificar foram floresta e lavoura de arroz, devido à semelhança espectral que as florestas sombreadas possuem com as lavouras de arroz irrigadas.

Land cover is a spatial information of great relevance for a variety of models for estimating sediment yield and to measure the potential of the landscape carbon sequestration. The classification of land cover by the supervised method requires training areas, these areas must be representative of each class of land cover. For the classification decision tree (DT) algorithm, the complexity of DT, results in different values of accuracies for thematic maps. Thus, the objective of this study was to estimate the minimum sample density in a DT model which would allow to discriminate land cover classes, evaluate the size of the generated DT model, as well as, identify the more difficult land cover class to be mapped. Satellite images from RESOURCESAT-1as well as spectral indices were used in the study. The minimum sample density varied between 0.15 and 0.30% of the total area for each class, this sampling interval allowed better results than 80% for kappa index. The smallest grouping of observations in the same terminal leaf was 45 observations. In this study the most difficult land use classes to be mapped were forest and rice crops due to spectral similarity of shaded forests with irrigated rice crops.

Informações digitais tornam possível um elevado grau de redundância das informações disponíveis para o ajuste de modelos preditores aplicados ao Mapeamento Digital de Solos (MDS). Entre esses modelos, a técnica de Árvores de Decisão (AD) tem aplicação crescente, em razão da sua potência no tratamento de grandes volumes de dados. Objetivou-se com este trabalho avaliar o impacto do volume de dados utilizados para gerar os modelos por AD, na qualidade dos mapas de solos gerados pela técnica de MDS. Uma área de estudo com 889,33 km² foi escolhida na região do Planalto Médio do Rio Grande do Sul. As relações solo-paisagem foram obtidas a partir de reambulação da área de estudo e delineamento das unidades de mapeamento em cartas topográficas de escala 1:50.000. Seis covariáveis preditoras ligadas aos fatores de formação do solo, relevo e organismos, juntamente com os conjuntos de dados de um, três, cinco, 10, 15, 20 e 25 % do volume total de dados, foram usadas para gerar os modelos preditivos por AD no programa WEKA. Neste estudo, densidades de amostragem menores do que 5 % resultaram em modelos com menor poder de capturar a complexidade da distribuição espacial do solo da área estudada. Amostragens entre cinco e 15 % conduziram a uma melhor relação entre o volume de dados a ser manipulado e a capacidade preditiva dos modelos gerados. Dados coletados no campo indicaram acurácia dos mapas preditos próxima a 70 %, para os modelos oriundos dessas densidades de amostragem.

Digital information generates the possibility of a high degree of redundancy in the data available for fitting predictive models used for Digital Soil Mapping (DSM). Among these models, the Decision Tree (DT) technique has been increasingly applied due to its capacity of dealing with large datasets. The purpose of this study was to evaluate the impact of the data volume used to generate the DT models on the quality of soil maps. An area of 889.33 km² was chosen in the Northern region of the State of Rio Grande do Sul. The soil-landscape relationship was obtained from reambulation of the studied area and the alignment of the units in the 1:50,000 scale topographic mapping. Six predictive covariates linked to the factors soil formation, relief and organisms, together with data sets of 1, 3, 5, 10, 15, 20 and 25 % of the total data volume, were used to generate the predictive DT models in the data mining program Waikato Environment for Knowledge Analysis (WEKA). In this study, sample densities below 5 % resulted in models with lower power of capturing the complexity of the spatial distribution of the soil in the study area. The relation between the data volume to be handled and the predictive capacity of the models was best for samples between 5 and 15 %. For the models based on these sample densities, the collected field data indicated an accuracy of predictive mapping close to 70 %.

O solo é cada vez mais reconhecido como tendo um importante papel nos ecossistemas, assim como para a produção de alimentos e regulação do clima global. Por esse motivo, a demanda por informações relevantes e atualizadas em solos é crescente. Pesquisadores em ciência do solo estão sendo demandados a gerar informações em diferentes resoluções espaciais e com qualidade associada dentro do que está sendo chamado de Mapeamento Digital de Solos (MDS). Devido ao crescente número de trabalhos relacionados ao MDS, faz-se necessário reunir e discutir as principais características dos estudos relacionados ao mapeamento digital de classes de solos no Brasil, o que irá possibilitar uma perspectiva mais ampla dos caminhos, além de nortear trabalhos e demandas futuras. O mapeamento de classes de solos empregando técnicas de MDS é recente no país, com a primeira publicação em 2006. Entre as funções preditivas utilizadas, predomina o emprego da técnica de regressões logísticas. O fator de formação relevo foi empregado na totalidade dos estudos revisados. Quanto à avaliação da qualidade dos modelos preditivos, o emprego da matriz de erros e do índice kappa têm sido os procedimentos mais usuais. A consolidação dessa abordagem automatizada como ferramenta auxiliar ao mapeamento convencional passa pelo treinamento dos jovens pedólogos para a utilização de tecnologias da geoinformação e de ferramentas quantitativas dos aspectos de variabilidade do solo.

Soil is increasingly being recognized as having an important role in ecosystems, as well as for food production and global climate regulation. For this reason, the demand for relevant and updated soil information is increasing. Soil science researchers are being demanded to produce information in different spatial resolutions with associated quality in what is being called Digital Soil Mapping (DSM). Due to an increasing number of papers related to the DSM in Brazil, it is necessary to discuss the main characteristics of those studies related to the automated mapping of soil classes, which will enable a broader perspective of the subject and guide future works and demands. The mapping of soil classes using DSM techniques is recent in the country, the first publication in this topic occurred just in 2006. Among the predictive functions the predominant is logistic regression. The soil formation factor relief was used in all studies reviewed. Quality of predictive models was evaluated employing error matrix and kappa which were the most common procedures. The consolidation of this automated approach as an auxiliary tool to the conventional soil mapping will demand training of young soil scientists to use geoinformation technologies and quantitative tools to handle aspects of soil variability.