Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / The traditional soil analysis many techniques are used in order to determine the physical and chemical properties of the soil. The radiometry appears as a promising alternative technique in the analysis of soil properties. This technique has demonstrated great potential for identification and quantification of certain properties of the soil. It is a non-destructive and non-polluting tool, with the ability to collect data on large spatial dimensions with relative speed. The radiometry may in cases be simpler than the traditional analysis of the soil and on various occasions, more accurately. The main objective of this study was to determine pedotransfer functions to soil properties based on radiometric data. It was observed that the heterogeneity of the soil decreases the accuracy of the models, however it was possible to construct prediction functions for the content of sand, silt, clay and soil organic matter from the radiometry with a level of prediction models acceptable. Considering that, in the prediction of soil properties using radiometry, there is no reagents spending and less resource invested beyond the analysis time shorter than the traditional analysis, the results were promising. / Na análise tradicional do solo muitas técnicas são utilizadas na tentativa de determinar suas propriedades físicas e químicas. A radiometria aparece como uma técnica alternativa e promissora na análise de propriedades do solo. Essa técnica tem demonstrado grande potencial na identificação e quantificação de determinadas propriedades do solo. Trata-se, de uma ferramenta não destrutiva, não poluidora, com capacidade de coleta de dados em grandes dimensões espaciais com relativa velocidade. A radiometria pode, em muitos casos, ser mais simples do que a análise tradicional do solo e em várias ocasiões, mais precisa. O principal objetivo desse trabalho foi determinar funções de pedotransferência para as propriedades do solo tendo como base os dados da radiometria. Observou-se que a heterogeneidade do solo diminui a precisão dos modelos, porém foi possível construir funções de predição para o teor de argila, areia, silte e matéria orgânica do solo a partir da radiometria com um nível de predição dos modelos aceitável. Considerando que, na predição das propriedades do solo utilizando a radiometria, não há gastos com reagentes e menos recursos investidos além do tempo de análise menor que a análise tradicional, os resultados apresentados foram promissores.

Submitted by Jorge Silva (jorgelmsilva@ufrrj.br) on 2017-07-18T18:30:23Z No. of bitstreams: 1 2015 - Helena Saraiva Koenow Pinheiro.pdf: 14533188 bytes, checksum: 58cff5581549af698fe42ba33bd8aa71 (MD5) / Made available in DSpace on 2017-07-18T18:30:23Z (GMT). No. of bitstreams: 1 2015 - Helena Saraiva Koenow Pinheiro.pdf: 14533188 bytes, checksum: 58cff5581549af698fe42ba33bd8aa71 (MD5) Previous issue date: 2015-07-30 / CAPES / CNPq / FAPERJ / Quantitative soil-landscape models represent a new trend in soil surveys. In this regard, the various digital mapping techniques are applied to predict the natural patterns of occurrence of soil types. The objective of this study was to apply digital mapping techniques to predict soil classes and attributes in a watershed, with wide range of landscape conditions, in Rio de Janeiro State, in Brazil. The approach was based on tacit soil knowledge, regarding the choice of landscape attributes that represent the variability of soil-forming factors in the region. In regard to construct the predictive models, terrain variables were generate from the digital elevation model, geology map and remote sensing data. Ten terrain attributes were created on softwareArcGIS Desktop v. 10, such as altimetry, slope, curvature, parental material map, topographic compound index and euclidean distance of hydrography. In the software ERDAS Imagine v.9 were generated three indices derived from remote sensing data (Landsat 5 TM). They are: clay minerals, iron oxide and vegetation index normalized difference - NDVI. To represent the landscape forms was generated map the "geomorphons" maps, the GRASS-GIS program. To provide enough datato predict soil properties, additional terrain variables were derived from a digital elevation model (DEM) generated in the software SAGA-GIS. The work development was organized into three steps, presented as chapters. The first chapter comprised bibliography review and presents the context of the study. The detailed analysis of soil-landscape relationships, considering the variability of environmental attributes and characteristics of pedo-enviroments are performed on the second chapter. The predominant soils in the area were Ferralsols, Acrisols, Gleysols, Cambissolos, Fluvisols and Regosols. The third chapter presented the application of the landform maps (?geomorphons?) as a covariate to pretic soil classes by neural network approach. The fourth chapter targets the application of trees-based models (decision trees and random forest) to predict soil classes. The evaluation of the inferred products to represent the soil classes was performed based on statistical indices (kappa, overall), generalization of soil classes and validation with control samples. The best performance was observed for the random forest model that showed better values to statistical indices and better generalization of mapping units. The fifth chapter comprised the prediction of soil texture components on topsoil layer by using multiple linear regressions and regression trees. The analyses indicated better performance by using regression trees algorithm to all soil attributes (sand, silt, and clay), independent of the database (harmonized or original). All predictive models were implemented in R software. Additional research is needed to select an appropriated set of predictive covariates; as so, collect more soil samples to use as input to models and also validate of the final products. Soil survey research is important in the actual context once can enhance the information generated by the soil surveys, as well as to obtain useful information to the final users, as example of the maps that represent the spatial variability of soil texture components. / Modelos solo-paisagem quantitativos representam uma nova tend?ncia nos levantamentos de solos. Neste sentido, as diferentes t?cnicas de mapeamento digital s?o aplicadas para prever os padr?es naturais de ocorr?ncia de classes de solo. O objetivo deste trabalho foi a aplica??o de geotecnologias no mapeamento de classes e atributos dos solos em uma bacia hidrogr?fica, que apresenta grande varia??o de condi??es de paisagem, no Estado do Rio de Janeiro, Brasil. A abordagem foi baseada em conhecimento pedol?gico t?cito, culminando na escolha de atributos da paisagem que representem a variabilidade dos fatores de forma??o de solos na regi?o. Na constru??o do modelo solo-paisagem foram gerados no programa de computa??o ArcGIS Desktop v. 10, atributos relacionados a pedog?nese na ?rea em estudo, como geologia altimetria, declividade, curvatura, ?ndice topogr?fico composto e dist?ncia euclidiana de hidrografia. No programa ERDAS Imagine v.9 foram gerados tr?s ?ndices derivados de dados de sensoriamento remoto (Landsat 5 TM). S?o eles: clay minerals, iron oxide e ?ndice de vegeta??o por diferen?a normalizada ? NDVI. Para representar as formas do relevo foi gerado mapa com as dez formas mais comuns do relevo (?geomorphons?), no programa GRASS-GIS. Adicionalmente, a predi??o de atributos do solo contou com co-vari?veis derivadas do modelo digital de eleva??o (MDE) geradas no programa SAGA-GIS. O trabalho de tese foi dividido em etapas, apresentadas na forma de cap?tulos. O primeiro cap?tulo apresenta a revis?o de literatura espec?fica de contextualiza??o do trabalho. O estudo das rela??es solo-paisagem e da variabilidade dos atributos do terreno, a caracteriza??o das unidades de mapeamento com base no levantamento de campo, constituem o segundo cap?tulo. Os solos predominantes na ?rea foram: Latossolos, Argissolos, Gleissolos, Cambissolos, Neossolos Fl?vicos e Lit?licos. O terceiro cap?tulo tratou do uso do mapa de formas da paisagem (?geomorphons?) como vari?vel preditora para o mapeamento de classes de solos, por abordagem de redes neurais artificiais. O quarto cap?tulo teve como objetivo a aplica??o de modelos baseados em ?rvores (?rvores de decis?o e random forest) para a predi??o de classes de solos. A avalia??o dos produtos inferidos para classes de solos foi baseada em ?ndices estat?sticos (kappa, exatid?o global), generaliza??o das classes de solos e valida??o com amostras de controle. O melhor desempenho foi observado para o modelo random forest que apresentou valor superior para os ?ndices estat?sticos e melhor generaliza??o das unidades de mapeamento. O quinto cap?tulo compreendeu a predi??o da composi??o da textura na camada superficial do solo atrav?s de regress?es lineares m?ltiplas e ?rvores de regress?o. As an?lises indicaram desempenho superior do algoritmo de ?rvores de regress?o, para todos os atributos testados (areia, silte, argila), utilizando dados harmonizados ou originais. Todos os modelos preditivos foram aplicados no programa R. An?lises adicionais s?o necess?rias para ajudar a definir conjunto de co-vari?veis preditoras adequado, assim como a coleta de mais amostras de solo, tanto para o processo de modelagem como para valida??o dos produtos. Trabalhos dessa natureza s?o importantes no contexto global de melhor aproveitamento das informa??es geradas em levantamento de solos, assim como para obten??o de mapas de car?ter pr?tico, como ? o caso da distribui??o espacial de atributos dos solos.

Indiana University-Purdue University Indianapolis (IUPUI) / Up-to-date and accurate information on soil properties is important for precision farming and environmental management. The spatial information of soil properties allows adjustments of fertilizer applications to be made based on knowledge of local field conditions, thereby maximizing agricultural productivity and minimizing the risk of environmental pollution. While conventional soil sampling procedures are labor-intensive, time-consuming and expensive, remote sensing techniques provide a rapid and efficient tool for mapping soil properties. This study aimed at examining the capacity of hyperspectral reflectance data for mapping soil organic matter (SOM), total nitrogen (N) and total phosphorus (P). Soil samples collected from Eagle Creek Watershed, Cicero Creek Watershed, and Fall Creek Watershed were analyzed for organic matter content, total N and total P; their corresponding spectral reflectance was measured in the laboratory before and after oven drying and in the field using Analytical Spectral Devices spectrometer. Hyperion images for each of the watersheds were acquired, calibrated and corrected and Hyperion image spectra for individual sampled sites were extracted. These hyperspectral reflectance data were related to SOM, total N and total P concentration through partial least squares (PLS) regressions. The samples were split into two datasets: one for calibration, and the other for validation. High PLS performance was observed during the calibration for SOM and total N regardless of the type of the reflectance spectra, and for total P with Hyperion image spectra. The validation of PLS models was carried out with each type of reflectance to assess their predictive power. For laboratory reflectance spectra, PLS models of SOM and total N resulted in higher R2 values and lower RMSEP with oven-dried than those with field-moist soils. The results demonstrate that soil moisture degrades the performance of PLS in estimating soil constituents with spectral reflectance. For in-situ field spectra, PLS estimated SOM with an R2 of 0.74, N with an R2 of 0.79, and P with an R2 of 0.60. For Hyperion image spectra, PLS predictive models yielded an R2 of 0.74 between measured and predicted SOM, an R2 of 0.72 between measured and predicted total N, and an R2 of 0.67 between measured and predicted total P. These results reveal slightly decreased model performance when shifting from laboratory-measured spectra to satellite image spectra. Regardless of the spectral data, the models for estimating SOM and total N consistently outperformed those for estimating total P. These results also indicate that PLS is an effective tool for remotely estimating SOM, total N and P in agricultural soils, but more research is needed to improve the predictive power of the model when applied to satellite hyperspectral imagery.

Yes / Archaeological research at Stonehenge (UK) is increasingly aimed at understanding the dynamic of the wider archaeological landscape. Through the application of state-of-the-art geophysical techniques, unprecedented insight is being gathered into the buried archaeological features of the area. However, applied survey techniques have rarely targeted natural soil variation, and the detailed knowledge of the palaeotopography is consequently less complete. In addition, metallic topsoil debris, scattered over different parts of the Stonehenge landscape, often impacts the interpretation of geophysical datasets. The research presented here demonstrates how a single multi-receiver electromagnetic induction (EMI) survey, conducted over a 22 ha area within the Stonehenge landscape, offers detailed insight into natural and anthropogenic soil variation at Stonehenge. The soil variations that were detected through recording the electrical and magnetic soil variability, shed light on the genesis of the landscape, and allow for a better definition of potential palaeoenvironmental and archaeological sampling locations. Based on the multi-layered dataset, a procedure was developed to remove the influence of topsoil metal from the survey data, which enabled a more straightforward identification of the detected archaeology. The results provide a robust basis for further geoarchaeological research, while potential to differentiate between modern soil disturbances and the underlying sub-surface variations can help in solving conservation and management issues. Through expanding this approach over the wider area, we aim at a fuller understanding of the human–landscape interactions that have shaped the Stonehenge landscape.

No description available.

Soil is one of the most precious resources on Earth because of its role in storing and recycling water and nutrients essential for life, providing a variety of ecosystem services. This vulnerable resource is at risk from degradation by erosion, salinity, contamination and other effects of mismanagement. Information from soil is therefore crucial for its sustainable management. While the demand for soil information is growing, the quantity of data collected in the field is reducing due to financial constraints. Digital Soil Mapping (DSM) supports the creation of geographically referenced soil databases generated by using field observations or legacy data coupled, through quantitative relationships, with environmental covariates. This enables the creation of soil maps at unexplored locations at reduced costs. The selection of an optimal scale for environmental covariates is still an unsolved issue affecting the accuracy of DSM. The overall aim of this research was to explore the effect of spatial scale alterations of environmental covariates in DSM. Three main targets were identified: assessing the impact of spatial scale alterations on classifying soil taxonomic units; investigating existing approaches from related scientific fields for the detection of scale patterns and finally enabling practitioners to find a suitable scale for environmental covariates by developing a new methodology for spatial scale analysis in DSM. Three study areas, covered by detailed reconnaissance soil survey, were identified in the Republic of Ireland. Their different pedological and geomorphological characteristics allowed to test scale behaviours across the spectrum of conditions present in the Irish landscape. The investigation started by examining the effects of scale alteration of the finest resolution environmental covariate, the Digital Elevation Model (DEM), on the classification of soil taxonomic units. Empirical approaches from related scientific fields were subsequently selected from the literature, applied to the study areas and compared with the experimental methodology. Wavelet analysis was also employed to decompose the DEMs into a series of independent components at varying scales and then used in DSM analysis of soil taxonomic units. Finally, a new multiscale methodology was developed and evaluated against the previously presented experimental results. The results obtained by the experimental methodology have proved the significant role of scale alterations in the classification accuracy of soil taxonomic units, challenging the common practice of using the finest available resolution of DEM in DSM analysis. The set of eight empirical approaches selected in the literature have been proved to have a detrimental effect on the selection of an optimal DEM scale for DSM applications. Wavelet analysis was shown effective in removing DEM sources of variation, increasing DSM model performance by spatially decomposing the DEM. Finally, my main contribution to knowledge has been developing a new multiscale methodology for DSM applications by combining a DEM segmentation technique performed by k-means clustering of local variograms parameters calculated in a moving window with an experimental methodology altering DEM scales. The newly developed multiscale methodology offers a way to significantly improve classification accuracy of soil taxonomic units in DSM. In conclusion, this research has shown that spatial scale analysis of environmental covariates significantly enhances the practice of DSM, improving overall classification accuracy of soil taxonomic units. The newly developed multiscale methodology can be successfully integrated in current DSM analysis of soil taxonomic units performed with data mining techniques, so advancing the practice of soil mapping. The future of DSM, as it successfully progresses from the early pioneering years into an established discipline, will have to include scale and in particular multiscale investigations in its methodology. DSM will have to move from a methodology of spatial data with scale to a spatial scale methodology. It is now time to consider scale as a key soil and modelling attribute in DSM.

A escassez de informações do solo que permitam o seu uso adequado, seja para fins agrícola, ambiental ou de projeto urbanos, pode ser minimizada com soluções provenientes do desenvolvimento de novas tecnologias. Nesse sentido, o presente estudo teve como objetivo aplicar duas estratégias digitais para obtenção de mapas de solos preliminares em áreas onde não foram realizados levantamentos pedológicos convencionais. As estratégias foram executadas com base em variáveis ambientais que estabelecem relações entre ocorrência de solos e suas posições na paisagem. A área de estudo compreendeu o município de Barra Bonita-SP, totalizando 11.072 ha. Para uso na predição dos solos pela técnica de Redes Neurais Artificiais (RNA) foram utilizadas as variáveis: declividade, elevação, perfil de curvatura, plano de curvatura e índice de convergência derivados de um Modelo Digital de Elevação (MDE), além das informações de geologia e das superfícies geomórficas identificadas na região. Na primeira estratégia, por meio de uma análise de agrupamento (Fuzzy k-médias) das variáveis, foram escolhidas cinco áreas chaves distribuídas na área de estudo, nas quais foi realizado levantamento de solos de nível semidetalhado para reconhecimento das unidades de mapeamento. Na estratégia 2, elaborou-se um mapa de solos de nível detalhado a partir de dados pré-existentes de apenas uma área chave, localizada no centro da região. Com a identificação das unidades de mapeamento foram gerados arquivos de treinamento e testes das redes neurais. Utilizou-se o simulador JavaNNS e o algoritmo de aprendizado backpropagation. Conjuntos de variáveis ambientais foram testados, avaliando a importância de cada variável na discriminação dos solos. A rede que exibiu melhor desempenho do índice Kappa foi utilizada para generalização de suas informações, obtendo os mapas digitais de solos. Pela aplicação de tabulação cruzada foram analisadas as correspondências espaciais entre os mapas digitais e um mapa convencional nível semidetalhado da região. Foram coletados pontos de referência para validar o desempenho dos mapas digitais. De acordo com a posição na paisagem e material de origem subjacente, notou-se tendência na ocorrência das classes de solos nas áreas chaves mapeadas. A mesma disposição dos solos foi observada nas classificações digitais. Os atributos do terreno elevação e declividade exibiram maior influência na distinção entre os solos pelas redes neurais em ambas as estratégias. A comparação com pontos de referência mostrou que o mapa digital produzido com base em unidades de mapeamento provenientes de abordagem convencional detalhada teve um desempenho superior (81,8% de concordância) ao mapa baseado em levantamento pedológico de nível semidetalhado (72,7%). Este estudo mostrou que a obtenção de mapas digitais de solos, com uso de variáveis ambientais que expressem a relação solo-paisagem, pode contribuir para a geração de informações preliminares do solo em locais não mapeados, a partir de unidades de mapeamento obtidas em áreas adjacentes. / The scarcity of land information to enable its proper use, whether for agricultural, environmental and urban design, can be minimized by solutions from the development of new technologies. Accordingly, this study aimed to apply two strategies to obtain digital maps of soil in areas where no preliminary surveys were carried out conventional pedological. The strategies were implemented based on environmental variables that establish relations between the occurrence of soils and their positions in the landscape. The study area comprised the municipality of Barra Bonita, SP, totaling 11,072 ha. For use in the prediction of soil by the technique of Artificial Neural Networks (ANN) were used variables: slope, elevation, profile curvature, plan curvature and convergence index derived from a Digital Elevation Model (DEM), in addition to information geology and geomorphic surfaces identified in the region. In the first strategy, through a cluster analysis (Fuzzy k-means) of variables, we selected five key areas distributed in the study area, soil survey being conducted semi-detailed level at these sites for recognition of the map units. In strategy 2, a map was drawn up detailed level of soil from pre-existing data of only one key area, located in the center of the region. Identifying the map units were generated files for training and testing of neural networks. Was used the simulator JavaNNS and learning algorithm \"backpropagation. Sets environmental variables were tested by assessing the importance of each variable to predict soil. The network showed better performance for the Kappa index was used to generalize their information, obtaining the digital soil maps. By applying cross tabulation analyzed the spatial correspondence between the digital maps and a conventional map of the region. Reference points were collected to validate the performance of digital maps. According to the position in the landscape and the underlying source material, was noticed a tendency of occurrence of soil classes in key areas mapped. The same arrangement was observed in the soil classifications digital. The attributes of the terrain elevation and slope exhibited a greater influence on the distinction between the soil by the neural networks in both strategies. The comparison with reference points showed that the digital map produced based on mapping units from the conventional approach detailed outperformed (81.8% agreement) to the map based on pedological survey of semi-detailed level (72.7 %). This study showed that to obtain digital maps of soils, use of environmental variables that express the soillandscape relationship, may contribute to the generation of information preeliminares soil in areas not mapped from map units obtained from adjacent areas.

A escassez de informações do solo que permitam o seu uso adequado, seja para fins agrícola, ambiental ou de projeto urbanos, pode ser minimizada com soluções provenientes do desenvolvimento de novas tecnologias. Nesse sentido, o presente estudo teve como objetivo aplicar duas estratégias digitais para obtenção de mapas de solos preliminares em áreas onde não foram realizados levantamentos pedológicos convencionais. As estratégias foram executadas com base em variáveis ambientais que estabelecem relações entre ocorrência de solos e suas posições na paisagem. A área de estudo compreendeu o município de Barra Bonita-SP, totalizando 11.072 ha. Para uso na predição dos solos pela técnica de Redes Neurais Artificiais (RNA) foram utilizadas as variáveis: declividade, elevação, perfil de curvatura, plano de curvatura e índice de convergência derivados de um Modelo Digital de Elevação (MDE), além das informações de geologia e das superfícies geomórficas identificadas na região. Na primeira estratégia, por meio de uma análise de agrupamento (Fuzzy k-médias) das variáveis, foram escolhidas cinco áreas chaves distribuídas na área de estudo, nas quais foi realizado levantamento de solos de nível semidetalhado para reconhecimento das unidades de mapeamento. Na estratégia 2, elaborou-se um mapa de solos de nível detalhado a partir de dados pré-existentes de apenas uma área chave, localizada no centro da região. Com a identificação das unidades de mapeamento foram gerados arquivos de treinamento e testes das redes neurais. Utilizou-se o simulador JavaNNS e o algoritmo de aprendizado backpropagation. Conjuntos de variáveis ambientais foram testados, avaliando a importância de cada variável na discriminação dos solos. A rede que exibiu melhor desempenho do índice Kappa foi utilizada para generalização de suas informações, obtendo os mapas digitais de solos. Pela aplicação de tabulação cruzada foram analisadas as correspondências espaciais entre os mapas digitais e um mapa convencional nível semidetalhado da região. Foram coletados pontos de referência para validar o desempenho dos mapas digitais. De acordo com a posição na paisagem e material de origem subjacente, notou-se tendência na ocorrência das classes de solos nas áreas chaves mapeadas. A mesma disposição dos solos foi observada nas classificações digitais. Os atributos do terreno elevação e declividade exibiram maior influência na distinção entre os solos pelas redes neurais em ambas as estratégias. A comparação com pontos de referência mostrou que o mapa digital produzido com base em unidades de mapeamento provenientes de abordagem convencional detalhada teve um desempenho superior (81,8% de concordância) ao mapa baseado em levantamento pedológico de nível semidetalhado (72,7%). Este estudo mostrou que a obtenção de mapas digitais de solos, com uso de variáveis ambientais que expressem a relação solo-paisagem, pode contribuir para a geração de informações preliminares do solo em locais não mapeados, a partir de unidades de mapeamento obtidas em áreas adjacentes. / The scarcity of land information to enable its proper use, whether for agricultural, environmental and urban design, can be minimized by solutions from the development of new technologies. Accordingly, this study aimed to apply two strategies to obtain digital maps of soil in areas where no preliminary surveys were carried out conventional pedological. The strategies were implemented based on environmental variables that establish relations between the occurrence of soils and their positions in the landscape. The study area comprised the municipality of Barra Bonita, SP, totaling 11,072 ha. For use in the prediction of soil by the technique of Artificial Neural Networks (ANN) were used variables: slope, elevation, profile curvature, plan curvature and convergence index derived from a Digital Elevation Model (DEM), in addition to information geology and geomorphic surfaces identified in the region. In the first strategy, through a cluster analysis (Fuzzy k-means) of variables, we selected five key areas distributed in the study area, soil survey being conducted semi-detailed level at these sites for recognition of the map units. In strategy 2, a map was drawn up detailed level of soil from pre-existing data of only one key area, located in the center of the region. Identifying the map units were generated files for training and testing of neural networks. Was used the simulator JavaNNS and learning algorithm \"backpropagation. Sets environmental variables were tested by assessing the importance of each variable to predict soil. The network showed better performance for the Kappa index was used to generalize their information, obtaining the digital soil maps. By applying cross tabulation analyzed the spatial correspondence between the digital maps and a conventional map of the region. Reference points were collected to validate the performance of digital maps. According to the position in the landscape and the underlying source material, was noticed a tendency of occurrence of soil classes in key areas mapped. The same arrangement was observed in the soil classifications digital. The attributes of the terrain elevation and slope exhibited a greater influence on the distinction between the soil by the neural networks in both strategies. The comparison with reference points showed that the digital map produced based on mapping units from the conventional approach detailed outperformed (81.8% agreement) to the map based on pedological survey of semi-detailed level (72.7 %). This study showed that to obtain digital maps of soils, use of environmental variables that express the soillandscape relationship, may contribute to the generation of information preeliminares soil in areas not mapped from map units obtained from adjacent areas.

La compréhension de la répartition spatiale des sols et leur cartographie est un enjeu important tant les services écosystémiques rendus par les sols ont un rôle fondamental dans les enjeux agro-environnementaux actuels. A l’échelle nationale, les données pédologiques sont fournies via des cartographies au 1 :250 000 des types de sols (Référentiel Régional Pédologique, RRP) dont la résolution est devenue insuffisante pour répondre à ces enjeux. Placés dans un contexte de cartographie numérique des propriétés des sols à l’échelle régionale (Languedoc-Roussillon) caractérisé par une grande étendue (27 236 km²) et une faible densité de données sur les sols ( 1 observation/13.5 km2), les travaux de thèse ont eu pour objectif de réaliser une nouvelle infrastructure de données pédologiques régionale satisfaisant les spécifications édictées dans le projet international GlobalSoilMap et répondant aux besoins des utilisateurs de la région.Dans un premier temps, plusieurs approches connues de cartographie numérique des sols utilisant les diverses données pédologiques issues du RRP ont été appliquées et comparées entre elles. Les meilleurs résultats ont été obtenus par des approches de régression krigeage utilisant les profils avec analyses de sol existant dans le RRP. Pour le pH, le carbone organique et les variables de texture (argile, limon, sable) les performances de prédiction se sont avérés modérées mais suffisantes pour permettre la production de cartes informatives (R2 entre 0.2 et 0.7). En revanche les propriétés de sol avec une trop faible densité de profils et/ou variant sur des distances trop courtes (Eléments grossier, Profondeur, CEC) n’ont pu être prédites .Dans un deuxième temps, des méthodologies ont été proposées et testées pour mieux estimer les incertitudes de prédictions de propriétés de sol. Concernant les incertitudes locales, des progrès par rapport à l’utilisation de la régression krigeage ont été obtenus avec l’utilisation d’arbres de régression quantile. Ces incertitudes locales ont pu d’autre part être propagées dans les calculs d’indicateurs de sol caractérisant des entités géographiques de la région (exemple : commune). Enfin une troisième étape a été consacrée à la mise en production effective de la nouvelle infrastructure de données pédologique régionale permettant une diffusion des cartes obtenues dans cette thèse vers les utilisateurs.Les résultats de la thèse permettent de démontrer la faisabilité d’une approche de cartographie numérique des propriétés de sols à l’échelle régionale qui pourra être généralisée sur le territoire français. Bien que certains verrous méthodologiques restent à lever (ex : modèles de prédiction pour données censurées, covariable « lithologie »), la faible densité des observations pédologiques stockées actuellement en bases de données représente le facteur limitant majeur qui devra être levé dans l’avenir pour obtenir des cartes numériques de propriétés de sol à des précisions acceptables et incertitudes connues. / Depicting and mapping the soil variability is an important issue since the ecosystem services provided by soils play an important role in solving the current agro-environmental challenges. At the French national scale, the pedological data are currently provided by regional soil databases (« Référentiel Régionaux Pédologiques », RRP) at 1:250,000. However they provide soil information at a spatial resolution that is too coarse for addressing these challenges. This thesis undertakes a Digital Soil Mapping approach at the regional scale in a region (Languedoc-Roussillon) characterized by a great extent (27 236 km ²) and a low density of soil observations (1 observation/13.5 km2). The goal is to produce a new regional infrastructure of pedological data that could satisfy the specifications enacted in the international project GlobalSoilMap and that meets the needs of the local end-users. In a first step, several known approaches of digital soil mapping using the various pedological data available in the RRP were applied and compared. The best results were obtained by a regression-kriging approach using the legacy measured soil profiles of the RRP. For the pH, organic carbon and the variables of texture (clay, silt, sand) the performances of prediction were of moderate quality but sufficient to allow the production of informative maps (R2 between 0.2 and 0.7). Conversely the soil properties with a too low density of profiles and/or that varied within too short distances (coarse fragment, soil Depth, CEC) could not be predicted. In a second step, methodologies were proposed and tested for better estimating uncertainties of predictions of soil properties. Concerning local uncertainties, a progress compared to the use of Regression Kriging was obtained with the use of Quantile Regression Tree. These local uncertainties could in addition be propagated in calculations of soil indicators characterizing the geographical entities of the area (example: districts). Finally a third stage was devoted to the setting in effective production of the new regional infrastructure of pedological data, which allowed the diffusion of the maps obtained in this thesis towards the users. The results of the thesis demonstrate the feasibility of a digital soil mapping approach at the regional scale that could be generalized over the French territory. Although some methodological obstacles have to be addressed (ex: models of prediction for censored data, soil covariate “lithology”), the low density of the pedological observations currently stored in regional databases represents the major limiting factor, which will have to be addressed in the future to obtain digital maps of soil properties with acceptable and known precision.

Much older soils data and soils information lies idle in libraries and archives and is largely unused, especially in developing countries like Kenya. We demonstrated the usefulness of a stepwise approach to bring legacy soils data ‘back to life’ using the 1980 <i>Reconnaissance Soil Map of the Busia Area</i> <i>(quarter degree sheet No. 101)</i> in western Kenya as an example. Three studies were conducted by using agronomic information, field observations, and laboratory data available in the published soil survey report as inputs to several digital soil mapping techniques. In the first study, the agronomic information in the survey report was interpreted to generate 10 land quality maps. The maps represented the ability of the land to perform specific agronomic functions. Nineteen crop suitability maps that were not previously available were also generated. In the second study, a dataset of 76 profile points mined from the survey report was used as input to three spatial prediction models for soil organic carbon (SOC) and texture. The three predictions models were (i) ordinary kriging, (ii) stepwise multiple linear regression, and (iii) the Soil Land Inference Model (SoLIM). Statistically, ordinary kriging performed better than SoLIM and stepwise multiple linear regression in predicting SOC (RMSE = 0.02), clay (RMSE = 0.32), and silt (RMSE = 0.10), whereas stepwise multiple linear regression performed better than SoLIM and ordinary kriging for predicting sand content (RSME = 0.11). Ordinary kriging had the narrowest 95% confidence interval while stepwise multiple linear regression had, the widest. From a pedological standpoint, SoLIM conformed better to the soil forming factors model than ordinary kriging and had a narrower confidence interval compared to stepwise multiple linear regression. In the third study, rules generated from the map legend and map unit descriptions were used to generate a soil class map. Information about soil distribution and parent material from the map unit polygon descriptions were combined with six terrain attributes, to generate a disaggregated fuzzy soil class map. The terrain attributes were multiresolution ridgetop flatness (MRRTF), multiresolution valley bottom flatness (MRVBF), topographic wetness index (TWI), topographic position index (TPI), planform curvature, and profile curvature. The final result was a soil class map with a spatial resolution of 30 m, an overall accuracy of 58% and a Kappa coefficient of 0.54. Motivated by the wealth of soil agronomic information generated by this study, we successfully tested the feasibility of delivering this information in rural western Kenya using the cell phone-based Soil Explorer app (<a href="https://soilexplorer.net/">https://soilexplorer.net/</a>). This study demonstrates that legacy soil data can play a critical role in providing sustainable solutions to some of the most pressing agronomic challenges currently facing Kenya and most African countries.<div><p></p></div>