BR102017017235A2 - ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE ROUTINE - Google Patents

Abstract

The field of the invention is focused on agribusiness, and comprises the provision of a new method, that is, a process, and its system involving product (hardware + software) to improve the chemical and physical analysis procedures of soil fertility in Large-scale routine, through a technology package dedicated to soil analysis, combining: - the use of a vis-nir spectrophotometer, called specsolo-scan with its digital platform; - multivariate analysis chemometric algorithms and; - integrated cloud-based software to produce analytical results. For routine application of the infrared spectroscopy technique to determine soil chemical and physical attributes, it is necessary to construct calibration models for each of the attributes that will be analyzed in the soil, associating the result of the wet analysis, performed according to with the reference methods, with the corresponding spectrophotometer of each soil sample, acquired through the spectrophotometer visible from the invention, or specsolo-scan.

Description

“ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN A LARGE SCALE ROUTINE”

FIELD OF INVENTION [0001] The present invention deals with “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE ROUTINE”, whose field of application is aimed at agribusiness.

[0002] In particular, the product of the invention can be used (i) by agronomic laboratories that provide services in soil fertility analysis; (ii) by sectors that use soil analysis information to guide their clients (rural producers) on the chemical management of soil fertility and plant nutrition (eg cooperatives, agronomic consultancies, fertilizer and pesticide industries, agricultural resales) , agricultural groups, sugar and alcohol plants); or (iii) by rural producers.

BACKGROUND OF THE INVENTION

Importance of Soil Analysis [0003] Soil has a very important property, which is the ability to retain and yield, under ionic form, certain chemical elements that are vital for the development of living beings. This characteristic is basically exercised by the soil colloids, be they organic (“humus”) or mineral (“clay”) in nature, due to the presence of surface charges, usually negative, thus giving the soil the capacity adsorb cations (Daniel Vidal Perez, Embrapa Solos).

[0004] Soil is a renewable natural resource that plays a fundamental role in agricultural productivity, as it contains essential nutrients for plants in its composition. A fertile soil has a great capacity to supply water and nutrients to plants, but their fertility can vary a lot,

Petition 870170088564, of 11/16/2017, p. 10/65

2/44 on a single farm. Therefore, the farmer needs to know the soil of his crop, which is only possible with the specific analysis (E mb ra pa site: https://www.embrapa.br/busca-de-noticias/-/noticia/17162564 / innovative technology-analyzes-soils-in-just-30-seconds).

[0005] The planning of fertilization and liming must be carried out based on accurate information on the level of fertility of the plots to be cultivated, which is done through the analysis of soils, followed by the interpretation of the results based on the knowledge of the areas. requirements of the species to be cultivated which is made by an Agronomist.

[0006] Soil analysis is a standard procedure for any and all crops and is carried out by a network of public and private laboratories.

[0007] The soil analysis also indicates the level of acidity of the soil through the pH index, which is the reference used to calculate the need for its correction through liming or plastering. The optimum pH index for each cultivated species, associated with the characteristics of the acidity concealer to be used, is applied to appropriate formulas, through which the amount to be applied is calculated. (Site

Cnptia / Embrapa: https: //www.agencia.cnptia.embrapa.br/Repositorio/AG01 2 2982005 81532.html).

[0008] Soil analysis is one of the best management practices (MPM) that continues to be important in both developed and developing countries because it is agronomically correct, profitable and environmentally responsible.

[0009] Soil analysis should be one of the most important management practices for crop production and environmental protection in the future. It will certainly be listed among the best management practices (MPM) to be used by agronomists from industries and universities, consultants and farm administrators for the benefit of farming customers. Soil analysis should be a planning tool and support service for the

Petition 870170088564, of 11/16/2017, p. 11/65

3/44 handling for fertilizer suppliers to use for the benefit of their customers. The environmental benefits of improved management of soil resources and fertilizer materials are remarkable when nutrient MPM are adapted to specific fields or areas of the farmer.

[0010] Statistics on fertilizer use and crop production indicate that the soil fertility of many properties may be declining due to deficit in nutrient management. The consequences of "mining" (more removal than application) of soil nutrients may not become apparent for several years. (IPNI, International Soil Fertility Manual, 1998).

[0011] Therefore, the agricultural practice of analyzing the soil is fundamental to improve crop productivity, increase the farmer's income, in addition to contributing to the sustainable use of inputs. Because of this, it is extremely important that the analysis is carried out with quality by the laboratory, since limestone and fertilizers are major investments and are a significant part of the production cost. In addition, the efficient use of these products is critical to the overall profits of the activity and environmental protection.

STATE OF THE ART [0012] The discovery of the spectral region of the near infrared is credited to the German Astronomer based in England, Sir Friedrich Wilhelm Herschel (Herschel, 1800). The work entitled “Experiments on the Refrangibility of the Invisible Rays of the Sun” was published in 1800 in the scientific journal “PhilosophicalTransactions of the Roayl Societyof London”. Herschel's discovery happened accidentally through an experiment considered simple today: sunlight was refracted in the colors of the rainbow with the aid of a prism, and calibrated thermometers were placed in each color to measure their temperatures. Herschel was surprised to find that a thermometer that was positioned just after the red region, that is, the invisible region, had a higher temperature than the visible portion of the spectrum. Given this fact, the scientist

Petition 870170088564, of 11/16/2017, p. 12/65

4/44 carried out a series of experiments involving thermometers to verify that there was heat production due to the incidence of an invisible portion of sunlight located close to red, known today as the Near Infrared region.

[0013] In addition to Herschel, other scientists contributed to the understanding of electromagnetic radiation and spectroscopy - study of the interaction of electromagnetic radiation with matter. The first near-infrared spectrum was measured photographically by Abney and Festing in 1881 in the 700 to 1200 nm region (Emil Ciurczackand III Drennen, 2002). In the early 20th century (around 1900), Coblentz used a prism made of saline material to build a primitive infrared spectrophotometer. During the Second World War (1939-1945) there was a great demand for the production of synthetic rubber, and the first industrial infrared spectrophotometers were built to meet this demand. However, these instruments operated in the middle infrared range, and later research showed that this spectral range was more suitable for elucidating structures than for quantifying substances.

[0014] With the explosion of the area of organic synthesis in the 1950s, practically every laboratory had a medium infrared spectrophotometer used for the identification of pure substances and characterization of materials. At that time, there was also the emergence of instruments that operated in the ultraviolet and visible range (UV-Vis) to complete the medium infrared range, and very little was invested in the production of instruments that operate in the NIR range (780 to 2500 nm ). However, these UV-Vis instruments came from the factory with the NIR region available, which made it possible to start the pioneering work developed by Karl Norris, an employee of the United States Department of Agriculture, who started a series of innovative experiments applying the NIR spectroscopy to measure chemical properties in agricultural products (Emil W. Ciurczakand III Drennen, 2002). Therefore, Karl Norris is credited with pioneering the

Petition 870170088564, of 11/16/2017, p. 13/65

5/44 establishment of NIR spectroscopy as an analytical tool as we know it today. However, NIR spectroscopy as an analytical tool only reached popularity among chemists from the 1990s onwards, driven by the development of analytical instrumentation and computers.

[0015] The region of the near infrared is characterized by vibrational transitions of first, second and third overtones (overtones) and by bands of combination of fundamental transitions coming from CH, OH, NH and SH connections. Due to the nature of these vibrations, the NIR spectrum is characterized by broad and overlapping bands, of low intensity when compared to the medium infrared. These characteristics make the NIR spectrum difficult to interpret univariate (band by band), requiring the use of multivariate methods of data analysis called by chemists Chemistry.

[0016] There are countless possibilities for the application of NIR spectroscopy in several areas of science, with emphasis on practical applications in the development of technologies aimed at meeting the demands of the agricultural sector, including soil analysis. Current research indicates that NIR spectroscopy is the most promising alternative for determining soil fertility parameters in total or partial replacement to traditional wet methods. A search of recent publications in the literature using the terms “Near infrared Spectroscopy” and “Soil” as topics in the Web of Science database, shows that in the last 3 (three) years (2016, 2015, 2014 and 2013) more than 400 (four hundred) works were published proposing NIR spectroscopy as an analytical tool for the analysis of soil attributes. About 94% of these works were published in the English language, with an emphasis on the fact that 95% are scientific articles and only 0.5% patent registrations.

Petition 870170088564, of 11/16/2017, p. 14/65

6/44 [0017] Thus, it is important to define the state of the art a brief review of the literature on the use of NIR spectroscopy for soil analysis and the main chemometric methods commonly used in the treatment and validation of the data, which can be verified in the descriptions of the techniques presented below.

Near infrared spectroscopy applied to soil analysis [0018] Most applications of infrared spectroscopy in Soil Science have been carried out in the medium and near infrared region associated or not with the visible (Vis / NIR, 400-2500 nm) , in qualitative and quantitative analyzes (Madari et al., 2006; Mccarty et al., 2002). The radiation in the infrared region is energetic enough to cause vibrational energy transitions, which basically results in a chemical and physical fingerprint of the soil sample, as shown in the graph shown in Figure 2 (Fidencio and Poppi, 2002; Stenberg et al. , 2010).

[0019] The graph above shows the average spectrum of soil samples representative of Brazil. The spectra were acquired in a dispersive type VisNIR spectrophotometer (NIRS XDS, Metrohm®) in the 400 to 2500 nm range. The bands classified as first, second, third overtones and combination bands were assigned according to work published by Stemberg et al. (2010).

[0020] Infrared spectroscopy techniques (near and medium) are being increasingly used in the quantification of various soil attributes, both those related to chemical composition (for example, organic carbon content, cation exchange capacity, pH) and mineralogical, as well as the physical parameters of soils (Canasveras et al., 2010; Demattê et al., 2004; Fidencio and Poppi, 2002; Fontán et al., 2010; Mouazen et al., 2010; Nocita et al., 2011; Rossel and Behrens, 2010; Stenberg et al., 2010; Vohland et al., 2011). The effectiveness of its widespread use as a routine analytical technique in soil laboratories is still in the ripening stage. At the

Petition 870170088564, of 11/16/2017, p. 15/65

7/44 however, Fuentes et al. (2012) affirm that the calibrations of success among the NIR spectra are increasing in the literature and, for example, the contents of one or more attributes of the soil; calibrations already sufficient to promote NIR as a substitute for conventional methods of soil analysis. However, they warn of the need to verify the robustness of these models against new samples and the use of new instruments.

[0021] The growing interest in NIR spectroscopy as an instrumental technique as an alternative to soil analysis can be justified by the countless and remarkable advantages that this technique presents in relation to conventional analyzes. Fuentes et al. (2012) argue that the fact that NIR technology is non-destructive free of undesirable residues and environmental impacts - besides being low cost, fast, requiring little handling of the samples when combined with Chemometrics (or multivariate analysis) - is among its main benefits. However, the interpretation of the NIR spectra is not immediate, since in its spectral region, overtone and combination bands are observed, which are not very intense and overlapping, therefore, univariate analysis becomes impracticable (as shown in the graph above). There are countless studies found in the literature using chemometric methods associated with NIR spectroscopy to determine soil attributes. Some recent works are used in Table 1, in which several methods of spectral pre-processing and multivariate calibration are employed for this purpose.

Table 1: Literature review comparing the numbers of samples used in the calibration (cal) / validation (val) sets, the spectral pre-processing methods, the RMSEP and R ² values.

Spectral Range Number Spectral Preprocessing RMSEP R ² Calibration Reference

Multivariate samples

371 cal / 20 val 350-2500 1 ^a Saviztky-Galoy derivative + SNV PLS 0.498 (g100g ^-1 ) 0.83 (Minasny et al., (Nm) (2011)

Petition 870170088564, of 11/16/2017, p. 16/65

8/44

104 cal / 25 val 1100-2500 1 ^the Saviztky-Galoy derivative PLS 0.32% (3.2 g Kg ^-1 ) 0.85 (Cambule et (nm) et. al., 2012) 306 cal / 50val 300-1700 MSC + 1 ^the Saviztky-Galoy derivative PLS 0.48% (4.8gKg ^-1 ) 0.74 (Mouazen (nm) et., al., 2007) 4184 cal / CV ¹ 350-2500 1st Saviztky-Galoy derivative BRT ² 9.0 (g Kg ^-1 ) 0.82 (Brown et (nm) et al., 2006) 295 cal / 60 val 1100-2000 MSC MPLS ³ (1.2 (g Kg ^-1 ) - (Fuentes (nm) et al., 2012) 360 cal / 154 val 350-2500 Saviztky-Galoy Straightening with PLS 2.88 (g Kg ^-1 ) 0.86 (Sarkhot et (nm) 3rd order polynomial et al., 2011) 168 cal / CV 453-2448 Standardization + BPNN ⁴ 0.68% (6.8g Kg ^-1 ) 0.84 (Mouazen (nm) 1st Saviztky-Galoy derivative et al., 2010) 19969 cal / 8731 ⁵ 400-2 5 00 1st Saviztky-Galoy derivative LOCAL-PLS 3.9 (g Kg ^-1 ) 0.79 (Nocita et (nm) al., 2014 288 cal / CV 350-2500 Visual noise removal PLS (0.35% (3.5g Kg ^-1 ) 0.57 (Summers (nm) instrumental et al., 2011)

¹ Cross validation (Cross Validation, CV).

² BRT — BoostedRegressionTress: non-linear method of multivariate regression.

³ MPLS - ModifledPartialLeastSquares.

⁴ BPNN - Back Propagation Neural Network.

⁵ Result presented only for one of the external groups evaluated in the work of Stemberg and collaborators (2010).

Chemometrics [0022] Several authors claim that NIR spectroscopy and Chemometrics coexist in symbiosis. Due to the fact that NIR spectroscopy has characteristics that limit univariate analysis of the spectra, it benefits from Chemometrics to become an increasingly robust tool for the identification and quantification of several parameters in

Petition 870170088564, of 11/16/2017, p. 17/65

9/44 different matrices. In turn, Chemometrics benefits from the type and the high amount of spectral information generated by NIR spectroscopy for the development and application of new methods (McClure, 1994; Pasquini, 2003).

[0023] Chemometrics involves the application of statistical and computational mathematical methods to plan, investigate and forecast data sets of chemical interest (Wold, 1995). The term was introduced in 1971 by Prof. SvanteWold (Umeã University) who defined the art of extracting the relevant chemical information from data produced in chemical experiments such as Chemometrics, in analogy to Biometrics and Econometrics (Wold, 1995). In 1974 Prof. SvanteWold founded, together with Prof. Bruce R. Kowalski (Universityof Washington), the International Chemistry Society (International Chemometrics Society), a pioneer institution in the dissemination of this new science. In 1980, Prof. Kowalski taught the first Chemistry course in Brazil, at the Chemistry Institute of Unicamp, at the invitation of Prof. Roy Edward Bruns, retired professor at this institution (Neto et al., 2006).

[0024] The emergence of Chemometrics in the 70s was driven by the development of computerized instrumental methods for chemical analysis, which made it possible to obtain large amounts of data much more quickly and efficiently (Neto et al., 2006). This significantly increased the number of variables that can be measured in a single sample. A notable example is the absorption intensity over more than one wavelength, which is routinely recorded over a single spectrum. Given the possibility of obtaining instrumental responses from multiple variables, there was also a need to extract reliable results and relevant information quickly and efficiently, which came to be done very efficiently by chemometric methods (Ferreira et al., 1999) .

Petition 870170088564, of 11/16/2017, p. 18/65

10/44 [0025] In the 1980s, two journals dedicated to the area appeared: Journal of Chemometrics, Chemometrics Intelligent and Laboratory Systems, covering fundamental and methodological aspects of research in Chemometrics. In addition, many applications have started to be published in other Analytical Chemistry journals, such as: Applied Spectroscopy, Analytical Chemistry, Analytical Chimica Acta, Talanta, etc. The first scientific paper published in Brazil on the subject is from 1985 by Bruns and Faigle (Bruns and Faigle, 1985). In this work, the authors introduced general aspects about theory and practice in Chemometrics, in addition to encouraging the creation of disciplines in the undergraduate and graduate courses of Brazilian universities. Also in 1985, the first Chemistry course was held as a graduate course in Brazil. The course was taught in the postgraduate program in Chemistry at the State University of Campinas (UNICAMP) by Prof. Roy Eduard Bruns and had only two students (Neto et al., 2006), a reality completely different from what is observed today.

[0026] Although Chemometrics affects all branches of Chemistry, Analytical Chemistry was the area most benefited due to the development of analytical instrumentation and its association with computers (Bruns and Faigle, 1985). Examples in ultraviolet-visible (UV-Vis), near and medium infrared, mass spectrometry (Mass Spectrometry - MS), nuclear magnetic resonance (Nuclear Magnetic Resonance, NMR), atomic absorption / emission and liquid / gas chromatography, are examples analytical techniques that offer the possibility of generating data characterized as multivariate.

[0027] There are many chemometric methods and their applications depend on the nature of the problem you want to solve or the type of information you want to obtain. In general, the methods that form the basis of Chemometrics can be classified as pattern recognition, multivariate calibration, multivariate signal resolution and design of experiments. The application of these methods basically requires some

Petition 870170088564, of 11/16/2017, p. 19/65

11/44 fundamental steps for the correct interpretation of data, such as pre-processing of instrumental responses, selection of variables, removal of outliers and the calculation of some figures of merit, to ensure the validity of the models built.

[0028] In the next subsections of this report, some of the main chemometric methods used as a basis for this work will be briefly introduced, such as Principal Component Analysis (PCA) and regression by partial least squares (Partial Least Squares - PLS) . The aspects related to the validation of the results will be presented in the section corresponding to the theme.

Principal component analysis [0029] The PCA is a method that allows the reduction of the dimensionality of the data through the representation of the data set in a new system of axes, called Principal Components (Principal Component-PC), allowing the visualization of the multivariate nature data in a few dimensions. In the original space, the samples are points located in an n-dimensional space, with ‘n’ equal to the number of variables. With the reduction of dimensionality provided by the PCA, the samples become points located in spaces of reduced dimensions defined by the PCs, for example, two or three dimensional. Mathematically, in PCA the matrix X is decomposed into a product of two matrices, called scores (T) and weights (P), plus an error matrix (E) (Wold and Sjostrom, 1998), as shown in Equation 1:

X = TP '+ E (1).

[0030] The scores refer to the similarity between the samples and represent their coordinates in the system of axes formed by the Main Components (PC). The CP is formed by the linear combination of the original variables and the coefficients of this combination are called weights. The weights are the cosines of the angles between the original variables and the PCs, representing,

Petition 870170088564, of 11/16/2017, p. 20/65

12/44 therefore, how much each original variable contributes to a given PC. The first main component (PC1) is drawn in the direction of the greatest variation in the data set; the second (PC2) is drawn orthogonally to the first, in order to describe the largest percentage of the variation not explained by PC1 and so on. The evaluation of the weights allows us to understand which variables contribute the most to the groupings observed in the score graph. Through the joint analysis of the scores and weights graph, it is possible to verify which variables are responsible for the differences observed between the samples. The number of PCs to be used in the PCA model is determined by the percentage of explained variance. Thus, a number of components are selected in such a way that the largest percentage of the variation present in the original data set is captured (Wold, 2002). Several algorithms are available to perform the PCA and four of them appear frequently in the literature: (Non linear Iterative Partial Least Squares NIPALS), decomposition by singular values (Singular Value Decomposition - SVD), which use the data matrix X, POWER and decomposition by eigenvalues (Eigen value Decompositon - EVD) that work with the cross product matrix X'X. SVD and EVD extract the Main Components simultaneously, while NIPALS and POWER calculate the PC sequentially (Andersson, 2009; Geladi and Kowalski, 1986).

[0031] SVD (Equation 2) is based on the linear algebra theorem that states that an X matrix (mxn), 'm' columns and 'n' lines, can be transformed into a product of three matrices U, S, V '(' subscript means transposed) w which have specific properties, where: matrices U and V are square and orthonormal; the matrix S is a diagonal rectangular matrix containing singular values diagonally and all elements outside the diagonal equal to zero (Wold, 2002). In this case, weights are given by matrix V and scores by: T = US

X = USV '(2).

Petition 870170088564, of 11/16/2017, p. 21/65

13/44 [0032] EVD is a method similar to SVD used to calculate eigenvalues and eigenvectors of a matrix. Both SVD and EVD are mathematically easy methods to use in Matlab, because in this software there are internal functions for the decomposition of singular values and for estimating eigenvectors and eigenvalues of a matrix.

[0033] NIPALS is a method commonly used for the calculation of Principal Components of a data set, in which the vectors of weights and scores are calculated iteratively, one at a time, in decreasing order of importance. The iterative process, for the first main component, is started with a first estimate of scores, which may be the column of X that has the greatest variance. Using these scores, weights are calculated as: p = tX / t't, being normalized to length equal to one. After that, the scores are calculated as: t = Xp / p'p. These score values are compared with the previous ones and if they are different (within a pre-established criterion), the weights are recalculated as shown above. This process continues until the scores are similar or a number of iterations have been performed. After convergence, the product tp 'is subtracted from X, obtaining residue E (Equation

3):

E = X - tp '(3).

[0034] The process continues to the next main component, replacing E with X (Andersson, 2009; Wold, 2002).

Regression by partial least squares (PLS) [0035] Certainly PLS is next to the PCA one of the most prominent methods used in Chemometrics, being considered the multivariate calibration method used as a reference for this type of application. In PLS, the variation in spectral response is related to the attributes (or properties) of interest through a multivariate regression (Geladi and Kowalski, 1986). PLS uses PCA to reduce

Petition 870170088564, of 11/16/2017, p. 22/65

14/44 dimensionality of the data set for later correlation between the spectra (matrix X) and the properties of interest (vector y) (Geladi and Kowalski, 1986). The property of interest is often the concentration of an analyte, but not limited to it, and it may even be physical-chemical properties, such as density and viscosity, which are related to the sample composition. Matrix X and vector y are decomposed by PCA using equations 4 and 5:

X = TP '+ E (4);

y = Uq '+ f (5);

Where: P and q are the weights of X and y, respectively; T and U are the scores of X and y, respectively, and E and f represent the residual matrices of X and y, respectively. The matrix of scores T is estimated as a linear combination of X with the weighted coefficients powder W (called weights);

T = XW (6);

From W, the regression coefficients of the PLS model can be estimated by:

b _PLS = W (P 'W) ^_1 q (7);

The PLS linear model can be represented by:

^y = ^X - ^b PLS ⁽⁸⁾ ·

Preprocessing [0036] The data preprocessing step is fundamental to the success of multivariate analysis. The main objectives of applying pre-processing techniques are to eliminate information that is not chemically relevant and make the data matrix better conditioned for analysis, enabling the subsequent exploratory analysis of the data set efficiently. There is a vast literature available on the different methods of

Petition 870170088564, of 11/16/2017, p. 23/65

15/44 data processing in spectroscopy (Xu et al., 2008). Normalize the spectra, center the data on the mean, derive and smooth using the Savitzky-Golay algorithm and apply the multiplicative scattering correction (Multiplicative Scatter Correction - MSC) are some of the most applied methods (Rinnan et al., 2009).

[0037] Centering the data on the mean (Equation 9) consists of averaging the intensities for each wavelength and subtracting each of the intensities from the mean value. In this way, each variable will have a zero average, that is, the coordinates are moved to the center of the data, allowing differences in the relative intensities of the variables to be easier to perceive.

Xjj (centered on the mean) = x ^ · (original) - Xij (mean) (9).

[0038] Baseline displacement and inclination can be corrected by deriving the spectra. Smoothing methods are used to mathematically reduce noise, thereby increasing the signal / noise ratio. In these methods, a window is selected, which contains a number of variables. The points in the window are then used to determine the value at the center point of the window and, thus, the window size directly influences the smoothing result. In the Savitzky-Galoy method, a low-order polynomial is adjusted to the window points and used to recalculate the central point (Savitzky and Golay, 1964).

[0039] Multiplicative scattering correction is a transformation method used to compensate for additive and / or multitiplicative effects in spectral data. This method removes the influence of physical effects on the spectra, such as particle size, roughness and opacity, which do not bring chemical information about the samples and introduce spectral variations such as the shift from the baseline. To make the correction, the MSC method assumes that each spectrum is determined by

Petition 870170088564, of 11/16/2017, p. 24/65

16/44 chemical characteristics of the sample added to the unwanted physical characteristics. Equation 10 describes the MSC's working principle:

_x = ^[x TFE (ortgtnai) + t] ^(k (transformed ') ⁽⁾ · where: x _(k (original) and x _(k (transformed) are the absorbance values before and after correction with the MSC in k wavelengths; ai and bi are constants estimated from a least squares regression of an individual spectrum x _(k against an average spectrum x of the calibration set at all wavelengths or in a subset, following Equation 11:

X (k = / + b (X + etk ⁽ 11 ⁾ ;

where: e _(k corresponds to all other effects in the spectra that have not been modeled (Rinnan et al., 2009).

Outlier removal [0040] Outliers (or anomalies) is the term used to designate anomalous samples that may be present in the calibration and validation sets, used in the construction of models in multivariate calibration. Typically, these anomalous samples behave differently than other samples in the data set, and their presence in the calibration set can result in models with low predictive capacity, that is, that produce high error values in prediction (Naes and Martens, 1984; Valderrama, 2005).

[0041] In the present work, the identification of outliers in the calibration and validation sets was carried out following the recommendations of the standard A16M-05 of ASTM and of Martens and Naes (1984) (Table 2). Therefore, they were evaluated based on extreme leverage, residuals not modeled in the spectral data and residuals not modeled in the dependent variable. At

Petition 870170088564, of 11/16/2017, p. 25/65

17/44 main equations used in the calculations of outliers are not described in Table 2 (ASTM E1655-05, 2012; Valderrama, 2005).

[0042] Leverage (Table 2) is the distance of a sample to the mean of the center of the data set. In the case of NIR spectra, it is a measure of the similarity of a given spectrum (spectral profile) in relation to the other spectra present in the mean. Therefore, samples with high leverage values are potential outliers in the data set and can be eliminated according to a given arbitrary criterion (Naes, 1989).

[0043] The identification of anomalies in relation to the non-modeled residues in the spectral data is obtained by comparing the total residual standard deviation (s (ê)) and the residual standard deviation of a sample i (s (ê ₍ )), defined in Table 2. If a sample shows s (ê ₍ )> 2s (ê) it is removed from the calibration set (Valderrama, 2005).

[0044] Regarding residues not modeled on the dependent variable, or outliers are identified by comparing the square root of the mean calibration error (Root Mean Square Error of Calibration - RMSEC) with the absolute error of that sample (Table 2). If the sample shows an absolute error (y ₍ - y ₍ ) greater than (3 x RMSEC), it is defined as an outlier (Valderrama, 2005).

[0045] The removal of outliers in the validation set is also performed through calculations of leverage and non-modeled residues in the spectral data, however, in addition, a test can be performed that considers the non-modeled residues of the spectral repeatability of samples with at least three levels of concentration with six replicates for each level. This test is based on the mean square root of the spectral residues (Root Mean Square Residuals - RMSSR), which, for a sample, is defined in Table 2. If the value of the RMSSR of a validation sample shows a higher value of the RMSSRlimit (Table 2), this sample must be removed from the validation set (Valderrama, 2005):

Petition 870170088564, of 11/16/2017, p. 26/65

18/44

Table 2: Equations used in outlier calculations

Outliers Equation Leverage hi = ti (T 'T) ^_1 ti (12) Unmodelated waste (spectra) ^{S (B0 _} nJ - J - Amax (n, J) Σ i = 1 ^{J Σ (χ} ϋ ^—x i, j ^{) 2} h = 1 (13) Residual standard deviation J ^{S (e} ' ^{) 2} = nJ - J— η, Ν ^X ' j = 1 (14)

n — A —1 n

^e . ^y i ^{) 2} ( ¹⁵ )

Absolute error vs RMSEC

Abs error (y - y ₍ ) = 3 X RMSEC =

RMSSR = J

J

Xj) ^{2 (34)}

Anomalous samples in validation

X (16)

T are the scores for all calibration samples, t ₍ is the vector of scores for a particular sample. _ / Is the number of spectral variables and n is the number of samples in the calibration. VMSSfí _max is the maximum observed value for calibration samples, VMSSR _{Cflí (i)} is the square root of the mean of the spectral residues for a replicate of each level that are included in the calibration set and fíMSSfí _repiy ( _í) is the square root of the mean of the spectral residues for the mean of six replicates of each level that will not be included in the calibration set (ASTM E1655-05, 2012; Valderrama, 2005).

Results Validation

Petition 870170088564, of 11/16/2017, p. 27/65

19/44 [0046] The proposition or development of an analytical method necessarily requires a careful and systematic evaluation of its performance and robustness in the conditions in which it will be applied. This process of analytical adequacy in which the requirements specified in the method are suitable for an intended use is called validation (VIM, 2008).

[0047] The validation of an analytical method can be attested by determining parameters known as figures of merit (Lorber, 1986; Valderrama, 2005; Valderrama et al., 2009). In Analytical Chemistry, the main merit figures commonly used in a validation process are: accuracy, precision, sensitivity, analytical sensitivity, selectivity, signal / noise ratio, detection limit, quantification limit, confidence intervals and tests for bias or bias ( Souza et al., 2016; Valderrama et al., 2009).

[0048] The determination of some of these parameters in multivariate calibration models is estimated in a very similar way to the multivariate calibration methods. However, this similarity cannot be observed for parameters such as confidence intervals or uncertainty, linearity, sensitivity, signal / noise ratio and selectivity (Souza et al., 2016; Valderrama et al., 2009).

[0049] Figures of merit such as linearity, sensitivity, signal / noise ratio and selectivity, can only be estimated by calculating the analytical signal (Net Analyte Signal - NAS) for the property of interest to be quantified by the multivariate method (Lorber et al., 1997; Valderrama et al., 2009). The NAS is the fraction of the analytical signal related to the property that is measured by the “free” multivariate method of the signal of the other compounds present in the sample (Lorberet al., 1997; Valderrama et al., 2009). With a scalar value that represents only the analyte signal for i calibration samples, and no longer a spectrum (vector), a pseudo-univariate regression model, by least squares, is constructed between the vector containing

Petition 870170088564, of 11/16/2017, p. 28/65

20/44 the NAS values and the vector containing the y values (Faber, 2000; Lorber et al., 1997; Valderrama, 2005).

[0050] In the literature there are several works addressing theoretical aspects regarding figures of merit in multivariate analysis (Lorber, 1986) and application of figures of merit in NIR spectroscopy data (Rambo et al., 2013; Souza et al., 2016 ; Valderrama et al., 2009). Specifically in Portuguese, Valderrama et al. (2009) published an article on the state of the art of figures of merit in multivariate calibration, which addresses practical and theoretical aspects in an accessible way. In the specific case of the use of figures of merit in the validation of NIR spectroscopy for soil analysis, Bellon-Maurel and collaborators (2010) published an article that presents a critical assessment of quality indicators in multivariate calibration models. However, several figures of merit mentioned above and the application of the NAS, were not addressed in that work.

[0051] A detail of the validation of the analytical methodology employed in the present patent application can be found in the article published by one of its authors in the following reference: (Souza et al., 2016).

[0052] Regarding patent references, it should be noted that, until now, the term “Near Infrared Spectroscopy” appears as a topic in 2599 patents inserted in the Derwent Innovations Index database. Of this total, 29 documents have the term “Soil”.

[0053] Hubert and his collaborators filed a patent describing the invention of a method for analyzing spectral data for the selection of a calibration model. In this method the spectra of a given substance are related to a physical or chemical parameter of the substance, in a predetermined range of concentration of that parameter, following the following steps: a) the spectral data of a substance were captured and related to its respective physical parameters - chemicals in a given concentration range; b) several calibration models were created

Petition 870170088564, of 11/16/2017, p. 29/65

21/44 using spectral data and physical-chemical parameters using resampling; c) the tolerance intervals for each reference level and calibration model were calculated and; d) the tolerance intervals were displayed at each of the reference levels over the predetermined interval for each calibration model. In this way, a possibility of spectral data analysis was proposed aiming at an automated selection of calibration.

[0054] Jiliang China University has also filed some patents on near infrared spectroscopy, among which can be cited:

- CN106096656, published on 11/09/2016, which describes a method of analysis of infrared spectroscopy close to the ground, based on scarce representation and BP neural network technique;

- CN205157419, published on 04/13/2016, which describes a soil nutrient detection device based on near infrared spectroscopy;

- CN20515741, published on 10/02/2016, which describes a device for the detection of soil nutrients based on near-visible infrared spectroscopy;

- CN105319172, published on 10/02/2016, which describes a device for detecting soil nutrients based on near-visible infrared spectroscopy;

- CN204832023, published on 02/12/2015, which describes a Device for rapid detection of soil moisture based on near infrared spectroscopy;

- CN204630922, published on 09/09/2005, which describes an organic system for the detection of organic matter and moisture based on the technique of near infrared spectroscopy.

Petition 870170088564, of 11/16/2017, p. 30/65

22/44 [0055] We highlight from the related documents some to be particularly cited, such as the document CN106124449-A, by Wang R. and others, which describes a process for analyzing and predicting the depth of the ground by near infrared spectroscopy involving obtaining of training and pre-processing samples, performing spectrometry, acquiring spectroscopic data, building a prediction model and scanning soil samples.

[0056] The document CN105784629-A, by Huang et al., Describes the Detection of a stable carbon isotopic relationship of the soil, that is, delta-13 carbon value using medium infrared spectroscopy, involving the acquisition of original spectra and the establishment of a quantitative relationship between the treated spectra and the stable carbon isotopic relationship.

[0057] The document CN105699314-A, by Kang H. et al., Describes a Method to detect the stable carbon isotopic relationship using infrared spectroscopy, involving the establishment of a quantitative relationship model according to an isotopic relationship. stable carbon content of calibrated soil samples.

[0058] Document CN105486663-A, by Kang H. et al., Describes a soil nutrient detection device based on visible and near infrared spectroscopy technology, having a fiber optic probe connected to the controller through the spectrograph and GPS module and LCD module connected to the controller.

[0059] Document WO2011051166-A1, by Janik LJ and others, extended to several countries, describes a Device for determining polycyclic aromatic hydrocarbons in general. The ground comprises a unit for exposing the sample of diffuse reflection infrared spectroscopy, and a detection unit for recording spectroscopic parameter of sample and computing unit.

Petition 870170088564, of 11/16/2017, p. 31/65

23/44 [0060] The classic wet methods are the most used for physical-chemical analysis of soil fertility. They incorporate products and processes employed by the man of the art, and in terms of process, it is worth mentioning that the methods for chemical analysis of soil samples involve 3 (three) main steps:

- Preparation of the sample: dry, remove, pass through the first 2 mm mesh and remove the amount of sample (volume or weight) specific for analysis of each attribute of interest, according to the Methods Manual;

- Extraction: add extracting solution (chemical reagent) and;

- Analytical Determination: sample analysis by several methods, such as Flame Atomic Absorption Spectrometry (FAAS), Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES), Colorimetry, Titulometry, among others.

[0061] The most widespread classic methods for physical analysis (granulometry) of the soil are based on the application of pre-treatments to remove cementing and flocculating agents; dispersion of the soil sample through the use of a combination of chemical processes and mechanical breakdown; and quantification of soil fractions by sieving, for coarse and fine sand fractions, and by sedimentation, for silt and clay fractions.

[0062] In spite of currently known technologies, including the cited patent references, with regard to products available on the market, nearby infrared spectrophotometers (NIR) can be highlighted, which have generalized applications for any type of matrix or sample. These instruments can be classified, according to the optical analysis system, into interferometric and dispersive. Interferometric instruments are those that use the Fourier transform to perform spectroscopic measurements and generate the infrared spectrum. Dispersive equipment makes use of diffraction grids

Petition 870170088564, of 11/16/2017, p. 32/65

24/44 to perform such measurements. Commercial equipment of both classes is available on the market.

[0063] The system of acquisition of spectra (analysis) of solids can present at least 3 (three) types of design: fixed, rotating or translational. In the case of fixed systems, spectra are acquired at a single point in the sample. In rotating systems, the sample is usually analyzed on a support similar to a petri dish with diameters ranging from 10 mm to 90 mm. In this case, a rotating system (called a spinner in English) causes the sample to rotate continuously throughout the process of spectra acquisition. At the end of the scan, a spectrum is generated which is the average of all spectra acquired during the sample scan. The translational system can work in a similar way to the rotational one, however, the supports are usually rectangular. Normally, the sample holder undergoes translation and a spectrum is acquired at a given fixed number of points. At the end of the process, a spectrum is obtained that represents the average of all the points analyzed in the sample. These devices analyze solid samples in the form of dust or grain in diffuse reflectance mode, and only one sample can be analyzed at a time.

[0064] The sample acquisition systems presented above can analyze the sample from below or from above, that is, the location of the infrared radiation beam may be below or above the sample carriers. In the case of equipment that reads below the sample holders, the optical path or distance between the beam and the sample is easily kept constant, as the surface of the sample holders, usually quartz or glass, acts as a border between the sample and the radiation beam. However, in commercial spectrophotometers that acquire the spectrum over the sample, the detector is fixed and the distance between the beam and the sample can vary depending on the amount of sample placed in the sample holder and the irregularity on the sample surface. when arranged in the container.

Petition 870170088564, of 11/16/2017, p. 33/65

25/44

INCONVENIENTS OF THE PRODUCTS AND / OR PROCESSES OF THE STATE OF THE TECHNIQUE [0065] In general, current methods for physical-chemical analysis of soil fertility require several different tests to analyze different characteristics of the soil, are laborious, can take a significant time for their execution, demand an expressive number of specialized labor, make use of several chemical reagents and produce residues that can generate negative environmental impacts.

[0066] In the case of Total Organic Carbon (TOC) analysis in the soil, the analysis laboratories usually employ the Walkley & Black method, which is based on the oxidation of carbon forms by dichromate (Cr2O7 ^2- ). In addition to the high cost and high time in the acquisition of data for determining organic carbon by this method, a toxic reagent is used in the analysis, which, at the end of the process, generates a chemical residue highly harmful to the environment and human health.

[0067] It is worth noting that the complexity in the official methodologies for soil analysis is related to the various stages for carrying out the test. Consequently, for a laboratory to guarantee the quality of the analysis result, it is necessary to implement several controls in its process, as each step contributes to the final uncertainty of the result. In addition, it is extremely important to observe the quality of the water and reagents used in the test, to have a preventive maintenance plan and equipment calibration and to constantly invest in the training of its employees. It is desirable that laboratories seek to implement quality management systems, such as ISO / IEC 17.025, as a way to improve process control and analysis quality.

[0068] Another important point that must be considered is that, in most cases, the analytical capacity of the soil laboratory is directly related to the amount of human resources employed

Petition 870170088564, of 11/16/2017, p. 34/65

26/44 in the process. In addition, the demand for soil analysis in agriculture is seasonal, that is, there are periods of the year when the demand for analysis is intense, to the detriment of other periods. Therefore, it is a very difficult task for laboratories to adequately size their analytical capacity and allocate the appropriate resources for this, since when the analytical capacity is undersized, the laboratory may fail to serve the market and reduce its billing potential, while allocate many resources and due to some market nuance there are no samples, the loss can be very great.

[0069] All of these factors contribute to hamper the implementation and quality operation of soil analysis laboratories in remote regions with an agricultural pole, where the availability of specialized labor and suppliers of reagent inputs is limited; the water quality is often not suitable for use in the test; access to service providers for equipment maintenance and calibration is scarce.

[0070] Regarding the NIR spectrophotometers available in the current market, it is worth mentioning that no near infrared spectrophotometer (NIR) with a knowledge bench dedicated to the matter, has a solution dedicated to soil analysis, with an associated digital platform, cloud-based (web), for forecasting and making results available. In general, this equipment has complex software for the construction of calibration models that require constant intervention by analysts, and the lack of specialized labor for multivariate data processing is recurrent. These software have limited resources for carrying out the calculations and exploratory analyzes required in multivariate calibration; these systems are not very intuitive, allowing the analyst to make gross errors that can compromise the quality of the results.

[0071] With regard to data modeling, services provided commercially, most often by the companies that sell the products themselves

Petition 870170088564, of 11/16/2017, p. 35/65

27/44 NIR equipment, is limited to offering customers the so-called “ready-made calibration curves”. This type of approach is highly technically discouraged, as any variation in the matrix composition can seriously affect the calibration model causing systematic errors and impairing the accuracy of the analyzes. In some cases, these companies offer a service of recalibration of models embedded in the system using customer data. However, this service is provided inefficiently because it does not have a systematized software solution or digital platform for multivariate data processing. Typically, these companies send the data to specialists based in their headquarters, located mainly in Europe and the United States, for recalibration to be performed. It should be noted that this type of approach has contributed negatively to the popularization, acceptance and understanding of NIR spectroscopy for soil analysis and other matrices.

[0072] Another drawback inherent in the infrared instruments available on the market, is that they allow the analysis of only one sample at a time. Considering the high investment for the acquisition of a commercial infrared spectrophotometer and, in general, the low added value of soil analysis services, if the NIR equipment does not have a high operational performance, its application in soil analysis routines is not viable .

THE INVENTIVE SOLUTION OF THE INVENTION METHOD IN RELATION TO THE STATE OF THE TECHNIQUE [0073] The new technology dedicated to chemical and physical analyzes of soil fertility by Vis-NIR spectroscopy (called SpecSolo) greatly simplifies the analytical procedure, limiting itself to just two phases:

- Sample preparation: dry, remove, pass through the first 2 mm mesh and transfer the amount of soil sample sufficient to cover the bottom of the sample container (plastic cup), approximately 150g;

Petition 870170088564, of 11/16/2017, p. 36/65

28/44

- Analytical Determination: collection of the sample spectrum with the Vis-NIR spectrophotometer of the invention, which automatically sends the information to a database hosted in the clouds, where the calibration models (mathematical models) are stored, which will produce the analytical results and make it available for consultation on the portal, in less than 30 (thirty) seconds. It also has the advantage of analyzing different soil attributes with just one spectral reading: dozens of soil fertility parameters can be analyzed simultaneously in a non-destructive, fast and economical way.

[0074] The controls necessary to assess the correct functioning of the Vis-NIR spectrophotometer and guarantee the quality of the analytical result are made through the use of a white reference standard (Spectralon), with known reflection around 0.99 across the spectrum visible, and a standard soil sample, with known analytical results. Before starting the process of analyzing soil samples in the tray, the spectrophotometer needs to be calibrated by reading Spectralon. This calibration aims to equalize the energy bundles, within a limit of 0 to 100%. If the Spectralon reading is approved, the autosampler will start reading the second control, which is the standard soil sample. The analytical result of the standard sample is predicted in the database and evaluated using the online Control Chart. Only if the standard sample spectrum is approved does the equipment proceed with the analysis of the entire tray.

[0075] Thus, the solution presented democratizes the realization and access to soil analysis technology, as it: (i) eliminates the need for specialized labor in the process; (ii) significantly reduces the operational cost of the analysis, enabling a reduction in the final price of the service, allowing the farmer to intensify the number of samples collected in the field, aiming at greater representativeness and better knowledge of the production area; (iii) reduces the analytical determination time to less than 30 (thirty) seconds, which meets the market demands for

Petition 870170088564, of 11/16/2017, p. 37/65

29/44 faster and more accurate information; (iv) standardization of the analytical process, providing greater repeatability and quality in the analysis results; (v) the possibility of implantation in remote areas, such as in agricultural centers, promoting greater access by rural producers to the use of analysis technology, still with the benefit of facilitating logistics and reducing the time for sending samples to the laboratory; (vi) eliminating the use of chemical reagents that generate toxic waste highly harmful to the environment and human health, in addition to reducing the bureaucracy in the process of implementing soil analysis laboratories, as it does not require the handling of hazardous products and the controlled disposal of waste.

[0076] Another differential of the technology of the invention is the high operational efficiency and low demand for dedicated labor in the process, enabling its application in large-scale routines. As the charging model for the use of the technology of the invention is per sample analyzed (by "click"), the fixed cost for maintaining the laboratory analytical structure will be drastically reduced, starting to have mainly variable cost (when performing the analysis). In this way, all work to plan the laboratory's analytical capacity to meet the intrinsic seasonality of the agricultural market will disappear, since the technology allows a small number of employees to be able to produce large volumes of soil samples per day.

[0077] Considering the determination of Total Organic Carbon (TOC) in the soil to be a routine analysis in agronomic laboratories and the growing demand for this type of analysis due to global initiatives to monitor the amount of carbon in the soil with Low-Level Agriculture programs Carbon, it is necessary to develop new high-performance analytical technologies that are not based on dichromate oxidation. The IPCC Intergovernmental Panelon Climate Change, recognizes the analysis of Total Organic Carbon in the soil by the dry combustion method using an elemental CHN analyzer. However, this method is not only very expensive

Petition 870170088564, of 11/16/2017, p. 38/65

30/44 does not have an adequate yield for application in laboratory routines in order to meet market demands. For this reason, new technology presented here for soil analysis by Vis-NIR spectroscopy becomes even more relevant.

[0078] According to the Agency for Toxic Substances and Disease Registry (ATSDR), located in Atlanta, USA, chromium is a natural element in plants, rocks, soils and is present in different forms ( Cr (0), Cr (III) and Cr (VI), Cr (III) being naturally occurring and Cr (0) and (VI) generally results from industrial processes, with hexavalent chromium at 18 ° in the priority list of dangerous substances published in 2007 (ATSDR, 2010).

[0079] According to NBR 10.004, of the Brazilian Association of Technical Standards, the wastes containing chromium (VI) are classified as hazardous, or Class I (ABNT, 2004) and require appropriate forms for transport and packaging.

[0080] There is no official data to confirm this information, but considering the estimate by Embrapa Solos that in Brazil at least 4 (four) million soil analyzes are performed; that the Walkley & Black method establishes for the TOC analysis the use of 10mL potassium dichromate + 10mL concentrated sulfuric acid + 200mL of distilled water + 10mL concentrated phosphoric acid (Total Solution containing Chromium (VI) used per sample: 240mL); COT analysis is a basic routine analysis, being required in all analytical services; therefore, at least 900,000 liters of solution containing Chromium (VI) are generated in Brazil from soil analysis.

THE INVENTIVE SOLUTION OF THE PRODUCT OF THE INVENTION IN RELATION TO THE STATE OF THE TECHNIQUE - HOW IT SOLVES THE RELATED INCONVENIENTS [0081] The great innovative reference of the automated Vis-NIR spectrophotometer, called SpecSolo-Scan, is the fact that it has been

Petition 870170088564, of 11/16/2017, p. 39/65

31/44 custom developed for soil fertility analysis in large-scale routines. It has a tray with the capacity to accommodate 40 (forty) soil samples at a time and an automatic sampler (autosampler) for sequential analysis of the samples. The equipment can also be associated with a conveyor that carries the samples, as an alternative to the 40 (forty) sample tray.

[0082] The spectrum acquisition system of the SpecSoloScan spectrophotometer of the invention is made on top of the sample, that is, the infrared radiation beam is coupled to the autosampler above the sample holders (glass / analysis vessel). In addition, the autosampler is equipped with a height sensor that adjusts according to the amount of sample placed in the container, ensuring that the optical path, or the distance between the beam and the sample, is always the same between samples, standardizing and improving the quality of spectra collection and, consequently, the robustness of the model and the quality of the result.

[0083] The identification of the samples by the system is performed by reading the bar code affixed to the sample container by means of a reader that is attached to the autosampler. Therefore, the samples can be placed in any position of the tray (or mat) that the equipment will be able to identify them at the time of collecting the spectrum, minimizing the analyst's intervention and the chances of errors.

[0084] The spectrophotometer of the invention (SpecSolo-Scan) is integrated into a software platform dedicated to soil analysis, called the Digital SpecSolo Platform, which controls the equipment and manages the entire analysis routine, from receiving the sample to making it available analytical results on the portal. The Digital SpecSolo platform consists of the following modules: SpecSolo-Manager, SpecSolo-Tray, SpecSolo-Lims, SpecSoloServices and SpecSolo-Toolbox.

OBJECTIVES OF THE INVENTION

Petition 870170088564, of 11/16/2017, p. 40/65

32/44 [0085] The objective of the present invention is to provide a new method, that is, a process, and the respective system involving the product (hardware + software) in order to improve the procedures for chemical and physical analysis of soil fertility, through a technological package dedicated to soil analysis, a package that combines the use of a Vis-NIR spectrophotometer, called SpecSolo-Scan, efficient chemometric algorithms and an integrated cloud-based software to produce the analytical results.

[0086] Another objective of the invention aims to democratize the realization and access to soil analysis technology, significantly reducing its operational cost, reducing the analysis time to less than 30 seconds, dispensing with the use of chemical reagents that generate toxic waste and enabling its implantation in remote areas, as in agricultural centers. It has the advantage of analyzing soil samples in a non-destructive, fast and economical way: dozens of soil fertility parameters can be analyzed simultaneously in just a few seconds.

DESCRIPTION OF THE DRAWINGS [0087] The invention will be described below in a non-limiting embodiment, and for the better understanding of the technology developed, references can be made to the attached drawings, in which they are represented:

FIGURE 1: Perspective view of the equipment dedicated to soil analysis by Vis-NIR spectroscopy, according to the invention, called SpecSolo-Scan;

FIGURE 2: Shows a graph of radiation in the energetic infrared region enough to cause vibrational energy transitions, which basically results in a chemical and physical fingerprint of the soil sample;

FIGURE 3: Illustrative view of the components of the technology of the invention;

FIGURE 4: Illustrates the interaction between the factors that make up the SpecSolo technology;

Petition 870170088564, of 11/16/2017, p. 41/65

33/44

FIGURE 5: Diagram showing the algorithm for the process of building the multivariate calibration model and forecasting the results;

FIGURE 6: Shows the flowchart of the analysis process with the product (SpecSolo-Scan spectrophotometer) according to the invention.

DETAILED DESCRIPTION OF THE INVENTION [0088] The “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE ROUTINE”, object of this Invention Patent Application, comprises the offering of a new method, that is, , a process, and respective system involving product (hardware + software) in order to improve the procedures for chemical and physical analysis of soil fertility, through a technological package dedicated to soil analysis, according to its own routine, a package that combines the use of a VisNIR spectrophotometer, called SpecSolo-Scan with its digital platform, chemometric algorithms for multivariate analysis and an integrated cloud-based software to produce the analytical results.

[0089] Thus, according to the invention, for routine application of the infrared spectroscopy technique to determine the chemical and physical attributes of the soil, it is first necessary to build calibration models for each of the attributes that will be analyzed in the soil , associating the result of the wet analysis, carried out according to the reference methods, with the corresponding spectrum of each soil sample, acquired through the Inv-NIR spectrophotometer of the invention, or SpecSoloScan.

[0090] On the other hand, the equipment (1) dedicated to soil analysis by VisNIR spectroscopy according to the invention, presents differentials in relation to the other near infrared spectrophotometers available on the market, listed below:

Petition 870170088564, of 11/16/2017, p. 42/65

34/44

- presence of an automatic sampler (autosampler) for sequential analysis of soil samples arranged in tray (2), with capacity for at least 40 (forty) positions;

- proximity sensor coupled to the autosampler to standardize the distance of the spectrum collection, that is, the distance between the beam of infrared radiation and the sample;

- presence of a barcode reader coupled to the autosampler for identification and registration of the sample in the tray.

[0091] With a command through the software that controls the equipment, the acquisition of spectra of the soil samples begins in all 40 (forty) positions available in the support. The scanning time of a sample for spectrum collection is less than 10 (ten) seconds.

[0092] The Digital SpecSolo platform is the technological package that integrates with the SpecSolo-Scan equipment (1) the systems for managing soil samples and calculating the analytical results, consisting of the following modules:

- SpecSolo-Manager. software for management and access control of partners accredited by SpecSolo;

- SpecSolo-Tray: software for controlling SpecSolo-Scan equipment and organizing samples for analysis;

- SpecSolo-Lims: software for managing soil samples and analytical services to be performed. Manages the entire analytical routine from the arrival of the sample to the availability of the result;

- SpecSolo-Services: services portal for integration (communication) of the SpecSolo Digital Platform with its own modules or third-party modules;

Petition 870170088564, of 11/16/2017, p. 43/65

35/44

- SpecSolo-Toolbox: proprietary software for data modeling and prediction of results, where the spectral pre-processing routines, construction and choice of the best models, validation and prediction of results were automated using efficient chemometric algorithms. The system is hosted in the cloud where the calibration models responsible for performing the multivariate analysis calculations necessary to generate the analytical results of each of the calibrated chemical and physical attributes are stored, from the spectra generated by the SpecSolo-Scan instrument. Therefore, according to the invention, the analyst does not require specific knowledge of chemometrics to treat and interpret the results generated since the system performs this task automatically. The robustness of the models is evaluated daily through trained algorithms and the updating of these models, or “recalibration”, happens automatically by inserting new analytical and spectral data of relevant soil samples into the system.

[0093] Table 3 shows the main features of each module that makes up the invention's digital platform, as follows:

Resource Module Registration of the accredited partner (SpecSolo client) on the platform SpecSolo-Manager Access control and privileges of the SpecSolo accredited partner SpecSolo-Manager Registration of samples with the respective requests for analytical services and printing of sample identification labels SpecSolo-Lims Record of receipt of the sample, by reading the bar code attached to the packaging of the SpecSolo-Lims

Petition 870170088564, of 11/16/2017, p. 44/65

36/44

sample Identify the sample for analysis using its barcode SpecSolo-Tray Command the equipment to collect spectra from the samples SpecSolo-Tray Associate the spectrum collected from the sample with its identification (sample ID) SpecSolo-Tray Store the spectra collected from the samples in a local database and send a copy of this spectrum, through SpecSolo-Services to the Cloud-hosted SpecSolo-Toolbox SpecSolo-Tray Calculate analytical results using calibration models SpecSolo-Toolbox SpecSolo-Lims automatically captures, through SpecSolo-Services, the analysis result generated by SpecSolo-Toolbox and makes available for consultation on the SpecSolo portal SpecSolo-Lims and SpecSolo-Services Printing test reports with analytical results SpecSolo-Lims Make the analytical result available for consumption by other applications through SpecSoloServices APIs SpecSolo-Services Management of services performed by access control by client SpecSolo-Manager Financial management of analytical services consumed SpecSolo-Manager

Petition 870170088564, of 11/16/2017, p. 45/65

37/44

by the accredited partner SpecSolo Provision of electronic payment methods for analytical services consumed by the accredited partner SpecSolo SpecSolo-Manager

[0094] As for the multivariate analysis algorithms, to be addressed sequentially to the digital platform, they are stored in the SpecSoloToolbox and make use of multivariate calibration methods to generate accurate and reliable results.

[0095] The invention contemplates a 'business model' based on SpecSolo technology, which consists of accrediting local partners so that they can offer the soil analysis service to farmers in their region. SpecSolo accredited partners - customers - can be selected from: (i) agronomic laboratories that provide services in soil fertility analysis; (ii) cooperatives, agronomic consultants, fertilizer and pesticide industries, agricultural resellers, agricultural groups, sugar and alcohol plants, among others, that use soil analysis information to guide their customers (rural producers) on the chemical management of the soil fertility and plant nutrition; or (iii) by rural producers.

[0096] The SpecSolo accredited partner will have an initial cost for the acquisition of the SpecSolo-Scan equipment, in addition to a variable cost depending on the number of soil samples processed through the platform. Therefore, the business model is based on charging a fee - value - per sample analyzed using the SpecSolo platform.

[0097] Regarding the method, or process of the invention, and for routine application of the infrared spectroscopy technique to determine the chemical and physical attributes of the soil, it is initially necessary to build calibration models for each of the attributes that will be

Petition 870170088564, of 11/16/2017, p. 46/65

38/44 analyzed in the soil, associating the result of the wet analysis, carried out according to the reference methods, with the corresponding spectrum of each soil sample, acquired through the SpecSolo-Scan equipment (1). For this, spectra were collected from more than 100,000 representative soil samples from all over the Brazilian territory and associated with the analytical results obtained by wetting these samples. Through treatment of multivariate data - chemometry - the calibration models were built for each of the 36 (thirty-six) parameters that were analyzed by the reference method in the more than 100,000 representative samples. The treatment of multivariate data was carried out with the aid of the SpecSoloToolbox software, where the spectral pre-processing routines, construction and choice of the best models, validation and prediction of results were optimized through efficient chemometric algorithms. Therefore, in a number “n” of spectra collected in a number of at least thousands of samples of a representative soil of a given location, through association with analytical results obtained by wetting these samples, when applying multivariate data treatment - chemometry - the calibration models are obtained for each of the dozens of parameters analyzed by the reference method.

[0098] After the construction of the calibration models, the user (or SpecSolo accredited partner) will be able to use the SpecSolo platform to carry out routine soil fertility analyzes.

[0099] For the construction of the calibration models, some sequential procedures must be observed, namely:

- Collection of the infrared spectrum of representative soil samples (> 100,000) using the SpecSolo-Scan equipment (1). These samples had previously been dried at 40 ° C, ground, passed through a 2 mm mesh sieve and analyzed by the reference method for the following parameters: P (Resin, mg / dm ³ ), P (mehlich, mg / dm ³ ), P (remainder,

Petition 870170088564, of 11/16/2017, p. 47/65

39/44 mg / dm ³ ), COS (oxidation, mg / dm ³ ), MOS (oxidation, mg / dm ³ ), pH (H2O), pH (CaCl2), pH (SMP), K (resin, mmolc / dm ³ ), Ca (resin, mmolc / dm ³ ), Mg (resin, mmolc / dm ³ ), Na (Mehlich, dm ³ ), H ³ + Al ³ (calculation, mmolc / dm ³ ), Al ³ (KCl , mmolc / dm ³ ), CTC (calculation, mmolc / dm ³ ), SB (calculation, mmolc / dm ³ ), V% (%), m% (%), S (calcium phosphate, mg / dm ³ ) , B (hot water, mg / dm ³ ), Cu (DTPA, mg / dm ³ ), Cu (mehlich, mg / dm ³ ), Fe (DTPA, mg / dm ³ ), Fe (Mehlich, mg / dm ³⁾ ), Mn (DTPA, mg / dm ³ ), Mn (mehlich, mg / dm ³ ), Zn (DTPA, mg / dm ³ ), K in CTC (calculation,%), Na in CTC (calculation,%), Al in CTC (calculation,%), H in CTC (calculation,%), Ca / K (calculation), Mg / K (calculation), Clay (HMS + NaOH, g / Kg), Silt (HMFS + NaOH, g / Kg), Total Sand (HMS + NaOH, g / Kg);

- The SpecSolo-Toolbox associates the analytical data of each sample, obtained by the reference method, with its respective infrared spectrum, creating a robust database;

- Through the processing routines and chemometric algorithms embedded in the SpecSolo-Toolbox, the best multivariate calibration models were created and validated for each of the 36 (thirty-six) parameters;

- After modeling, the SpecSolo-Toolbox software saves the calibration models to be used in forecasting the concentrations of the new soil samples analyzed in the infrared.

[0100] Figure 3 illustrates the interaction between the factors that make up the SpecSolo technology, that is:

- results of soil attributes obtained by wet methods (or reference method) (3);

- NIRS spectra (SpecSolo-Scan) (4);

- database (5) and;

- algorithms (6) (SpecSolo-Toolbox).

Petition 870170088564, of 11/16/2017, p. 48/65

40/44 [0101] On the other hand, Figure 4 illustrates the flowchart of operation of the algorithm for building the models and multivariate calibration and forecasting of results.

[0102] According to the flowchart in Figure 4, the process of building the models and multivariate calibration begins with the receipt and registration of the samples (7), followed by preparation in the air-dried fine earth (TFSA) condition (8). The samples are analyzed by NIR spectroscopy (9) and by the Reference Methods (10) and used in the construction of the calibration model (11). Then, the presence of outliers in the calibration model (12) is verified, in case they return to repeat the spectral measurements (9) and reference (10). The outiliers are evaluated based on the graph of Leveragevs residues in X (11B) and Y (11C), if not, proceed to the stage of spectral pre-processing (11D). The available pre-processing methods (13) are: Center on Average, Derivatives, MSC, SNV and Self-Scaling. After pre-processed (13), the data are used in the construction of the calibration (14) and validation (15) models. The calibration model is evaluated using the following figures of merit (16): Analytical sensitivity, signal / noise ratio, selectivity, accuracy (RMSEC and RMSEP), precision and determination coefficient (R2, cal and val) (17). After the construction and evaluation of the calibration model, the forecasting step for new samples begins, that is, use in routine soil analysis by NIR spectroscopy (18). The samples provided for in the routine are evaluated for the existence of outliers (19); if so, the outliers are added to the calibration model in order to update it (11); if not (20), the results are released (21) in three different ways (22): supplied individually (23), by strip (24) or by recommendation for fertilization and liming (25).

[0103] The present invention stands out, therefore, for bringing real solutions when it comes to chemical and physical analyzes of soil fertility by Vis-NIR spectroscopy, in order to simplify in an extraordinary way

Petition 870170088564, of 11/16/2017, p. 49/65

41/44 analytical procedure, limited, as a general rule, to two main steps:

- Sample Preparation: dry, grind, pass through a sieve and transfer the amount of soil sample sufficient to cover the bottom of the sample container (plastic cup), approximately 150g and;

- Analytical Determination: collection of the sample spectrum with the Vis-NIR spectrophotometer (1), which automatically sends the information to a cloud-based service server, where the calibration models that will produce the analytical results are stored and made available for consultation at portal, in less than 30 (thirty) seconds. It also has the advantage of analyzing different soil attributes with just one spectral reading: dozens of soil fertility parameters can be analyzed simultaneously in a non-destructive, fast and economical way.

[0104] Table 4 details each step of the analysis process with SpecSolo, while Figure 5 illustrates the flowchart of this process, as follows:

Stage No. of blocks in Flow of Fig. 5 description Benefits To use soil analysis services Partner Accreditation 26 and 27 SpecSolo requires accreditation beforehand. This stage of accreditation of the partner on the SpecSolo Digital Platform is carried out by the project's internal team through the system's SpecSolo-Manager Module, and at this moment the access rules are defined

Petition 870170088564, of 11/16/2017, p. 50/65

42/44

and partner privileges. Client register 28 Once accredited, the partner will be able to register their clients (farmers), through the SpecSoloLimes Module, and provide a login and password for them to have access to the platform. Management registration customers through digital platform Sample registration 29, 30, 32, 33, 34, 43 The registration of samples in the system can be done by the accredited partner or by your client (farmer). Both will use the same interface for registration, the SpecSolo-Lims Module. Registration by the Farmer: the farmer accesses the digital platform with his login and password and performs the registration of the samples with the respective requests for analytical services and defines the form of payment for the service. In this step, he generates the identification tags (barcode or QRCode) for printing to fix on the packages that he will place the collected soil samples for sending for analysis. Registration by the Accredited Partner: in the case of the Accredited Partner receiving samples without registration, he must proceed with the process of registering the samples with the respective analytical services and define the form of collection. In this step, the system will generate for printing the job labels (barcode) to be attached to the analysis package. Registration sample can be done online by yourself farmer (end user), streamlining the operational process in the routine and avoiding registration errors. 31 Registration of Sample Receipt Registered by the Farmer: if the Accredited Partner receives samples already Register of sample receipt by

Petition 870170088564, of 11/16/2017, p. 51/65

43/44

registered by the farmer, through the SpecSolo-Lims Module, he will read the identification code fixed on the package and generate for printing a work label (bar code) to be fixed on the analysis package. code reader (bars or QRCode), avoiding typing and streamlining the process. Sample preparation 36 Sample preparation consists of drying the sample in an oven at 40 ° C, grinding the sample in a hammer mill, passing the sample through a 2 mm mesh sieve (air-dried fine earth - TFSAR) and transferring the sample to the container analysis (plastic cup) identifying this cup with the job label (barcode). The analysis cups should then be distributed in the tray (2) of the equipment (1) SpecSolo-Scan, which has 40 positions. Spectrum Collection, Data Treatment and Publication of Results. 37 To start the analysis process, the Accredited Partner must only activate the SpecSolo-Scan equipment, through the software integrated with the SpecSolo-Tray equipment. Once the SpecSolo-Scan is activated, before starting the analysis process of the samples in the tray, it will calibrate the spectrophotometer using Spectralon. If the reading is approved by the system, the autosampler passes the second control, which is the reading of the standard sample, provided by SpecSolo for quality control of the process. The spectrum collected from the standard sample is automatically evaluated by a Control Chart and, if approved, the autosampler starts collecting data. Fast, accurate analysis, no need for reagents, reduced operating cost, results made available through Digital platform.

Petition 870170088564, of 11/16/2017, p. 52/65

44/44

spectra of the samples distributed in the tray, and in each position it also captures the sample identification by reading the bar code fixed on the cups. This entire process is controlled by the SpecSolo-Tray Module. 38 and 39 After the readings are finished, the system stores the spectra in a local database and sends a copy of the spectrum, through SpecSolo-Services to the Cloud-hosted SpecSolo-Toolbox. 40 The SpecSolo-Toolbox is the module responsible for performing the calculations of the analytical results through the previously built calibration models. 41.42.43 SpecSolo-Lims automatically captures, through SpecSolo-Services, the analytical result of the sample and makes it available for consultation on the SpecSolo Portal, which can be printed by the partner or by the customer (farmer). Management Finance 45 Closing the analytical services, the SpecSolo-Manager Module makes all financial management of the analytical services consumed by the Accredited Partner and provides payment methods electronic. Transparency and convenience

[0105] For the sake of clarity, and for a better definition of the product use parameters required according to the invention, it should be noted that the infrared spectra are acquired at ambient temperature and pressure, in the range of 400 to 2500nm.

Claims (5)

1) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE SCHEDULE”, comprises the provision of a new method, that is, a process, and the respective system involving the product (hardware + software ) in order to improve the procedures for chemical and physical analysis of soil fertility in large-scale routines, characterized by a technological package dedicated to soil analysis, which combines: - the use of a Vis-NIR spectrophotometer, called SpecSolo-Scan with the respective digital platform; - chemometric algorithms for multivariate analysis; - integrated cloud-based software to produce analytical results. 2/5 transfer the amount of soil sample sufficient to cover the bottom of the sample container), approximately 150g and; - Analytical Determination: collection of the sample spectrum with the Vis-NIR spectrophotometer (1), which automatically sends the information to a cloud-based service server, where the calibration models that will produce the analytical results are stored and made available for consultation at portal, in less than 30 (thirty) seconds; analyzes different soil attributes with just one spectral reading: dozens of soil fertility parameters can be analyzed simultaneously. 2) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN A LARGE SCALE ROUTINE”, according to claim 1, characterized by, for routine application of the infrared spectroscopy technique for determining chemical and physical attributes of the soil, it is necessary to build calibration models for each of the attributes that will be analyzed in the soil, associating the result of the wet analysis, carried out according to the reference methods, with the corresponding spectrum of each sample. of soil, acquired through the Inv-NIR spectrophotometer of the invention, or SpecSoloScan. 3/5 with the scanning time of a sample for spectrum collection less than 10 (ten) seconds. 6) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE SCHEDULE”, according to claim 1, characterized by the Digital SpecSoloser platform the technological package that integrates the SpecSolo-Scan equipment (1) systems for managing soil samples and calculating analytical results, consisting of the following modules: - SpecSolo-Manager. software for management and access control of partners accredited by SpecSolo; - SpecSolo-Tray: software for controlling SpecSolo-Scan equipment and organizing samples for analysis; - SpecSolo-Lims: software for managing soil samples and analytical services to be performed. Manages the entire analytical routine from the arrival of the sample to the availability of the result; - SpecSolo-Services: services portal for integration (communication) of the SpecSolo Digital Platform with its own modules or third-party modules; - SpecSolo-Toolbox: proprietary software for data modeling and prediction of results, where the spectral pre-processing routines, construction and choice of the best models, validation and prediction of automated results using chemometric algorithms. 7) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE ROUTINE”, according to claim 6, characterized in that the system is hosted in the cloud where the responsible calibration models are stored for performing the multivariate analysis calculations necessary to Petition 870170088564, of 11/16/2017, p. 56/65 3) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN A LARGE SCALE SCHEME”, according to claims 1 and 2, characterized by presenting, as a general rule, two main steps: - Sample Preparation: dry the sample in an oven at 40 ° C, grind the sample in a hammer mill, pass the sample through a 2 mm mesh sieve and Petition 870170088564, of 11/16/2017, p. 54/65 4/5 generate the analytical results, for each of the chemical and physical attributes calibrated, from the spectra generated by the SpecSolo-Scan instrument. 8) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE ROUTINE”, according to claims 1 and 2, characterized by, for the construction of calibration models, some sequential procedures must be observed, namely: - Collection of the infrared spectrum of soil samples representative of the region where the technology will be used through the SpecSolo-Scan equipment (1); - The SpecSolo-Toolbox associates the analytical data of each sample, obtained by the reference method, with its respective infrared spectrum, creating a robust database; - Through the processing routines and chemometric algorithms embedded in the SpecSolo-Toolbox, the best multivariate calibration models are created and validated for each of the analyzed parameters; - After modeling, the SpecSolo-Toolbox software saves the calibration models to be used in forecasting the concentrations of the new soil samples analyzed in the infrared. 9) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN A LARGE SCALE ROUTINE”, according to claims 1 and 8, characterized by after the construction of the calibration models, the user - or SpecSolo accredited partner - be able to use the SpecSolo platform to perform routine soil fertility analyzes. 10) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN A LARGE SCALE SCREEN”, in accordance with claim 9, characterized by the ‘model of Petition 870170088564, of 11/16/2017, p. 57/65 4) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE ROTINE”, according to claim 1, characterized in that the equipment (1) dedicated to soil analysis by VisNIR spectroscopy the CARACTERISTICS: - presence of an automatic sampler (autosampler) for sequential analysis of soil samples arranged in tray (2), with capacity for at least 40 (forty) positions; - proximity sensor coupled to the autosampler to standardize the distance of the spectrum collection, that is, the distance between the beam of infrared radiation and the sample; - presence of a barcode reader coupled to the autosampler for identification and registration of the sample in the tray; - optionally to the tray, an automated conveyor can be used for feeding. 5) “ANALYSIS OF CHEMICAL AND PHYSICAL ATTRIBUTES OF SOIL FERTILITY BY VIS-NIR SPECTROSCOPY FOR USE IN LARGE SCALE SCHEDULE”, according to claim 1, characterized by the acquisition of the software that controls the equipment. spectra of soil samples in all 40 (forty) positions available in the support, Petition 870170088564, of 11/16/2017, p. 55/65 5/5 business' consists of accrediting local partners to offer soil analysis services to farmers in their region. The 'business model' in question foresees an initial investment, by the accredited partner, for the acquisition of SpecSolo-Scan (1) and collection of fee - value - depending on the number of soil samples processed through the SpecSolo platform, or that is, by analyzed sample (by “click”).