nueva página del texto (beta)
Inglés (pdf)
Artículo en XML
Referencias del artículo
Traducción automática
Enviar artículo por email
Citado por SciELO 
Similares en
SciELO 
Permalink
Introduction
The significance of soil organic matter has recently been recognized as a natural process of carbon storage that may help to mitigate climate change (Viscarra-Rossel et al. 2008, Powlson et al. 2011, Ontl and Schulte 2012). This has led to the organization of a worldwide network to develop large databases of soil organic carbon inventories (Paz and Etchevers 2016). However, the traditional method of analysis of organic matter in soils is relatively expen- sive, requires intensive laboratory work, and is definitively time-consuming polluting waste are produced from the analysis. A far more practical technique for determining a soil’s composition is the indirect method of studying the soil’s color for indications of its composition (Levin et al. 2005, Viscarra-Rossel et al. 2008, Cortés et al. 2015, Hausmann et al. 2016).
The soil color is the physical property of primary consideration in the identification of soil types (Spielvogel et al. 2004), soil ethnopedological classes (Bautista and Zink 2010, Sánchez-Hernández et al. 2018), and orders or primary groups of soils (IUSS Working Group WRB 2015). The study of soil color has also been widely used in the research of soil genesis (Kumaravel et al. 2010), as well as for the identification of fertile soils (Schulze et al. 1993, Leirana-Alcocer and Bautista 2014) and automated identification of soil horizons (Zhang and Hartemink 2019).
Some compounds that give color to the soil are minerals and organic matter. For example, the colors vary with the presence of iron oxides (Torrent et al. 1983, Schwertmann 1993, Levin et al. 2005, Viscarra et al. 2008), soluble salts such as calcium carbonate, gypsum and others (Sánchez et al. 2004), heavy metals (Cortés et al. 2015, Marín et al. 2018, Delgado et al. 2019) and organic carbon (Torrent et al. 1983, Bédidi et al. 1992, Schwertmann 1993, Lindbo et al. 1998, Viscarra-Rossel et al. 2008, Vodyanitskii and Savichev 2017).
Colorimeters for the analysis of solid samples, such as soil, have been manufactured in recent years. At the same time, several color systems have been developed that can be expressed numerically, as CIE-RGB, CIE-L*a*b* and CIE-XYZ (Leirana-Alcocer and Bautista 2014, Cortés et al. 2015, Aguilar et al. 2013, Marín et al. 2018). These color measurement systems allow mathematical relationships to be established with other soil properties (Leirana-Alcocer and Bautista 2014, Levin et al. 2005, Cortés et al. 2015, Marín et al. 2018, Delgado et al. 2019). The CIE-L*a*b* parameters are useful in obtaining the optimum redness index (Kirillova et al. 2014), which is helpful in determining the presence and contribution of Fe-oxides on percentage (Vodyanitskii and Savichev 2017). L* represents the contrast ranging from black to white (0-100) and a* and b* are chromatic coordinates, a* being the variance from red to green and b* that from yellow to blue (CIE 1978).
Simon et al. (2020) concluded that the relation between the color of soil and other properties as organic matter, texture, soil chemical composition, and particle size are variables; thus it is necessary to develop precise predictive models under soil specific properties of each place. Shields et al. (1968), indicated that the concentration and nature of the or- ganic carbon in the organic matter generate several colorations. According to Chen et al. (2018), the coordinates a y b of a system of colors CIE L*a*b*, correlate intimately with soil organic carbon concentration.
In the tropical karst areas of Mexico on the Yucatan peninsula, there are large areas with soils of contrasting colors, which vary between white limestone rock and black organic matter (Bautista et al. 2003, Bautista 2021, Fragoso et al. 2017). To improve the accuracy of soil organic carbon inventories, it will be necessary to analyze thousands of soil samples; for this reason, it will be essential to generate models for estimating soil organic matter with proxy technologies. Hypothetically, the physical and chemical characteristics of a karstic soil don’t impede the development of robust models to predict the organic matter concentration from soil color properties. Thus, the aim of this study was to explore the use of soil color parameters in order to estimate the organic matter in soils from a karstic zone in the Yucatan Peninsula in México.
Materials and methods
The study zone is in Chetumal (Quintana Roo), the south Yucatan Peninsula at the southeastern part of Mexico. This region is a large limestone plain with a tropical climate, where the Leptosols dominate the poorly developed karst plains, although there are also other soil groups such as Gleysols, Vertisols, Phaeozems, and Luvisols (Bautista et al. 2011, Fragoso et al. 2017). On the peninsula of Yucatán, the dominant vegetation is the low and subdeciduous tropical forest and medium and perennial tropical forest.
Fifty soil samples were collected; these samples were airdried in the shade and sieved with a 2 mm mesh. The soil samples were selected from a set of samples considering that the colors should be between the white color where limestone predominates and the black color due to the high percentage of organic matter in the soil. The chemical analyzes performed on the soil samples were: pH (Lean 1982), organic matter was measured using the wet oxidation method with potassium dichromate (Nelson and Sommers 1982), exchangeable cations Ca, Mg, Na and K with ammonium acetate (Okalebo et al. 1993).
An X Ray Diffraction (XRD) analysis was performed to identify minerals present in a soil sample. As calcite was the dominant mineral and prevents the identification of residual minerals, 10% HCl was added to destroy the carbonates. A soil sample was placed on a silicon sample holder coated with silicone grease suitable for XRD; subsequently, they were analyzed on a Siemens D-5000 diffractometer, Bragg-Brentano Mode, with a monochromatic Cu tube (l = 1.5418 Å), a step time of 3 seconds, step size 0.02 degrees, at 34KV and 25 mA.
The organic matter index
The color of the soil samples was analyzed using a Konica Minolta CR-5 reflectance and transmission colorimeter. The color parameters were obtained using the system CIE-L*a*b* and CIE-XYZ defined by the International Commission on Illumination (CIE). The use of the CIE-L*a*b* simplifies and strengthens statistical calculations (Vodyanitskii and Savichev 2017) and the CIE-XYZ is the base of the transmission to any other color space (Viscarra- Rossel et al. 2006).
In the CIE-XYZ model, X is the color red, which varies from 0 to 0.9505, Y is the color green, which varies from 0 to 1.0, and Z is the color blue, which varies from 0 to 1.089 (Kirillova et al. 2014). The redness index was introduced by Barron and Torrent (1986) to estimate the percentage content of hematite in soils, but it should simply be called "color index" like this in general, because it includes all the parameters of the CIE-L*a*b* color system and therefore can be associated with the materials and minerals that give the soil its color (Bautista et al. 2003; 2005),
where RI = redness index, L* = luminosity, a* = coordinates of red/green, and b* = coordinates of yellow/blue. In this paper we will refer to this as the organic matter index (OMI).
The relationship between the soil organic matter (SOM) and the OMI is established under the assumption of a possible curvilinear behavior (Schulze et al. 1993), and proposes an adjustment based on the power regression analysis with the two terms.
where a, b and c are the coefficients that should be found with a confidence limit of 95%; however, another adjustment is proposed with a logarithmic regression analysis (Viscarra-Rossel et al. 2008).
where a and b are the coefficients that should be found for both equations with a confidence limit of 95% using regression analysis. A cross validation between the SOM and the equations obtained (Eq. 2, Eq 3) was applied in order to verify the viability to obtain the percentage of SOM.
Formation of soil sample groups by color
The soil samples were separated into color groups using color parameters following the methodology established by Cortes et al. (2015). The first step was to transform the parameters of CIE-XYZ system to the RGB system to rescale the XYZ triplets and subsequently make use of the square matrix 3 x 3 for the standard illuminant D65 at 2° (Viscarra-Rossel et al. 2006).
The CIE-RGB was used to make the cluster analysis and separate the soil samples into groups based on the similarities and differences between the color parameters with a k-means clustering (Matlab 2020a).
Based on the study of Vodyanitskii and Savichev (2017), we use the brightness (L), redness (a) and yellowness (b) to make a linear regression analysis for each group obtained.
The parameters of color form a rectangular matrix MATN,4, and the SOMN,1 is a vector of the soil organic matter.
The vector of the coefficients X4,1 is determined by the method of least squares with the Moore-Penrose pseudo-inverse algorithm.
The present study used the measure of error referred to as the K-factor (ΔSOM/SOM1/2) (Vodyanitskii and Savichev 2017). The multiple determination coefficient R2 was applied to observe whether the estimated SOMe was accurate about the original SOM. The number of the samples should be quite large (N>8) and the content of the SOM greater than 0.4% to obtain a strong correlation (Vodyanitskii and Savichev 2007).
Finally, the SOM and OMI by color group of soils were compared using the Kruskal-Wallis test, as it is the best method to compare population in which there is no gaussian distribution of the data. Kruskal and Wallis test (1952)evaluates the hypothesis that the median of each group is equal; it combines the data of every group and orders them from least to greatest, and subsequently calculates the average range for the data of each group.
Results
The organic matter index
Descriptive statistics provided the values of the variation of the SOM and the OMI (Table 1) in all of the samples, making it possible to observe the variation of the SOM in its concentrate between 2.11 ± 1.30 and that of the OMI between 18.80 ± 18.28 with a median similar to the mean that indicate that the samples belong to the same group of soil samples. Other chemical properties of soils also show a wide range of variance, such as the CEC ranging from 18.9 to 34.7 cmol kg−1 (Table 1).
Table 1 Chemical properties of soils
| pH | EC | Ca | Mg | K | Na | CE C | SOM | SOMI | |
| ------ cmol/kg ------ % | |||||||||
| X | 7.8 | 0.59 | 36.1 | 2.6 | 0.2 | 0.4 | 25.2 | 2.1 | 18.8 |
| Sd | 0.2 | 0.13 | 6.5 | 0.7 | 0.1 | 0.4 | 4.0 | 1.3 | 18.2 |
| Max | 8.3 | 1.19 | 48.9 | 4.7 | 0.4 | 1.2 | 34.7 | 5.1 | 92.4 |
| Min | 7.4 | 0.39 | 24.2 | 1.6 | 0.0 | 0.0 | 18.9 | 0.0 | 2.7 |
EC = electrical conductivity; CEC = cation exchange capacity; SOM = soil organic matter; SOMI = soil organic matter index
The SOM vs. OMI has two mathematical adjust by power and logarithmic fit with a small deviation that ensured the correlation between both parameters with an R2>0.86 and a R2>0.87 (Table 2, Figure 1) respectively. The relation that exists between SOM vs SOMI provided an equation to estimate the soil organic matter (SOMe) and when is made a crossvalidation between the SOM (measured) and the SOMe was obtained a R2>0.85 and the RMSE 0.50% which indicates a clear association.
Table 2 Regression analysis, equation for correlation and cross-validation
| Regression analysis | R2 | SOM vs. SOMe | Cross-validation (R2, RMSE) |
| SOM = 9.75(SOMI)0.101 - 10.5 | 0.86 | f(x) = 0.86x + 0.30 | 0.85, 0.51 |
| SOM = 1.26Ln(SOMI) - 1.05 | 0.87 | f(x) = 0.85x + 0.31 | 0.85, 0.52 |
R2 = the square correlation between the response values and the predicted response values; SOM = Soil’s Organic Matter; SOMe = Estimated Soil Organic Matter; SOMI = organic carbón index RMSE = Root Mean Squared Error
Figure 1 Fit analysis between the SOM and the SOMI: a) Power regression analysis and cross-validation, b) Logarithmic adjustment and cross-validation
The multiple linear correlations between the CIE-L*a*b parameters and the SOM provided good results for each group, especially for Groups I, II, and V. This further emphasizes the relationship that exists between the three parameters of color in the CIE-L*a*b* system with the SOM, providing equations to estimate the soils organic matter estimated in the Groups (SOMeg).
Groups II and V did not meet the first requirements of the number of samples N> 8 for a suitable adjustment, but despite this, the results present a substantial correspondence.
Formation of soil sample groups by color
The full set of samples was organized into five groups divided by color: Group I with soils of a pinkish white color, Group II with soils of a brownish grey color, Group III with soils of a grey color, Group IV with soils of a greyish brown color, and Group V with soils of a dark grey color (Table 3, Figure 2). These soil colors have significative difference for the SOM and OMI observed in Figure 3 by the box plot.
Table 3 Descriptive statistics of the soil organic carbon in the five color groups.
| Groups/Colors | I Pinkish white | II Brownish grey | III Grey | IV Greyish brown | V Dark grey |
| Mean SOM (%) | 0.55 | 1.66 | 2.36 | 2.96 | 4.44 |
| Maximum | 1.21 | 2.22 | 3.11 | 3.56 | 5.12 |
| Minimum | 0.16 | 0.48 | 0.03 | 1.83 | 2.69 |
| Standard deviation. | 0.27 | 0.73 | 0.79 | 0.57 | 1.17 |
| N | 13 | 6 | 16 | 11 | 4 |
| n = sample number. |
Figure 2 Color groups of soil samples. L* = luminosity, a* = coordinates of red/green, and b* = coordinates of yellow/blue (CIE 1978)
These five groups were independently related with their SOM by a multiple linear system. The correlation of each of the groups is well defined, especially for Group V, where the system that it conformed was quite outstanding by the mathematical analysis due to the matrix SOM4×4 being quadratic, which ensured the solution of the system. The amounts of the SOM have the major percentage for the group V and gradually decrease until the group I (Table 4).
Table 4 Linear regression between the soil’s organic carbon and parameters of color.
| Group | SOM (%) | SOM = XL L* +Xa a* + Xb b* + X0 | SOMe (%) | R2 | ||
| I (pinkish white) | 0.55 | 0.82 | 3.41 - 0.043L* - 0.16a* + 0.084b* | 0.54 | 0.80 | 0.92 |
| 0.28 | 0.28 | |||||
| II (brownish grey) | 1.66 | 2.39 | 12.96 - 0.15L* + 0.83a* - 0.34b* | 1.66 | 2.31 | 0.89 |
| 0.93 | 1.01 | |||||
| III (grey) | 2.36 | 3.15 | 5.75 + 0.01L* - 0.14a* - 0.36b* | 2.56 | 3.20 | 0.85 |
| 1.57 | 1.92 | |||||
| IV (greyish brown) | 2.96 | 3.53 | 12.18 - 0.24L* - 6.67a* + 2.14b* | 2.79 | 3.26 | 0.83 |
| 2.39 | 2.32 | |||||
| V (dark grey) | 4.44 | 5.61 | 50.53 - 0.75L* - 16.7a* + 3.33b* | 4.47 | 5.65 | 1.0 |
| 3.27 | 3.30 | |||||
SOM = the soils’ organic matter, SOMe = soils’ organic matter estimated by Group, R2 = the square coefficient of correlation
The multiple linear correlations between the CIE-L*a*b parameters and the SOM provided good results for each group, especially for Groups I, II, and V. This further emphasizes the relationship that exists between the three parameters of color in the CIE-L*a*b* system with the SOM, providing equations to estimate the soils organic matter estimated in the Groups (SOM).
Groups II and V did not meet the first requirements of the number of samples N> 8 for a suitable adjustment, but despite this, the results present a substantial correspondence.
Discussion
The organic matter index
The value of the correlation between SOM (measured) and SOMe (estimated) in this study was acceptable compared to that obtained by Stiglitz et al. (2017). They used the soil color parameters as a predictor of the SOM developing a prediction model, using the soil depth, L * and a * as independent variables in dry soils obtaining values of R2= 0.7978 and RMSE = 0.0819. In contrast, in soils wet R2 = 0.7254 and RMSE = 0.1536. These results suggest that the soil color is efficient for the rapid determination of SOM. However, they warn that the high iron contents, carbonates, depth, and the humidity of the soil are variables that can negatively affect the model’s predictive capacity. To improve the accuracy of color measurement with electronic equipment we recommended taking into account: a) the size of the particle (Sánchez et al. 2004); b) soil moisture (Domínguez et al. 2012); c) particulate organic matter.
A better fit was obtained between SOM and SOMe because the parameter L (luminosity) plays an important role due to the colors of the soil samples vary from white (Calcite) to black (humidified organic matter) (Figure 2).
The content of exchangeable cations (Ca, Mg, Na, and K), the pH value, electrical conductivity, and the cation exchange capacity are typical of soils developed on limestone (Bautista et al. 2011).
Calcite and, to a lesser extent, quartz are the minerals that appear in the diffractograms. Once the Calcite is eliminated with HCl, other minerals appear, such as Tosudite (white, light yellow, light green), Hematite (red), Dickite(white), Boehmite (white), and Goethite (brown) (Figure 4); in addition, the sample turns darker in color because both the Calcite and the Quartz are white, which gives the soil sample more luminosity. As occurs in Leptosols of karst origin of the Yucatan peninsula (Bautista et al. 2011). The color of the mineral fraction of the soil must be taken into account because, in some cases, it is the one that dominates the color of the soil, mainly iron minerals (Barron y Torrent, 1986, Schwertmann1993, Schulze et al. 1993).
Formation of soil sample groups by color
According to Simon et al. (2020), SOM has a great soil darkening capacity, which even masks the white colors of minerals. This property can be adequately predicted through color due to the strong re lationship between the color and nature of the soil or ganic matter. Humic acids with higher carbon richness had darker colors (Shields et al. 1968), which explains the dark gray color of group V and the higher percentage of organic matter.
In the opinion of Chen et al. (2018), the CIE L * a * b * color system alone can predict the percentage of SOM, although multiple linear regression analysis can marginally improve the prediction; the b* coordinate correlates negatively with the concentration of SOM, mainly expresses the yellowish colors, so it is related to the low concentration of SOM; the a * coordinate, on the other hand, exhibits a stronger correlation with brownish colors, while the L * coordinate shows a low correlation with the concentration of the SOM.
The five equations obtained (Table 4) for each group can be used to estimate organic matter in large collections of soil samples in karst areas; however, further equations must be generated for soil samples with other colors such as reds, yellows and browns that also exist in karstic zones from peninsula of Yucatan (Bautista et al 2003, Bautista et al. 2005, Bautista et al. 2011).
The association between soil color and organic matter is widely known (Schulze et al. 1993); the mathematical model proposed between the color components (L*a*b) and the percentage of organic matter is relevant in this study. In this same sense, other mathematical models have been proposed (Spielvogel et al. 2004, Domínguez et al. 2012, Stiglitz et al. 2017, Chen et al. 2018); however, they are very different because the soils are also different in mineralogy and the type of organic matter and particle size.
Conclusions
The SOMI allowed to estimation the soil’s organic matter using both Power and Logarithmic fit. The grouping of soil samples by color allowed to describe a linear relationship between the soil color and its organic matter percentage, which improved the efficiency of this proxy technique. The darker or lighter colors as dark grey and pinkish-white, showed a higher level of R2, concerning other colors as brownish grey, grey, and greyish brown. Thus, the correlation sequence of color groups is V (dark grey)> I (pinkish white) >II (brownish-grey) >III (grey) >IV (greyish brown). For the karstic conditions of the Yucatan Peninsula, the study of soil color (SOMI and color parameters) may be considered useful for the estimation of the organic matter in large collections of soil samples, even in samples with low SOM content. Furthermore, this technique is much cheaper and less time-consuming compared to the traditional environmentally harmful waste products
