ABSTRACT Most Brazilian municipalities do not have regulated areas for solid waste disposal in civil construction. Usually, residues are disposed of vacant lots and dumps, posing risks to the population health and the environment. Soils are the primary means for the disposal or recycling of waste, highlighting the importance of well-characterized soils and their respective landscape. This study aimed to establish a land suitability system for solid residues in civil construction and apply such information in a case study in Southeastern Brazil. An unprecedented digital soil map with a resolution of 30 m was created using the random forest classifier algorithm and soil field prospection information. A guide listing favorable soil-landscape attributes that most prevent soil erosion, water bodies or water table contamination was elaborated and discussed. Thus, such information was linked through a suitability system to classify areas with potential for receiving waste on a daily volume basis as follows: large size: >500 m³ day-1, medium size: >100 m³ day-1 and <300 m³ day-1, and small size <100 m³ day-1. Topography and soil depth were the most limiting factors of the areas in the case study. The proposed attributes as criteria for the suitability system complement the current state legislation. A total of 236 ha closer to the urban perimeter connected by roads in good condition were classified as suitable for managing medium- and small-scale daily volume, whose destination might reduce transportation and installation costs in the study area. Usually dumps environment wellcharacterized well characterized landscape Brazil 3 soillandscape erosion discussed Thus follows 500 >50 day1, day1 day 1, 1 100 >10 day- 300 <30 <10 day1. 1. legislation 23 smallscale scale area 50 >5 10 >1 <3 <1 2 5 > <

ABSTRACT Electromagnetic sensors are widely used to monitor soil water content (θ); however, site-specific calibrations are necessary for accurate measurements. This study compares regression models used for calibration of soil moisture sensors and investigates the relation between soil attributes and the adjusted parameters of the specific calibration equations. Undisturbed soil samples were collected in the A and B horizons of two Ultisols and two Inceptisols from the Mantiqueira Range in Southeastern Brazil. After saturation, the Theta Probe ML2X was used to obtain the soil dielectric constant (ε). Several readings were made, ranging from saturation to oven-dry. After each reading, the samples were weighted to calculate θ (m3 m–3). Fourteen regression models (linear, linearized, and nonlinear) were adjusted to the calibration data and checked for their residue distribution. Only the exponential model with three parameters met the regression assumptions regarding residue distribution. The stepwise regression was used to obtain multiple linear equations to estimate the adjusted parameters of the calibration model from soil attributes, with silt and clay contents providing the best relations. Both the specific and the general calibrations performed well, with RMSE values of 0.02 and 0.03 m3 m–3, respectively. Manufacturer calibration and equations from the literature were much less accurate, reinforcing the need to develop specific calibrations.

ABSTRACT Rainfall erosivity (R factor) is one of the six factors of the Universal Soil Loss Equation, being calculated based on the product of rainfall kinetic energy multiplied by its 30-minute maximum intensity. However, the lack of detailed and reliable rainfall data in many parts of the world has driven the use of other methods to estimate rainfall erosivity based on daily, monthly or annual data. These methods still need to be assessed to determine if their estimates are consistent with the standard method for calculating rainfall erosivity. This study aimed to select a consistent method for such replacement in Brazilian conditions without access the rainfall intensity data. The tested methods included: modified Fournier, MF; modified Fournier by Zhang, MF-Z; modified Fournier by Men, MF-M; Rainfall Disaggregation, RD; TRMM Satellite with modified Fournier coefficient, TRMM-F; and TRMM Satellite with monthly rainfall, TRMM-M. The rainfall data were obtained from the USP Meteorological Station, referring to the period from 2009 to 2015. The analyses were performed according to the Additive Main effects and Multiplicative Interaction (AMMI) model and Scott-Knott statistical tests. Considering the 1:1 line, all methods had a good adjustment, presenting similar behavior in relation to the standard method. The methods behaved differently for monthly and annual periods. The MF method proved to be capable of consistently replacing the standard method in all aforementioned situations. Considering the driest period, any method can be used. For annual rainfall erosivity estimation, the RD, MF, TRMM-F and TRMM-M methods can be applied; highlighting that the TRMM-based methods are optimal for locations without on-site rain gauges. Additionally, it was computed that the modified Fournier by Men and the modified Fournier by Zhang underestimated and overestimated the rainfall erosivity, respectively.

Abstract Portable X-ray fluorescence (pXRF) spectrometry offers valuable information for prediction models of soil fertility attributes spatial variation, although this approach is yet scarce in tropical regions. This study aims to predict and build spatial variability maps of soil pH, remaining phosphorus (P-Rem), soil organic matter (SOM) and sum of bases (SB) using pXRF results through stepwise multiple linear regression (SMLR) and Random Forest (RF) in a highly variable tropical area. Composite samples from soil A horizon were collected at 90 points throughout the campus of the Federal University of Lavras, Minas Gerais, Brazil, for pH, P-Rem, SOM, SB and pXRF analyses. RF predictions showed the highest accuracies, especially for P-Rem and SB (R² values of 0.66 and 0.55, respectively). Attributes that showed higher R² in punctual predictions also exhibited higher R² in spatial predictions. Data obtained from pXRF in tandem with RF can be used to assist prediction models for soil fertility attributes, consequently enabling the digital mapping of such attributes and helping to improve the knowledge about the spatial variability of such attributes in soils of tropical climate. This technique can therefore assist in the identification and orientation of adequate management practices in tropical agricultural practices.

RESUMO Há uma crescente demanda por mapas de solo, já que informações detalhadas de solos estão diretamente relacionadas com agricultura, urbanização e conservação ambiental. Porém, há escassez de mapas de solo em grande escala em países tropicais em desenvolvimento como o Brasil. Apesar de existirem mapas para pequenas áreas, grandes regiões possuem mapas pouco detalhados. Considerando-se a importância de se buscarem alternativas para superar a falta de informações detalhadas de solo, o objetivo deste trabalho foi extrapolar mapas locais para áreas maiores semelhantes que necessitam de informações mais detalhadas. A bacia do rio Anhumas, no município de Itajubá, sudeste do Brasil, foi manualmente mapeada e este mapa foi utilizado para predizer a distribuição de solos para todo o município. Primeiro, o modelo de predição foi testado na mesma bacia e alcançou acurácia global de 67% e coeficiente Kappa de 0,62. Depois, os resultados foram usados conjuntamente com dados da bacia José Pereira para predizer as classes de solo de todo o município, alcançando acurácia global de 54% e coeficiente Kappa de 0,40. Dados de material de origem em baixa resolução confundiram os modelos; as predições obtiveram melhores resultados quando esta variável foi removida. O mapa de solos de Minas Gerais apresenta apenas a classe de solo Argissolo e suas associações com outros solos para a área. O mapa confeccionado por este trabalho, porém, identificou mais classes de solo. O mapeamento de áreas representativas e sua extrapolação para áreas maiores constitui alternativa promissora para superar a escassez de mapas detalhados de solo.

ABSTRACT There is an ever-growing need for soil maps, since detailed soil information is directly related to agricultural activities, urbanization and environmental protection. However, there is a lack of large-scale soil maps in developing tropical countries such as Brazil. Albeit there are soil maps for small areas, large regions usually have undetailed maps. Considering the importance of finding low-cost alternatives to overcome the lack of detailed soil information, the main objective of this work was to manually create a local soil map and extrapolate it to similar larger areas that lack detailed soil information. The Anhumas River Basin, in the municipality of Itajubá, southeast Brazil, was manually mapped and this map was used to predict soils distribution for the entire municipality. First, the prediction model was tested in the same basin and provided sufficient results, achieving 67% global accuracy and 0.62 Kappa coefficient. Second, the resulting map was used together with the soil map of the larger José Pereira Basin to map the entire municipality, achieving 54% global accuracy and 0.40 Kappa coefficient. Low resolution parent material information was found to confuse models; maps showed better results when this variable was removed. The Minas Gerais soil map presents general mapping units only for the Acrisol class and its associations with other soil classes in the area. The soil map predicted by this work identified more soil classes. Mapping representative areas and extrapolating these maps to larger similar areas constitute a promising alternative to overcome the lack of detailed soil maps.

RESUMO Latossolos são os principais solos do Brasil e classicamente ocorrem em superfícies geomórficas suavizadas e estáveis. No entanto, no Sul de Minas Gerais, Latossolos têm sido observados em relevo forte ondulado, uma condição rara para estes solos nesta região fisiográfica. Este trabalho teve como objetivo acessar a dinâmica elementar e o relevo com uso de espectrometria de fluorescência de raios-X portátil (pXRF), em associação à geomorfologia, para entender tal ocorrência. Os solos estudados estão localizados no município de Muzambinho, Estado de Minas Gerais, Brasil. Perfis de Latossolo foram caracterizados no terço superior (P1), terço médio (P2) e terço inferior (P3) da encosta. Após a descrição morfológica, amostras de solo foram coletadas nos horizontes A, AB e Bw dos perfis. Foram realizadas análises físicas, químicas e espectrométricas dos solos em laboratório. Os teores de Ti e Fe (pXRF) tiveram alta correlação com as frações texturais dos solos. Foram encontrados altos graus de intemperismo-lixiviação dos Latossolos, embora ocorram em encosta íngreme onde solos mais jovens (Cambissolos) seriam esperados. A concordância de altitude dos topos observada nas áreas circundantes mais elevadas indica um paleorrelevo plano, ajudando a explicar esta ocorrência anormal. Os valores dos índices de intemperismo dos solos diminuíram ao longo da encosta, indicando o enriquecimento de sílica por escoamento superficial e fluxo de água subsuperficial nas posições mais baixas da paisagem. As análises de pXRF ajudaram na caracterização destes solos velhos em topografia forte ondulada. O equipamento forneceu bases para confirmação dos seus altos graus de intemperismo, além de detectar evidências do pré-intemperismo do regolito dessa paisagem muito antiga.

ABSTRACT Oxisols are the main soils in Brazil and they classically occur in stable and gentle geomorphic surfaces. However, in the south of Minas Gerais State, Oxisols have been observed under strongly undulated relief, a very rare condition for these soils in this physiographic region. Therefore, this work aimed to assess the elemental dynamics and relief enhanced with portable X-ray fluorescence (pXRF) spectrometry, associated to geomorphology, to understand such occurrence. The studied soils are located in Muzambinho municipality, Minas Gerais State, Brazil. Oxisol profiles were characterized in the upper third (P1), middle third (P2) and lower third (P3) of the hillslope. After morphological description, soil samples were collected in the A, AB and Bo horizons of the profiles. Physical, chemical and pXRF analyses of the soils were carried out in the laboratory. Ti and Fe content (pXRF) had a high correlation with the soil textural fractions. High weathering-leaching degrees were determined, although they occur on a steep slope where younger soils (Inceptisols) would be expected in this region. The altitudinal concordance of summits observed in the more elevated surrounding areas indicates a flat paleorelief, helping to explain this unusual occurrence. The soils’ weathering degree values decreased down the hillslope, indicating silica enrichment through runoff and subsuperficial water flux in the lower landscape positions. The pXRF analyses assisted the characterization of these old soils occurring in steep relief. It provided bases for confirming their high weathering degree indexes and detected evidences of the pre-weathering of the regolith of this very old landscape.

RESUMO Mundialmente, a degradação do solo devido à erosão hídrica tem sido um dos maiores problemas ambientais nas regiões tropicais, especialmente sob as severas condições dos Andes. Este estudo objetiva detalhar o risco do solo à erosão hídrica numa bacia hidrográfica produtora de água, localizada nos Andes colombianos (Bacia do rio Combeima), usando o modelo RUSLE e discutindo como reduzir os processos erosivos nas condições ambientais da região (forte declividade, variabilidade climática, diferentes classes e propriedades de solos, e usos da terra). O modelo RUSLE foi aplicado com auxílio de SIG para estimar as taxas de erosão atual e potencial na bacia a partir de sua calibração para regiões montanhosas, identificando áreas mais vulneráveis à erosão hídrica. A taxa de erosão atual em 50,5% da área total da bacia é maior que 25 t ha-1 ano-1, o que significa que a bacia apresenta um risco muito alto à erosão hídrica, com 30,4% de sua área com risco severo (> 100 t ha-1 ano-1). Foi possível concluir que os atuais usos da terra e os sistemas de manejo do solo não têm sido efetivos na redução da erosão hídrica, principalmente em áreas com topografia muito declivosa, demonstrando a necessidade de planejamento sustentável para melhor controlar as perdas de solo, água, carbono orgânico, nutrientes e outros elementos (não-nutrientes) nesta região dos Andes tropicais.

ABSTRACT Globally, soil degradation by water erosion has become one of the major environmental problems in tropical regions, especially under the severe environmental conditions of the Andes. This study aims to detail the soil risk to degradation by water erosion in a water producer basin located in the Tropical Colombian Andes (Combeima River basin), applying the RUSLE model, discussing how to minimize the erosion processes under its environmental conditions (steep slope, climatic variability, soil classes and properties diversity, and alternative land uses). RUSLE was applied with the support of GIS to estimate current and potential risk to soil erosion in the basin, allowing the identification of areas more prone to degradation. It was found that currently, 50.5% of the basin’s area presents, on average, annual soil losses greater than 25 Mg ha-1 yr-1, meaning a very high risk to water erosion, with 30.4% showing a severe risk (> 100 Mg ha-1 yr-1). It was possible to conclude that the current land uses and soil management systems have not been effective in mitigating soil erosion, mainly when situated in steep topography. Therefore, it is necessary sustainable planning for the conservation of soil, water, organic carbon, plant nutrients, and other elements (not-nutrients) in this tropical Andes region.

ABSTRACT: In Southeast Brazil, the change of grape harvest from wet summer to dry winter through double-pruning management has improved the quality of wines, currently denominated winter wines. In order to better understand the influences of soil, macroclimate, and vineyard management in winter wines, we investigated seven vineyards in the states of Minas Gerais (Três Corações – TC, Três Pontas – TP, Cordislândia – COR, São Sebastião do Paraíso – SSP and Andradas – AND) and São Paulo (Itobi – ITO and Espirito Santo do Pinhal – PIN) during three consecutive growing seasons. The vineyards are located in warm temperate zones and grouped in four soil types: Acrudox in TC, AND and SSP, Hapludox in TP; Hapludult in AND and PIN; Eutrudept in ITO. The high clay content (> 35 %) observed in all soil types, associated to low evapotranspiration demand, avoided the occurrence of severe water stress, as observed by the high values of leaf and stem water potential, stomatal conductance, photosynthesis, and transpiration. Differences in vigor were more related to vineyard management and did not affect grape composition. Among vineyards, parameters for berry quality from ITO, such as sugar and acidity, were more associated to high soil sand content and winter temperature. No significant differences were found in anthocyanins and total phenols of berries among vineyards, suggesting that the high thermal range and low precipitation during autumn-winter, historically observed in all municipalities, seemed to be the main factor for improvement of phenolic compounds.

RESUMO A espectrometria portátil de fluorescência de raios-X (pXRF) tem sido útil em todo o mundo para determinar o teor dos elementos no solo em condições de campo e de laboratório. No entanto, os resultados obtidos em campo podem ser influenciados por vários fatores, como umidade (U), textura (T) e matéria orgânica do solo (MOS). Assim, o objetivo deste trabalho foi criar modelos matemáticos lineares para a conversão dos teores dos elementos obtidos por pXRF em campo para resultados obtidos em laboratório, i.e., na Terra Fina Seca ao Ar (TFSA), utilizando U, T e MOS como variáveis auxiliares, uma vez que elas influenciam as leituras. As análises com pXRF foram realizadas em 12 perfis de solo com diferentes materiais de origem, seguidas por coleta de 59 amostras. Leituras com pXRF foram realizadas também em laboratório em amostras de TFSA. Os dados de pXRF obtidos em campo foram utilizados sozinhos ou combinados aos dados de U, T e MOS como variáveis auxiliares, para criar modelos de regressão linear para predição dos resultados de pXRF em TFSA. A acurácia dos modelos foi calculada pelo método leave-one-out cross-validation. À exceção de elementos mais leves, as leituras de campo com pXRF subestimaram o teor total dos elementos. Modelos de predição incluindo T apresentaram maior acurácia na predição de Al2O3, SiO2, V, Ti e Zr, enquanto a predição dos teores de Fe e K2O foi insensível à adição das variáveis auxiliares. A melhora relativa (MR) nos modelos de predição foi maior nas predições de SiO2 (T+MOS: MR = 22,29%), V (U+T: MR = 18,90%) e Ti (T+MOS: MR = 11,18%). Este trabalho demonstrou que é possível a correção dos dados de pXRF obtidos em campo através de modelos de regressão linear.

ABSTRACT Portable X-ray fluorescence (pXRF) spectrometry has been useful worldwide for determining soil elemental content under both field and laboratory conditions. However, the field results are influenced by several factors, including soil moisture (M), soil texture (T) and soil organic matter (SOM). Thus, the objective of this work was to create linear mathematical models for conversion of Al2O3, CaO, Fe, K2O, SiO2, V, Ti and Zr contents obtained by pXRF directly in field to those obtained under laboratory conditions, i.e., in air-dried fine earth (ADFE), using M, T and SOM as auxiliary variables, since they influence pXRF results. pXRF analyses in field were performed on 12 soil profiles with different parent materials. From them, 59 samples were collected and also analyzed in the laboratory in ADFE. pXRF field data were used alone or combined to M, T and SOM data as auxiliary variables to create linear regression models to predict pXRF ADFE results. The models accuracy was assessed by the leave-one-out cross-validation method. Except for light-weight elements, field results underestimated the total elemental contents compared with ADFE. Prediction models including T presented higher accuracy to predict Al2O3, SiO2, V, Ti and Zr, while the prediction of Fe and K2O contents was insensitive to the addition of the auxiliary variables. The relative improvement (RI) in the prediction models were greater in predictions of SiO2 (T+SOM: RI=22.29%), V (M+T: RI=18.90%) and Ti (T+SOM: RI=11.18%). This study demonstrates it is possible to correct field pXRF data through linear regression models.

ABSTRACT Hydrosedimentological modeling is a useful tool to predict the water dynamic in a basin and for water resources management. This study aimed to i) evaluate the ability of Soil and Water Assessment Tool (SWAT) to model sediment load and continuous monthly streamflow in the Mortes River Basin (MRB) in Southeastern Brazil; ii) estimate the sediment yield spatially distributed by sub-basins; iii) estimate the sediment load export to the Funil Hydroelectric Power Plant reservoir (FHPP), located in the MRB outlet. For the sensitivity analysis, calibration, and uncertainty analysis of the model, a semi-automatic calibration in SWAT-CUP version 5.1.6 software with the “Sequential Uncertainty Fitting” algorithm was used. To evaluate the ability of SWAT to reproduce the continuous MRB monthly streamflow and sediment load, statistical indexes, and graphical analyses were used to compare the simulated and observed data. For the sediment evaluation, a spatial and temporal comparison of sediment yield maps was used as well as the sediment yield observed in sub-basins, aiming to identify the areas with a more significant contribution to the sediment generation in the basin. The results demonstrated that SWAT performed satisfactorily in simulating both monthly sediment load and streamflow. For discharge calibration, 99 % of the measured data were bracketed by the 95 % prediction uncertainty (95PPU), and for validation, 97 % of the data were bracketed by the 95PPU, which indicates proper bracketing of the measured data within model prediction uncertainty. Uncertainty analysis indicated that 95PPU could capture 78 % of the sediment loads measured during the calibration and 72 % of the measured data during the validation period at MRB. The hydrologic response unit with pasture and Argissolos (Ultisols), Neossolos Litólicos (Entisols), and Cambissolos (Inceptisols) combined with undulated relief were the main areas responsible for the highest sediment contributions. The sediment load delivered to the reservoir from its filling 2002 to 2015 was estimated in 6,682,704 m3 (16,706,761 Mg) (density of 2.5-Mg m-3) which value corresponded to 2.6 % of storage capacity (water plus sediment) in 14 years. These results are strategic since to become feasible identifying priority areas for soil and water conservation practices as well as useful information for water resources planning and management in the studied basin.

ABSTRACT Sorption of selenate (SeO 4 2- ) and selenite (SeO 3 2- ) is poorly understood in Brazilian agroecosystems, especially in soils from agricultural areas containing different contents of competing anions, such as sulfate (SO 4 2- ). This study aimed to assess the sorption behavior of selenate and selenite at different soil layers of a tropical soil treated with different rates of agricultural gypsum (thus, containing different contents of sulfate), collected under a coffee plantation. Soil samples from an experimental area where phosphogypsum has been previously applied at different rates (0, 7, 14, and 56 t ha -1 ) were taken at the following soil layers: 0.15-0.25, 0.35-0.45, and 1.25-1.35 m. Adsorption experiments were carried out adding 20 mL of solutions containing 100 and 500 μg L -1 of selenate and 10 and 15 mg L -1 of selenite to 2 grams of soil. Desorption experiments were also performed using a soil:solution ratio of 1:10. Adsorption of selenate increased with soil depth and decreased upon increasing sulfate contents in the soil, by contrast, selenite was consistently adsorbed at higher contents - when compared with selenate - at any soil depth and its sorptive behavior was not affected by the presence of sulfate. Furthermore, selenite was less desorbed than selenate under all conditions. In conclusion, selenite is much more retained in tropical soils and less available to plants than selenate. Also, although sulfate has shown to be able to hinder selenate retention, it has no substantial effect on the sorption behavior of selenite in tropical agroecosystems.

ABSTRACT Sulfuric acid digestion analyses (SAD) provide useful information to environmental studies, in terms of the geochemical balance of nutrients, parent material uniformity, nutrient reserves for perennial crops, and mineralogical composition of the soil clay fraction. Yet, these analyses are costly, time consuming, and generate chemical waste. This work aimed at predicting SAD results from portable X-ray fluorescence (pXRF) spectrometry, which is proposed as a “green chemistry” alternative to the current SAD method. Soil samples developed from different parent materials were analyzed for soil texture and SAD, and scanned with pXRF. The SAD results were predicted from pXRF elemental analyses through simple linear regressions, stepwise multiple linear regressions, and random forest algorithm, with and without incorporation of soil texture data. The modeling was developed with 70 % of the data, while the remaining 30 % was used for validation through calculation of R2, adjusted R2, root mean square error, and mean error. Simple linear regression can accurately predict SAD results of Fe2O3 (R2 0.89), TiO2 (R2 0.96), and P2O5 (R2 0.89). Stepwise regressions provided accurate predictions for Al2O3 (R2 0.87) and Ki - molar weathering index (SiO2/Al2O3) (R2 0.74) by incorporating soil texture data, as well as for SiO2 (R2 0.61). Random forest also provided adequate predictions, especially for Fe2O3 (R2 0.95), and improved results of Kr - molar weathering index (SiO2/(Al2O3 + Fe2O3)) (R2 0.66), by incorporation of soil texture data. Our findings showed that the SAD results could be accurately predicted from pXRF data, decreasing costs, time and the production of laboratory waste.

ABSTRACT Through the lack or non-use of conservationist criteria for adequate land use and management, the scarcity of natural resources becomes ever more evident. This study aimed to analyze the origin of the sediments in the Posses catchment, municipality of Extrema, state of Minas Gerais, Brazil, throughout the fingerprinting technique and portable X-ray fluorescence. Samples from soils under agriculture, pasture, and roads; and from the subsoil of theses land uses were taken in a widespread and representative manner from the entire Posses catchment. Lag deposits and river bed sediment samples were collected downstream from the catchment outlet. A total of 45 geochemical elements were analyzed in the samples by a portable X-ray fluorescence device (pXRF). The outlier test, Kruskal-Wallis test, multivariate discriminant analysis, and a mixing model were used to estimate the contribution of each source in relation to the sediments that arrive at the mouth of the catchment. The elements selected as geochemical tracers were Sr, Al2O3, Ba, Rb, Ti, Fe, and Zn, which combined correctly discriminated 81 % of the sediment sources. The largest and smallest proportion of sediment from the Posses catchment outlet comes from rural roads and agriculture, respectively. The contribution of the subsoil was higher for lag deposits or lower for river bed sediments, than the pasture. There was a low degree of uncertainty (<8 %) for predictions made by the model employed. The types of use, selected as potential sediment sources in the Posses catchment, are adequately discriminated through the geochemical tracers quantified through the pXRF. The fingerprinting technique estimates that the contributions to outlet sediments are dominated by rural roads, following by subsoil or pasture (depending on the type of sediment evaluated) and by agriculture. The sediment sampling strategies used in this study provided similar results for the period studied. Our results showed the potential of the fingerprinting technique and the pXRF for use as tools by the program of Payment for Environmental Services in the monitoring of catchment areas.

ABSTRACT In the tropics, water erosion is one of the most important factors leading to the degradation and deterioration of agricultural land. Olive orchards have a low canopy coverage, especially during the first years after planting, due to the low density of olive trees. Given the fast expansion of olive orchards in Brazil, this study aimed to evaluate the effect of cover vegetation on soil and water losses under natural rainfall. In addition, it was assessed the crop performance and the vegetation cover index in different management systems in olive orchards. The study was carried out in soil erosion plots, where water and sediment were sampled and measured over two crops season, under the following treatments: in the first season, bare soil with olive cultivation (OBS); olive trees intercropped with spontaneous vegetation (OSV); olive trees intercropped with jack beans (OJB); olive trees intercropped with millet (OM) and, as a control, only bare soil (BS). In the second season, the OM treatment was replaced by olive trees intercropped with sunn hemp (OSH). On bare soils, soil loss was the highest reaching 303.9 Mg ha-1 yr-1 and where the surface runoff amounted to 484.8 mm yr-1. However, in the absence of competition for resources with other crops, olive trees performed best under this system. The olive orchards planted in shallow and sloping soils without cover crops showed unsustainable soil loss, crusting, and sealing in the superficial soil layer, which can progress quickly for soil degradation in the future. The efficiency in the reduction of loss in relation to bare soil was 4.11 and 12.93 % for the soil loss and 12.15 and 25.17 % for water loss, respectively, for olive with spontaneous vegetation and olive with jack beans. Cover crops combined with olive trees, and reconciled with the crop performance aspects of cultivation in tropical regions, is of great relevance for improving sustainability, especially regarding the reduction of soil and water losses due to water erosion.

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