SÃO PAULO STATE UNIVERSITY – UNESP JABOTICABAL CAMPUS QUANTIFICATION OF LITHOPEDOGENIC IRON OXIDES BY DIFFUSE REFLECTANCE SPECTROSCOPY AND MAGNETIC SUSCEPTIBILITY FOR MAPPING PURPOSES Laércio Santos Silva Agronomic Engineer 2020 SÃO PAULO STATE UNIVERSITY – UNESP JABOTICABAL CAMPUS QUANTIFICATION OF LITHOPEDOGENIC IRON OXIDES BY DIFFUSE REFLECTANCE SPECTROSCOPY AND MAGNETIC SUSCEPTIBILITY FOR MAPPING PURPOSES Laércio Santos Silva Advisor: Prof. Dr. José Marques Júnior Co-advisor: Prof. Dr. Vidal Barrón Lopés de Torre Thesis presented to the College of Agricultural and Veterinarian Sciences – UNESP, Jaboticabal Campus, as partial fulfillment of the requirements for obtaining the title of Doctor in Agronomy (Soil Science). 2020 AUTHOR DETAILS LAÉRCIO SANTOS SILVA (soil scientist)- born in Barreiros - Pernambuco, studied Agronomic Engineering at the Federal Rural University of Pernambuco - Recife - UFRPE, from 2009 to 2014. During that time, Dr. Laércio Santos Silva developed scientific studies at the Department of Soil Science and at the Empresa Brasileira de Pesquisa Agropecuária (Embrapa - UEP Recife), in the areas of soil and water management and conservation, soil physics, fertility and mineralogy, geoprocessing of soil data, and plant improvement. He also attended a master's degree in the Graduate Programs in Agronomy (Soil Science) of University Estadual Paulista (Unesp Jaboticabal – São Paulo). His PhD was taken on soil mineralogy in the Universidade Estadual Paulista, with studies Fundação de Amparo à Pesquisa do Estado de São Paulo — FAPESP (process: 2017 / 01704-2) in the period from 2018 to 2020. He attended part of his PhD at the University of Córdoba-ES under the guidance of the renowned Doctor and Scientist Professor Vidal Barrón Lopés de Torre. Scientist Laércio Santos Silva develops studies in the following alignments: applied mineralogy, mapping, pedometry and photochemistry of NOx gases in soils. Silva, L.S “Nordestino sim, sim senhô”. "The difference between the traditional and the quantitative approach ceases to exist when technology and formal knowledge come together to produce relevant information for society". unknown author Dedicated To God in the first place, for having given me the gift of life, in addition to allowing the completion of another stage of my life and for many other achievements. ACKNOWLEDGMENTS To my parents Amauri Bernardo da Silva and Laudicéia dos Santos Silva, the reason for my life, who have tirelessly accompanied me, always providing support in difficult times and celebrating in each small and big achievement. To the Faculdade de Ciências Agrárias e Veterinárias, Universidade Estadual Paulista “Júlio de Mesquita Filho”, Campus de Jaboticabal (FCVA – Unesp), and to the post-graduation program “Ciência do Solo”, by reception and learning opportunities. To advisor and mentor Prof. Dr. José Marques Júnior by his years of interaction, advice, exchange of experiences and opportunities for personal and scientific evolution. My scientific father. To my co-advisor Prof. Vidal Barrón Lopés de Torre from whom I learnt a lot. It was an honor to work with you and an unspeakable opportunity for my personal, scientific, and professional life. To Department of Agronomy (Soil Science) of the University of Córdoba (UCO), Spain, especially to Prof. Dr. José Torrent for the moments of conversations about Fe oxides, which contributed to the innovation and improvement of this study. To scientist José María Méndez for his contribution to laboratory analysis. I thank all the members of the Soil Characterization for Specific Management Purposes group (CSME), faithful friends who accompanied me during the research and the realization of this work: Romário Pimenta Gomes, Vinicius Augusto Filla, Luiz Gustavo Rosa Campos, Renato Nery Malmegrim Junior, Luis Fernando Vieira da Silva, Luigi Cicilini Benedini Moura, Angelina Pedro Chitlhango, Daniel De Bortoli Teixeira, Ludmila de Freitas, Ivanildo Amorim, Renato Eliotério de Aquino. To the examination panel members of the thesis defense: Prof. Dr. Antônio Carlos Saraiva da Costa, Prof. Dr. Rogério Teixeira de Faria, Prof. Dr. Gener Tadeu Pereira, and Zigomar Menezes de Souza, who dedicated the evaluation of the thesis, timely and constructive criticisms that contributed to the improvement of this thesis. My sincere gratitude to all. I thank the São Paulo Research Foundation (FAPESP), Brazil for funding this study (Doctorate fellowship no. 2017/01704-2 and BEPE, Spain, no. 2018/15694-1). i SUMMARY Page ABSTRACT ................................................................................................................ iii RESUMO.................................................................................................................... iv ABBREVIATIONS LIST .............................................................................................. v TABLES LIST ............................................................................................................ vi FIGURES LIST .......................................................................................................... vii CHAPTER 1 – General considerations .................................................................... 1 1.1 Introduction ......................................................................................................... 1 1.2 Literature review .................................................................................................. 5 1.2.1 Iron oxides in tropical soils ............................................................................. 5 1.2.2 Spatial variability of iron oxides ...................................................................... 8 1.2.3 Quantification methods of iron oxides .......................................................... 10 1.2.3.1 X-ray diffraction (XRD) .......................................................................... 10 1.2.3.2 Diffuse reflectance spetroscopy (DRS) ................................................. 11 1.2.3.2.1 Theoretical basis of DRS in soil ....................................................... 11 1.2.3.2.2 Vis-NIR-MIR and applications in pedology ....................................... 13 1.2.3.4 Magnetic susceptibility (χ) ..................................................................... 17 1.3 General material and methods .......................................................................... 22 1.3.1 Site and geomorphological background ....................................................... 22 1.3.2 Soil sampling ................................................................................................ 24 1.3.3 Laboratory analysis ...................................................................................... 24 1.3.3.1 Conventional soil analysis ..................................................................... 24 1.3.3.2 Mineralogical analysis ........................................................................... 25 1.3.3.2.1 X-ray diffraction ................................................................................ 25 1.3.3.2.2 Diffuse reflectance spectroscopy ..................................................... 26 1.3.3.3 Selective dissolution in 1.8 mol L-1 H2SO4............................................. 26 1.3.3.4 Magnetic susceptibility measurements ................................................. 27 1.3.4 Statistical and geostatistical analysis ........................................................... 28 1.3.4.1 Spatial variability ................................................................................... 29 1.3.4.2 Validation of spectral maps ................................................................... 30 1.4 References ........................................................................................................ 30 ii CHAPTER 2 – Spatial variability of iron oxides in soils from brazilian sandstone and basalt................................................................................................................. 46 ABSTRACT ............................................................................................................. 46 2.1 Introduction ....................................................................................................... 47 2.2 Results and discussion ...................................................................................... 48 2.2.1 Physical and chemical properties of soil profiles .......................................... 48 2.2.2 Identification of iron oxides ........................................................................... 51 2.2.2.1 X-ray diffraction (XRD) .......................................................................... 51 2.2.2.2 Diffuse reflectance spectroscopy .......................................................... 53 2.2.2.3 XRD versus DRS for characterizing iron oxides in surface soils ........... 56 2.2.3 Geostatistical analysis and mapping of iron oxides ...................................... 58 2.3 Conclusions ....................................................................................................... 62 2.4 References ........................................................................................................ 62 CHAPTER 3 – Estimation and mapping of magnetic minerals of lithopedogenic origin in Brazilian soils ........................................................................................... 69 ABSTRACT ............................................................................................................. 69 3.1 Introduction ....................................................................................................... 70 3.2 Results and discussion ...................................................................................... 71 3.2.1 Soil attributes ............................................................................................... 71 3.2.2 Magnetic susceptibility ................................................................................. 73 3.2.3 Estimating the contents in magnetic minerals with different methods .......... 76 3.2.4 Magnetic signature in the spatial distribution of magnetic oxides ................. 80 3.3 Final considerations and applications ................................................................ 84 3.4 Conclusions ....................................................................................................... 84 3.5 References ........................................................................................................ 85 APPENDIX A. Supplementary information .......................................................... 91 iii QUANTIFICATION OF LITOPEDOGENIC IRON OXIDES BY DIFFUSE REFLECTANCE SPECTROSCOPY AND MAGNETIC SUSCEPTIBILITY FOR MAPPING PURPOSES ABSTRACT - The Western Paulista Plateau (WPP) corresponds to approximately 13 million hectares (~ 50% of the state of São Paulo) and stands out in the production of citrus in the country (80% of the national production) and significant participation in the production of sugar and alcohol. The geological and geomorphological diversity of the region are the main cause for the spatial variability of Fe oxides, mainly goethite (Gt), hematite (Hm), maghemite (Mh) and Magnetite (Mt). Thus, recognizing the spatial variability of these oxides makes possible to identify the potential of agricultural areas, allowing the establishment of minimum management areas and the strategic planning of agricultural activities, minimizing environmental risks and optimizing operating costs. An alternative for developing the agricultural practices in these areas is the identification of the variability of soil attributes. Therefore, this study aimed to characterize the spatial variability of Fe oxides in WPP soils, using X-rays diffraction (XRD), diffuse reflectance spectroscopy (DRS) and magnetic susceptibility (χ). Sustained by sandstone (Vale do Rio do Peixe Formation — VRP) and basaltic (Serra Geral Formation — SG) rocks, WPP soils are distributed in three stages of landscape evolution: slightly, moderately and highly dissected. A total of 300 samples, collected at a depth of 0.0–0.2 m and representative of the geological and geomorphological compartments, were characterized by XRD, DRS and χ. The results were evaluated through spatial variability, relating the mineralogical data with the geomorphometric map, through statistical and geostatistical analysis. The lithological contrast between sandstone and basalt and the degree of dissection of the landscape control the spatial variability of Hm, Gt, Mh and Mt, in environments with low and high levels of Fe oxides. The similarity between maps obtained by conventional technique (XRD) and indirect methods (DRS and χ) highlights the efficiency and reliability of DRS and χ in the spatial characterization of soil Fe oxides in large areas, under complex soil-landscape relationships, in less time and investment cost. Keywords: goethite, hematite, maghemite, magnetite, pedometry, spatial variability iv QUANTIFICAÇÃO DE ÓXIDOS DE FERRO LITO-PEDOGÊNICOS POR ESPECTROSCOPIA DE REFLETÂNCIA DIFUSA E SUSCETIBILIDADE MAGNÉTICA PARA FINS DE MAPEAMENTO RESUMO - O Planalto Ocidental Paulista (POP) corresponde a aproximadamente 13 milhões de hectares (~ 50% do estado de São Paulo) que se destaca na produção do cultivo de citros do país (80% da produção nacional) e significativa participação na produção de açúcar e álcool. A diversidade geológica e geomorfológica da região é responsável pela variabilidade espacial dos óxidos de Fe, principalmente, goethita (Gt), hematita (Hm), maghemita (Mh) e magnetita (Mt). Assim, o reconhecimento da variabilidade espacial desses óxidos torna possível a identificação de áreas com diferentes potenciais agrícolas, permitindo estabelecimento de áreas mínimas de manejo e o planejamento estratégico das atividades agrícolas, minimizando riscos ambientais e otimizando os custos operacionais. Uma alternativa para o avanço agrícola nessas áreas é a identificação da variabilidade dos atributos do solo. Este estudo objetivou caracterizar a variabilidade espacial dos óxidos de Fe nos solos do POP, utilizando Difratometria de raios-X, Espectroscopia de Reflectância Difusa (ERD) e Suscetibilidade Magnética (χ). Sustentados por rochas areníticas e basálticas, os solos do POP estão distribuídos em três estágios de evolução da paisagem: pouco, intermediariamente e altamente dissecada. Um total de 300 amostras, coletadas na profundidade de 0,0 - 0,2 m e representativas dos compartimentos geológicos e geomorfológicos, foram caracterizadas por DRX, ERD e χ. Os resultados foram avaliados no estudo da variabilidade espacial, relacionando os dados mineralógicos com o mapa geomorfométrico, por meio de análise estatística e geoestatística. O contraste litológico arenito-basalto e o grau de disseção da paisagem controlam a variabilidade espacial de Hm, Gt, Mh e Mt, em ambientes com baixos e altos teores de óxidos de Fe. A semelhança entre os mapas obtidos por técnica convencional (DRX) e métodos indiretos (ERD e χ) enaltece a eficiência e confiabilidade da ERD e χ na caracterização espacial de óxidos de Fe do solo em grandes áreas, sob complexas relações solo- paisagem, em menor tempo e custo de investimento. Palavras-chave: goethita, hematita, maghemita, magnetita, pedometria, variabilidade espacial v ABBREVIATIONS LIST a Range Hd Highly dissected Ana Anatase C0 Nugget effect C0 + C1 Sill DCB Dithionite–citrate–bicarbonate DRS Diffuse reflectance spectroscopy Fed Free-crystallinity iron Feo Low-crystallinity iron DSD Degree of spatial dependence H2SO4 Sulfuric acid Hm Hematite Gb Gibbsite Gt Goethite Kt Kaolinite LVef Rhodic Eutrudox PVA Typic Kandiudalf ME Mean error Mh Maghemite Mt Magnetite Qz Quartz Pd Slightly dissected POP Planalto Ocidental Paulista Md Moderately dissected R2 Coefficient of determination RT Rietveld RMSE Root mean square error SDE Sandard deviation of the error WPP Western Paulista Plateau χ Magnetic susceptibility χlf Magnetic susceptibility in low frequency χhf Magnetic susceptibility in high frequency χfd% Frequency-dependent magnetic susceptibility XDR x-ray diffraction vi TABLES LIST CHAPTER 2 Page Table 1. Oxide contents as determined by sulfuric digestion, sand, silt, and clay contents, weathering indices, and soil color of six typical soils profiles in slightly, moderately, and highly dissected units of SG (basalt) and VRP (sandstone) ........... 50 Table 2. Descriptive statistics for goethite (Gt) and hematite (Hm) as determined by XRD in 200 soils samples in slightly (Sd), moderately (Md), and highly dissected (Hd) units of the Serra Geral (SG) and Vale do Rio Peixe (VRP) formations .................... 53 Table 3. Descriptive statistics for Hm and Gt contents as determined by XRD and estimated by DRS in 200 soil samples from the Western Paulista Plateau. PG ........ 56 CHAPTER 3 Table 1. Descriptive statistics for various attributes of soils from WPP classified according to parent material ...................................................................................... 72 Table 2. Descriptive statistics for soil magnetic susceptibility in WPP basalt–sandstone lithological contrasts .................................................................................................. 73 Table 3. Descriptive statistics for maghemite (Mh) and magnetite (Mt) as determined with different methods in basalt–sandstone soils from WPP ..................................... 77 Table 4. Models adapted to the experimental semivariograms ................................. 80 vii FIGURES LIST CHAPTER 1 Page Figure 1. Landscape evolution model (geomorphogenesis) in Western Paulista Plateau (WPP) ............................................................................................................. 9 Figure 2. (a) Scheme of diffuse reflectance spectroscopy equipment and (b) type of reflectance when interacting with soil particles. Adapted from Gomes (2017) .......... 13 Figure 3. Hypothetical schematic analysis of diffuse reflectance spectroscopy of soil sample and respective attributes in the regions of the visible (Vis - 400 to 780 nm) and near infrared (VIR - 800 to 2500 nm) (a) and mid infrared (MIR - 2500 to 25000 nm) (b). OM – organic matter, Qz - quartz as sand, OC - organic compounds, Ca - calcite, Kt - Kaolinite, Gb - Gibbsite, Hm - hematite, Gt - goethite, OH - characteristics of mineral water.. ........................................................................................................... 14 Figure 4. Diffuse reflectance spectroscopy for the main pedogenic Fe oxides. Model presented by Torrent and Barrón (2008). .................................................................. 15 Figure 5. Magnetic behavior in the presence of an external magnetic field. Adapted from Barrón (2020), personal communication. .......................................................... 18 Figure 6. Geological maps (Fernandes et al., 2007) (a) and landscape dissection units (Vasconcelos et al., 2012) (b) representative soil profiles in the Western Paulista Plateau: 1, 2 and 3 - Rhodic Eutrudox (LVef), 4 and 5 - Rhodic Hapludox (LVd) and 6 - Typic Kandiudalf (PVA) (Soil Survey Staff, 2010) ................................................... 23 CHAPTER 2 Figure 1. XRD patterns for six profiles representing the basaltic soils in the Serra Geral formation (SG) and sandstone soils in the Vale do Rio do Peixe formation (VRP). Sd, Md, and Hd denote slightly, moderately, and highly dissected units, respectively. Gt, goethite. Hm, hematite. Qz, quartz. Ana, anatase. Mh, maghemite. NaCl, sodium chloride ...................................................................................................................... 52 Figure 2. Diffuse reflectance spectra for six soil profiles typical of three different dissection levels in Serra Geral (SG) and Vale do Rio Peixe (VRP) formations, Western Paulista Plateau. ....................................................................................................... 55 Figure 3. Regression models for (a) hematite as determined by X-ray diffraction (XRD) and estimated by diffuse reflectance spectroscopy (DRS); and (b) goethite as determined by XRD and estimated by DRS .............................................................. 57 Figure 4. Variograms for hematite and goethite as obtained from XRD and DRS data for soils from the lithostratigraphic units of the Western Paulista Plateau. ................ 58 viii Figure 5. Spatial patterns for hematite as estimated by XRD (a) and DRS (b), and validation between the two techniques (c). Spatial patterns for goethite as estimated by XRD (d) and DRS (e), and validation between the two techniques (f). Geological maps (g) and landscape dissection units (h). ............................................................ 60 CHAPTER 3 Figure 1. Correlations of low-frequency magnetic susceptibility (χLF), percent frequency-dependent susceptibility (χFD) in the air-dried fine earth (ADFE) fraction and various soil properties. Fed, iron in crystalline iron oxides as extracted with sodium dithionite–citrate–bicarbonate. Feo, iron in poorly crystalline iron oxides as extracted with ammonium oxalic acid........................................................................................ 76 Figure 2. Regression between low-frequency magnetic susceptibility (χLF) and the contents in magnetic minerals as determined with various methods. (a) Maghemite from χLF after extraction with DBC. (b) Magnetite from remaining χLF. (c) Maghemite from χLF in oxide-concentrated clay fraction after extraction with NaOH. (d) Maghemite by sulfuric digestion. (e) and (f) Maghemite by X-ray diffraction spectroscopy without and with Rietveld refinement, respectively. ............................................................... 78 Figure 3. Spatial patterns of magnetic minerals in air-dried fine earth and oxide- concentrated clay. (a) Low-frequency magnetic susceptibility. (b) Maghemite as determined from χLF after extraction with DCB. (c) Magnetite as determined from remaining low-frequency susceptibility. (d) Maghemite as determined from χLF. (e) Maghemite as determined by sulfuric digestion. (f) and (g) Maghemite as determined by X-ray diffraction spectroscopy without and with Rietveld refinement, respectively .................................................................................................................................. 82 1 CHAPTER 1 - General considerations 1.1 Introduction Brazilian agribusiness boosts the Brazilian economy, with an emphasis on the global production of soybean, sugar cane, coffee and beef. Modern and globalized agriculture has been marked by digital technology and Big Data connected by software linked to agricultural equipment that optimize production during the entire production stage. Among the sectors that most advanced in technology can be mentioned the improvement of plant genetics (Carrer et al., 2010) and the agricultural machinery industry, with sensors and the development of “intelligent” machines. Paradoxically, or in a situation of constant inertia, soil mappings are found. Analogous to this, contemporary agriculture resembles a “Ferrari” car (technological evolution) on a “pothole” road (gaps left by current soil mapping). Brazil pays a high price because it don't know better his soils (Polidoro et al., 2016), making difficult to prevent natural problems such as floods and landslides that plague urban areas and agricultural areas during the rainy season, bringing irreparable disruption to society Brazilian. With a focus on agricultural environments, the lack of information on soil attributes prevents the development of suitable agricultural practices. The result of this is the increase in desertified areas, compaction and intense erosion processes, making the soil unsuitable for food production. The history of the first mapping of Brazilian soils dates back to the 1940s, so that Brazilian pedology was imminent (Santos et al., 2011). In 1981, the first exploratory soil maps were produced on a scale of 1: 5,000,000. Over the years, public agencies under the Ministry of Agriculture and state institutes such as Campinas (IAC) and the National Land Survey and Conservation Service (SNLCS) have been tasked with updating existing ones and drawing up new maps. A major project was active in this process of diagnosing the potential of natural resources in the Brazilian territory: RADAMBRASIL. Operated between 1970 and 1985, RADAMBRASIL was a challenging and audacious project, responsible for mapping 2.5 million km2 of Brazilian territory by aerial radar images, captured by plane. About 70% of the efforts were 2 concentrated in the Amazon region and the remaining 30% in the northeastern and pre-Amazon regions. In 2001, IBGE in partnership with Embrapa launched a new approach to Brazilian soils, with updated information from the one developed in 1981. The improvements were only taxonomic in nature, with no change in scale (Embrapa, 2011). Therefore, depending on the user's interest and with a focus on agriculture, the current soil mappings are outdated, without accuracy, full of technical terminologies that, sometimes, the information contained is difficult to interpret. For modern agriculture, the level of detail of the maps on the scale of 1: 5,000,000 is inefficient for the development of precision agriculture, in which the lowest cost with agricultural inputs and the increase in production is sought. Due to the need to know the soil better, in 2016, a new soil survey proposal was launched coordinated by Embrapa, called the National Soil Program of Brazil (Pronasolos). The goal is to build more detailed maps on a scale of 1:25,000, 1:50,000, 1:100,000, that meets government needs. For Planasolos, the information from the 1:25,000 maps may assist decision making in the agricultural sector. This would be the first step in overcoming the impeding gaps inherited from the pioneering soil surveys carried out in 1981 and 2001. The prediction for completion is 30 to 40 years and will cost to governmental budget a total of R$ 3 billion, with return of R$ 40 billion for the country in a decade. The high monetary cost and the long term of accomplishment are discouraging in the face of the current technological scenario that Brazilian agriculture is experiencing. Possibly, the methodologies used are the main reason for increasing costs and delivery time to users. It is in this panorama that alternative methods can be the key to assist conventional methods, generating accurate and reliable results, in a timely manner and with a lower investment cost (Carvalho et al., 2013; Silva et al., 2020). The pioneering spirit of the state of São Paulo in mapping soil attributes by indirect methods inspires and attests to their potential in mapping soil attributes (Siqueira et al., 2010; Marques Jr et al., 2015; Peluco et al., 2015; Camargo et al., 2016; Silva et al., 2020; Demattê et al., 2020). The research group of Professor Dr. José Alexandre Melo Demattê from the University of São Paulo (USP) has been betting on the spectral library for characterization and design of soil surveys based on soil 3 color, using diffuse reflectance spectroscopy. In 2010, through magnetic susceptibility, Professor Antônio Carlos Saraiva da Costa's Mineralogy group prepared the magnetism map for soils in the state of Paraná. The Soil Characterization for Specific Management Purposes group (CSME), under the coordination of Professor Dr. José Marques Júnior, has numerous works published in high impact scientific journals, demonstrating the efficiency and reliability of indirect methods - magnetic susceptibility and reflectance spectroscopy diffuse - in the spatial characterization of soil attributes. In addition, the CSME over 20 years of existence has formed and stimulated new soil scientists with affinities and mastery of mathematical models (Siqueira et al., 2010; Teixeira et al., 2017; 2018) that help to describe processes of soil, reducing the subjectivity of traditional models of mapping soil attributes. Scientific publications confirm the acceptance of indirect methods and mathematical modeling by the scientific community. In fact, the main research funding agency in the state of São Paulo, FAPESP, has credited confidence in this new problem-solving horizon. For example, funding for research and training of masters and doctors with the purpose of generalizing the information on spatial variability of soil attributes can be mentioned: Lívia Arantes Camargo (nº: 17 / 01790-6), Diego Silva Siqueira (nº: 11 / 06053-3; 08 / 07693-3 and 04 / 09552-7), Angélica de Souza Bahia (nº: 13 / 17552-6), Kathleen Fernandes (nº: 17 / 05477-0 and 15 / 20692-0) and Laércio Santos Silva (nº: 17 / 01704-2 and 18 / 15694-1). The products of the aforementioned dissertations and theses were detailed maps of soil attributes, in the soil-landscape context and lithological complexity, providing detailed information on soil attributes for teaching, research and extension purposes, tactical and operational planning of agricultural activities, governance of soils and more effective public policy decisions. Twenty years ago, the CSME group approached quantitative mathematical, statistical and geostatistical methods to Soil Science in an attempt to describe, analyze and interpret soil data. This quantitative character of interpreting the soil and the processes involved was called pedometry (Webster, 1994). There are reports that pedometry emerged in the face of criticism of conventional soil surveying methods, as they are very qualitative and subjective (McBratney et al., 2000). Thus, studying the soils by metrics has made it possible to recognize, for example, the spatial pattern in 4 a more objective way, as shown by some studies developed in Brazil (Camargo et al., 2008ab; Siqueira et al., 2010; Oliveira Jr et al. , 2011; Camargo et al., 2013; Barbieri et al., 2013; Marques et al., 2014; 2015; Peluco et al., 2015; Camargo et al., 2016; Teixeira et al., 2017; Teixeira et al., 2018; Silva et al., 2020). It is evident that the adoption of pedometry and indirect methods of soil mapping yield faster returns and reduce costs when compared to conventional methods (Bahia et al., 2015; Camargo et al., 2016; Demattê et al., 2020). However, in addition to the promising methods of soil mapping, it is still necessary to select “key attributes”. These are attributes that coordinate, and that the behavior of others can be gauged from them - basic principle of pedotransfer function modeling (PTFs). The PTFs allows predicting more complex attributes from another strongly correlated, easily measured and obtained at lower costs (Budiman et al., 2003; Lagacherie and McBratney, 2007). The mineralogy of the soil, specifically, of the clay fraction, has gained prestige as a key attribute for the prediction and characterization of the spatial pattern of covariate attributes (Barbieri et al., 2013; Peluco et al., 2015; Camargo et al., 2016). Among the minerals of the clay fraction, Fe oxides are preferred candidates in prediction studies, as they control the physical and chemical properties of the soil (Camargo et al., 2013). Other factors confirm Fe oxides as strategic minerals, such as: (I) they persist in the soil and record environmental changes in color and crystallographic properties (Fernandes et al., 2004; Silva et al., 2020); (II) markers of environmental processes and source material (Long et al., 2015; Silva et al., 2020); paleoclimatic (Maher et al. 2003) and palioenvironment (Torrent et al., 2010; Wang et al., 2016); (III) some exhibit magnetic properties, (IV) reflect the pedoenvironmental conditions that were formed, reasons that praise its function as a natural pedoindicator (Cornell and Schwertmann, 2003). In order to confirm the potential of applying diffuse reflectance spectroscopy (DRS) and magnetic susceptibility (χ) in the quantification and mapping of iron oxides - goethite, hematite, maghemite and magnetite - as well as the accuracy of these techniques in the prediction of the respective minerals in large areas, some hypotheses have become relevant in the present study: (I) The spectral signature and magnetic susceptibility of soil samples can be used to estimate the content of antiferromagnetic (hematite and goethite) and ferrimagnetic 5 (magnetite and maghemite) minerals with good precision and accuracy for WPP soils when compared with the traditional X-rays diffractometry technique. (II) The spatial variability of the hematite, goethite, maghemite and magnetite contents are influenced by the geomorphological compartments at different stages of evolution. (III) The geomorphological environments identified by the geomorphometric signature technique coincide with different environments for the formation of hematite, goethite and maghemite pedogenetic minerals. (IV) Maps of soil minerals obtained by DRS and χ and information on geomorphological compartments at different stages of evolution can assist in the identification of areas with different agricultural potentials; (V) DRS and χ techniques can be applied reliably in large and complex areas. In view of the statement of the problem addressed, the study had the general purpose of: Characterizing (identifying and quantifying) the spatial variability of litho-pedogenic Fe oxides in the soils of the Western São Paulo Plateau, using X-rays diffraction, diffuse reflectance spectroscopy and magnetic susceptibility. This purpose was developed into two specific objectives: Chapter 2. We assessed the efficiency of DRS for estimating the spatial variability in Gt and Hm in the framework of soil–landscape relationships in the Western Paulista Plateau (Brazil) to facilitate the development of ancillary methods for mapping large areas. Chapter 3. We assessed the potential of magnetic susceptibility (χ) for predicting spatial variability in the magnetic minerals maghemite (Mh) and magnetite (Mt) in soils developed on sandstones and basalts in relation to alternative determination methods. 1.2 Literature review 1.2.1 Fe oxides in tropical soils Fe oxides, a generic nomenclature to define Fe oxides and oxidroxides (Melo and Aleonni, 2009), are important components of the mineralogical assembly of soils, especially those in the last stage of evolution (Ker, 1997; Santos, 2018). Among the oxides of Fe occurring in the soil, such as ferrihydrite, lepdocrocyte, schwertmannite, specifically hematite (Hm = α-Fe2O3) and goethite (Gt = α-FeOOH) are the most 6 abundant (Schwertmann and Taylor, 1989 ), being present in soils with low and high levels of total Fe (Campos et al., 2007; Metri et al., 2008; Camargo et al., 2013; Bahia et al., 2015; Carvalho Filho et al., 2015 ; Camêlo et al., 2017; Poggere et al., 2018; Silva et al., 2020). In the soil, Fe oxides are products of the dissolution of primary (lithogenic) and secondary (pedogenic) minerals that contain iron (Fe+2) in their crystalline structure (Cornell and Schwertmann, 1996). Free in solution, the Fe+2 and / or Fe+3 ions can combine with other species in the soil (O-, OH-, H+) and form Fe oxyhydroxides, like goethite and hematite (Schulze, 1989; Sposito, 1989). Although they are formed from Fe ions and normally coexist in the soil, the occurrence of one is to the detriment of the other (Schwertmann and Taylor, 1989; Kämpf and Curi, 2000). In addition, the concentration of these oxides greatly depends on the factors and processes of soil formation, especially rock. Unlike goethite, hematite has its preferential formation in soils derived from material rich in Fe, such as basalt rocks (Schwertmann, 1985; Bigham et al., 2002). Inda Junior and Kämpf (2005) reported that, in addition to the source material, the formation and stability of these minerals are conditioned by the pedoenvironmental characteristics (temperature, humidity, organic matter content, pH, Eh, among others), characteristics that provide great variation in color, in the form and in the very constitution of Fe oxides (Schwertmann and Carlson, 1994). Reasons that give Fe oxides the function of environmental pedoindicators, as they reflect the conditions of pedogenesis under which they would have been formed (Fitzpatrick and Schwertmann, 1982). Some soils, especially those derived from basic rocks, may have magnetic properties, due to the presence of magnetite (Mt = FeO.Fe₂O₃) and maghemite (Mh = γ-Fe2O3) (Costa et al., 1999; Cornell and Schwertmann, 2003; Schaefer et al., 2008). When these minerals occur simultaneously in the soil, maghemite can be formed from the direct oxidation of magnetite (Barrón and Torrent, 2002). The transformation of goethite, hematite and ferrihydrite into maghemite is common in soils that have suffered from burning, in the presence of organic compounds (Mullins, 1977; Schwertmann and Cornell, 1991). According to Costa et al. (1999), although Mh occurs in lower concentrations in the soil, it can constitute up to 40% of the total iron oxide 7 content. However, there is no evidence in the maghemite literature as the only existing Fe oxide, always occurring associated with hematite (Costa and Bigham, 2009). The content of total Fe oxides (in the form of Fe2O3) is a diagnostic attribute of the Brazilian Soil Classification System (SiBCS) in the differentiation of soil classes and, consequently, in the evaluation of the degree of weathering (Santos et al., 2018). This Fe is obtained in air-dried fine earth after sulfuric digestion (H2SO4) (Donagema et al., 2011). Although it is common to call it "total Fe2O3", it would be better to call it "pseudototal", since sulfuric digestion little attacks minerals in the coarse fractions (sand and silt) that may contain Fe in its structure, being effective in the clay fraction. However, as the tropical soils are well evolved, it can be said that the totality of Fe in these is found in the clay fraction. Considering the Fe2O3 content, it is possible to classify the soil in low levels of Fe (hypoferric; Fe2O3 < 8%), medium (mesoferric; 8% < Fe2O3 ≤ 18%), high (ferric; 18% < Fe2O3 ≤ 36%) and very high (ferric; Fe2O3 > 36%). This classification is commonly used for the Latosol and Nitosol classes. High levels of Fe2O3 are usually associated with soils from volcanic eruptive rocks. The studies by Carvalho Filho et al. (2015) recorded levels above 70% in Ferruginous dolomite soil, that is, in 1 kg of soil, 700 g are Fe2O3 - such high values were unknown in Brazilian literature. Another importance of Fe2O3 is as an indication of the degree of weathering of the soil [Ki = 1.7 ×% SiO2 /% Al2O3, Kr = 1.7 ×% SiO2 / (% Al2O3 + 0.6325 ×% Fe2O3)], whose values separate oxidic soils (very weathered) ; Ki and Kr <0.7) of those kaolinitic (slightly weathered; Ki and Kr > 0.7). Another important aspect of the soil is the color, indispensable in SiBCS, considered the basis for surveying Brazilian soils (Santos et al., 2018). In general, color is a safe pedoindicator morphological property of the presence and type of Fe oxides, even in low concentration (Fernandes et al., 2004), of covariate attributes (Camargo et al., 2016) and of organic matter. For example, more yellowish soils (2.5Y and 7.5RY) indicate goethite as the main Fe oxide, while hematite gives the soil a more reddish color (5R and 2.5YR) (Ker, 1997; Bigham et al., 2002; Santos et al., 2013). In turn, maghemite has a hue ranging from 2.5YR to 5YR, commonly associated with hematite with a brighter hue (2.5YR - 5YR) (Schwertmann, 1993; Costa and Bigham, 2009). 8 Easy to identify, color expresses the interaction of soil formation factors and processes (Torrent and Barrón, 1993; Santos et al., 2018), providing valuable information on the environment in which the soil would have been formed. Among some examples, the color helps to differentiate the horizons of the soil profile and may indicate a pedoenvironment with poor internal drainage, whose absence of oxygen promotes the reduction of Fe3+ — Fe2+, a formation process marked by the expression of mottled (plinthite) to the gray color from soil. Easily determined, based on a letter from Munsell (Munsell, 1994), color is used for order differentiation. As a legacy of the rock, the variation in the color of the soil may indicate a difference from the source material (Silva et al., 2020). This would be one of the reasons why color is a diagnostic attribute in other soil classification systems, such as Chinese (Chinese Soil Taxonomy, 2001) and American (Soil Survey Staff, 2003). 1.2.2 Spatial variability of Fe oxides Studies of soil-landscape relationship show that the formation and spatial variability of iron oxides are conditioned by landscape shapes (geomorphology), by interfering in the distribution of water in the soil, in promoting chemical reactions and in the transport of solids or materials in solution (Marques Jr and Lepsch, 2000; Schoorl et al., 2000; Ghidin et al., 2006). The studies by Camargo et al. (2008a) and Montanari et al. (2010) found a predominance of goethite in the concave forms of the landscape, while hematite had its formation favored in linear and convex. This is because in linear and convex pedoenvironments, water infiltration is facilitated, conditioning a drier environment and higher temperatures, that is, oxidizing conditions that favor the formation of hematite. In concave pedoforms, the soils remain moist for a longer time and tend to accumulate organic matter, providing reducing conditions that provide the formation of goethite (Schwertmann, 1985). Geomorphology, mentioned here as the degree of dissection of the landscape (Figure 1), also determines the formation of iron oxides, as reported by Coventry et al. (1983) and Williams and Coventry (1979) on North Carolina (USA) soils. In a cause- effect relationship, these authors reported that in the poorly dissected pedoenvironments of the landscape, that is, where weathering was more active 9 (pedogenesis prevailed over morphogenesis), the soils were deeper with good drainage conditions, that is, in environments oxidants predominated the hematite with better crystallinity in the clay fraction. In turn, goethite was the main iron oxide in highly dissected pedoenvironments, where the soils were shallower (lithic contact), with the presence of water stagnation for a certain period of time, configuring reducing conditions. In this regard, Curi and Franzmeier (1984) state that this occurs because goethite is formed in the first stages of weathering of primary minerals, therefore it accumulates in young soils or in the horizons near the rocks. Figure 1. Landscape evolution model (geomorphogenesis) in Western Paulista Plateau (WPP). Studies on spatial variability of mineralogical attributes, their relationship with chemical and physical soil attributes and the productivity of different crops were carried out in areas of varying dimensions from 1 to 500 ha in soils originating from Basalt and Sandstone (Montanari et al., 2010; Marques Jr et al., 2012; Camargo et al., 2013, 2014; Dantas et al., 2014; Peluco et al., 2015; Bahia et al., 2015; Camargo et al. , 2016). However, in large territorial areas, detailed knowledge of the soil is difficult. Among the causes are the natural variability of the conditioning factors of the spatial variability of the soil (geology, landscape shapes, among others) (Legros, 2006), lack of experienced professionals (Demattê et al., 2007), lack of government incentives (Bazaglia Filho, 2012) and deficit in the formation of staff in the geostatistics area. The delimitation of homogeneous areas may correspond to future soil series, however the use of conventional methods of mapping soil attributes would make the establishment of specific management areas unfeasible. Soil properties are generally 10 described in detail by a pedologist and then validated and elaborated by chemical analysis (Ben-Dor et al., 2008). In the field, diagnostic descriptions, mainly associated with soil color, are subject to problems such as subjectivity. Laboratory chemical analyzes are time-consuming, costly and require complex sample treatment procedures, which are common problems for both the Brazilian soil system and other countries, such as China (Chinese Soil Taxonomy, 2001) and the United States (Soil Survey Staff, 2003). For the reasons presented, pedometry has been well accepted among modern soil researchers, who recognize the obstacles and the high cost of traditional mapping methods. Although pedometric methods have a strong scientific basis, with a statistical and mathematical basis, the presence of an experienced pedologist is indisputable, in order to interpret the important qualitative parameters in pedometric models, minimizing inferences of modeling, erroneous conclusions, or emotive of metric models (Mendonças-Santos, 2007). 1.2.3 Methods of quantification of Fe oxides 1.2.3.1 X-ray diffraction (XRD) Traditionally, X-ray diffraction (XRD) is considered the standard method for characterizing the mineral phases of the soil (Whitting and Alardice, 1986). However, considering the soil as a mixture of inorganic and organic particles, the concentrations and degree of crystallinity of minerals are sometimes impediments in the evaluation of the crystalline phases, and can generate misinterpretations (Silva et al., 2020). For this reason, soil analysis by XRD is accompanied by a series of procedures, including the separation of the sand, silt and clay fractions, the removal of minerals such as kaolinite and gibssite, and the concentration of oxides (Kämpf and Schwertmann, 1982; Balsam et al., 2014). These procedures are essential for the safe assessment of the mineral phases of the soil, avoiding misinterpretations. However, the laboratory processing time (Bahia et al., 2015; Silva et al., 2020) and the interpretation and cost of the analyzes make the characterization of iron oxides in large areas unfeasible, due to the need for a large number of samples. 11 Other limitations can be mentioned: (I) overlapping diffraction peaks; (II) low sensitivity to low Fe levels in soils inherited from the source material itself; (III) high isomorphic substitution of Fe for Al (Scheinost et al., 1998); (IV) imperfect crystalline and reduced size of the mineral (Jenkins and Snyder, 1996; Kämpf and Curi, 2000; Fabris et al., 2009). In general, XRD equipment is found in major research centers, due to the high cost of acquiring the equipment and the need for a professional to operate and interpret the data. In view of these obstacles, alternative techniques have been gaining ground in the mapping of soil minerals at a detailed level, allowing for precise inferences, with low cost of characterizing the minerals for the purpose of soil mapping and in relatively faster time compared to conventional methods, such as the XRD. 1.2.3.2 Diffuse reflectance spectroscopy (DRS) 1.2.3.2.1 Theoretical basis of DRS in soil Diffuse reflectance spectroscopy (DRS) is an indirect method of analysis for quantitative applications (Minasny and McBratney, 2008; O’Rourke et al., 2016;). The basic theoretical principle of the technique is based on radiation (incident light) and its interaction with soil constituents (Stenberg et al., 2010). Depending on the constituents of the soil, radiation promotes vibrations of the individual molecular bonds, allowing the absorption of light to varying degrees, due to the difference between the two energy levels. The soil sample submitted to the light source produces a characteristic spectrum, which, for Salvaggio and Miller (2001), allows it to be used for analytical purposes. The resultant of absorption at a given wavelength (λ), that is, the frequencies at which light is absorbed on the ground, is given in percentage reflectance [(R%): A = log (1 / R)]. R can be understood as the ratio between the radiant flow reflected by incident radiant flow (Wyszecki and Stiles, 1982). In turn, maximum reflectance is obtained when a body lets practically all the light falling on it, and on the ground it never occurs (Torrent and Barrón, 2007). The light, when striking the surface of a body, follows different paths, depending on the size of particles, angle of reflection, free space 12 between the particles and degree of compaction in the preparation of the soil sample. The way the reflected light acts or interacts with soil particles occurs at any wavelength, discriminating the regular (specular or mirrored) and / or diffuse reflectance. When the surface affected by the light has flat (smooth) faces, or when the size of the particles is relatively uniform (the angles of incidence and reflection are the same), the impacted light follows a unidirectional path, receiving the name of regular reflectance, situation in that the size of the surface particles is less than the wavelength of the incident radiation. There is the opposite behavior for diffuse reflectance, which predominates on a rough surface, due to the distinct and random size of the particles, in such a way that the incident light disperses in different angular directions. In a situation of diffuse reflectance, the wavelength of the incident radiation is less than the size of the surface particles. The radiation reflected in all directions provides more information about the soil in relation to the regular, being more affective in the characterization of soil attributes and processes (Torrent and Barrón, 2007). Figure 2. (a) Scheme of diffuse reflectance spectroscopy equipment and (b) type of reflectance when interacting with soil particles. Adapted from Gomes (2017). 13 The ability of matter color to absorb more or less light at different wavelengths (Barrón et al., 2000), applies very well to soil, since it is a mixture of mineral and organic particles that absorb and scatter the incident light (Barrón and Torrent, 1986; Fernandes et al., 2004). The interaction of light with soil properties, resulting in a spectral curve, aroused scientific interest in applying the applications of DRS in the characterization and estimation of soil attributes, in order to be used in laboratory analysis and even in the field. In the soil, some properties control the reflectance of the spectral curve (Hunt et al., 1971; Janik and Keeling, 1996): organic matter, humidity, concentration and type of Fe oxides and relative proportions of sand, silt and clay (Demattê et al., 2012; Dotto et al., 2014; Coblinski et al., 2020). 1.2.3.2.2 Vis-NIR-MIR and applications in pedology The attributes of the soil can be identified in three sensitive regions, defined as: Visible (Vis) - characterized by electronic excitations due to the high radiation energy in the wavelength of 400 to 780 nm (Figures 3a and 3b). Information about soil color, Fe content and composition, water molecules and organic matter (Sherman and Waite, 1985; Mortimore et al., 2004) are recognized in this region. According to Hunt et al. (1971) the visible region is typical of opaque minerals typical of basic rocks such as goethite, hematite, magnetite and ilmenite (Figure 3a). Goethite is characterized in the 425 to 450 nm ranges and, for hematite, in the 545 to 590 nm ranges. The presence of these oxides in the soil reduces reflectance, due to the charge transfer between Fe ions and oxygen (Sellitto et al., 2009). Another indication of Fe oxides in the soil is the more pronounced concavity (Bahia et al., 2015). Demattê et al. (2012) reported the increase in reflectance in soil samples after removal of organic matter, validating the energy absorption capacity of organic matter (Gmur et al., 2012). When the proportions of hematite and goethite are greater than that of maghemite, this has not been considered because it does not influence the spectral curve. This is because the reflectance pattern of maghemite occurs in an intermediate 14 range to hematite and goethite, as illustrated by Torrent and Barrón (2008) in Figure 4. Considering the coexistence of these oxides, it can be said that the behavior of the Figure 3. Hypothetical schematic analysis of diffuse reflectance spectroscopy of soil sample and respective attributes in the regions of the visible (Vis - 400 to 780 nm) and near infrared (VIR - 800 to 2500 nm) (a) and mid infrared (MIR - 2500 to 25000 nm) (b). OM – organic matter, Qz - quartz as sand, OC - organic compounds, Ca - calcite, Kt - Kaolinite, Gb - Gibbsite, Hm - hematite, Gt - goethite, OH - characteristics of mineral water. 15 spectral curve is governed by minerals hematite and goethite, main soil color conditioners. Inda et al. (2013) report that maghemite has expression in the 410 - 445 nm band, the same as that of goethite. Thus, for low levels of maghemite, its influence on reflectance patterns is usually disregarded. Figure 4. Diffuse reflectance spectroscopy for the main pedogenic Fe oxides. Model presented by Torrent and Barrón (2008). The region from 780 to 2500 nm is called near infrared (NIR) or short – wave infrared (SWIR) (Figure 3a). Typical range of OH molecular vibrations, phyllosilicates, sorosilicates, hydroxides, sulfates, amphiboles, carbonates, soil water molecules and organic matter (Clark, 1999; Viscarra Rossel et al., 2006a). Water molecules have a strong influence on Vis-NIR, being found in several regions, but their presence is defined around 1400 to 1900 nm (Liu et al. 2002). Important minerals in the soil such as kaolinite and gibbsite are also characterized in Vis-NIR, however, unlike Fe oxides, they do not have a strong influence on the reflectance intensity, the presence of which is characterized by features in the form of valleys around 1880 at 2300 nm. 16 The mid – infrared (MIR - 2500 at 25000 nm) has as its fundamental principle the electronic transition of atoms, longer vibrations of molecules and crystals, depends on a frequency (Figure 3b). Electronic excitation sets up the formation of peaks characteristic of soil constituents (Stenberg et al., 2010). For example, strong Si-O, Al- O, Fe-O peaks, HO, HC and HN bonds in the MIR region are related to the presence of silicates, Fe and Al quartz oxides, calcium carbonate and organic compounds (Nguyen et al., 1991; Cañasveras et al., 2010). In quantitative pedology, DRS has gained more and more applications. The fact that it is a non-destructive technique and does not require prior preparation of the sample under study, draws the attention of soil scientists, especially when working with a large database (large number of samples). Another reason is its use in pedotransfer function studies, combined with chemometric programs, such as ParLeS (Viscarra- Rossel, 2008) making it possible to correlate spectral information with the mineral composition of the soil. ParLeS has principal component analysis (PCA), NIPALS algorithm (Martens and Naes, 1989) and regression analysis of partial least squares (PLSR) (Geladi and Kowaslski, 1986), making it possible to calibrate, validate and obtain the best model of prediction of atributes. Because it is a fast, less expensive, non-destructive and simple operation technique (Bahia et al., 2015; Camargo et al., 2016; Silva et al., 2020), DRS is promising in pedology. In addition, this technique allows the simultaneous characterization of many soil attributes with agronomic and environmental relevance (McBratney et al., 2002, 2006; Viscarra Rossel et al., 2006, 2010; Cañasveras et al., 2010; Kodaira and Shibusawa, 2013), in addition to being adaptable for use in the field (Viscarra Rossel and Mcbratney, 1998). Several countries have adopted this technique for and quantification of soil minerals and ore areas, such as Spain (Torrent & Barrón, 2002), China (Chen et al., 2002; Hu et al., 2013), Australia (Viscarra Rossel et al., 2010) and Brazil (Fernandes et al., 2004; Carioca et al., 2011; Silva et al., 2020). In 1993, Torrent and Barrón proposed the use of DRS to quantify hematite and goethite, based on the Kubelka-Munk function (1931). By this method, the spectra coming from the samples' DRS are transformed into absorbance spectra, and with certain treatment the optical band-gap values are calculated. The application of the Kubelka-Munk function in the quanfication of goethite and hematite was applied by 17 Bahia et al. (2015), Camargo et al. (2016) and by Silva et al. (2020) in soils, compared with the results obtained by XRD. These authors found very close values for hematite and goethite between DRS and XRD, with R2 values above 60% for Gt and 80% for Hm. In all of these cases, DRS color saturation was observed in soils with high concentrations of Fe2O3 and Hm, but not enough to compromise the quality of the results compared to those of the XRD. The need to meet the growing demand for detailed information on tropical soils, especially mineralogical attributes, has increasingly used the use of indirect techniques, such as Australia, the country where the spectral signature for orbital sensors was obtained to build the hematite and goethite map of the entire federation (769,902,400 hectares) (Viscarra Rossel et al., 2010). In Brazil, incipient studies in small areas have already been developed using the spectral signature by DRS (Almeida et al., 2003; Bahia et al., 2015; Peluco et al., 2015; Camargo et al., 2016). 1.2.3.4 Magnetic susceptibility (χ) The χ identifies Fe-bearing minerals. It turns out that Fe is an external transition element, in which its most energetic electrons in the 3d sublevel are incomplete, giving it a magnetic moment with orbital motion (Thompson and Oldfield, 1986). The χ intensity is a result of how the molecular and atomic structure of a substance is organized (Thompson and Oldfield, 1986). Therefore, the magnetic moment depends on the arrangement of the electrons (e-) inside, and each electron has a magnetic moment associated with its spin. This occurs when a subatomic particle is subjected to a magnetic field it assumes a unilateral and, or varied direction, with a direct effect on the magnetic expression, which depending can be null or positive (Thompson and Oldfield, 1986; Cornell and Schwertmann, 2003). In a more simplistic way, magnetism can be understood, as the property that some substances and or minerals have to be attracted by a magnet (Resende et al., 1988; Ferreira et al., 1994). In the soil, Fe oxides such as magnetite, maghemite and ferrimagnetic ferrihydrite are the main holders of magnetism and, based on this property, it is possible to recognize five magnetic groups (Thompson and Oldfield, 18 1986): Ferromagnetic, Diamagnetic, paramagnetic, ferrimagnetic and antiferromagnetic (Figures 5a-5e). The ferromagnetic group is a particular case, since there is no representative in the mineral form, only for pure substances: Fe, Co and Ni (Melo and Aleonni, 2009). These substances have the electron spins always aligned even when in the absence of an external magnetic field and have extreme χ values around 27,600,000 × 10-8 m3 kg-1. Figure 5. Magnetic behavior in the presence of an external magnetic field. Adapted from Barrón (2020), personal communication. Classified as diamagnetic (Figure 5e) are those minerals that do not have χ (χ <0), that is, they are not attracted by a magnet, so they assume negative χ values. As an example of some representatives of this group, kaolinite, gibssite [Al2Si2O5 (OH)4] and quartz (SiO2), calcite (CaCO3), albite (NaAlSi3O8) and apatite [Ca5 (PO4)3] are mentioned. They are minerals made up of atoms and or molecules in which all the electrons are paired, causing an inversion of the electron orbital movement, which gives zero magnetic moment. Paramagnetics define minerals with positive χ (χ > 0), but weak. This occurs when the electron spins in the presence of an external magnetic field, however, as it is an imposed magnetization, it is not permanent. Representatives of paramagnetic minerals are ilmenite (FeTiO3) and lepidocrocyte (γ - FeOOH). Certain minerals have a strongly aligned magnetic moment, but with unequal opposing forces controlled by the crystalline structure (Figure 5b), called ferrimagnetic 19 minerals. This category includes the main magnetic minerals in the soil of lithogenic origin, such as magnetite and titanomagnetite (Fe2TiO4), and pedogenic, such as maghemite, (Dearing et al., 1994). Coexisting in the soil, the magnetic signal of maghemite is masked by magnetite, which prevents it from identifying its origin, that is, lithogenic, pedogenic or mixture origin. In general, they are common minerals in basic soils, for example, basaltic rocks and itabirite (Costa et al., 1999; Silva et al., 2010; Camêlo et al., 2017; Pogerre et al., 2018; Silva et al., 2020). Common in tropical soils, especially in advanced weathering stages, goethite and hematite (Cornell and Schwertmann, 2003; Melo and Aleonni, 2009) are antiferromagnetic minerals (Figure 5d). The representatives of this group have the spins of identical magnetic moments aligned in opposite directions, giving positive or zero net magnetization. According to Peters and Dekkers (2003), the magnetic signal of these minerals varies from 0.13 to 3.83 × 10-6 m3 kg-1 for goethite, and from 0.46 to 5.92 × 10-6 m3 kg-1 for hematite, which may explain the χ in soils with low levels of Fe3O3, in which the formation of maghemite and magnetite is disadvantaged. Although the contribution of hematite and goethite in the magnetic signal of the soil is modest when compared to maghemite and magnetite, it should not be ruled out, especially in soils with low levels of Fe2O3 (e.g. sandstone soils), as goethite and hematite may be the only source magnetic field (Siqueira et al., 2010; Camargo et al., 2016). In addition to the intrinsic characteristic of the mineral, the magnetic signal of the soil is controlled by the type of rock, climate, vegetation, relief and anthropic activity (Grimley et al., 2004; Hanesch and Scholger, 2005). Fontes et al. (2000) analyzed the χ of the sand, silt and clay fractions of ten Brazilian soils, and concluded that the magnetic behavior of the soils was coordinated by the nature of the source material. A similar result was found by Camêlo et al. (2017) in soils of different origins: basalt, tufite, itabirite, amphibolite and gabbro. Pedological studies indicate an increase in χ with the degree of weathering in well-drained soils (Camêlo et al., 2017; Poggere et al., 2018), but this is not always the case. The enrichment of magnetic minerals with the weathering of the soil seems to occur in soils formed from the same source material (Hanesch and Scholger, 2005). In the soil, the particle size of the magnetic minerals and the phenomenon of isomorphic substitution also contribute to different and wide χ values (Batista et al., 20 2010). The influence of particle size on soil χ can be inferred by the frequency dependent on magnetic susceptibility (χfd%), which, according to Dearing (1994), assumes values between 0 to 14%. Based on the size of particles, magnetic minerals can be classified into multidomains (MD: Ø between 100 - 1000 μm), pseudo-domain (PSD), single stable domain (SSD) and superparamagnetic (SP: Ø <1 μm). Very large particles have multiple magnetization zones, a typical case of magnetite, being characterized by magnetic multidomains. Small particles, less than 1 μm, characterize single-domain magnetic minerals, called superparamagnetic, such as maghemite. Thus, χfd% suggests the origin of magnetic minerals in the soil. Another common phenomenon in soil is the isomorphic substitution in Fe oxides (Cornell and Schwertmann, 2003). In general, in an advanced stage of pedogenesis it is common to exchange Fe in the crystalline structure of magnetic oxides by diamagnetic elements, such as Al (aluminum), providing a χ decay, a finding clarified by Batista et al. (2010). There is little information in the literature about maghemite, and the limit of Al and Fe isomorphic substitution is not yet known; such information would be of great value to elucidate the participation of maghemite in the mechanism of adsorption of ionic species from the soil and, consequently, for the definition of management zones. Large variations in the isomorphic substitution capacity in maghemite, from 16 to 26 mol% of Fe by Al in highly weathered soils, were found by Fontes and Weed (1991). In synthetic samples, Gillot and Rousset (1990) reported up to 66 mol% of Al in maghemite prepared from organic precursors, while Wolska and Schwertmann (1989) proposed a limit of 10 mol% for samples prepared from inorganic precursors. However, the isomorphic substitution of Fe for Al in the structure of the magnetic oxides does not only occur with Al, with a cation exchange record for Cd and Zn (Barista et al., 2008). Therefore, the replacement of Fe by diamagnetic elements reduces the magnetic signal from the soil. The magnetic signal from the soil can also be affected by agricultural activities (Barrios et al., 2017). It turns out that Fe oxides are sensitive to environmental changes (Cornell and Schwertmann, 2003). In fact, Camargo et al. (2016) found high χfd% values, above 14% in sugarcane areas, which in the past were burned to facilitate 21 harvesting. Values of this magnitude, for Dearing (1994), are strong indications of a predominance of superparamagnetic minerals, such as maghemite. Oxidation of ferrihydrite from fire in the presence of organic matter, or antiferromagnetic minerals such as goethite and hematite, becomes maghemite (Schwertmann and Fechter, 1984; Anand and Gilkes, 1987). This would be the main thesis for the high values of χ in soils of Terra Preta de Indio da Amazônia, Brazil (Oliveira, 2017), with its genesis intertwined with the practice of fire by indigenous peoples. χ has facilitated the recognition of specific management areas, assisting in the tactical and operational decisions of agricultural activities (Siqueira et al., 2010; Camargo et al., 2013; 2016; Marques Jr et al., 2015; Peluco et al., 2015; Barbosa et al., 2019) and as a digital marker of environmental quality (Hanesch et al. 2001; Jordanova et al. 2004). Barbosa et al. (2019) used χ to predict soil erodibility of a basalt-sandstone transect, aiming at a less expensive mapping. The results of Peluco et al. (2015) ratified the potential of χ in recognizing specific areas for application of phosphorus at a varied rate. Reasons such as (I) low equipment acquisition cost; (II) easy operation; (III) non-destructive technique (that is, it does not require chemical extractors); (IV) does not require prior preparation of the samples; (V) providing instant results; (VI) ease of measurements in the laboratory and in the field, (VII) and (VIII) are measures that complement many other types of environmental analysis. These advantages make χ promising in the management of agricultural activities. The proposal of the spectral signature by DRS and χ is complementary to the results of conventional analyzes, such as XRD (Silva et al., 2020), dissolution of Fe (Mehra and Jackson, 1960; Schwertmann, 1964; McKeague and Day, 1966), spectroscopy de Mössbauer (Murad and Johnston, 1987), which are routinely used in the analysis of Fe oxide minerals in a smaller amount. DRS results are also complementary to magnetic signature results (Grimley and Vepraskas, 2000; Grimley et al., 2004; Siqueira et al., 2010; Marques Jr et al., 2014) and orbital sensor spectroradiometry (Viscarra Rossel et al., 2009, 2010, 2015) and proximal (Demattê et al., 2004; 2007; 2015). Given the current scenario of modern and globalized agriculture, the technological advent of sensors makes it feasible to know the soil on a detailed scale, 22 responding to some gaps left by traditional soil mapping that limit the development of precision agriculture. Although, pioneering studies in Brazil (Camargo et al., 2008ab; Siqueira et al., 2010; Marques Jr et al., 2015; Peluco et al., 2015; Teixeira et al., 2018; Barbosa et al., 2019) have raised the potential of DRS and χ, so far they have only been implemented in smaller areas. Therefore, it is substantial to prove simpler techniques, of low cost and with known accuracy to provide opportunities for mapping soil attributes at a detailed level in large areas. 1.3 General material and methods 1.3.1 Site and geomorphological background The study area was located in the Western Paulista Plateau, which spans about 13 million hectares (roughly 45% of the São Paulo state, Brazil). Also, their locations spanned the climatic spectrum of the area. Based on the classification of Thornthwaite (1948), the climate is tropical with dry winters (C2rA′a′) in the north and northwest; humid temperate with hot summers (B4rB4′a) in the south; and temperate humid with dry winters and hot summers (B2rB3′a) in the east and southeast. The natural vegetation consists of Atlantic forest species in the west, and Savannahs in the east and southeast. Geologically, the Plateau dates from the higher Cretaceous (88–65 million years) and was formed largely (57%) from sandstones in the Vale do Rio do Peixe formation (VRP), Bauru Group, over the basaltic spills (15%) of the Serra Geral formation (SG), and other sedimentary formations (27%) (Fernandes et al., 2007) (6a). The landform map was elaborated according to the method proposed by Vasconcelos et al. (2012), based on the geomorphometric signature. The method aims to classify subtle changes in landforms, reducing the subjectivity of identifying compartments by conceptual landscape models (Troeh, 1965; Daniels et al., 1971), applicable at several scales (Teixeira et al., 2018). Information from the Shuttle Radar Topography Mission (SRTM) was used for elaborating the geomorphometric signatures. The SRTM Digital Elevation Model (DEM) is a regular grid with a spatial resolution of 03 arcsec (~90 m) and a vertical accuracy of 15 m (Smith and Sandwell, 2003). 23 Figure 6. Geological maps (Fernandes et al., 2007) (a) and landscape dissection units (Vasconcelos et al., 2012) (b) representative soil profiles in the Western Paulista Plateau: 1, 2 and 3 - Rhodic Eutrudox (LVef), 4 and 5 - Rhodic Hapludox (LVd) and 6 - Typic Kandiudalf (PVA) (Soil Survey Staff, 2010). By interpreting the landform intensity map of Fernandes et al. (2007) for the Western Paulista Plateau, Vasconcelos et al. (2012) succeeded in identifying three different levels of landscape dissection, namely: slightly dissected (Sd), moderately dissected (Md), and highly dissected (Hd) (Figure 6b). The identification of dissection levels was supplemented with field observations intended to confirm the mutual relationships between components of the physical environment. Sd units characterize the most stable landscapes, soft wavy and flat, convex and linear-convex relief, predominating at the top of the landscape, with an altitude of 450–610 m, and occupying an extension of 20,365 ha. Md units dominate the landscape, with an extension of 92,209 ha, found at altitudes of 431–529 m, and soils in evolution processes. Spanning approximately 21,629 ha, Hd units are depressed areas, where drainage networks are located, with the occurrence of less weathered soils and water table fluctuation. It has steep relief, of concave-convex type and V-shaped valleys, whose altitude varies from 250 to 550 m. 24 1.3.2 Soil sampling Soil surveys were carried out and six profiles representing the geological diversity and landscape dissection units were selected (Figure 6) in order to examine the variation of the total iron oxide (Fe2O3) content with depth as an indicator of lithological contrast. The trenches were dug at altitude from 250 m to the greatest in the Plateau (610 m). The soils are mainly Latosols (Oxisols) or Argisols (Ultisols) (Soil Survey Staff, 2010; Santos et al., 2018). The state highway map of the Brazilian roads and traffic department was used to construct a sampling plan with the aid of the ET GeoWizards tool in the software ArcView 9.3. A total of 300 samples were collected from the 0.0–0.2 m soil layer at representative points with minimal anthropic interference (Figure 6a). However, with routine laboratory analyzes, some samples were lost or contained an insufficient quantity for specific analyzes. Thus, 200 soil samples were used for the elaboration of the second chapter, and 291 for the third chapter. The shortest and longest distance between sampling points was 10 and 60 km, respectively. The points were well distributed in space to ensure that the samples would be representative of lithology and landscape dissection units (Figure 6). The number of sampling points used was based on the experience gathered in previous geostatistical studies on WPP (Marques Jr et al., 2015; Teixeira et al., 2018). 1.3.3 Laboratory analysis 1.3.3.1 Conventional soil analysis Soil samples were passed through 2 mm sieves (air-dried fine earth, ADFE) from the studied profiles (B horizons) were analyzed for Si, Al, and Fe by total analysis with sulfuric acid (H2SO4, ratio 1:1). The total contents thus determined were used to calculate Ki and Kr (Santos et al., 2018). The Ki =1.7 × %SiO2 / %Al2O3 and Kr = 1.7 × %SiO2 / (%Al2O3 + 0.6325 × %Fe2O3) indices were used to indicate the development stage of soils: more weathered–leached soils show low indices (Donagema et al., 2011; Silva et al., 2020). Crystalline free iron (Fed) was extracted with sodium dithionite–citrate– bicarbonate (DCB) at 25 °C for 16 h and determined according to 25 Mehra and Jackson (1960). Finally, poorly crystalline iron oxides (Feo) were extracted with ammonium oxalic acid and quantified according to McKeagne and Day (1966). Particle size distribution was determined by using a 0.1 M NaOH solution under slow stirring as a dispersant, and the clay fraction was quantified with the pipette method (Donagema et al., 2011). 1.3.3.2 Mineralogical analysis 1.3.3.2.1 X-ray diffraction Clay fraction of the 300 samples was used to quantify Hematite (Hm), goethite (Gt) and maghemite (Mh) by powder X-ray diffraction (XRD). The concentrated Fe- oxides fraction were obtained by boiling the clay in 5 mol L−1 NaOH (Norrish and Taylor, 1961), whereas kaolinite (Kt), gibbsite (Gb) and the Gb/(Gb + Kt) ratio were estimated after removing iron oxides from the clay fraction (Mehra and Jackson, 1960). Samples were diffracted on a Bruker D8 Advance instrument using Cu Kα radiation. The proportions of Gt, Hm and Mh in the iron oxide concentrate were calculated from the areas of the (110), (012) and (220) peaks, respectively, as well as the difference Fed– Feo. The areas for the (012) and (220) peaks were multiplied by a factor of 3.5 because there was overlap by any other peak. The total area was taken to be the combination of (110), 3.5 times (012) and 3.5 times (220) (Equations 1–3). Finally, the Kt/(Kt + Gb) ratio was calculated from the areas of the reflection peaks (002) for gibbsite and (001) for kaolinite. Gt% = ( area Gt(110) total area ) × 100 (1) Hm% = ( area Hm(012)×3.5 total area ) × 100 (2) Mh% = ( area Mh(220)× 3.5 total area ) × 100 (3) The proportions of Gt, Hm and Mh were also determined by applying the Rietveld refinement as implemented in the software Powder Cell v. 2.4. The files with 26 the crystal structures of the oxides were obtained from the American Mineralogist Crystal Structure Database (Downs and Hall-Wallace, 2003). Spectra were adjusted by using a polynomial function for the baseline and reflections with the pseudo-Voigt model, and the width at half-height was calculated from the angular parameters U, V and W: WHH = (Utan2 θ + Vtan θ + W)½. The refinement quality was assessed in terms of Rexp and goodness of fit (GOF) (Young et al., 1995). Thus, Rexp values < 10 and GOF values < 2 were deemed good, whereas Rexp values > 10 but < 20 and GOF values > 2 but < 5 were judged acceptable. 1.3.3.2.2 Diffuse reflectance spectroscopy Alternatively, samples were also analyzed by DRS. For this purpose, an amount of 1 g of air-dried fine earth (ADFE) was ground to constant color in a mortar and placed in a cylindrical holder of 16 mm in diameter. Reflectance measurements (R) were obtained with a Lambda 950 UV/Vis/NIR spectrometer coupled to an integrating sphere of 150 mm in diameter. Spectra were recorded at 0.5 nm intervals over the wavelength range of 250–2500 nm (Vis-NIR), using an integration time of 2.43 nm s−1 and Halon (PTFE) as a blank. The spectral data were used to calculate the second derivative of the Kubelka- Munk function over the wavelength range of 380–710 nm (Kubelka and Munk, 1931). A spline procedure involving 30 data points was used to estimate the Hm/(Hm + Gt) ratio according to Scheinost et al. (1998). This procedure allowed comparing DRS potential in the estimation of Gt and Hm obtained by XRD. According to Torrent and Barrón (2002), this procedure allows Gt and Hm contents lower than 0.1% – which is one order of magnitude lower than the limit for the XRD technique – to be determined. 1.3.3.3 Selective dissolution in 1.8 mol L-1 H2SO4 Triplicate 0.1 g portions of clay previously treated with boiling 5M NaOH to concentrate the iron oxides were digested with 1.8 mol L–1 H2SO4 at 75 °C for 2 h (Schwertmann and Fetcher, 1984; modified by Costa et al., 1999). In order to selectively dissolve Fe from Mh (Mh-H2SO4), the extraction procedure was optimized 27 with the aid of low-frequency magnetic susceptibility (χlf) measurements (Costa et al., 1999); thus, digestion was finished when χlf decreased to 5% of its initial value. Iron thus extracted was quantified by atomic absorption spectroscopy and the results multiplied by a factor of 1.43 to calculate the amount of Mh-Fe2O3. 1.3.3.4 Magnetic susceptibility measurements Magnetic susceptibility (χ) was measured at a low frequency (lf, 0.47 kHz) and a high frequency (hf, 4.7 kHz), using a Bartington MS2 instrument for measurements in (1) air-dried fine earth [χlf-ADFE], (2) the clay fraction and (3) the (silt + sand) fraction. Dual frequency measurements allowed us to calculate the frequency-dependent percent magnetic susceptibility (χfd%, Eq. 4) (Dearing, 1999), a proxy for the presence of single, multiple and superparamagnetic minerals of the magnetic domain. Samples with very high χfd values (≥ 14%) were reanalyzed for confirmation. χfd% = ( χlf − χhf χlf ) × 100 (4) Where, χlf, χhf and χfd% are the low-frequency, high-frequency and percent frequency- dependent magnetic susceptibility, respectively. Triplicate samples of ADFE were also extracted with DCB for 16 h (Mehra and Jackson, 1960). This was followed by centrifugation, removal of the supernatant and washing with deionized water, the process being repeated three times at 3000 rpm for 10 min. The centrifuged residue was used to measure χlf and the difference in χlf before and after extraction with DCB was assumed to correspond to maghemite (Mh-χlf-DCB). A conversion factor of 763 × 10–6 m3 kg–1 was used in Equation 5 to estimate Mh in g kg–1 (Peters and Dekkers, 2003). An identical procedure was used to quantify magnetite (McKeague and Day, 1966), the content in which was estimated from the magnetic susceptibility remaining in the residual (sand + silt) fraction [χlf-rem], using a conversion factor of 1000 × 10–6 m3 kg–1 in Equation 6. 28 Mh = ( χlf ADFE − χlf rem 763 × 10−6 ) × 10 (5) Mt = ( χlf rem × 100 1000 × 10−6 ) × 10 (6) Where, χlf and χlf-rem are the low-frequency magnetic susceptibility of the ADFE fraction before and after extraction, respectively. Maghemite was also determined in the oxide-concentrated clay fraction obtained by boiling in 5 mol L–1 NaOH, the content thus obtained being designated Mh- χlf-NaOH: Mh = ( χlf concentrated clay 763 × 10−6 ) × 10 (7) Where, χlf is the low-frequency magnetic susceptibility of the oxide-concentrated clay fraction. 1.3.4 Statistical and geostatistical analysis The analytical results were initially used to calculate means, maximum and minimum values, medians, standard deviations, coefficients of variation, asymmetry, and kurtosis. The means for landscape dissection units were compared by the Tukey’s test (5%), and Gt and Hm contents were estimated from XRD patterns, and the DRS alternative procedure was subjected to regression analysis. XRD and DRS results were compared in terms of the following parameters: coefficient of determination (R2), mean error (ME, Eq. 8), standard deviation of the error (SDE, Eq. 9), residual prediction deviation (RPD, Eq. 10), and root mean square error (RMSE, Eq. 11). The RPD is the ratio of the standard deviation of the original data by the RMSE of cross validation predictions. RPD values were classified according to Chang et al. (2001) as excellent (RPD > 2), acceptable (1.4 < RPD < 2) or unreliable (RPD < 1.4). Usually, accurate 29 models have high R2 and RPD values but low RMSE and SDE values. These parameters were calculated according to Viscarra Rossel et al. (2006). 𝑀𝐸 = 1 𝑁 ∑ (�̂�𝑖 − 𝑦𝑖)𝑛 𝑖=1 (8) 𝑆𝐷𝐸 = ∑ (�̂�𝑖−𝑦𝑖)2𝑛 𝑖=1 𝑁−1 (9) 𝑅𝑃𝐷 = 𝑆𝐷 𝑅𝑀𝑆𝐸⁄ (10) 𝑅𝑀𝑆𝐸 = √ 1 𝑁 ∑ (�̂�𝑖 − 𝑦𝑖)2𝑛 𝑖=1 (11) Where, N is the number of estimated values, �̂�𝑖 is the predicted value, and 𝑦𝑖 is the observed value; ME < 0 indicates that observed values are underestimated by predicted values, and ME > 0 indicates overestimated values. 1.3.4.1 Spatial variability Spatial variability of Hm and Gt values obtained with the two techniques was assessed geostatistically. The experimental semivariance was calculated according to Oliver and Webster (2014) criteria from the following equation: γ ^ (h)= 1 2N(h) ∑ [Z(xi)-Z(xi+h)]2N(h) i=1 (12) Where, γ ^ (h) is the semivariance at distance h, N(h) is the number of pairs used to calculate h, Z(xi) is the value of the attribute Z at position xi, and Z(xi + h) is the value of the attribute Z at a distance h from xi. The spherical mathematical model used was fitted to the variograms in terms of the number of pairs used to estimate the semivariance, the sum of the square of residuals (SQR), the presence of a sill in the variograms (Oliver and Webster, 2014), 30 and the coefficient of determination (R2). Once variograms were modeled by GS+ software (Robertson, 1998), the values corresponding to the unknown points were estimated by ordinary kriging and the maps processed with Surfer (1999) software. The degree of spatial dependence (DSD) was estimated from the ratio of the nugget effect (C0) to the sill (C0 + C1). An attribute was assumed to have high, moderate or low DSD if its C0/(C0 + C1) ratio was lower than 25%, 25–75% or higher than 75%, respectively (Cambardella et al., 1994). 1.3.4.2 Validation of spectral maps The spatial patterns for Gt and Hm obtained from XRD and DRS results were compared through traits. The percent variability at a local scale was calculated from the following equation: Δ (%)= DRS-XRD XRD ×100 (13) A Δ value of up to 20% for XRD values was taken to be acceptable here. This criterion was assumed based on the degree of similarity between DRS and XRD maps and geology (Camargo et al., 2016; Silva et al., 2020). Positive and negative Δ values were assumed to represent over- and underestimation, respectively, in an attribute. This parameter allowed us to classify Hm and Gt estimates as acceptable or unacceptable. 1.3 References Almeida JA, Torrent J, Barrón V (2003) Cor de solo, formas do fósforo e adsorção de fosfatos em Latossolos desenvolvidos de basalto do extremo-sul do Brasil. Revista Brasileira de Ciência do Solo 27:985-1002. Anand RR, Gilkes RJ (1987) The association of maghemite and corundum in Darling Range laterites, Western Australia. Australian Journal of Soil Research 35:303– 311. 31 Bahia ASRS, Marques Jr J, Siqueira DS (2015) Procedures using diffuse reflectance spectroscopy for estimating hematite and goethite in Oxisols of São Paulo, Brazil. Geoderma 5:150–156. Balsam W, Ji J, Renock D, Deaton BC, Williams E (2014) Determining hematite content from NUV/Vis/NIR spectra: Limits of detection. American Mineralogist 99:2280–2291. Barbieri DM, Marques Jr J, Pereira GT, La Scala Jr N, Siqueira DS, Panosso AR (2013) Comportamento dos óxidos de ferro da fração argila e do fósforo adsorvido, em diferentes sistemas de colheita de cana-de-açúcar. Revista Brasileira de Ciência do Solo 37:1557–1568. Barbosa RS, Marques Jr J, Barrón V, Martins Filho MV, Siqueira DS, Peluco RG, Camargo LA, Silva LS (2019) Prediction and mapping of erodibility factors (USLE and WEPP) by magnetic susceptibility in basalt-derived soils in northeastern São Paulo state, Brazil. Environmental Earth Sciences 78:1-12. Barrón V, Mello JWV, Torrent J (2000) Caracterização de óxidos de ferro em solos por Espectroscopia de Reflectância Difusa. In.: Novais RF, Alvarez VH, Schaefer CEGR (Eds.) Tópicos em ciência do solo. Viçosa: Sociedade Brasileira de Ciência do Solo, p. 139-162. Barrón V, Torrent J (1986) Use of the Kubelka-Munk theory to study the influence of iron oxides on soil color. Journal of Soil Science 37:499-510. Barrón V, Torrent J (2002) Evidence for a simple pathway to maghemite in earth and mars soils. Geochimica et Cosmochimica Acta 66:2801-2806. Batista MA, Costa ACS, Bigham JM, Santana H, Zaia DAM, Souza Junior IG (2010). Mineralogical, Chemical, and Physical Characterization of Synthetic Al- substituted Maghemites (γ-Fe2–xAlxO3). Clays and Clay Minerals 58:451–461. Bazaglia Filho O (2012) Comparação entre mapas de solos obtidos pelos métodos convencional e digital numa área complexa. 190 f. Dissertation (MSc in Soils and Plant Nutrition) – Escola Superior de Agricultura Luiz de Queiroz/USP, Piracicaba. Ben-Dor, E, Heller D, Chudnovsky A (2008) A novel method of classifying soil profiles 32 in the field using optical means. Soil Science Society of American Journal 72:1113−1123. Bigham JM, Fitzpatrick RW, Schulze D (2002) Iron oxides In.: Dixon JB, Schulze DG. Soil mineralogy with environmental applications. Madison: Soil Science Society of America, p. 323-366. (Book Series, 7). Budiman M, McBratney AB, Mendonça-Santos ML, Santos HG (2003) Revisão sobre funções de pedotransferência (PTFs) e novos métodos de predição de classes e atributos do solo. Rio de Janeiro: Embrapa Solos, 50 p. (Documento 45). Camargo LA, Marques J Jr, Pereira GT (2013) Mineralogy of the clay fraction of Alfisols in two slope curvatures: III – Spatial variability. Revista Brasileira de Ciência do Solo 37:295–306. Camargo LA, Marques Jr J, Barrón V, Alleoni LRF, Barbosa RS, Pereira GT (2015) Mapping of clay, iron oxide and adsorbed phosphate in Oxisols using diffuse reflectance spectroscopy. Geoderma 251–252:124–132. Camargo LA, Marques Jr J, Pereira GT, Alleoni LRF, Bahia ASRS, Teixeira DDB (2016) Pedotransfer functions to assess adsorbed phosphate using iron oxide content and magnetic susceptibility in an Oxisol. Soil Use and Management 32:172–182. Camargo LA, Marques Jr J, Pereira GT, Bahia ASRS (2014) Clay mineralogy and magnetic susceptibility of Oxisols in geomorphic surfaces. Scientia Agricola 71:244- 256. Camargo LA, Marques Jr J, Pereira GT, Horvat RA (2008a) Variabilidade espacial de atributos mineralógicos de um Latossolo sob diferentes formas do relevo: II - correlação espacial entre mineralogia e agregados. Revista Brasileira de Ciência do Solo 32:2279–2288. Camargo LA, Marques Jr J, Pereira GT, Horvat RA (2008b) Variabilidade espacial de atributos mineralógicos de um Latossolo sob diferentes formas de relevo. I- Mineralogia da fração argila. Revista Brasileira de Ciência do Solo 32:2269–2277. Cambardella CA, Moorman TB, Novak JM, Parkin TB, Karlen DL, Turco RF, Konopka AE (1994) Field-scale variability of soil properties in Central Iowa Soil. Soil Science Society of America Journal 58:1501-1508. 33 Camêlo DL, Ker JC, Fontes MPF, Corrêa MM, Costa ACS, Melo VF (2017) Pedogenic iron oxides in iron-rich Oxisols developed from mafic rocks. Revista Brasileira de Ciência do Solo 41:e0160379. Campos MCC, Marques Jr J, Pereira GT, Montanari R, Camargo LA (2007) Relações solo-paisagem em uma litossequência arenito-basalto na região de Pereira Barreto, SP. Revista Brasileira de Ciência do Solo 31:519-529. Cañasveras JC, Barrón V, Del Campillo MC, Torrent J, Gómez JA (2010) Estimation of aggregate stability indices in Mediterranean soils by diffuse reflectance spectroscopy. Geoderma 158:78–84. Carioca AC, Costa GM, Barrón V, Ferreira CM, Torrent J (2011) Aplicação da espectroscopia de reflectância difusa na quantificação dos constituintes de bauxita e de minério de ferro. Revista Escola de Minas 64:199–204. Carrer H, Barbosa AL, Ramiro DA (2010) Biotecnologia na agricultura. Estudos Avançados 24:149–164. Carvalho CCN, Nunes FC, Antunes MAH (2013) Histórico do levantamento de solos no Brasil: da industrialização brasileira à era da informação. Revista Brasileira de Cartografía 65:681-694. Carvalho Filho A, Inda AV, Fink JR, Curi N (2015) Iron oxides in soils of different lithological origins in Ferriferous Quadrilateral (Minas Gerais, Brazil). Applied Clay Science 118:1-7. Chen J, Ji J, Balsam W, Chen Y, Liu ZA (2002) Characterization of the Chinese loess– paleosol stratigraphy by whiteness measurement. Palaeogeography, Palaeoclimatology, Palaeoecology 183:287–297. Chinese Soil Taxonomy Research Group, Institute of Soil science, Chinese Academy Sciences, Cooperative Research Group on Chinese Soil Taxonomy (2001). Keys to Chinese Soil Taxonomy. 3rd ed. Hefei: Chinese Science and Technological University Press. Coblinski JA, Giasson É, Demattê JAM, Dotto AC, Costa JJF, Vašát R (2020) Prediction of soil texture classes through different wavelength regions of reflectance spectroscopy at various soil depths. Catena 189:104485. 34 Cornell RM, Schwertmann U (2003) The Iron Oxides: Structure, Properties, Reactions, Occurrence, and Uses. Weinheim: Wiley VCH. 664 p. Cornell RM, Schwertmann, U (1996) The Iron Oxides: Structure, Properties Reactions Occurrence and Uses. New York: Weinheim-VCH. 573 p. Costa ACS, Bigham JM, Rhoton FE, Traina SJ (1999) Quantification and characterization of maghemite in soils derived from volcanic rocks in southern Brazil. Clays and Clay Minerals 47:466–473. Coventry RJ, Taylor R, Fitzpatrick RW (1983) Pedological significance of the gravel in some red and grey earth of central North Queensland. Australian Journal of Soil Research v. 21:219–240. Curi N, Franzmeier DP (1984) Toposequence of Oxisols from the central plateau of Brazil. Soil Science Society of America Journal 48:341-346. Daniels RB, Gamble EF, Cady JG (1971) The relation between geomorphology and soil morphology and genesis. Advances in Agronomy 23:51-87. Dantas JS, Marques Jr J, Martins Filho MV, Resende JMA, Camargo LA, Barbosa RS (2014) Gênese de solos coesos do Leste Maranhense: relação solo-paisagem. Revista Brasileira de Ciência do Solo 38:1039–1050. Dearing, JA (1994) Environmental magnetic susceptibility: using the Bartington MS2 system. England: British Library. 104 p. Dearing JA (1999) Environmental magnetic susceptibility: Using the Bartington MS2 system. 2nd ed. Keniloworth: Chi Publishing. 54 p. Demattê JAM, Araújo SR, Fiorio PR, Fongaro CT, Nanni MR (2015) Espectroscopia VIS-NIR-SWIR na avaliação de solos ao longo de uma topossequência em Piracicaba (SP). Revista Ciência Agronômica 46:679– 688. Demattê JAM, Galdos MV, Guimarães R, Genú AM, Nanni MR, Zullo Jr J (2007) Quantification of tropical soil attributes from ETM+/Landsat-7 data. International Journal of Remote Sensing 8:3813–3829. 35 Demattê JAM, Genú AM, Fiorio PR, Ortiz JL, Mazza JA, Leonardo HCL (2004) Comparação entre mapas de solos obtidos por sensoriamento remoto espectral e pelo método convencional. Pesquisa Agropecuária Brasileira 39:1219–1229. Demattê JAM, Terra F, Quartaroli CF (2012) Spectral behavior of some modal soil profiles from São Paulo State, Brazil. Bragantia 71:413-423. Donagema GK, Campos DVB, Calderano SB, Teixeira WG, Viana JHM (2011). Manual de métodos de análise do solo. 2nd ed. Rev. Rio de Janeiro: Embrapa Solos. Dotto AC, Dalmolin RSD, Araújo Pedron F, Caten A, Ruiz LFC (2014) Mapeamento digital de atributos: granulometria e matéria orgânica do solo utilizando espectroscopia de reflectância difusa. Revista Brasileira de Ciência Do Solo, 38:1663–1671. Downs, RT and Hall-Wallace, M (2003) The American Mineralogist Crystal Structure Database. American Mineralogist 88, 247-250. Available on: (accessed on: Feb. 10, 2020). Fabris JD, Viana JHM, Schaefer CEGR, Wypych F, Stucki JW (2009) Métodos físicos de análises em mineralogia do solo. In.: Alleoni LRF, Melo VF (Eds.) Química e mineralogia dos solos. Viçosa: Sociedade Brasileira de Ciência do Solo, p. 611–694. Fernandes LA, Castro AB, Basilici G (2007) Seismites in continental sand sea deposits of the Late Cretaceous Caiuá Desert, Bauru Basin, Brazil. Sedimentary Geology 199:51–64. Fernandes RBA, Barrón V, Torrent J, Fontes MPF (2004) Quantificação de óxidos de ferro de Latossolos brasileiros por espectroscopia de refletância difusa. Revista Brasileira de Ciência do Solo 28:245–257. Ferreira SAD, Santana DP, Fabris JD, Curi N, Nunes Filho E, Coey JMD (1994) Relações entre magnetização, elementos traços e litologia de duas seqüências de solos do estado de Minas Gerais. Revista Brasileira de Ciência do Solo 18:167-174. Fidêncio PH, Ruisánchez I, Poppi RJ (2001) Application of artificial neural networks to the classification of soils from São Paulo state using near-infrared spectroscopy. Analyst 126:2194–2200. 36 Fitzpatrick RW, Schwertmann U (1982) Al-substituted goethite – an indicator of pedogenic and other weathering environments in South Africa. Geoderma 27:335- 347. Fontes MPF, Oliveira TS, Costa LM, Campos AAG (2000) Magnetic separation and evaluation of magnetization of Brazilian soils from different parent materials. Geoderma 96:81-99. Fontes MPF, Weed SB (1991) Iron oxides in selected Brazilian Oxisols: I. Mineralogy. Soil Science Society of America Journal 55:1143-1149. Geladi P, Kowalski BR (1986) Partial least-squares regression: a tutorial. Analytica Chimica Acta 185:1–17. Ghidin AA, Melo VF, Lima VC, Lima JMJC (2006) Topossequências de Latossolos originados de rochas balsáticas no Paraná. I - Mineralogia da fração argila. Revista Brasileira de Ciência do Solo 30:293–306. Gillot B, Rousset A (1990) On the limit of aluminum substitution in Fe3O4 and g-Fe2O3. Physica Status Solidi (a) - Applied Research 118: K5-K8. Gmur S, Vogt D, Zabowski D, Moskal LM (2012) Hyperspectral Analysis of Soil Nitrogen, Carbon, Carbonate, and Organic Matter Using Regression Trees. Sensors 12:10639–10658. Gomes, RP (2017) Gênese, mineralogia e dinâmica do fósforo nos solos do Planalto Ocidental Paulista. 76 f. Dissertação (Mestrado em Ciência do Solo) – Unesp, Jaboticabal. Grimley DA, Arruda NK, Bramstedt MW (2004) Using magnetic susceptibility to facilitate more rapid, reproducible, and precise delineation of hydric soils in the midwestern USA. Catena 58:83–213. Grimley DA, Vepraskas MJ (2000) Magnetic Susceptibility for Use in Delineating Hydric Soils. Soil Science Society American Journal 64:2174–2180. Hanesch M, Scholger R (2005) The influence of soil type on the magnetic susceptibility measured throughout soil profiles. Geophysical Journal International, 161:50–56. 37 Hanesch M, Scholger R, Dekkers MJ (2001) The application of fuzzy c-means cluster analysis and non-linear mapping to a soil data set for the detection of polluted sites, Physics and Chemistry of the Earth 26:885–891. Hu P, Liu Q, Torrent J, Barrón V, Jin C (2013) Characterizing and quantifying iron oxides in Chinese loess/paleosols: Implications for pedogenesis. Earth and Planetary Science Letters 369-370:271–283. Hunt GR, Salisbury JW, Lenhoff CJ (1971