Search

The spatiotemporal dynamic of rainfall erosivity is essential for environmental studies and guidance to control erosion. The purpose of this study is to assess rainfall erosivity (monthly and annual), testing machine learning algorithms aided by a covariate bank for spatial prediction of rainfall erosivity in Southeastern Brazil. The modeling tested Random Forest-RF, Cubist, Support Vector Machine, Earth, and Linear Model, associated with 154 covariates (topographic, climatic, and vegetation data). However, we apply the cut-off correlation function (findcorrelation) and feature selection algorithm (Recursive Feature Elimination—RFE) to select strong covariates. Our results show that the RF algorithm was more efficient in modeling (R² values between 0.29 and 0.82), whit the best metrics in the low rainfall period (winter). The modeling showed fluidity by selecting only 43 significant covariates due to the findcorrelation and RFE functions. The most important and frequent covariates in spatial modeling were coordinates, water deficit, topographical, and climatic data. In general, the spatial results show that the dynamics of rainfall erosivity is strongly affected by factors of air mass circulation, relief, and geographic position. Our approach is promising as it is a method capable of estimating rainfall erosivity in unsampled areas, capturing information from significant spatial covariates. [ABSTRACT FROM AUTHOR]

Geophysical sensors combined with machine learning algorithms were used to understand the pedosphere system and landscape processes and to model soil attributes. In this research, we used parent material, terrain attributes, and data from geophysical sensors in different combinations to test and compare different and novel machine learning algorithms to model soil attributes. We also analyzed the importance of pedoenvironmental variables in predictive models. For that, we collected soil physicochemical and geophysical data (gamma-ray emission from uranium, thorium, and potassium; magnetic susceptibility and apparent electric conductivity) by three sensors (gamma-ray spectrometer, RS 230; susceptibilimeter KT10, Terraplus; and conductivimeter, EM38 Geonics) at 75 points and analyzed the data. The models with the best performance (R2 0.48, 0.36, 0.44, 0.36, 0.25, and 0.31) varied for clay, sand, Fe2O3 , TiO2 , SiO2 , and cation exchange capacity prediction, respectively. Modeling with the selection of covariates at three phases (variance close to zero, removal by correction, and removal by importance) was adequate to increase the parsimony. The results were validated using the method "nested leave-one-out cross-validation". The prediction of soil attributes by machine learning algorithms yielded adequate values for field-collected data, without any sample preparation, for most of the tested predictors (R2 values ranging from 0.20 to 0.50). Also, the use of four regression algorithms proved to be important since at least one of the predictors used one of the tested algorithms. The performance values of the best algorithms for each predictor were higher than those obtained with the use of a mean value for the entire area comparing the values of root mean square error (RMSE) and mean absolute error (MAE). The best combination of sensors that reached the highest model performance was that of the gamma-ray spectrometer and the susceptibilimeter. The most important variables for most predictions were parent material, digital elevation, standardized height, and magnetic susceptibility. We concluded that soil attributes can be efficiently modeled by geophysical data using machine learning techniques and geophysical sensor combinations. This approach can facilitate future soil mapping in a more time-efficient and environmentally friendly manner. [ABSTRACT FROM AUTHOR]

Geophysical sensors combined with machine learning algorithms have been used to understand the pedosphere system, landscape processes and to model soil attributes. In this research, we used parent material, terrain attributes and data from geophysical sensors in different combinations, to test and compare different and novel machine learning algorithms to model soil attributes. Also, we analyzed the importance of pedoenvironmental variables in predictive models. For that, we collected soil physico-chemical and geophysical data (gamma-ray emission from uranium, thorium and potassium, magnetic susceptibility and apparent electric conductivity) by three sensors, gamma-ray spectrometer - RS 230, susceptibilimeter KT10 - Terraplus and Conductivimeter - EM38 Geonics) at 75 points and, we performed soil analysis afterwards. The results showed varying models with the best performance (R2 > 0.2) for clay, sand, Fe2O3, TiO2, SiO2 and Cation Exchange Capacity prediction. Modeling with selection of covariates at three phases (variance close to zero, removal by correction and removal by importance), demonstrated to be adequate to increase the parsimony. The prediction of soil attributes by machine learning algorithms demonstrated adequate values for field collected data, without any sample preparation, for most of the tested predictors (R2 ranging from 0.20 to 0.50). Also, the use of four regression algorithms proved important, since at least one of the predictors used one of the tested algorithms. The performances of the best algorithms for each predictor were higher than the use of a mean value for the entire area comparing the values of Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). The best combination of sensors that reached the best model performance to predict soil attributes were gamma-ray spectrometer and susceptibilimeter. The most important variables were parent material, digital elevation model, standardized height and magnetic susceptibility for most predictions. We concluded that soil attributes can be efficiently modelled by geophysical data using machine learning techniques and geophysical sensors combinations. The technique can bring light for future soil mapping with gain of time and environment friendly. [ABSTRACT FROM AUTHOR]

• Pioneer proximal radiometric and magnetic susceptibility survey in Antarctica. • Geophysical variables related to periglacial processes and pedogenesis. • Soil geophysical data helped explain pedogenesis and periglacial processes. • Geophysical data affected by cryoturbation in Turbic Cryosols. • Qualitative analysis of geophysical variables identified soil types in depth. Maritime Antarctica is still pedologically unexplored by proximal geophysical sensors. In-depth soil geophysical characterizations in depth can provide comprehensive information regarding how weathering, pedogenesis, and periglacial processes operate in periglacial environments, with advantages like non-invasive methodology, fast and precise data acquisition, practical equipment operation, and cost-effectiveness. This research aimed to carry out the first proximal gamma-spectrometric and magnetic geophysical characterization of soil profiles in maritime Antarctica and to relate the geophysical variables analyzed with periglacial processes, weathering, pedogenesis, geology, and landscape dynamics. The study was carried out in the Keller Peninsula (King George Island, Maritime Antarctica) with the pedological characterization of soil profiles, using two proximal geophysical sensors (gamma spectrometer and susceptibility meter) that quantify, respectively, soil radionuclides uranium, thorium and potassium as well as soil magnetic susceptibility. For that, twenty soil profiles were described and characterized, and soil samples were collected for physico-chemical analysis. In addition, readings were carried out with the two geophysical sensors in each soil profile, through soil horizons up to the lithic contact and/or permafrost top. We performed exploratory statistical analyses in R software as follows: visualizing the geophysical variables in soil depth for each profile and soil class, obtaining descriptive statistics among soil classes, analyzing geophysical variables by geology and soil classes, performing Spearman correlation between geophysical variables and soil physico-chemical attributes, and applying principal component analysis to pedological classes, geophysical variables, and soil physico-chemical attributes. The results showed that radionuclide contents increased with soil profile depth in all profiles and soil classes analyzed, except for Cryosols where permafrost is present, and cryoturbation was more pronounced. The potassium content was high in all soil profiles, evidencing the low degree of chemical weathering and leaching. The soil magnetic susceptibility values tended to decrease in depth, mainly where permafrost was present. The gravel content contributed more to the levels of radionuclide content. The clay content in some soil classes present contributed secondarily to uranium and thorium contents. Soil profiles located on marine terraces were influenced by periglacial reworking and deposition of different materials, showing increases in magnetic susceptibility values with soil depth by the absolute accumulation of ferrimagnetic minerals. Principal component analyses efficiently separated Regosols from other soil classes by soil κ related to lithology, whereas Cambisols and other soil classes were separated by radionuclide contents related to incipient pedogenesis. These findings may give new insights into understand this inhospitable continent. [ABSTRACT FROM AUTHOR]

• The chemical weathering intensity in Maritime Antarctica is modeled. • A geophysical survey and lithological characteristics as input data. • Proximal geophysical sensors and terrain attributes are combined for modeling. • Use of leave-one-out-cross-validation method in modeling. • Periglacial processes control the distribution of geophysical variables. The chemical weathering intensity in Antarctica is underestimated. As the chemical weathering intensity increases, hydrological, geochemical and geophysical changes occur in the different environmental spheres and at their interfaces through reactions and energy flows. Thus, once chemical weathering rates are understood and estimated, they can be used to predict and assess changes and trends in different environmental spheres. Few studies on the chemical weathering intensity have been performed in Antarctica. We used radiometric and magnetic properties associated with terrain attributes and the chemical degree of alteration of the igneous rock to model the chemical weathering intensity in Maritime Antarctica by using machine learning. Then, we related the chemical weathering intensity and geophysical variables with periglacial processes. To do this, gamma-spectrometric and magnetic readings were carried out using proximal-field sensors at 91 points located on different lithologies in a representative area of Maritime Antarctica. A qualitative analysis of chemical alteration for the different lithologies was carried out based on field observations and rock properties, and the levels of the chemical weathering degree were established. The geophysical data associated with terrain attributes were used as input data in the modeling of the weathering intensity. Then, the levels of the rock weathering degree were used as the "y" variable in the models. The results indicated that the C5.0 algorithm had the best performance in predicting the weathering intensity, and the most important variables were eTh, 40K, 40K/eTh, 40K/eU, the magnetic susceptibility and terrain attributes. The contents of radionuclides and ferrimagnetic minerals in different lithologies, concomitantly with the intensity at which chemical weathering occurs, determine the contents of these elements. However, the stability and distribution of these elements in a cold periglacial environment are controlled by periglacial processes. The chemical weathering intensity prediction model using gamma-spectrometric and magnetic data matched the in situ estimate of the chemical degree of alteration of the rock. The pyritized andesites showed the highest intensities of weathering, followed by tuffites, diorites, andesitic basalts and basaltic andesites, and the lowest weathering intensity was shown by undifferentiated marine sediments. This work highlighted the suitability of using machine learning techniques and proximal-field sensor data to study the chemical weathering process on different rocks in these important and inhospitable areas of the cryosphere system. [ABSTRACT FROM AUTHOR]

Physical and chemical remediation techniques used in contaminated areas are expensive and damaging to the soil structure. Biological alternatives, such as phytoremediation, are economical and applicable to large areas. The main limitation of phytoremediation is identifying plants that are both capable of stabilizing and/or absorbing metals from soil and adapted to edaphoclimatic conditions of the contaminated areas. The objective of this study is to evaluate the ability of plant species adapted to Brazilian semi-arid conditions to grow in soils contaminated with Pb. A greenhouse experiment was carried out in a 4 × 5 factorial: four plant species (M. oleifera, P. juliflora, A. peregrina, and U. ruziziensis) and five Pb concentrations in soil (0.0; 0.52; 1.05; 2.10, and 4.20 g kg−1). All species grew at all Pb levels, but only P. juliflora and A. peregrina did not exhibit significant reductions in most growth variables. U. ruziziensis , despite showing reductions in growth variables, was the species with the highest dry matter accumulation in both shoots and roots, in addition to accumulating higher amounts of Pb. We conclude that the species P. juliflora, A. peregrina and U. ruziziensis are more suitable for cultivation in soils containing high levels of Pb. [Display omitted] • Species adapted to the semiarid can be used in phytoremediation programs. • M. oleifera, U. ruziziensis, A. peregrine, P. juliflora showed low Pb translocation factor. • A. peregrine and P. juliflora had lower toxic effects related to increasing doses of Pb. • U. ruziziensis have the capacity to accumulate levels of Pb above 1400 mg kg−1. [ABSTRACT FROM AUTHOR]

Detecting and characterizing continuous changes on Earth's surface has become critical for planning and development. Since 2016, Planet Labs has launched hundreds of nanosatellites, known as Doves. Despite the advantages of their high spatial and temporal resolution, these nanosatellites' images still present inconsistencies in radiometric resolution, limiting their broader usability. To address this issue, a model for radiometric normalization of PlanetScope (PS) images was developed using Multispectral Instrument/Sentinel-2 (MSI/S2) sensor images as a reference. An extensive database was compiled, including images from all available versions of the PS sensor (e.g., PS2, PSB.SD, and PS2.SD) from 2017 to 2022, along with data from various weather stations. The sampling process was carried out for each band using two methods: Conditioned Latin Hypercube Sampling (cLHS) and statistical visualization. Five machine learning algorithms were then applied, incorporating both linear and nonlinear models based on rules and decision trees: Multiple Linear Regression (MLR), Model Averaged Neural Network (avNNet), Random Forest (RF), k-Nearest Neighbors (KKNN), and Support Vector Machine with Radial Basis Function (SVM-RBF). A rigorous covariate selection process was performed for model application, and the models' performance was evaluated using the following statistical indices: Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Lin's Concordance Correlation Coefficient (CCC), and Coefficient of Determination (R2). Additionally, Kruskal–Wallis and Dunn tests were applied during model selection to identify the best-performing model. The results indicated that the RF model provided the best fit across all PS sensor bands, with more accurate results in the longer wavelength bands (Band 3 and Band 4). The models achieved RMSE reflectance values of approximately 0.02 and 0.03 in these bands, with R2 and CCC ranging from 0.77 to 0.90 and 0.87 to 0.94, respectively. In summary, this study makes a significant contribution to optimizing the use of PS sensor images for various applications by offering a detailed and robust approach to radiometric normalization. These findings have important implications for the efficient monitoring of surface changes on Earth, potentially enhancing the practical and scientific use of these datasets. [ABSTRACT FROM AUTHOR]

Background and Aims: To enhance Brazilian sugarcane production sustainably, crop simulation models have been utilized. However, due to the lack of reliable information, particularly concerning soil variability, these models have shown limited performance for specific analyses. This study aims to evaluate Digital Soil Mapping (DSM) as an alternative for filling soil data gaps in crop modeling and to assess the influence of these products on prediction uncertainties. The study site is located in Piracicaba region, Southern Brazil. Methods: The framework was: (i) a legacy soil data were utilized, and equal-spline equations were applied to standardize the dataset.; (ii) a machine learning (ML) algorithm was used to predict soil attributes and their uncertainties; (iii) pedotransfer functions were applied to obtain soil hydrological properties; (iv) DSSAT/CANEGRO crop model was used to estimate sugarcane yield; (iv) a legacy soil map (LSM), SoilGrids (SG) and a map of attributes derived from regional DSM (RDSM) were compared; (v) a Monte Carlo Simulation (MCS) was conducted with the RDSM maps to evaluate the impact of uncertainties in the estimation of sugarcane yield. Results: The DSM proved to be a reliable source for use in crop models, reaching similar results to field data. The sugarcane yield map emphasized the model's sensitivity to soil attributes, with texture and depth significantly impacting yield estimations. Conclusion: In this sense, coupling DSM and crop modeling is a feasible way to improve yield estimates, especially in countries with limited soil databases. Highlights: • Crop simulation models have limited application due to the lack of soil data. • Digital Soil Mapping was coupled to a sugarcane simulation model to fill the gap of soil information. • Soil attributes and their uncertainties were predicted on a 250-m grid using machine learning algorithm. • A spatially-explicit DSSAT/CANEGRO model was able to represent variations in sugarcane yield at the regional scale; • Sugarcane yield was strongly affected by soil variability and its uncertainties; • Our finds indicate the importance of detailed soil databases and their impact on yield predictions. [ABSTRACT FROM AUTHOR]

Maritime Antarctica (M.A.) contains the most extensive and diverse lithological exposure compared to the entire continent. This lithological substrate reveals a rich history encompassing lithological, pedogeomorphological, and glaciological aspects of M.A., all influenced by periglacial processes. Although geophysical surveys can detect and provide valuable information to understand Antarctic lithologies and their history, such surveys are scarce on this continent and, in practice, almost non-existent. In this sense, we conducted a pioneering and comprehensive gamma-spectrometric (natural radioactivity) and magnetic susceptibility (κ) survey on various igneous rocks. The main objective was to create ternary gamma-ray and κ maps using machine learning algorithms, terrain attributes, and a nested-leave-one-out cross-validation method. Additionally, we investigated the relationship between the distribution of natural radioactivity and κ to gain insights into pedogeomorphological and periglacial processes and dynamics. For that, we used proximal gamma-spectrometric and κ data in different lithological substrates associated to terrain attributes. The geophysical variables were collected in the field from various lithological substrates, by use field portable equipment. The geophysical variables were collected in the field from various lithological substrates using portable equipment. These variables, combined with relief data and lithology, served as input data for modeling to predict and spatially map the content of radionuclides and κ by random forest algorithm (RF). In addition, we use nested-LOOCV as a form of external validation in a geophysical data with a small number of samples, and the error maps as evaluation of results. The RF algorithm successfully generated detailed maps of gamma-spectrometric and κ variables. The distribution of radionuclides and ferrimagnetic minerals was influenced by morphometric variables. Nested-LOOCV method evaluated algorithm performance accurately with limited samples, generating robust mean maps. The highest thorium levels were observed in elevated, flat, and west beach areas, where detrital materials from periglacial erosion came through fluvioglacial channels. Lithology and pedogeomorphological processes-controlled thorium contents. Steeper areas formed a ring with the highest uranium contents, influenced by lithology and geomorphological-periglacial processes (rock cryoclasty, periglacial erosion, and heterogeneous deposition). Felsic rocks and areas less affected by periglacial erosion had the highest potassium levels, while regions with sulfurization-affected pyritized-andesites near fluvioglacial channels showed the lowest potassium contents. Lithology and pedogeochemical processes governed potassium levels. The κ values showed no distinct distribution pattern. Pyritized-andesite areas had the highest levels due to sulfurization and associated pyrrhotite, promoting iron release. Conversely, Cryosol areas, experiencing freezing and thawing activity, had the lowest κ values due to limited ferrimagnetic mineral formation. Lithology and pedological-periglacial processes in Cryosols played a significant role in controlling κ values. In regions characterized by diverse terrain attributes and abundant active and intense periglacial processes, the spatial distribution of geophysical variables does not reliably reflect the actual lithological composition of the substrate. The complex interplay of various periglacial processes in the area, along with the morphometric features of the landscape, leads to the redistribution, mixing, and homogenization of surface materials, contributing to the inaccuracies in the predicted-spatialized geophysical variables. [ABSTRACT FROM AUTHOR]

Understanding the weathering intensities can provide answers for environmental issues, soil, and geoscience studies. Recently, geophysical approaches and machine learning techniques have been applied in soil science to access weathering. This study aimed to model weathering intensity using combined data from geophysical sensors, satellite images, and morphometry associated to machine learning algorithms. We also we evaluated the efficiency of nested-leave one out cross-validation applicability in a small geodata set evaluated the importance of covariates and the resulting weathering intensity map in relation to pedogeomorphological processes. Our study focused on a 184-ha area in southwest Brazil, where we conducted soil analysis at 71 sites. We applied principal component analysis and determined the ideal number of clusters to determine the classes of weathering intensity. We used six geophysical sensor parameters, including equivalent uranium, equivalent thorium, potassium 40, magnetic susceptibility, and soil apparent electrical conductivity, along with the weathering index to create the clusters. Then, four machine learning algorithms were used to infer different weathering intensities in soils formed from the same parent materials. To validate our results, we used the nested-leave-one-out-cross-validation ("nested-LOOCV") method, which is suitable for small datasets. Our findings showed that the random forest model performed the best with three clusters as the ideal number. We also found that the geophysical data, clusterization, and machine learning algorithm contributed significantly to identifying different weathering intensities. The results indicated that weathering operated at different intensities on both the diabase/Rhodic-Nitisols and the siltite/metasiltite Rhodic and Xanthic Lixisols areas, with the highest intensities occurring in the west Xanthic Lixisols and the lowest intensities occurring in the Rhodic and Lixisols in the east area. The siltite/metamorphosed siltite and Lixisols areas presented moderate weathering rates. We found that the all-geophysical variables used were related and affected by weathering intensity, which contributed to the modeling and clusterization processes. • Combined use of geophysical sensors in modeling soil weathering intensity. • Nested Leave One Out Cross-Validation for external validation in small dataset. • Use pedoenvironmental covariates and satellite image to determine ideal number of clusters. • Geophysical data, machine learning and clustering identified different weathering intensities. • Use of clusters in modeling processes. [ABSTRACT FROM AUTHOR]

• We propose a hybrid model to predict soil depth. • Soil depth prediction by mechanistic, empirical, and hybrid models was compared. • The mechanistic model performed better for shallow soils. • The Empirical model was more suitable for deeper soils. • The hybrid model showed the best performance. Soil thickness is an important property as it influences the landscape dynamics, partakes part in hydrologic and geomorphologic processes, and controls water saturation and soil moisture, which are directly related to agricultural production. However, soil thickness data are difficult to obtain in situ , especially in areas with deep soils (>2 m). In this study, we developed and compared three models to predict soil thickness. First, we developed a mechanistic model which uses physical equations from a landscape evolution model applied to a digital elevation model (DEM) (30 m spatial resolution). We evaluated the inclusion of parameters derived from the soil parent material, including erosion and sediment deposition. Second, we developed an empirical model using terrain derivatives obtained from a 30-m DEM, using a Random Forest algorithm. This model was calibrated using 1,362 soil thickness data collected in field as right censored data. We implemented a hybrid model using the residual from the mechanistic model as a dependent variable in the empirical model. The models were validated with 214 soil observation points collected in field as right censored data and 12 data points with real data. The result was added back to predictions of the mechanistic model. For all models, we verified coherence with a soil map at 1:100,000. The models were also evaluated considering changes in the spatial resolution. The mechanistic model was improved when parent material parameters were added. The mechanistic models performed better in areas with shallow soils (<1 m), whereas the empirical model was better in predicting deeper soils and was more coherent with a soil class map. Model performance could be further improved when updating DEM data to a 5-m resolution. As expected, the hybrid model could combine the model performances and improve the predictions. However, the predictions remained poor for shallow soils. [ABSTRACT FROM AUTHOR]

Intensive cropland expansion for an increasing population has driven soil degradation worldwide. Modeling how agroecosystems respond to variations in soil attributes, relief and crop management dynamics can guide soil conservation. This research presents a new approach to evaluate soil loss by water erosion in cropland using the RUSLE model and Synthetic Soil Image (spectroscopy technique), which uses time series remotely sensed environmental, agricultural and anthropic variables, in the southeast region of São Paulo State, Brazil. The availability of the open-access satellite images of Tropical Rainfall Measuring Mission (TRMM) and Landsat satellite images provided ten years of rainfall data and 35 years of exposed soil surface. The bare soil surface and agricultural land use were extracted, and the multi-temporal rainfall erosivity was assessed. We predict soil maps' attributes (texture and organic matter) through innovative soil spectroscopy techniques to assess the soil erodibility and soil loss tolerance. The erosivity, erodibility, and topography obtained by the Earth observations were adopted to estimate soil erosion in four scenarios of sugarcane (Saccharum spp.) residue coverage (0%, 50%, 75%, and 100%) in five years of the sugarcane cycle: the first year of sugarcane harvest and four subsequent harvesting years from 2013 to 2017. Soil loss tolerance means 4.3 Mg ha−1 exceeds the minimum rate in 40% of the region, resulting in a total soil loss of ~6 million Mg yr−1 under total coverage management (7 Mg ha−1). Our findings suggest that sugarcane straw production has not been sufficient to protect the soil loss against water erosion. Thus, straw removal is unfeasible unless alternative conservation practices are adopted, such as minimum soil tillage, contour lines, terracing and other techniques that favor increases in organic matter content and soil flocculating cations. This research also identifies a spatiotemporal erosion-prone area that requests an immediately sustainable land development guide to restore and rehabilitate the vulnerable ecosystem service. The high-resolution spatially distribution method provided can identify soil degradation-prone areas and the cropland expansion frequency. This information may guide farms and the policymakers for a better request of conservation practices according to site-specific management variation. [ABSTRACT FROM AUTHOR]

The potential of using spectroscopy for the quantification of soil attributes through its spectral signature is widely documented in the literature. However, a protocol to support formal soil classification systems combining spectral data has not been established. This research proposed a protocol for soil profile classification by combining spectral data from the near visible, shortwave infrared (Vis-NIR-SWIR) and mid infrared (MIR). For this purpose, we used 15 soil profiles located in the Pernambuco State, Brazil. A quantitative analysis between soil attributes and spectral curves was performed for the selection of bands with the best correlations (method 1). In addition, the recursive feature elimination (RFE) function was used for the selection of discriminant bands between soil profiles (method 2). The results of this research indicated that the combined use of spectra is efficient to successfully grouping Ferralsol, Gleysol, and Acrisol. The integrated use of sensors, pedometric techniques, and the expertise of soil scientists can lead to an advanced understanding of soil science. [ABSTRACT FROM AUTHOR]

Soil apparent electrical conductivity (ECa) is related to various soil attributes and processes. This research aimed to understand the relationship between ECa and paedogenesis, lithology and attributes. The study area is located in S~ao Paulo State, Brazil. Soil samples were collected for physical-chemical analysis from 79 locations (0–20 cm layer). A geophysical field-portable equipment (Geonics EM38-MK2 conductivity meter) was used to measure soil ECa. For that, four toposequences were selected on landscape. Sixteen soil profiles were allocated in the landscape for soil characterization and classification. The ECa values were measured across toposequences, soil types and lithology variations, at the same 79 locations used for physico-chemical characterization. The statistical analysis and Kruskal–Wallis rank sum test and Cluster analysis by k-means and Wilcoxon–Mann–Whitney test was performed to verify differences in ECa values over lithology, relief and soil types. The values of ECa measurements were related to soil attributes variation. The results showed that lithology strongly affects ECa values, mainly in diabase-derived soils, followed by metamorphosed siltstone. In fluvial sediments, the ECa exhibits different behaviours. In less evolved soils, the lithology contributed more to ECa than paedogenesis, as the opposite occurs in more evolved ones. The ECa decreases downslope and correlates with lithology alteration and soil types. This study demonstrates that soil ECa can be a tool to differentiate between various soil types and lithological transitions across the landscape, supporting the soil survey and mapping at finer scales. [ABSTRACT FROM AUTHOR]

Despite occupying a large area of the globe and being the next agricultural frontier, sandy soils are seldom explored in scientific studies. Considering the high capacity of remote sensing in soil characterization, this work aimed to: (i) characterize sandy soils' profiles from proximal sensing; (ii) assess the ability of visible, near, and short-wave infrared (Vis-NIR-SWIR) as well as mid-infrared (MIR) spectroscopy to distinguish soil classes of highly sandy content; (iii) quantify physical and chemical attributes of sandy soil profiles from Vis-NIR-SWIR and MIR spectroscopy as well as X-ray fluorescence (pXRF). Samples were described and collected from 29 sandy soil profiles. The 127 samples went under Vis-NIR-SWIR and MIR spectroscopy, X-ray fluorescence, and chemical and physical analyses. The spectra were analyzed based on "Morphological Interpretation of Reflectance Spectrum" (MIRS), Principal Components Analysis (PCA), and cluster methodology to characterize soils. The integration of different information obtained by remote sensors, such as Vis-NIR-SWIR, MIR, and Portable X-ray Fluorescence (pXRF), allows for pedologically complex characterizations and conclusions in a short period and with low investment in analysis and reagents. The application of MIRS concepts in the VNS spectra of sandy soils showed high potential for distinguishing pedological classes of sandy soils. The MIR spectra did not show distinct patterns in the general shapes of the curves and reflectance intensities between sandy soil classes. However, even so, this region showed potential for identifying mineralogical constitution, texture, and OM contents, assuming high importance for the complementation of soil pedometric characterizations using VNS spectroscopy. The VNS and MIR data, combined or isolated, showed excellent predictive performance for the estimation of sandy soil attributes (R2 > 0.8). Sandy soil color indices, which are very important for soil classification, can be predicted with excellent accuracy (R2 from 0.74 to 0.99) using VNS spectroscopy or the combination of VNS + MIR. [ABSTRACT FROM AUTHOR]

The District of Mariana in Minas Gerais, Brazil, witnessed one of the largest natural disasters in history. It involved the deposition of thousands of cubic meters of mining waste on soils near tributaries of the Doce River, resulting in degraded areas and a new environmental landscape. Numerous studies have focused on changes in soil microbiota and physical-chemical attributes. However, studies quantifying and analyzing carbon flux (FCO 2) dynamics in the field using proximal sensors under varying temperatures, humidity conditions, and vegetation cover are scarce and nearly nonexistent. These studies could aid in monitoring the recovery of degraded areas and in understanding FCO 2 dynamics in such environments. The aim of this study was to quantify and assess FCO 2 dynamics using combined sensing techniques in diverse mining-affected areas, comparing different stages of recovery and their vegetation impact, and linking the results to pedoenvironmental factors. To achieve this, four areas were carefully selected and evaluated: affected pasture (AP), pasture (P) (non-affected), mix (MIX), and native trees (NT). In each area, readings were taken using an Infrared gas analyzer (IRGA) sensor, and temperature and soil moisture were measured. Additionally, soil samples were collected for chemical, physical, and biological characterization. Normalized Difference Vegetation Index (NDVI) assessments were conducted using satellites to evaluate vegetation cover. Non-parametric tests (Kruskal Wallis and Principal Component Analyses - PCA) were employed to evaluate differences between the areas and to assess the correlation among the environmental variables. The P area had the highest FCO 2 and total organic carbon (TOC), differing significantly from the other site areas. Conversely, AP had the lowest FCO 2 and was distinct from other sites. The MIX and NT areas had intermediate FCO 2 and TOC values, which were statistically similar to each other. PCA identified distinct patterns in FCO 2 , soil temperature, and moisture. FCO 2 was positively correlated with soil temperature and negatively correlated with moisture. There was a relationship between the biological variables microbial quotient (qMic), metabolic quotient (qCO 2), COT, and FCO 2. qMic reached the highest values in the MIX area, decreasing linearly from NT to P. Conversely, AP had the lowest qMic. qCO 2 had the highest value in AP and NT. The proximal IRGA sensor effectively quantified FCO 2 and differentiated mining tailing-affected areas. This assessment incorporated soil temperature, moisture sensors, vegetation index, and soil attribute data. The FCO 2 levels were higher in P areas and lower in AP areas. Elevated FCO2 levels were correlated with a high soil temperature and low humidity. In AP, qMic was low, and qCO 2 was high, indicating lower FCO 2 levels. Conversely, P exhibited a higher FCO2. Higher NDVI values were correlated with elevated FCO 2 in areas with specific vegetation cover during environmental recovery, while lower NDVI areas had lower FCO2 levels. • Application of IRGA to monitor environmental changes in tailing-affected areas. • Soil CO 2 flux dynamics and emissions were evaluated by sensing techniques. • Integration of pedoenvironmental variables with proximal and remote sensing techniques to enhance environmental monitoring. • Representative data on carbon flux dynamics in tailing-affected areas. • Technonosol CO 2 flux varies in tailing-affected areas by land cover. [ABSTRACT FROM AUTHOR]

Abstract Brazil has extensive forests and savannas on deep weathered soils and plays a key role in the discussions about carbon sequestration, but the distribution of soil organic carbon (SOC) stocks up to 1 m depth has not been investigated in Brazil using machine learning techniques. In this study, we applied a methodological framework to optimize the prediction of SOC stocks for the entire Brazilian territory and determine how the environmental heterogeneity of Brazil influences the SOC stocks distribution. We used a legacy dataset of 8227 soil profiles which consisted of 37,693 samples. For each profile, the vertical distribution of SOC and bulk density were interpolated to standard depths (0–5, 5–15, 15–30, 30–60 and 60–100 cm) using mass preserving equal-area quadratic splines. The covariates database was composed of 74 variables including bioclimatic (temperature and precipitation) data, soil and biome maps, vegetation indexes and morphometric maps derived from a digital elevation model, with a 1 km spatial resolution. To obtain the best prediction performance, we tested four machine learning algorithms: Random Forests, Cubist, Generalized Linear Model Boosting and Support Vector Machines. Random Forests showed the best performance in predicting SOC stocks for all depths, with the highest performance at 30–60 cm for training (R2 = 0.32) and validation (R2 = 0.33); hence, it was selected for the spatial prediction of SOC stocks. The most important covariates selected by Random Forests using the recursive feature elimination were: soil class, sum of monthly mean temperature (SAMT), precipitation, slope height and vegetation indexes (NDVI, GPP). In total, Brazilian soils store approximately 71.3 PgC within the top 100 cm, where the first 0–30 cm contains almost 36 PgC. Approximately 31% of the total SOC stocks (22.2 PgC) occurs in protected areas (2.6 million km2), which are not subjected to land use pressure and carbon losses. Although the Amazon biome has the highest amount of stored SOC (36.1 PgC), its soils do not represent a good potential for carbon accumulation. Among soil classes, the Luvisols showed the lowest SOC density (6.45 kg m−2) and the Histosols presented the highest values (14.87 kg m−2). More than 57% of the total SOC was found in nutrient-poor, deep-weathered Ferralsols and Acrisols, which are the dominant soils in Brazil. The presented methodological framework covers all steps of prediction process, building maps with known accuracy and has great potential to be used in future soil carbon inventories at large scales. Concerning conservation issues, the results highlight the importance of nature reserves for protecting SOC in the long-term. Highlights • We modelled and mapped the SOC stocks and uncertainty at 1 m soil depth for Brazilian territory. • Random Forest showed highest performance for SOC stocks prediction. • Soil class, climate variables and vegetation indexes most influence SOC stocks. • Amazon biome holds the highest total SOC stocks. • Protected areas hold about 31% of the total SOC stocks in Brazil. [ABSTRACT FROM AUTHOR]

• Orbital and proximal sensors predict sand subfractions proportions in sandy soils. • Particle sizes content were identified by proximal and orbital sensing. • SWIR and MIR regions are important for sandy soils study and their subfractions. • Bare topsoil reflectance, by orbital level, can identify sandy subfractions. • Orbital sensing could define potential uses for sandy soils. Sandy soils, which expressly cover 7% of the Earth's land surface, are known for their management complexity and their significant influence on the proportion of sand subfractions in terms of physical, chemical, and physical-hydric properties. Proximal and remote sensing techniques offer cost-effective ways to improve soil evaluation. To address this gap, this study evaluates the potential of different sensing techniques, including laboratory-based analysis using the FieldSpec 3 spectroradiometer equipment and satellite imagery, to characterize and estimate sandy soil texture fractions and subfractions. We first defined 216 samplings in Mato Grosso State, Brazil considering the pedology, geology, synthetic soil image (exposed soil), curvature and slope of the terrain. This sampling location include sandy soils with different proportions of sand subfractions and mineralogy within a region of agricultural expansion. The sensing data used in the study were composed of total element concentrations obtained by pXRF, Vis-NIR-SWIR, and MIR spectra. The satellite data were obtained from exposed soil images of Landsat 5 and Sentinel 2, and also from simulations of Landsat 5, Sentinel 2, and Terra (ASTER). The proximal and satellite data were descriptively analysed and used to estimate the levels of clay, total sand (TS), very coarse sand (VCS), coarse sand (CS), medium sand (MS), fine sand (FS) and very fine sand (VFS). It was observed, at both the proximal and the satellite levels, that the reflectance intensity of sandy soils is inversely proportional to the particle diameter of the predominant sand subfraction. Proximal level models were slightly more accurate in predicting the texture fractions of sandy soils compared to models based on satellite data (mean validation R2 0,45 and 0,40, respectively). In both models, SWIR and MIR stand out as key predictor variables. The results obtained in this study can be implemented to optimise the expansion of agricultural frontiers in sandy soils in other areas. [ABSTRACT FROM AUTHOR]

The continental margin of Northeast Brazil (NE Brazil) is the most seismogenic of the passive margin of South America and has several evidence of neotectonics. However, the duration and magnitude of neotectonic events cannot be uniformly replicated for a scale of ∼3000 km of extension. Therefore, this study aimed to understand the tectonic dynamics, especially from Neogene to recent, for a region little studied in this scope. We conducted multiscale analysis: landforms, sub-basins, drainage systems, and sedimentary outcrops. Applying a multi-technical approach: geomorphic indices (Stream Length-Gradient-Sl, Channel Steepness k sn , Basin Asymmetry Factor); use of Machine Learning Algorithms (ML) for spatialization of Sl and k sn indices; and OSL-palynology chronology. The results indicate multiphase tectonic control occurring by reactivation of geological faults from the Cretaceous to the recent, affecting shear zones, Cretaceous Basins, and Miocene deposits (Barreiras Formation). In these areas, predictions by ML (Sl and k sn indices) showed high values associated with fault zones, which confine and direct drainage, and landforms conditioning ungraded river long-profiles and tilted sub-basins. Tectonic deformations also manifest at the outcrop scale, with strata showing deformations (folds, joints, and atypical dips). Furthermore, strata with marine palynomorphs (Miocene) and Pleistocene-Holocene sediments (Post-Barreiras) in an altitudinal position above ∼38 m above sea level suggest post-depositional movement. The multiscale and multi-technical approaches showed that tectonic activity acted in different periods. Notably, the spatial prediction of geomorphic indices with ML is potentially competent in regional analyzes of neotectonics. [Display omitted] • Geomorphic indices and sediment analysis explain passive margin neotectonics. • Machine learning algorithms enable spatialization of Sl and k sn geomorphic indices. • Neotectonics affected Miocene and Pleistocene sediments in northeastern Brazil. • Marine palynomorphs occur in sediments tens of meters above the present sea level. [ABSTRACT FROM AUTHOR]

• We apply the Random Forest algorithm to predict regolith thickness. • The most important covariates were the drainage density and east–west direction. • The average regolith thickness for the Quadrilátero Ferrífero was 125 m. • The model explained only about 40% of the regolith thickness variation. The Quadrilátero Ferrífero (QF) in Brazil is a region of great economic, social and environmental importance, involving conflicts of interest due to land use and heavily pressured by iron mining and urban sprawl. This is an area of environmental importance due to supporting rupestrian vegetation areas on ferruginous substrates and springs from important watersheds. Thus, studies that can bring more knowledge about this region becomes important to support future decisions based on technical information. We used Random Forests algorithm and several databases to model the regolith thickness of the entire QF region and we also created an individual model to predict regolith thickness in the lithostratigraphic unit Minas Supergroup that contains most of the drillhole samples. The regolith thickness modeled for the QF region presented an average of 125.32 m and R2 of 0.38, and the specific model for the Minas Supergroup also predicted the average regolith thickness of 125.39 m and R2 of 0.39. These values are in accordance with the average regolith thickness (124 m) data from the drillhole database obtained from exploratory programs for iron ore in the QF region. The most important predictive covariates included drainage density, east–west direction, terrain texture and vertical distance from drainage. This study is the first attempt to model the regolith thickness in this important region and the analysis of model uncertainty can orient future studies. [ABSTRACT FROM AUTHOR]

There are different methods for predicting streamflow, and, recently machine learning has been widely used for this purpose. This technique uses a wide set of covariables in the prediction process that must undergo a selection to increase the precision and stability of the models. Thus, this work aimed to analyze the effect of covariable selection with Recursive Feature Elimination (RFE) and Forward Feature Selection (FFS) in the performance of machine learning models to predict daily streamflow. The study was carried out in the Piranga river basin, located in the State of Minas Gerais, Brazil. The database consisted of an 18-year-old historical series (2000–2017) of streamflow data at the outlet of the basin and the covariables derived from the streamflow of affluent rivers, precipitation, land use and land cover, products from the MODIS sensors, and time. The highly correlated covariables were eliminated and the selection of covariables by the level of importance was carried out by the RFE and FFS methods for the Multivariate Adaptive Regression (EARTH), Multiple Linear Regression (MLR), and Random Forest (RF) models. The data were partitioned into two groups: 75% for training and 25% for validation. The models were run 50 times and had their performance evaluated by the Nash Sutcliffe efficiency coefficient (NSE), Determination coefficient (R2), and Root of Mean Square Error (RMSE). The three models tested showed satisfactory performance with both covariable selection methods, however, all of them proved to be inaccurate for predicting values associated with maximum streamflow events. The use of FFS, in most cases, improved the performance of the models and reduced the number of selected covariables. The use of machine learning to predict daily streamflow proved to be efficient and the use of FFS in the selection of covariables enhanced this efficiency. • The Forward Feature Selection (FFS) was more advantageous for selecting covariables. • In most cases, the FFS selected fewer covariables to predict daily streamflow. • The machine learning models with covariable selection were efficient in predicting streamflow. • The most important covariable was the streamflow of affluent rivers lagged in time. [ABSTRACT FROM AUTHOR]

Focused on supporting the planning and management of water resources in a watershed domain, this study aimed to assess and compare the performance of multiple approaches for streamflow regionalization. The applied methods were based on linear interpolation, regional regressions (RR), and kriging. The dependent variables used were low-flows of reference (Q 7.10 , Q 95 , and Q 90) and the long-term mean streamflow (Q mld). The independent variables used by the RR method were drainage area, mean annual precipitation, mean precipitation in the dry and rainy seasons, besides the flow equivalent to the precipitated volume minus the 750 mm abstraction (Peq750). The RR's equations were best fitted using the potential model, and three hydrologically homogeneous regions (RHH) were identified. Meanwhile, the exponential model provided the best fit to the experimental semivariograms in the geostatistical analysis. The methodologies' performance was estimated by the following metrics: mean relative error, percent bias, root mean square error, Nash-Sutcliffe coefficient, and coefficient of determination. The evaluated methodologies had similar performance for the low-flows and long-term mean streamflow, despite the lower accuracy by using the kriging methods for the Q 7.10 at the RHH III. All approaches performed better for Q mld , since it is not an extreme variable. Although all methods have shown good results, the adoption of kriging and linear interpolation must always be linked with a careful analysis of factors such as the density and spatial distribution of stream gauge stations in the watershed. • Very similar performance between the traditional and geostatistical approaches. • The methodologies showed better prediction for the long-term mean streamflow. • Specific flow approach was mostly affected by differences in areas size involved. • The specific flow and geostatistical approaches need careful analysis beforehand. • The geostatistical methods showed to be promising for poorly gauged basins. [ABSTRACT FROM AUTHOR]

This study aims to assess different machine learning approaches for streamflow regionalization in a tropical watershed, analyzing their advantages and limitations, and to point the benefits of using them for water resources management. The algorithms applied were: Random Forest, Earth and linear model. The response variables were the three types of minimum streamflow (Q 7.10 , Q 95 and Q 90), besides the long-term average streamflow (Q mld). The database involved 76 environmental covariates related to morphometry, topography, climate, land use and cover, and surface conditions. The elimination of covariates was performed using two processes: Pearson's correlation analysis and importance analysis by Recursive Feature Elimination (RFE). To validate the models, the following statistical metrics were used: Nash-Sutcliffe coefficient (NSE), percent bias (PBIAS), Willmott's index of agreement (d), coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE) and relative error (RE). The linear model was unsatisfactory for all response variables. The results show that nonlinear models performed well, and their covariate of greatest predictive importance was flow equivalent to the precipitated volume, considering the subtraction of an abstraction factor of 750 mm (Peq750). Generally, the Random Forest and Earth models showed similar performances and great ability to predict the minimum streamflow and long-term average streamflow assessed, constituting powerful and promising alternatives for the streamflow regionalization in support to the management and integrated planning of water resources at the level of river basins. • Peq750 was the most significant covariate for Random Forest and Earth. • Random Forest and Earth exhibited good performance for streamflow prediction. • Random Forest is more recommended for predicting Q 7.10 and Q 90. • Random Forest and Earth are promising alternatives for streamflow regionalization. [ABSTRACT FROM AUTHOR]

• The conversion of pasture into secondary forests (SF) improves soil chemical quality. • SF have highest values of macroporosity, soil water infiltration and carbon storage. • Macropores had greatest contribution to soil water infiltration in the SF. • Meso- and micropores had greatest effect on soil water infiltration in the pasture. • Saturated hydraulic conductivity was smaller in the western SF than eastern SF. The recovery of soil hydraulic properties after the conversion of pasture areas into forests is still poorly studied. The aim of this research was to evaluate the effect the vegetation has on the hydraulic properties of soil, as well as the physical and chemical properties of the soil in pasture and secondary forest areas, located in a hill with east and west sun exposed faces. The study was carried out in two 15-year-old secondary forest areas and two pasture areas. Undeformed soil samples were collected in the upper third of each slope using steel rings at depths of 0–10 cm, for laboratory analyses of the physical properties and the saturated hydraulic conductivity of the soil. Additional samples were collected to evaluate the total organic carbon, the carbon storage and the chemical characteristics of the soil. A tension infiltrometer was used to determine the hydraulic properties and the contribution of the macropores to the infiltration of water into the soil. The use of a unmanned aerial vehicle helped to identify the soil cover at different locations and different faces of sun exposure. Secondary forests showed the higher values of macroporosity, soil water infiltration, and carbon storage than pasture areas. Macropores had greatest contribution to soil water infiltration in the secondary forests, whereas the mesopores and micropores had greatest effect in the pasture areas. The high content of organic matter found in the forests resulted in soils with better chemical quality. The saturated hydraulic conductivity and the carbon storage of the soil were smaller in the western forest than eastern forest. These results indicate that in order to increase water flow and water quality it is required to encourage the use of vegetative practices such as silvopastoral systems in pastures and the revegetation of the upper third of the hills. [ABSTRACT FROM AUTHOR]