Mapping a hydrophysical soil property through a comparative analysis of local and global scale approximations

Current available soil information allows building baselines to improve research, such as sustainable resource management; however, its use requires analysis of accuracy and precision that describes specific variables on local and global scales. Therefore, this study evaluated differences in the spatial distribution of water retention capacity (WRC) of the soil at a depth of 0.3 m, calculated from local general soil surveys and the global gridded soil information system (SoilGrids), using detailed or semi-detailed soil surveys as a reference, in two regions of Colombia (A and B). The qualitative and statistical analyses evaluated differences in WRC surfaces generated by the information sources. Neither information sources described WRC accurately, achieving correlations between −0.15 and 0.49 and average absolute errors between 9.65 and 19.52 mm for zones A and B, respectively. However, studies on the local scale remain within the ranges observed in the most detailed local studies. The use of products on the global scale is subject to regional analyses; nevertheless, they can be included as a covariate in digital soil mapping studies on more detailed scales.(AU)

Mapping a hydrophysical soil property through a comparative analysis of local and global scale approximations

Current available soil information allows building baselines to improve research, such as sustainable resource management; however, its use requires analysis of accuracy and precision that describes specific variables on local and global scales. Therefore, this study evaluated differences in the spatial distribution of water retention capacity (WRC) of the soil at a depth of 0.3 m, calculated from local general soil surveys and the global gridded soil information system (SoilGrids), using detailed or semi-detailed soil surveys as a reference, in two regions of Colombia (A and B). The qualitative and statistical analyses evaluated differences in WRC surfaces generated by the information sources. Neither information sources described WRC accurately, achieving correlations between −0.15 and 0.49 and average absolute errors between 9.65 and 19.52 mm for zones A and B, respectively. However, studies on the local scale remain within the ranges observed in the most detailed local studies. The use of products on the global scale is subject to regional analyses; nevertheless, they can be included as a covariate in digital soil mapping studies on more detailed scales.

Open legacy soil survey data in Brazil: geospatial data quality and how to improve it

Spatial soil data applications require sound geospatial data including coordinates and a coordinate reference system. However, when it comes to legacy soil data we frequently find them to be missing or incorrect. This paper assesses the quality of the geospatial data of legacy soil observations in Brazil, and evaluates geospatial data sources (survey reports, maps, spatial data infrastructures, web mapping services) and expert knowledge as a means to fix inconsistencies. The analyses included several consistency checks performed on 6,195 observations from the Brazilian Soil Information System. The positional accuracy of geospatial data sources was estimated so as to obtain an indication of the quality for fixing inconsistencies. The coordinates of 20 soil observations, estimated using the web mapping service, were validated with the true coordinates measured in the field. Overall, inconsistencies of different types and magnitudes were found in half of the observations, causing mild to severe misplacements. The involuntary substitution of symbols and numeric characters with similar appearance when recording geospatial data was the most common typing mistake. Among the geospatial data sources, the web mapping service was the most useful, due to operational advantages and lower positional error (~6 m). However, the quality of the description of the observation location controls the accuracy of estimated coordinates. Thus, the error of coordinates estimated using the web mapping service ranged between 30 and 1000 m. This is equivalent to coordinates measured from arc-seconds to arc-minutes, respectively. Under this scenario, the feedback from soil survey experts is crucial to improving the quality of geospatial data.

Open legacy soil survey data in Brazil: geospatial data quality and how to improve it

Spatial soil data applications require sound geospatial data including coordinates and a coordinate reference system. However, when it comes to legacy soil data we frequently find them to be missing or incorrect. This paper assesses the quality of the geospatial data of legacy soil observations in Brazil, and evaluates geospatial data sources (survey reports, maps, spatial data infrastructures, web mapping services) and expert knowledge as a means to fix inconsistencies. The analyses included several consistency checks performed on 6,195 observations from the Brazilian Soil Information System. The positional accuracy of geospatial data sources was estimated so as to obtain an indication of the quality for fixing inconsistencies. The coordinates of 20 soil observations, estimated using the web mapping service, were validated with the true coordinates measured in the field. Overall, inconsistencies of different types and magnitudes were found in half of the observations, causing mild to severe misplacements. The involuntary substitution of symbols and numeric characters with similar appearance when recording geospatial data was the most common typing mistake. Among the geospatial data sources, the web mapping service was the most useful, due to operational advantages and lower positional error (~6 m). However, the quality of the description of the observation location controls the accuracy of estimated coordinates. Thus, the error of coordinates estimated using the web mapping service ranged between 30 and 1000 m. This is equivalent to coordinates measured from arc-seconds to arc-minutes, respectively. Under this scenario, the feedback from soil survey experts is crucial to improving the quality of geospatial data.(AU)

Preprocessing procedures and supervised classification applied to a database of systematic soil survey

Data Mining techniques play an important role in the prediction of soil spatial distribution in systematic soil surveying, though existing methodologies still lack standardization and a full understanding of their capabilities. The aim of this work was to evaluate the performance of preprocessing procedures and supervised classification approaches for predicting map units from 1:100,000-scale conventional semi-detailed soil surveys. Sheets of the Brazilian National Cartographic System on the 1:50,000 scale, Dois Córregos (Brotas 1:100,000-scale sheet), São Pedro and Laras (Piracicaba 1:100,000-scale sheet) were used for developing models. Soil map information and predictive environmental covariates for the dataset were obtained from the semi-detailed soil survey of the state of São Paulo, from the Brazilian Institute of Geography and Statistics (IBGE) 1:50,000-scale topographic sheets and from the 1:750,000-scale geological map of the state of São Paulo. The target variable was a soil map unit of four types: local soil unit name and soil class at three hierarchical levels of the Brazilian System of Soil Classification (SiBCS). Different data preprocessing treatments and four algorithms all having different approaches were also tested. Results showed that composite soil map units were not adequate for the machine learning process. Class balance did not contribute to improving the performance of classifiers. Accuracy values of 78 % and a Kappa index of 0.67 were obtained after preprocessing procedures with Random Forest, the algorithm that performed best. Information from conventional map units of semi-detailed (4th order) 1:100,000 soil survey generated models with values for accuracy, precision, sensitivity, specificity and Kappa indexes that support their use in programs for systematic soil surveying.

Preprocessing procedures and supervised classification applied to a database of systematic soil survey

Data Mining techniques play an important role in the prediction of soil spatial distribution in systematic soil surveying, though existing methodologies still lack standardization and a full understanding of their capabilities. The aim of this work was to evaluate the performance of preprocessing procedures and supervised classification approaches for predicting map units from 1:100,000-scale conventional semi-detailed soil surveys. Sheets of the Brazilian National Cartographic System on the 1:50,000 scale, Dois Córregos (Brotas 1:100,000-scale sheet), São Pedro and Laras (Piracicaba 1:100,000-scale sheet) were used for developing models. Soil map information and predictive environmental covariates for the dataset were obtained from the semi-detailed soil survey of the state of São Paulo, from the Brazilian Institute of Geography and Statistics (IBGE) 1:50,000-scale topographic sheets and from the 1:750,000-scale geological map of the state of São Paulo. The target variable was a soil map unit of four types: local soil unit name and soil class at three hierarchical levels of the Brazilian System of Soil Classification (SiBCS). Different data preprocessing treatments and four algorithms all having different approaches were also tested. Results showed that composite soil map units were not adequate for the machine learning process. Class balance did not contribute to improving the performance of classifiers. Accuracy values of 78 % and a Kappa index of 0.67 were obtained after preprocessing procedures with Random Forest, the algorithm that performed best. Information from conventional map units of semi-detailed (4th order) 1:100,000 soil survey generated models with values for accuracy, precision, sensitivity, specificity and Kappa indexes that support their use in programs for systematic soil surveying.(AU)

Soil class map of the Rio Jardim watershed in Central Brazil at 30 meter spatial resolution based on proximal and remote sensed data and MESMA method.

Geospatial soil information is critical for agricultural policy formulation and decision making, land-use suitability analysis, sustainable soil management, environmental assessment, and other research topics that are of vital importance to agriculture and economy. Proximal and Remote sensing technologies enables us to collect, process, and analyze spectral data and to retrieve, synthesize, visualize valuable geospatial information for multidisciplinary uses. We obtained the soil class map provided in this article by processing and analyzing proximal and remote sensed data from soil samples collected in toposequences based on pedomorphogeological relashionships. The soils were classified up to the second categorical level (suborder) of the Brazilian Soil Classification System (SiBCS), as well as in the World Reference Base (WRB) and United States Soil Taxonomy (ST) systems. The raster map has 30 m resolution and its accuracy is 73% (Kappa coefficient of 0.73). The soil legend represents a soil class followed by its topsoil color.

Soil type spatial prediction from Random Forest: different training datasets, transferability, accuracy and uncertainty assessment

Different uses of soil legacy data such as training dataset as well as the selection of soil environmental covariables could drive the accuracy of machine learning techniques. Thus, this study evaluated the ability of the Random Forest algorithm to predict soil classes from different training datasets and extrapolate such information to a similar area. The following training datasets were extracted from legacy data: a) point data composed of 53 soil samples; b) 30 m buffer around the soil samples, and soil map polygons excluding: c) 20 m; and d) 30 m from the boundaries of polygons. These four datasets were submitted to principal component analysis (PCA) to reduce multidimensionality. Each dataset derived a new one. Different combinations of predictor variables were tested. A total of 52 models were evaluated by means of error of models, prediction uncertainty and external validation for overall accuracy and Kappa index. The best result was obtained by reducing the number of predictors with the PCA along with information from the buffer around the points. Although Random Forest has been considered a robust spatial predictor model, it was clear it is sensitive to different strategies of selecting training dataset. Effort was necessary to find the best training dataset for achieving a suitable level of accuracy of spatial prediction. To identify a specific dataset seems to be better than using a great number of variables or a large volume of training data. The efforts made allowed for the accurate acquisition of a mapped area 15.5 times larger than the reference area.

Soil type spatial prediction from Random Forest: different training datasets, transferability, accuracy and uncertainty assessment

Different uses of soil legacy data such as training dataset as well as the selection of soil environmental covariables could drive the accuracy of machine learning techniques. Thus, this study evaluated the ability of the Random Forest algorithm to predict soil classes from different training datasets and extrapolate such information to a similar area. The following training datasets were extracted from legacy data: a) point data composed of 53 soil samples; b) 30 m buffer around the soil samples, and soil map polygons excluding: c) 20 m; and d) 30 m from the boundaries of polygons. These four datasets were submitted to principal component analysis (PCA) to reduce multidimensionality. Each dataset derived a new one. Different combinations of predictor variables were tested. A total of 52 models were evaluated by means of error of models, prediction uncertainty and external validation for overall accuracy and Kappa index. The best result was obtained by reducing the number of predictors with the PCA along with information from the buffer around the points. Although Random Forest has been considered a robust spatial predictor model, it was clear it is sensitive to different strategies of selecting training dataset. Effort was necessary to find the best training dataset for achieving a suitable level of accuracy of spatial prediction. To identify a specific dataset seems to be better than using a great number of variables or a large volume of training data. The efforts made allowed for the accurate acquisition of a mapped area 15.5 times larger than the reference area.(AU)

VIS-NIR spectral reflectance for discretization of soils with high sand content / Refletância espectral VIS-NIR para discretização de solos com alto teor de areia

The aim of this study is to evaluate the spatial distribution and relationships between the physicochemical attributes and radiometry of soils with high sand contents. One hundred surface horizon samples were collected for physicochemical and spectral analyses of the soil. The samples were selected spatially by the conditioned Latin hypercube method. The physicochemical analyses consisted of granulometry, organic carbon content, and iron oxides content, extracted using sodium dithionite-citrate-bicarbonate (DCB). The spectral response of the soils was analyzed in the 400 to 1000 nm range. The spectral curves were obtained from the samples of the surface horizons, which were categorized according to the attribute in question. The relationship between the soil physicochemical attributes and soil radiometry was evaluated through a Pearson's correlation. There was a tendency for the organic carbon content to decrease with an increase in soil depth, associated with the presence of soils with higher sand contents. For soils with iron contents lower than 80 g kg <->1, there was an increase in the reflectance along the spectrum, whereas for soils with contents between 80 and 160 g kg <->1, the reflectance decreased after 600 nm, with greater variation along the spectrum for soils with iron contents higher than 120 g kg <->1. The diffuse reflectance spectroscopy could potentially allow for granulometric distinction between the soils evaluated.(AU)

Este estudo objetivou avaliar a discretização espacial e relações entre atributos físico-químicos com a radiometria dos solos com altos teores de areia. Foram coletadas 100 (cem) amostras de horizontes superficiais para análises físico-químicas e espectrais do solo, selecionadas espacialmente pelo método do hipercúbico latino condicionado. As análises físico-químicas foram a granulometria, teores de carbono orgânico e de óxidos de ferro extraído por ditionito-citrato-bicarbonato de sódio (DCB). A resposta espectral dos solos foi analisada na faixa de 400 a 1000 nm. As curvas foram elaboradas a partir da média de reflectância espectral de cem amostras de horizontes superficiais categorizados de acordo com o atributo em questão. A relação entre os atributos físico-químicos do solo e a radiometria dos solos foi avaliada através da correlação de Pearson. Houve a tendência do decréscimo nos teores de carbono orgânico com a diminuição da altitude, associada a presença de solos com maiores teores de areia. Para os solos com teores de ferro inferiores 80 g kg <->1 houve um aumento da refletância ao longo do espectro, enquanto que solos com teores entre 80 a 160 g kg <->1 a refletância decresceu significativamente após os 600 nm, com maior variação nos solos com teores de ferro superiores a 120 g kg <->1. A espectroscopia de refletância difusa mostrou-se uma ferramenta com potencial de distinção granulométrica para os solos avaliados.(AU)

VIS-NIR spectral reflectance for discretization of soils with high sand content / Refletância espectral VIS-NIR para discretização de solos com alto teor de areia

The aim of this study is to evaluate the spatial distribution and relationships between the physicochemical attributes and radiometry of soils with high sand contents. One hundred surface horizon samples were collected for physicochemical and spectral analyses of the soil. The samples were selected spatially by the conditioned Latin hypercube method. The physicochemical analyses consisted of granulometry, organic carbon content, and iron oxides content, extracted using sodium dithionite-citrate-bicarbonate (DCB). The spectral response of the soils was analyzed in the 400 to 1000 nm range. The spectral curves were obtained from the samples of the surface horizons, which were categorized according to the attribute in question. The relationship between the soil physicochemical attributes and soil radiometry was evaluated through a Pearson's correlation. There was a tendency for the organic carbon content to decrease with an increase in soil depth, associated with the presence of soils with higher sand contents. For soils with iron contents lower than 80 g kg

Este estudo objetivou avaliar a discretização espacial e relações entre atributos físico-químicos com a radiometria dos solos com altos teores de areia. Foram coletadas 100 (cem) amostras de horizontes superficiais para análises físico-químicas e espectrais do solo, selecionadas espacialmente pelo método do hipercúbico latino condicionado. As análises físico-químicas foram a granulometria, teores de carbono orgânico e de óxidos de ferro extraído por ditionito-citrato-bicarbonato de sódio (DCB). A resposta espectral dos solos foi analisada na faixa de 400 a 1000 nm. As curvas foram elaboradas a partir da média de reflectância espectral de cem amostras de horizontes superficiais categorizados de acordo com o atributo em questão. A relação entre os atributos físico-químicos do solo e a radiometria dos solos foi avaliada através da correlação de Pearson. Houve a tendência do decréscimo nos teores de carbono orgânico com a diminuição da altitude, associada a presença de solos com maiores teores de areia. Para os solos com teores de ferro inferiores 80 g kg

Geomorphometric tool associated with soil types and properties spatial variability at watersheds under tropical conditions

The application of quantitative methods to digital soil and geomorphological mapping is becoming an increasing trend. One of these methods, Geomorphons, was developed to identify the ten most common landforms based on digital elevation models. This study aimed to make a quantitative assessment of the relationships between Geomorphons units, determined at three spatial resolutions and nine radii, and soil types and properties of two watersheds with different soil-landscape relationships in Brazil to help soil surveying and mapping under tropical conditions. The study was conducted at Lavrinha Creek (LCW) and Marcela Creek (MCW) watersheds, located in the state of Minas Gerais, Brazil. Spatial resolutions of 10, 20 and 30 m were the basis for generating Geomorphons at 9 radii of calculation for the watersheds. They were overlapped to detailed soil maps of the watersheds and a chi-square test was carried out to assess their relationship with soil types. Observation points were compared with the most highly correlated Geomorphons to also assess relationships with soil properties. Geomorphons with resolution of 30-m and radii of 20 and 50 cells, respectively for LCW and MCW, were more highly correlated with the variability of soil types, in accordance with the terrain features of these watersheds. The majority of observation points for each soil type was located in the same Geomorphon unit that was dominant when analyzing soil maps. There was less variability in soil properties between Geomorphon units, which was probably due to the highly weathered-leached stage of soils. Geomorphons can help to improve soil maps in tropical conditions when assessing soil variability due to its high correlation with tropical soil types variability.

Geomorphometric tool associated with soil types and properties spatial variability at watersheds under tropical conditions

The application of quantitative methods to digital soil and geomorphological mapping is becoming an increasing trend. One of these methods, Geomorphons, was developed to identify the ten most common landforms based on digital elevation models. This study aimed to make a quantitative assessment of the relationships between Geomorphons units, determined at three spatial resolutions and nine radii, and soil types and properties of two watersheds with different soil-landscape relationships in Brazil to help soil surveying and mapping under tropical conditions. The study was conducted at Lavrinha Creek (LCW) and Marcela Creek (MCW) watersheds, located in the state of Minas Gerais, Brazil. Spatial resolutions of 10, 20 and 30 m were the basis for generating Geomorphons at 9 radii of calculation for the watersheds. They were overlapped to detailed soil maps of the watersheds and a chi-square test was carried out to assess their relationship with soil types. Observation points were compared with the most highly correlated Geomorphons to also assess relationships with soil properties. Geomorphons with resolution of 30-m and radii of 20 and 50 cells, respectively for LCW and MCW, were more highly correlated with the variability of soil types, in accordance with the terrain features of these watersheds. The majority of observation points for each soil type was located in the same Geomorphon unit that was dominant when analyzing soil maps. There was less variability in soil properties between Geomorphon units, which was probably due to the highly weathered-leached stage of soils. Geomorphons can help to improve soil maps in tropical conditions when assessing soil variability due to its high correlation with tropical soil types variability.(AU)

Extrapolação das relações solo-paisagem a partir de uma área de referência / Extrapolation of soil-landscaperelationship from a reference area

O objetivo deste estudo foi aplicar uma técnica automatizada de mapeamento de solos a partir de modelos preditivos ajustados em uma área de referência e, posteriormente, aplicados em áreas próximas de mesmas relações solo-paisagem. Modelos lineares generalizados foram desenvolvidos utilizando-se de nove atributos de terreno, derivados de um Modelo Digital de Elevação, como covariáveis preditoras e classes de solos, obtidas em um levantamento convencional, como variáveis dependentes. Os modelos foram capazes de distinguir as três principais formas da paisagem local. Classes de solos de pedogênese intimamente ligada às covariáveis preditoras obtiveram os melhores resultados. O mapa de solos gerado apresentou uma reprodutibilidade de 46,12 por cento e uma exatidão de 21,06 por cento.

The aim of this study was to apply an automated technique of soil mapping from predictive models developed at a reference area, into nearby areas of the same soil-landscape relationships. Generalized linear models were developed using nine terrain attributes derived from a digital elevation model as covariate predictors, with soil classes, obtained from a conventional soil survey, as dependent variables. The models were able to distinguish the three main forms of the local landscape. Soil classes with pedogenesis intimately tied to the predictive covariates obtained the best results. The soil maps generated, showed a reproducibility of 46.12 percent and an accuracy of 21.06 percent.

Digital soilscape mapping of tropical hillslope areas by neural networks

Geomorphometric variables are applied in digital soil mapping because of their strong correlation with the disposition and distribution of pedological components of the landscapes. In this research, the relationship between environmental components of tropical hillslope areas in the Rio de Janeiro State, Brazil, artificial neural networks (ANN), and maximum likelihood algorithm (MaxLike) were evaluated with the aid of geoprocessing techniques. ANN and MaxLike were applied to soilscape mapping and the results were compared to the original map. The ANN architectures with seven and five neurons in the hidden layer produced the best classifications when using samples obtained systematically. When random samples were applied, the best neural net architectures were within 22 and 16 neurons in the hidden layer. In conclusion, the ANN can contribute to soilscape surveys, making map delineation faster and less expensive. The digital elevation model (DEM) and its derived attributes can contribute to the understanding of the soil-landscape relationship of tropical hillslope areas; the use of artificial neural networks and MaxLike is feasible for digital soilscape mapping. The systematic sampling method provided a global accuracy of 70 % and 65.9 % for the ANN and the MaxLike, respectively. When the random sampling method was applied, the ANN had a global accuracy of 69.6 %, and the MaxLike had an accuracy of 62.1 %, considering the total study area in relation to the reference map.

Extrapolação das relações solo-paisagem a partir de uma área de referência

The aim of this study was to apply an automated technique of soil mapping from predictive models developed at a reference area, into nearby areas of the same soil-landscape relationships. Generalized linear models were developed using nine terrain attributes derived from a digital elevation model as covariate predictors, with soil classes, obtained from a conventional soil survey, as dependent variables. The models were able to distinguish the three main forms of the local landscape. Soil classes with pedogenesis intimately tied to the predictive covariates obtained the best results. The soil maps generated, showed a reproducibility of 46.12% and an accuracy of 21.06%.

O objetivo deste estudo foi aplicar uma técnica automatizada de mapeamento de solos a partir de modelos preditivos ajustados em uma área de referência e, posteriormente, aplicados em áreas próximas de mesmas relações solo-paisagem. Modelos lineares generalizados foram desenvolvidos utilizando-se de nove atributos de terreno, derivados de um Modelo Digital de Elevação, como covariáveis preditoras e classes de solos, obtidas em um levantamento convencional, como variáveis dependentes. Os modelos foram capazes de distinguir as três principais formas da paisagem local. Classes de solos de pedogênese intimamente ligada às covariáveis preditoras obtiveram os melhores resultados. O mapa de solos gerado apresentou uma reprodutibilidade de 46,12% e uma exatidão de 21,06%.

Extrapolação das relações solo-paisagem a partir de uma área de referência

The aim of this study was to apply an automated technique of soil mapping from predictive models developed at a reference area, into nearby areas of the same soil-landscape relationships. Generalized linear models were developed using nine terrain attributes derived from a digital elevation model as covariate predictors, with soil classes, obtained from a conventional soil survey, as dependent variables. The models were able to distinguish the three main forms of the local landscape. Soil classes with pedogenesis intimately tied to the predictive covariates obtained the best results. The soil maps generated, showed a reproducibility of 46.12% and an accuracy of 21.06%.

O objetivo deste estudo foi aplicar uma técnica automatizada de mapeamento de solos a partir de modelos preditivos ajustados em uma área de referência e, posteriormente, aplicados em áreas próximas de mesmas relações solo-paisagem. Modelos lineares generalizados foram desenvolvidos utilizando-se de nove atributos de terreno, derivados de um Modelo Digital de Elevação, como covariáveis preditoras e classes de solos, obtidas em um levantamento convencional, como variáveis dependentes. Os modelos foram capazes de distinguir as três principais formas da paisagem local. Classes de solos de pedogênese intimamente ligada às covariáveis preditoras obtiveram os melhores resultados. O mapa de solos gerado apresentou uma reprodutibilidade de 46,12% e uma exatidão de 21,06%.

Digital soilscape mapping of tropical hillslope areas by neural networks

Geomorphometric variables are applied in digital soil mapping because of their strong correlation with the disposition and distribution of pedological components of the landscapes. In this research, the relationship between environmental components of tropical hillslope areas in the Rio de Janeiro State, Brazil, artificial neural networks (ANN), and maximum likelihood algorithm (MaxLike) were evaluated with the aid of geoprocessing techniques. ANN and MaxLike were applied to soilscape mapping and the results were compared to the original map. The ANN architectures with seven and five neurons in the hidden layer produced the best classifications when using samples obtained systematically. When random samples were applied, the best neural net architectures were within 22 and 16 neurons in the hidden layer. In conclusion, the ANN can contribute to soilscape surveys, making map delineation faster and less expensive. The digital elevation model (DEM) and its derived attributes can contribute to the understanding of the soil-landscape relationship of tropical hillslope areas; the use of artificial neural networks and MaxLike is feasible for digital soilscape mapping. The systematic sampling method provided a global accuracy of 70 % and 65.9 % for the ANN and the MaxLike, respectively. When the random sampling method was applied, the ANN had a global accuracy of 69.6 %, and the MaxLike had an accuracy of 62.1 %, considering the total study area in relation to the reference map.