Search

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES The development of large databases usually implies combining data collected for different purposes under different standards and methodologies, which often leads databases to suffer from disparate and inconsistent soil data. Despite the potential of visible and nearinfrared (Vis-NIR) spectroscopy to predict soil organic carbon (SOC) from those databases, the effectiveness and consistency among analytical methods used to produce the target data are seldom discussed. The main purpose of this research was to investigate the interplay among preprocessing techniques, model architectures, and especially the analytical methods used to produce the SOC target data. To accomplish it, we set up two specific objectives: i) evaluate the interplay among analytical methods, preprocessing techniques, and model architectures on SOC predictions, ii) assess whether this interplay can be translated into some form of hierarchy across validation metrics. In this PhD thesis, two chapters cover the topic where the above-mentioned objectives were met. Chapter I presents how changes in the analytical method (dry (SOCDC) and wet combustion with quantification by titrimetry (SOCWCt) and colorimetry (SOCWCc)) and the preprocessing techniques (smoothing (SMO), continuum removal (CRR), and Savitzky-Golay first derivative (SGD)) affect the empirical relationship captured by different machine learning algorithms (random forest, cubist, and partial least square regression (PLSR)). Cross-validation metrics were used to compare the parallel performance of 27 predictive models. The relationship between covariate matrix and target data is explored based on the variable importance. Chapter II shows how the interplay among those three factors can be translated into a hierarchy. A resampling technique was used to split the dataset into training and validation sets 100 times to achieve realistic performances and explore how the predictive performance changed as the training set changed. Conditional inference tree analysis was performed to evaluate how those three factors influenced global validation metrics. The predictive performance in both studies varied depending on the SOC analytical method, preprocessing technique, and model architecture employed. Among the three analytical methods tested, DC and WCt provided a higher correlation between SOC and spectra than WCc, and thus, resulted in higher models performance. The model architecture had a larger influence on the validation metrics over preprocessing techniques and analytical methods. PLSR models were more influenced by the analytical method, whereas the preprocessing technique influenced random forest and cubist more. Cubist models combined with CRR minimized the accuracy differences resulting from the employed SOC analytical methods. However, this combination resulted in overfitted model and high uncertainty on predictions. PLSR presented a more consistent performance than random forest and cubist. Overall, SOC data produced using different analytical methods in a training dataset significantly affected the prediction reliability, capability, and assessment. These results will be useful either to guide the analytical method selection for new projects or to manage already available databases. Besides that, they highlight the need for transparent and precise documentation over spectroscopy modeling to enable a fair comparison between publications. O desenvolvimento de grandes bancos de dados geralmente implica a combinação de dados coletados para diferentes propósitos sob diferentes padrões e metodologias, o que muitas vezes leva os bancos de dados a sofrer com dados de solo díspares e inconsistentes. Apesar do potencial da espectroscopia de infravermelho próximo e visível (Vis- NIR) para prever o carbono orgânico do solo (COS) a partir desses bancos de dados, a eficácia e a consistência entre os métodos analíticos usados para produzir os dados alvo raramente são discutidos. O objetivo principal desta pesquisa foi investigar a interação entre as técnicas de pré-processamento, arquiteturas de modelo e, especialmente, os métodos analíticos usados para produzir os dados alvo do COS. Para alcançá-lo, estabelecemos dois objetivos específicos: i) avaliar a interação entre métodos analíticos, técnicas de pré-processamento e arquiteturas de modelo nas predições de COS, ii) avaliar se essa interação pode ser traduzida em alguma forma de hierarquia entre as métricas de validação. Nesta tese de doutorado, dois capítulos abordam o tema onde os objetivos acima mencionados foram alcançados. O Capítulo I apresenta como mudanças no método analítico (seco (COSDC) e combustão úmida com quantificação por titulometria (COSWCt) e colorimetria (COSWCc)) e as técnicas de pré-processamento (suavização (SMO), continuum remoção (CRR) e primeira derivada de Savitzky-Golay (SGD) afetam a relação empírica capturada por diferentes arquiteturas de modelos (random forest, cubist e regressão de mínimos quadrados parciais (PLSR)). Métricas de validação cruzada foram usadas para comparar o desempenho paralelo de 27 modelos preditivos. A relação entre a matriz de covariável e os dados alvo é explorada com base na importância da variável. O Capítulo II mostra como a interação entre esses três fatores pode ser traduzida em uma hierarquia. Uma técnica de reamostragem foi usada para dividir o conjunto de dados em conjuntos de treinamento e validação 100 vezes para atingir desempenhos realistas e explorar como o desempenho preditivo mudou conforme o conjunto de treinamento mudou. A análise da árvore de inferência condicional foi realizada para avaliar como esses três fatores influenciaram as métricas de validação global. O desempenho preditivo em ambos os estudos variou dependendo do método analítico COS, da técnica de pré-processamento e da arquitetura do modelo empregada. Dentre os três métodos analíticos testados, DC e WCt proporcionaram maior correlação entre COS e espectros do que WCc e, portanto, resultaram em melhor desempenho dos modelos. A arquitetura do modelo teve maior influência nas métricas de validação sobre as técnicas de pré-processamento e métodos analíticos. Os modelos PLSR foram mais influenciados pelo método analítico, enquanto a técnica de pré-processamento influenciou mais os modelos random forest e cubist. Modelos cubists combinados com CRR minimizaram as diferenças de precisão resultantes dos métodos analíticos de COS empregados. No entanto, essa combinação resultou em um modelo sobreajustado e alta incerteza nas previsões. PLSR apresentou desempenho mais consistente do que random forest e cubist. No geral, os dados do COS produzidos com diferentes métodos analíticos em um conjunto de dados de treinamento afetaram significativamente a confiabilidade, capacidade e avaliação das predições. Esses resultados serão úteis para orientar a seleção de métodos analíticos para novos projetos ou para gerenciar bancos de dados já disponíveis. Além disso, eles destacam a necessidade de documentação transparente e precisa sobre a modelagem espectroscópica para permitir uma comparação justa entre as publicações.

Agriculture employs increasingly innovative techniques in search of optimize inputs, maximize profitability, and reduce environmental impact. An example of this is the emergence of Agriculture 4.0, in which sensors collect information through Proximal Soil Sensing. These methods, called photon-based methods, employ different electromagnetic radiation wavelengths to measure soil attributes and properties in situ or ex situ. National and international research institutions have produced knowledge and contributed to the training of professionals able to apply these new approaches in soil science. In this context, this review aimed to produce a synthesis of the main proximal soil sensing techniques and made it accessible to students, technicians, and researchers. A agricultura emprega técnicas cada vez mais inovadoras na busca por otimizar insumos, maximizar a lucratividade e diminuir o impacto ambiental. Exemplo disso é o despontar da agricultura 4.0, na qual sensores coletem informações através do Sensoriamento Proximal do Solo. Esses métodos, chamados photon-based methods, empregam distintos comprimentos de onda da radiação eletromagnética para mensurar atributos e propriedades do solo in-situ ou ex-situ. Instituições nacionais de pesquisa têm produzido conhecimento relevante e contribuído para a formação de profissionais aptos a aplicar essas novas abordagens em ciência do solo. Nesse contexto, esta revisão bibliográfica teve como objetivo verter as principais técnicas de sensoriamento proximal em uma síntese acessível para estudantes, técnicos e pesquisadores.

A agricultura emprega técnicas cada vez mais inovadoras na busca por otimizar insumos, maximizar a lucratividade e diminuir o impacto ambiental. Exemplo disso é o despontar da agricultura 4.0, na qual sensores coletem informações através do Sensoriamento Proximal do Solo. Esses métodos, chamados photon-based methods, empregam distintos comprimentos de onda da radiação eletromagnética para mensurar atributos e propriedades do solo in-situ ou ex-situ. Instituições nacionais de pesquisa têm produzido conhecimento relevante e contribuído para a formação de profissionais aptos a aplicar essas novas abordagens em ciência do solo. Nesse contexto, esta revisão bibliográfica teve como objetivo verter as principais técnicas de sensoriamento proximal em uma síntese acessível para estudantes, técnicos e pesquisadores.

The relationship between visible-near-infrared (Vis-NIR-SWIR) spectra and soil organic carbon (SOC) and the effects of preprocessing techniques on SOC predictive models have been shown in several studies. However, little attention has been given to the effect of analytical methods used to produce the SOC data used to calibrate those models. The predictive performance of Vis-NIR spectral models depends not only on the preprocessing technique and machine learning method but also on the analytical method employed to produce the SOC data. Our hypothesis is that some combinations of preprocessing and models may be more sensitive to laboratory (measurement) error than others. To test this hypothesis, we evaluated the leave-one-out cross-validation performance of three predictive models (Random Forest (RF), Cubist, and Partial Least Square Regression (PLSR)) calibrated using SOC data produced via three analytical methods (dry combustion (DC) and wet combustion with quantification by titration (WCt) and colorimetry (WCc)) and three Vis-NIR spectra preprocessing techniques (smoothing (SMO), continuum removal (CRR), and Savitzky-Golay first derivative (SGD)). The prediction performance varied among the models. DC and WCt provided a higher correlation between SOC and spectra than WCc, and thus, resulted in higher accuracy. The Cubist+CRR was ranked the best performing model, with an average of R2 = 0.81 and RMSE = 0.81% among analytical methods. Cubist+CRR also minimized the accuracy differences resulting from SOC analytical methods employed. The RF model had low accuracy and was unable to explain more than 46% of the variance. Overall, the analytical method significantly affects SOC predictions, and its impact may be larger than the preprocessing.

The Brazilian agricultural business and industry are committed to continue to increase their participation in the global biomass produc- tion. The focus on fertilizers, not considering other soil functions, is changing to a new, more holistic paradigm. Therefore, the productivity of Brazilian agriculture can still be increased by improving and adopting appropriate soil management practices for each biome, and encom- passing environment protection. We need to develop integrated man- agement systems for land uses and conservation adapted to the variability of soils. The lack of diffusion of information and detailed soil mapping in the country brings about significant differences in the development of these topics. The complexity of the Brazilian biomes leads to a high level of uncertainty regarding soil response to agricul- tural interventions, and also impacts on carbon and microbiological soil dynamics. Monitoring impacts and conservation of these biomes is crucial, their soils are rich in ecobiosystems and their degradation will have negative impacts on climate, water and soils, local and globally. Despite the importance of all disciplines, at first, to normalize the issues it is absolutely necessary to have detailed soil survey information to make the best decision on all areas, from management and planning of agriculture to the conservation of ecosystems.

Diffuse reflectance spectroscopy (DRS) has the potential to predict soil organic carbon (SOC). However, it is still little used as a matter of routine in soil laboratories in Brazil. The objective of this study was to make evaluations as to whether SOC predicted by spectral techniques can replace measurement by routine chemical methods with no loss in quality and be applied in the recommendation of nitrogen fertilizer as well as identifying the best prediction strategies to use. A data set containing 2,471 samples from six soil spectral libraries (SSL) was used to develop spectroscopic models for SOC content prediction, including consideration of sample stratification and preprocessing techniques. The SOC was quantified through the analytical-chemical methods of wet combustion with determination by titration, designated as the reference method (REM), and colorimeter, designated as the routine method (ROM in an independent data set). SOC contents predicted by the spectral analysis method (SAM) were compared to the REM and ROM results, converted to soil organic matter (SOM) and used for N recommendations. The best estimate for SOM content using the SAM was achieved through stratification of the SSL and application of the standard normal variate (SNV) preprocessing. The SOC predicted by spectral techniques proved capable of replacing the SOC measured by routine chemical methods with no loss of quality and supported by an appropriate nitrogen fertilizer recommendation, provided the models met the conditions and possessed the characteristics of the samples to be analyzed.

Based on the premise that the modeling and mining of soil and environmental data are capable of generating useful spatial information for use and planning in the forestry supply chain, we have established the following objectives: i) predict soil depth (SoD) in a topographically complex landscape via Digital Soil Mapping (DSM), ii) evaluate the potential of incorporating spatial data on SoD and topographic attributes in the prediction of the height of 30-year old Pinus taeda L., and iii) assess whether the global predictions of depth to bedrock (DTB) from Soil Grids is as effective as the local predictions for use in silviculture. The study was conducted in a 1.08-km2 Pinus taeda L. forest, in first rotation, 30 years old, and in the mountain region of Santa Catarina, Brazil. The dendometric (tree height) and pedologic (SoD) data were measured at 102 points and used to train random forest (RF) models by leave-one-out cross-validation (LOOCV). Nine topographic covariates derived from a digital elevation model were used to spatially predict SoD. For spatial prediction of tree height, the models were trained using three set of covariates: 1) four topographic covariates (model 1), 2) SoD map predicted by RF plus four topographic covariates (model 2), and 3) DBT plus four topographic covariates (model 3). The RF model could adequately describe SoD and the general characteristics of the distribution of data measured in a landscape with complex topography using terrain attributes as covariates. The model obtained R2 = 0.91 and RMSE = 0.17 m. The tree height was predicted with R2 up to 0.93 and RMSE = 0.82 m. SoD and elevation were the most important covariates for it. The SoD covariate stood out compared to the others, improving the fit of model 2, while DBT was not considered important in model 3. Our results showed that SoD played a critical role to predict the tree height. However, local predictions of SoD are needed to obtain accurate predictions of tree height. These products, generated by DSM, showed to be useful for establishing methodologies to guide the long-term soil and forest management practices.

The aims of this study were to analyze the pedological and topographic variables of two forest fragments in different stages of regeneration and verify the relationship with the physiognomic and community structure of the vegetation. The study was developed in the Edge of the Plateau of Rio Grande do Sul state (RS), Brazil. Ten plots of 10 x 20 m (0.46 ha) were established in a secondary forest fragment (FS) and 9 plots (0.18 ha) in the capoeira fragment (CA), using a systematic sampling method. It was used vegetation survey and measured pedological and topographic variables. The variables were selected by Principal Component Analysis (ACP), separated into groups by cluster analysis (AA) and compared to phytosociological variables through canonical correspondence analysis (ACC). The greatest species richness was identified in soils with higher nutritional quality, where the variables with the greatest influence on the distribution of the species were: depth, sand, pH and red (IV) index. On the contrary, in CA, the greatest species richness was associated to the more acidic and nutrient poor soils in which the silt, altitude and IV variables showed greater influence on the species distribution. These results highlight the importance of pioneer species in this stage of forest regeneration.

Including environmental covariates, when available, can be a valuable strategy for achieving higher soil predictive performance, but it is still unknown whether environmental data should be used either as covariates, combined with Vis-NIR spectra, to predict soil organic carbon (SOC) or as criteria to stratify a soil spectral library (SSL). We hypothesized that the performance of Vis-NIR spectroscopy in predicting SOC could be improved by the inclusion of auxiliary environmental data as covariates in Cubist models and thereby overcome the data stratification limitations. To test this, we evaluated six covariate sets, in which the following covariates were combined: spectral data, spectral classes, pedological data, and environmental data calibrated using Cubist models. An SSL composed of 2,461 samples from southern Brazil was used to calibrate models considering different sets of covariates. Model 1 included Vis-NIR reflectance (350–2500 nm), and Model 2 included Vis-NIR reflectance and spectral class. Model 3 included physiographic region, land-use/land-cover (LULC), climate, parent material, elevation, and clay content, while Model 4 included the parameters from Model 3 with the addition of Vis-NIR reflectance. Model 5 included Vis-NIR reflectance, spectral class, physiographic region, LULC, and soil textural class, while Model 6 included Vis-NIR reflectance, spectral class, physiographic region, LULC, climate, parent material, elevation, and clay content. The inclusion of environmental data as covariates along with Vis-NIR improved the predictive performance for SOC. Among the six covariate sets tested, the set including all covariates (Model 6) showed the best performance, improving the accuracy in prediction by 12% and reducing the prediction error (RMSE) by 22% compared to Model 1. Model 6 was the most accurate, and the input of environmental data as covariates is a promising strategy to achieve more accurate SOC predictions. Thus, covariates (e.g. elevation, clay content) that correlate with SOC improved the prediction accuracy of Vis-NIR spectroscopy models. The Cubist model was able to achieve similar accuracy and overcome the limitations presented by the stratification strategy documented in the literature by preventing the reduction of the sample size in the calibration of the models.

This study aimed to describe and compare the floristic composition and structure of tree and shrub vegetation, as well as identifying trends in the dynamics of species in forest fragments with different stages of regeneration. An inventory was conducted in the forest fragments, adopting the method of systematic sampling with sampling units (SUs) of fixed area, with dimensions of 20 x 10 m, distributed throughout the area. 23 SUs were allocated in the secondary forest fragment and 10 SUs in the capoeira fragment, which was held the vegetation survey three classes of inclusion: R1 (seedlings regeneration): h and ? 50cm DBH < 2 cm; R2 (established regeneration): DBH between 2 and 5 cm, and A (arboreal stratum): DBH ? 5 cm. 87 species were found belonging to 36 families. The floristic composition was similar in the arboreal stratum, showing Lithrea molleoides as the species most important in both fragments. The natural regeneration of forest was characterized by Psychotria leiocarpa and regeneration of capoeira by Myrcia oblongata. The differences observed between forests showed trends of changes in the woody component according to the natural dynamics of the species in the process of secondary succession.

More accurate models for the prediction of soil organic carbon (SOC) by visible-near-infrared (Vis-NIR) spectroscopy remains a challenging task, especially when the soil spectral libraries (SSL) is composed of soils with a high pedological variation. One proposition to increase the models accuracy is to reduce the SSL variance, which can be achieved by stratifying the library into sub-libraries. Thus, the main objective of this study was to evaluate whether the stratification of a SSL by environmental, pedological and Vis-NIR spectral criteria results in greater accuracy of spectroscopic models than to general models for prediction of SOC content. The performance of the models was evaluated considering the variance of soil components and sample number. In addition, we tested the effect of two spectral preprocessing techniques and two multivariate calibration methods on spectroscopic modeling. For these purposes, a SSL composed of 2471 samples from Southern Brazil was stratified based on i) physiographic region; ii) land-use/land-cover; iii) soil texture, and iv) spectral class. Two spectral processing techniques: Savitzky-Golay - 1st derivative (SGD) and continuum removed reflectance (CRR) and two multivariate methods (partial least squares regression - PLSR and Cubist) were used to fit the models. The best performances for the global and local models were achieved with the CRR spectral processing associated with the Cubist method. The stratification of the SSL in more homogeneous sample groups by environmental criteria (physiographic regions and land-use/land-cover) improved the accuracy of SOC predictions compared to pedological (soil texture) and Vis-NIR spectral (spectral classes) criteria. The reduction in the number of samples negatively affected the performance of models for sub-libraries with high pedological and spectral variation. Stratification criteria were proposed in a flowchart to assist in decision making in future studies. Our findings suggest the importance of sample balance across environmental, pedological and spectral strata, in order to optimize SOC predictions.

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.

Soil spectroscopy (Vis-NIR-SWIR 350?2500 nm) has known potential to predict clay content and soil organic matter (SOM), which are properties that determine soil quality. It is well known that data predicted from Vis-NIR-SWIR carry prediction errors resulting from modeling process. However, these errors are not spatially correlated and, even being numerically small, they can alter the spatial correlation of the data. Thus, the aim of the study was to evaluate wether clay and SOM data predicted by spectroscopy models preserve the spatial structure of data obtained by traditional chemical analyses. This was done considering a soil-spectral dataset from a small viticultural area (3 ha) in Southern Brazil. Soil samples were collected in 74 sites, in two depths (0.00?0.20 m and 0.20?0.40 m depth). Collected in 74 sites, totaling 148 soil samples. The Vis-NIR-SWIR data were used to train spectral models in raw form, and subjected to three pre-processing techniques (smoothing - SMO; Savitzky-Golay with first derivative - SGD; Binning - BIN). In the calibration step, three machine learning techniques were tested (Cubist ?CUB; Random Forest ?RF; Partial Least Squares Regression - PLSR). The spatial correlation of the data measured by traditional wet chemistry, as well the data predicted by spectroscopy, was analyzed clay and SOM maps were then generated by ordinary kriging for both data. Among the spectral models, the RF + SGD model presented the best cross-validation performance for clay, with R2cv = 0.95 and RMSEcv = 1.06%, and the PLSR + SGD for SOM, with R2cv = 0.98 and RMSEcv = 0.07%. The data predicted by Vis-NIRSWIR spectroscopy preserved the spatial structure of the data obtained by the traditional wet chemical analysis. However, our results suggest that spectral modeling was more effective in predicting SOM while spatial interpolation was more effective in predicting clay. Data predicted by Vis-NIR-SWIR spectroscopy with lower accuracy increased the residuals of the spatial predictions. The clay and SOM maps prodiced with spectroscopy predictions show similarities and feasible accuracy for the use and management of agricultural soils in vineyard areas. Made available in DSpace on 2022-03-03T12:03:59Z (GMT). No. of bitstreams: 1 1-s2.0-S2352009422000062-main.pdf: 3136849 bytes, checksum: 91920296772b2c263955f7c17b701363 (MD5) Previous issue date: 2022