Mid-Infrared Spectroscopy Supports Identification of the Origin of Organic Matter in Soils

Spectroscopic methods combined with statistics have recently gathered substantial interest in pedological studies. Near-infrared (NIR) spectroscopy has been utilized, for example, for reconstructions of the history and transformations of Chernozems, although no similar research was conducted based on mid-infrared (MIR). In this paper, the relevance of MIR spectroscopy was tested in studies on the origin/affinity of organic matter from chernozemic soils. Samples collected from three vegetation classes (grasslands, forests and arable lands) were investigated using MIR spectroscopy in order to create a statistical model, which was applied on buried profiles of unknown origin. The results showed a clear disjunction of vegetation classes. Samples of buried soil were placed in the space between all classes, indicating the relation to variable vegetation. Therefore, arable lands should not be omitted in paleoecological reconstructions, because we cannot exclude the cultivation of fertile soils before their burial. It was concluded that MIR methods may have similar applicability to NIR spectroscopy. Additionally, MIR spectra may also be discriminated according to the recognized soil type, which allows for direct reconstructions of the transformation trends in buried profiles.


Introduction
Reflectance spectroscopy has gathered significant interest in environmental studies, particularly in recent years. Spectroscopic methods have been applied for quantitative and qualitative analysis of soil components, such as humic substances, carbonates and clay minerals [1,2]. However, the main reason for its popularity is the possibility to analyze the entire soil sample at once to obtain information about its chemistry [3][4][5]. The main advantages of spectroscopy are rapid spectral measurements and minimal, nondestructive sample preparation, highlighted in contrast to destructive chemical analysis [6]. The most promising spectral methods, such as visible (VIS), mid-infrared (MIR) and nearinfrared (NIR) spectroscopies, create new possibilities in the process of assessing soil quality, mainly due to the complex insight into its physical and chemical properties [7]. These approaches could have a particular importance for large-scale inventories and long-term monitoring programs [8]. Moreover, reflectance spectroscopy has been widely adopted for the simultaneous prediction of a variety of soil attributes, such as texture, carbonate content, soil organic carbon and cation exchange capacity, all of which are particularly useful for soil cartography to increase the precision of agricultural treatments [9,10]. This approach requires calibration to relate the spectra with the measured properties of soils/materials. Furthermore, additional pre-mathematical processing is necessary due to an overlapping or low intensity of peaks from various chemical functional groups [11,12].
Absorbing groups in the IR electromagnetic spectrum describe the organic and mineral composition of the compartments in the soil mixture. Many papers provide detailed descriptions of selected ranges of VIS, NIR or MIR spectra that correspond to specific functional groups of components [13,14]. In the context of mid-infrared spectra studies, the presence of peaks between 3000 and 2820 cm -1 is highly related to C-H aliphatic chain vibrations, which are characteristic of soil organic matter (SOM). Other popular regions of interest reflect clay mineralogy such as kaolinite, smectite and illite (3900-3600 cm −1 and 1200-1000 cm −1 ) [15]. The bands at 1500-1400 cm −1 may be useful for the detection of carbonates in soils [14].
The more specific application of various ranges of absorption spectroscopy strongly depends on the aforementioned techniques' limitations. For example, due to significantly less sensitivity to background interference caused by soil mineralogy, NIR spectroscopy seems to be more suitable for predicting soil properties [16] or pollution concentrations [17]. Recently, a new application of NIR spectroscopy has been tested for distinguishing soil organic matter type (or origin) according to various kinds of vegetation. It is assumed that the single soil spectra contain "fingerprints" of plant materials, which may be used to identify the vegetation that provides the biomass for SOM actually present in the soil [18]. Following this assumption, researchers [19,20] attempted to reconstruct the vegetation history on the chernozemic soils in Central Europe. Due to the capacity of mineral and organic signature detection, the application of MIR is preferentially centered on clay mineral and humic and fulvic acid studies [21]. However, MIR is increasingly being introduced in similar research where NIR has been used or is even used simultaneously [22,23]. However, there is no research focused on paleoenvironmental reconstruction based on organic matter that uses this spectral range.
This research aimed to verify the usefulness of the MIR spectral range (4000-525 cm −1 ) in the process of differentiating SOM under various types of vegetation, as well as to examine the paleoenvironmental reconstruction of a few buried profiles of chernozemic soils. The investigations may contribute to the further development of reflectance spectroscopy methods that pursue answers to the process of self-preservation of Chernozems until modern times in Central Europe along with their possibility of transformation into other soil types (e.g., Phaeozems, Luvisols).

Soil Sample Collection
The study focused on the Chernozems and related soils developed from loess, having a similar texture of silt loam. Samples were collected from arable fields, forests and grasslands existing under the same land use for a long time (at least since the beginning of the 20th century) that is checked on available topographical maps [19]. The sampling sites were located in the patches of Chernozems preserved in the loess belt of South Poland (Wrocław Plain, Głubczyce Plateau, Miechów Upland, Hrubieszów Basin and Przemyśl Foothills) ( Figure 1). However, field inspection during the sampling has revealed a deep leaching of carbonates or a lack of mollic/chernic horizons in some profiles respective to the present temperate climate [23] or history of occupation and land use [24]. Therefore, some soils did not fulfil the criteria of Chernozems in terms of World Reference Base (WRB) classification [25] and were classified as Phaeozems or Luvisols, in line with recent reports [26][27][28].
Soil samples were collected from the topsoil humus layers (i.e., from the depth of 0-10 cm in forest and grassland soils, and from the plow layer of arable soils) to obtain the closest relationship between the spectral picture of organic matter present in soil samples and the current vegetation cover. The plant cover of the non-arable areas consisted of multi-species grasslands and, in case of forest, we limited the type of vegetation to deciduous trees only. In the forest composition, we discriminated three main groups of stands, according to Soil samples were collected from the topsoil humus layers (i.e., from the depth of 0-10 cm in forest and grassland soils, and from the plow layer of arable soils) to obtain the closest relationship between the spectral picture of organic matter present in soil samples and the current vegetation cover. The plant cover of the non-arable areas consisted of multi-species grasslands and, in case of forest, we limited the type of vegetation to deciduous trees only. In the forest composition, we discriminated three main groups of stands, according to prevailing tree species: oak-(Quercus robur L.), beech-(Fagus sylvatica L.) and hornbeam-/maple-/lime-(Carpinus betulus, Acer pseudoplatanus L., Tilia cordata Mill) dominated stands.
As a testing material for the created statistical model, three buried soils from SW Poland were chosen from the barrows (kurgans) of the Funnel Beaker culture formed during the Neolithic period [29][30][31][32]. All buried soils were covered by a thick overlying barrow mound (up to 100 cm), which has preserved their humus horizons from transformation into Luvisols that naturally occurred in a direct surrounding of the kurgans. Samples from approximately 20 cm-thick layers were collected starting from the uppermost buried A horizon, down to the transitional AB horizons (the sampling depths differed in particular kurgans).

Sample Preparation and Analysis
From each location that represented a different type of vegetation cover, 2 to 5 samples were taken. The final database contained 348 soil samples in total. Samples were airdried, crushed and sieved through a 2 mm mesh. For standard soil characteristics, the particle-size distribution was analyzed using the hydrometer method. The pH in a water suspension (1:5 v/v) was measured potentiometrically with a Mettler Toledo pH-meter, and the calcium carbonate content (equivalent) was determined by a volumetric method. The soil organic carbon (SOC) content was determined using a dry combustion method (CS-MAT analyzer, Bruker AXS Inc., Madison, USA). An estimation of SOC content allowed us to verify the correct selection of samples because it is recommended to build the spectral library using samples containing at least 0.3% SOC [19]. Before spectral measure- As a testing material for the created statistical model, three buried soils from SW Poland were chosen from the barrows (kurgans) of the Funnel Beaker culture formed during the Neolithic period [29][30][31][32]. All buried soils were covered by a thick overlying barrow mound (up to 100 cm), which has preserved their humus horizons from transformation into Luvisols that naturally occurred in a direct surrounding of the kurgans. Samples from approximately 20 cm-thick layers were collected starting from the uppermost buried A horizon, down to the transitional AB horizons (the sampling depths differed in particular kurgans).

Sample Preparation and Analysis
From each location that represented a different type of vegetation cover, 2 to 5 samples were taken. The final database contained 348 soil samples in total. Samples were air-dried, crushed and sieved through a 2 mm mesh. For standard soil characteristics, the particle-size distribution was analyzed using the hydrometer method. The pH in a water suspension (1:5 v/v) was measured potentiometrically with a Mettler Toledo pH-meter, and the calcium carbonate content (equivalent) was determined by a volumetric method. The soil organic carbon (SOC) content was determined using a dry combustion method (CS-MAT analyzer, Bruker AXS Inc., Madison, USA). An estimation of SOC content allowed us to verify the correct selection of samples because it is recommended to build the spectral library using samples containing at least 0.3% SOC [19]. Before spectral measurements, samples were additionally ground to pass a 0.1 mm mesh, which aimed to homogenize the material as thoroughly as possible. Then, samples were dried at 37 • C to remove water absorbed from the air.
MIR spectra were performed on approx. 0.5 g of material using a Nicolet iZ10 FT-IR spectrometer with the Smart iTX accessory supplied with a diamond plate (Thermo Scientific) in the range 4000-525 cm −1 with a resolution of 2 cm −1 . Recorded reflectance was transformed automatically into absorbance. Obtained spectra created a data matrix with 348 rows (samples) and 7200 columns (peaks), which were later averaged in blocks (consisting of 20 subsequent readings) to 340 columns that allowed for further discriminant analysis. Thus, the final resolution of spectral database was 40 cm −1 . Two-step mathematical pretreatment [19,33] was performed (Unscrambler 10.4 software, Norway) to maximize the differentiation between the target groups of samples. First, the spectra were standardized to zero mean and variance = 1 ( Figure 2B) to reduce the influence of organic matter quantity and the size of particles on the final spectral image. Then, the first derivative was applied ( Figure 2C), which, as expected, increased the quantity of information from functional groups of organic compounds.
analysis. Thus, the final resolution of spectral database was 40 cm −1 .
Two-step mathematical pretreatment [19,33] was performed (Unscrambler 10.4 software, Norway) to maximize the differentiation between the target groups of samples. First, the spectra were standardized to zero mean and variance = 1 ( Figure 2B) to reduce the influence of organic matter quantity and the size of particles on the final spectral image. Then, the first derivative was applied ( Figure 2C), which, as expected, increased the quantity of information from functional groups of organic compounds.
Ertlen [33] recommended the first or second derivative, respectively, to the Mahalanobis distances. The Mahalanobis distance, defined as the distance between means of canonical scores, was used to assess the discrimination between examined groups [34,35]. In the present study, no significant increase in Mahalanobis distances was observed between derived groups after the second derivative application; thus, further calculations were performed using the spectra after the standardization and application of the first derivative. Canonical variate analysis (CVA), was applied (using a Systat 13.2 software, Chicago) to (a) distinguish the groups of samples based on multiple variables, i.e., the transformed MIR spectra, (b) to measure the Mahalanobis distances between groups and (c) predict the affinity of samples of unknown type/origin to the distinguished groups [33]. Additionally, Ertlen [33] recommended the first or second derivative, respectively, to the Mahalanobis distances. The Mahalanobis distance, defined as the distance between means of canonical scores, was used to assess the discrimination between examined groups [34,35]. In the present study, no significant increase in Mahalanobis distances was observed between derived groups after the second derivative application; thus, further calculations were performed using the spectra after the standardization and application of the first derivative.
Canonical variate analysis (CVA), was applied (using a Systat 13.2 software, Chicago) to (a) distinguish the groups of samples based on multiple variables, i.e., the transformed MIR spectra, (b) to measure the Mahalanobis distances between groups and (c) predict the affinity of samples of unknown type/origin to the distinguished groups [33]. Additionally, basic properties of the distinguished groups were statistically characterized. SOC content and pH values had a normal distribution, checked by the Kruskal-Wallis test; thus, the mean values, standard errors and standard deviations were calculated to characterize the central tendency and data dispersion. The significance of differences between means was checked using the post-hoc LSD Fisher test at p < 0.05. All above mentioned calculations and related graphs were performed using the Statistica 13 package (USA).

Results
Although all soils under study were developed from similar parent material (loess) and presumably had a similar primary chernozemic origin, presently they differ in their physicochemical properties, mainly depending on land management. Arable soil had the highest, whereas forest soils-the lowest pH values, ca. 7 and 5, respectively ( Figure 3A). Contrastingly, the SOC content was significantly higher in the forest and grassland than in arable soils, with average values ca. 4.3% and 1.3%, respectively ( Figure 3B). Among the dominant soil types (i.e., reference soil groups in terms of WRB classification), the topsoil samples of Luvisols were acid, whereas the topsoil samples of Chernozems and Phaeozems were neutral or alkaline, with mean pH values of ca. 4.6 and 7.5, respectively ( Figure 3C). The SOC content was less differentiated, but the mean SOC values in Luvisols, ca. 3.7%, were significantly higher than the mean SOC in Chernozems and Phaeozems, ca.'2.1-2.5% ( Figure 3D). mean values, standard errors and standard deviations were calculated to characterize the central tendency and data dispersion. The significance of differences between means was checked using the post-hoc LSD Fisher test at p < 0.05. All above mentioned calculations and related graphs were performed using the Statistica 13 package (USA).

Results
Although all soils under study were developed from similar parent material (loess) and presumably had a similar primary chernozemic origin, presently they differ in their physicochemical properties, mainly depending on land management. Arable soil had the highest, whereas forest soils-the lowest pH values, ca. 7 and 5, respectively ( Figure 3A). Contrastingly, the SOC content was significantly higher in the forest and grassland than in arable soils, with average values ca. 4.3% and 1.3%, respectively ( Figure 3B). Among the dominant soil types (i.e., reference soil groups in terms of WRB classification), the topsoil samples of Luvisols were acid, whereas the topsoil samples of Chernozems and Phaeozems were neutral or alkaline, with mean pH values of ca. 4.6 and 7.5, respectively ( Figure 3C). The SOC content was less differentiated, but the mean SOC values in Luvisols, ca. 3.7%, were significantly higher than the mean SOC in Chernozems and Phaeozems, ca. 2.1-2.5% ( Figure 3D). The MIR spectra (standardized, first derivative) averaged for the distinguished landuse classes clearly differed in intensity and width of characteristic bands, in particular in the Section 900-1100 cm −1 (Figure 4). These differences have variable scales for particular bands and their individual analysis does not reflect, neither quantitatively nor qualitatively, the mutual relationship between classes. In contrast, the canonical variate analysis yielded a clear differentiation, without mutual overlapping, of the soil samples into classes according to land use ( Figure 5). The MIR spectra (standardized, first derivative) averaged for the distinguished landuse classes clearly differed in intensity and width of characteristic bands, in particular in the Section 900-1100 cm −1 (Figure 4). These differences have variable scales for particular bands and their individual analysis does not reflect, neither quantitatively nor qualitatively, the mutual relationship between classes. In contrast, the canonical variate analysis yielded a clear differentiation, without mutual overlapping, of the soil samples into classes according to land use ( Figure 5).
The Mahalanobis distances, used for quantification of the discrimination, exceeded the high values of 20-25 (Table 1). These high values related typically to one of two canonical variables, whereas they were much lower in relation to the other one (Table 1). Only the discrimination between arable and grassland soils was similarly strong in relation to both canonical variates ( Table 1). The scores of the samples from buried humus horizons (where the land use of soil before the burial was unknown) were, in general, in the same range as those of surface soils but not always within the space between the land-use classes ( Figure 5). However, the depth-related trends may be distinguished from an arrangement of the particular samples. In kurgans K2 and K3, the signal from the forest became stronger towards the buried soil surface. In kurgan K3, samples collected at a lower depth (115-140 cm) were placed closer to the grassland class, whereas in kurgan K2 the signal indicated arable lands. In both cases, the samples from a depth of 60-80 cm were located closer to the forest class. The score from the topsoil sample of buried soil in kurgan K1  The Mahalanobis distances, used for quantification of the discrimination, exceeded the high values of 20-25 (Table 1). These high values related typically to one of two canonical variables, whereas they were much lower in relation to the other one (Table 1). Only the discrimination between arable and grassland soils was similarly strong in relation to both canonical variates ( Table 1). The scores of the samples from buried humus horizons (where the land use of soil before the burial was unknown) were, in general, in the same range as those of surface soils but not always within the space between the landuse classes ( Figure 5). However, the depth-related trends may be distinguished from an arrangement of the particular samples. In kurgans K2 and K3, the signal from the forest became stronger towards the buried soil surface. In kurgan K3, samples collected at a lower depth (115-140 cm) were placed closer to the grassland class, whereas in kurgan K2 the signal indicated arable lands. In both cases, the samples from a depth of 60-80 cm were located closer to the forest class. The score from the topsoil sample of buried soil in kurgan  The Mahalanobis distances, used for quantification of the discrimination, exceeded the high values of 20-25 (Table 1). These high values related typically to one of two canonical variables, whereas they were much lower in relation to the other one (Table 1). Only the discrimination between arable and grassland soils was similarly strong in relation to both canonical variates ( Table 1). The scores of the samples from buried humus horizons (where the land use of soil before the burial was unknown) were, in general, in the same range as those of surface soils but not always within the space between the landuse classes ( Figure 5). However, the depth-related trends may be distinguished from an arrangement of the particular samples. In kurgans K2 and K3, the signal from the forest became stronger towards the buried soil surface. In kurgan K3, samples collected at a lower depth (115-140 cm) were placed closer to the grassland class, whereas in kurgan K2 the signal indicated arable lands. In both cases, the samples from a depth of 60-80 cm were located closer to the forest class. The score from the topsoil sample of buried soil in kurgan The canonical variate analysis applied to discriminate the samples among the classes related to soil type (reference soil group, i.e., Chernozems, Phaeozems and Luvisols) yielded a differentiation without mutual class overlapping but less prominent as compared to land-use classes ( Figure 6). The Mahalanobis distances reached values of 13-19 between Chernozems/Phaeozems and Luvisols (according to the 1st variate), whereas the distances according to the 2nd variate were much lower ( Table 1). The scores of the samples from buried humus horizons in the K3 kurgan indicated a poorly marked depth-related trend (i.e., samples from a lower depth are located closer to the Chernozem class, whereas samples from a shallower depth are located closer to the Phaeozem or Luvisols group). The sample from the deeper sub-layer of the buried humus horizon in kurgan K1 was close to Phaeozems, and the sample from an overlying sublayer was placed between Phaeozems and Luvisols. In general, the samples from soils buried under kurgans rarely had scores indicating their adjacency to Chernozems ( Figure 6). Furthermore, the canonical variate analysis was applied to discriminate the samples of forest soils in relation to dominant forest vegetation. Even if the broadleaf forests under study had mixed species composition, the three classes were clearly differentiated (i.e., beech-, oak-and hornbeam-/maple-/lime-dominated stands (Figure 7). The 1st variate strongly discriminated between oakand hornbeam-dominated forests (with the Mahalanobis distance reaching a value of 24, Table 1), whereas the 2nd variate clearly discriminated samples from beech-dominated forests from the other broadleaf stands (Mahalanobis distances of [12][13][14][15]. Although the majority of buried soils under study did not indicate their adjacency to the forest class, a clear discrimination into classes related to species composition was not possible. However, the connection of buried soil's organic matter with forest vegetation is not excluded, and further investigation on this field may require a larger set of data from burial profiles.

Discussion
The simple relationship of an origination of Chernozems and the presence of steppe vegetation and continental climate in Central Europe was recently questioned due to paleobotanic studies, which demonstrated the predominance of forest vegetation and strong human impacts during the Holocene. However, the other findings suggested that the early removal of forests and the expansion of pastures and arable lands since the Neolithic period could preserve or even restore some of these soils [28,29,36,37]. The determination of organic matter was successfully conducted by [19] in some buried Central European Chernozems using NIR spectroscopy. In this reconstruction, the bottom parts of the humus horizons predominantly had a signature of grasslands, whereas the spectra from topsoil horizons indicated stronger influences of broadleaf forests. Contrary to [19], we believe that such paleopedological reconstructions should involve the humus horizons of arable soils, in addition to grassland and forests soils, because we cannot exclude that the

Discussion
The simple relationship of an origination of Chernozems and the presence of steppe vegetation and continental climate in Central Europe was recently questioned due to paleobotanic studies, which demonstrated the predominance of forest vegetation and strong human impacts during the Holocene. However, the other findings suggested that the early removal of forests and the expansion of pastures and arable lands since the Neolithic period could preserve or even restore some of these soils [28,29,36,37]. The determination of organic matter was successfully conducted by [19] in some buried Central European Chernozems using NIR spectroscopy. In this reconstruction, the bottom parts of the humus horizons predominantly had a signature of grasslands, whereas the spectra from topsoil horizons indicated stronger influences of broadleaf forests. Contrary to [19], we believe that such paleopedological reconstructions should involve the humus horizons of arable soils, in addition to grassland and forests soils, because we cannot exclude that the fertile Chernozems were cultivated before their burial [37,38]. The spectroscopic signal from the humus horizons from arable, forest and grassland soils (all developed from loess and originally located in chernozemic zones) differed clearly in this study ( Figures 5 and 6), which allowed displaying the samples of unknown history in relation to three (instead of two) classes of vegetation, as previously tested by Ertlen [18]. The spectroscopic signals in the samples collected from sublayers of three soils buried under Neolithic kurgans in SW Poland suggested various scenarios (from the bottom towards topsoil layers of the buried soil). The change from grassland to forest vegetation (kurgan K3) is consistent with the findings of [19]. In contrast, the change from arable to mixed grassland/forest vegetation (kurgans K1 and K2) indicates that the period of early cultivation was replaced further with pastures or open forests, which also may indicate human occupation and intentional land management, but less intense [24].
The MIR spectra analyzed in relation to the recognized soil type (reference soil group in terms of WRB classification) yielded three well-discriminated classes (Chernozems, Phaeozems and Luvisols), which suggested the possibility of a direct reconstruction of the transformation trends in the buried soils ( Figure 5). The canonical scores of some soil samples did not fit the "space" created by the abovementioned three classes. This result suggests possible influences from the other physicochemical soil components important for pedogenic processes, such as carbonates and clay minerals [39]. However, the placement of some samples indicated the soil transformation (analyzed as the change from the bottom towards topsoil layers of the buried soil) from Chernozem through Phaeozem towards Luvisol (kurgan K3) and from Phaeozem towards Luvisol (kurgans K1 and K2), which is consistent with opinions regarding the transformations of chernozemic soils during the Holocene period [19,29,40,41]. Clear discrimination of the topsoil horizons of modern soils into non-overlapping classes referring to the land use and soil type has confirmed the potential of non-destructive spectroscopic methods in paleopedological studies. Unfortunately, the samples tested in the present research were obtained from three already available buried profiles; therefore, the presented findings are single case studies, and a larger number of buried soils must be investigated to draw more general conclusions.
In the present research, we tested the relevance of MIR spectroscopy in the studies on the origin/affinity of organic matter in humus horizons of chernozemic soils as well as the applicability of MIR spectroscopy for studies on the composition of transformation or organic substances [42][43][44]. However, the spectral range 4000-525 cm -1 was considered by some researchers to be too resourceful for efficient work due to the much stronger peaks in the entire spectrum compared to VIS and NIR spectroscopy [45][46][47]. Previous attempts of paleoenvironmental reconstructions were conducted using NIR spectroscopy [18,19]. However, the Mahalanobis distances between the identified classes, obtained in the present research, were relatively high and even exceeded the value of 20 (Table 1), which guaranteed the clear disjunction of classes ( Figures 5 and 6). Thus, despite the large variety of information in the MIR spectrum, MIR may differentiate samples in relation to analyzed variables after respective mathematical pretreatment and using multivariate discrimination analysis. It was, therefore, concluded that MIR spectroscopy may have similar applicability in the paleopedological and paleoecological reconstructions as NIR spectroscopy. As shown in Figure 7, MIR spectra discriminate clearly between topsoil samples in relation to dominant forest vegetation, which confirms the applicability of this technique for studies of forest soil transformation, as previously confirmed for NIR spectroscopy [48][49][50][51].

Conclusions
Our study showed that MIR application in the process of reconstruction of Chernozems' history may have similar applicability to previous studies using NIR spectroscopy. This conclusion results from the high Mahalanobis values and clear disjunction of vegetation classes presented in this study. However, we also stated that further paleopedological studies on Chernozems' history by using infrared spectroscopy should involve additional groups of arable soils samples, in addition to grassland and forests groups. Results acquired by the addition of a third set of data to the procedure support the theory that one preservation mechanism of Chernozems in the Holocene could be the introduction of early cultivation, pastures and open forests by human occupation. Additionally, the combination of canonical data of MIR spectra in relation to the soil reference groups and dominant forest vegetation showed the potential applicability of this technique in studies of soil transformation; for example, under forest ecosystems. However, it should be noted that minimizing background interference and using a large set of samples is crucial to draw conclusions with increased certainty.