Quick Analysis of Organic Amendments via Portable X-ray Fluorescence Spectrometry

International Journal of Environmental Research and Public Health Article Quick Analysis of Organic Amendments via Portable X-ray Fluorescence Spectrometry Rafael López-Núñez 1,* , Fátima Ajmal-Poley 1, JoséA. González-Pérez 1, Miguel Angel Bello-López 2and Pilar Burgos-Doménech 1 1Instituto de Recursos Naturales y Agrobiología de Sevilla (IRNAS-CSIC), Avda. Reina Mercedes 10, 41012 Sevilla, Spain; [email protected] (F.A.-P.); [email protected] (J.A.G.-P.); [email protected] (P.B.-D.) 2 Department of Analytical Chemistry, Faculty of Chemistry, Universidad de Sevilla, c/Prof. Garc í a Gonz á lez, s/n, 41012 Sevilla, Spain; [email protected] *Correspondence: [email protected] Received: 17 September 2019; Accepted: 30 October 2019; Published: 6 November 2019   Abstract: The determination of heavy metals in soils and organic amendments, such as compost, manure, biofertilizer, and sludge, generally involves the digestion of samples with aqua regia, and the determination of those in the solution using various techniques. Portable X-ray fluorescence (PXRF) has many advantages in relation to traditional analytical techniques. However, PXRF determines the total elemental content and, until now, its use for the analysis of organic amendments has been limited. The objective of this work is the calibration of a PXRF instrument to determine the aqua regia-soluble elemental contents directly in solid samples of organic amendments. Our proposal will avoid the digestion step and the use of other laboratory techniques. Using a training set of samples, calibration functions were obtained that allow the determination of the aqua regia-soluble contents from the PXRF readings of total contents. The calibration functions (obtained by multiple linear regression) allowed the quantitative determination of the aqua regia-soluble contents of Fe, K, P, S, Zn, Cu, Pb, Sr, Cr, and Mn, as well as the organic matter content and a semi-quantitative assessment of Al, Ca, V, Ba, Ni, and As contents. The readings of Si, Fe, Al, Ca, K, or S were used as correction factors, indicating that the calibrations functions found are truly based on the chemical composition of the sample matrix. This study will allow a fast, cheap, and reliable field analysis of organic amendments and of other biomass-based materials. Keywords: compost; sewage sludge; aqua regia-soluble content; heavy metal; urban garden; loss on ignition; organic matter determination; multiple linear regression 1. Introduction Urban agriculture is a growing activity in many cities on a worldwide scale [ 1 , 2 ]. Urban agriculture is practised in both developed and developing economies, although it can serve different purposes in different places, such as recreation or food security [ 3 ]. Under both approaches, there is a revitalised interest in research concerning urban agriculture from a variety of perspectives, including landscaping and urban planning [ 4 ], social and economic aspects [ 5 ], health and the environment [ 6 ], as well as technical subjects [7]. There is debate over the advantages and disadvantages of urban agriculture. There are many well-documented benefits of urban agriculture: subsistence survival, food sovereignty and security, reduced food transportation distance, carbon sequestration, a potentially reduced urban heat island effect, improved physical and mental health, improved aesthetics, community building, employment Int. J. Environ. Res. Public Health 2019,16, 4317; doi:10.3390/ijerph16224317 www.mdpi.com/journal/ijerph Int. J. Environ. Res. Public Health 2019,16, 4317 2 of 18 opportunities, improved local land prices, shortened supply chains, and, thus, reduced price differentials between producers and consumers, provision of habitats for wildlife, and waste recycling [8]. A major concern of urban agriculture lies in risks related to the eventual pollution of urban products. Fruits and vegetables grown in cities could be polluted by undesirable organic or inorganic substances [ 2 ], and there are several studies that have focused on trace elements found in urban fruits and vegetables. Metals can reach vegetables through different pathways, one of them being the use of polluted fertilizers, composts, sewage sludge, and livestock manure [ 9 ]. The reviews of Alloway [ 9 ] and Wei and Yang [10] reported on anthropogenic sources of metal contamination. The determination of heavy metals in soils and organic amendments generally involves the digestion of samples with aqua regia, and the determination of those in the solution using various analytical techniques such as ICP-OES, atomic absorption, and others. Urban gardeners generally lack the financial and technical requirements necessary to test metal contamination in their soils and amendments using these analytical techniques. Recently, improvements in X-ray tube and detector technologies have resulted in field-portable instruments, such as portable X-ray fluorescence (PXRF), that enable such uses as geologic investigations [ 11 ], as well as having environmental, pedological, and agronomic applications [ 12 ]. PXRF has many advantages in relation to traditional analytical techniques and is less expensive. PXRF analysis is non-destructive and very fast, with a simple or no sample treatment. The instrument could be used in the laboratory or in the field. The technique has a wide range of quantifications, having detection limits poorer than those of large laboratory-based XRF instruments or of ICP techniques, but it is sufficient for environmental monitoring of a large number of elements that can be determined simultaneously [ 12 , 13 ]. Despite lower detection limits of the chemical methods, sometimes these become more imprecise and inaccurate than XRF spectrometry, mainly due to the fact that the latter requires virtually no sample pretreatment [ 14 ]. There are approved protocols for using this instrument in soils and sediments [ 15 , 16 ]. Recently, Caporale et al. [ 17 ] evaluated the possibility of assessing the aqua regia-extractable content of metals from a multi-element profile of the soil using a PXRF analyser. These authors indicated that the discrepancy between the ‘pseudototal’ metal contents measured in soil samples by aqua regia and the ‘total’ contents measured by PXRF strongly depended on the properties of the element to be analysed, the chemical and mineralogical composition of the soil matrix, the organic matter content as well as of the type of acid used in the total extraction. Sharma et al. [ 18 ] applied multiple linear regression to a modeling dataset to establish the relationship between lab-determined soil Cation Exchange Capacity and eight variables from PXRF elemental data. They further improved the model using auxiliary input data (soil clay, pH, organic matter) as modeling variables; when using only clay, organic matter, and two elements (Cr, Ni) to make the prediction they obtained better regression coefficients. Despite the possibilities of this technique, until now its use for the analysis of compost and organic amendments has been limited to a few investigations [ 19 – 24 ], and the possibility for quantitative determination has been limited to a few elements. McWhirt et al. [ 22 ] and Havukainen et al. [ 23 ] found only eight elements that could be quantitatively determined in dry samples, and only qualitative results were obtained for a few more. They indicated that both the physical and the chemical properties of the samples affected measurement accuracy; significant influences due to moisture content, particle size, sample heterogeneity, positioning of samples, matrix effects (e.g., inter-element effects), calibration and reference standards, sample pre-treatment (e.g., drying, sieving, and homogenization), and the organic content of the sample were found. Nevertheless, these studies concluded that there is a great potential for widespread evaluation and definitive quantification of elemental concentrations, and that the lower cost, faster speed, and the portability of PXRF may outweigh the slight loss of accuracy, especially when the field evaluation of composted materials is a primary objective. We hypothesized that elemental data obtained from PXRF analyser can serve as a proxy for aqua regia-soluble contents of organic amendments. We also hypothesized that aqua regia-soluble contents Int. J. Environ. Res. Public Health 2019,16, 4317 3 of 18 could be reliably determined using multiple linear regression including correction terms based on the contents of major elements, such as Si and Fe. The objective of this work is the calibration of a PXRF instrument to determine the aqua regia-soluble elemental contents, directly from solid samples of organic amendments, avoiding previous digestion processes and the use of other laboratory techniques. Using a training set of samples, calibration functions based on multiple linear regression will be obtained that could allow the determination of the aqua regia-soluble contents from the PXRF readings of the total contents. To the authors’ knowledge, such an approach has not been used to date. Not even in the case of soils, in which the relationship between PXRF and aqua regia-extractable contents has been more studied, has this correction been proposed, despite Si being a frequent element in soils. 2. Materials and Methods 2.1. Field Portable X-ray Fluorescence (PXRF) Analysis The analyzer Niton XL3t 950s GOLDD +XRF (Thermo Scientific Inc., Billerica, MA, USA), with its laboratory stand, was used in this study. The analyzer was equipped with an X-ray tube with an Ag anode, operated at 50 keV, 200 µ A, and 2 W, with a geometrically optimized large Si drift detector (GOLDD). The analyzer can measure a range of elements, from Mg to U. Two factory calibration modes, known as the Mining (M) mode and the Soils (S) mode, were used for the measurements. In the M mode the following elements could be detected: Ba, Sn, Cd, Pd, Ag, Mo, Nb, Zr, Sr, Bi, As, Se, Hg, W, Pb, Zn, Cu, Ni, Co, Fe, Mn, Cr, V, Ti, Ca, Cl, K, S, P, Si, and Al. The S mode is recommended for soil metallic elements at low concentrations, and the following elements could be detected: Ba, Sb, Sn, Cd, Ag, Mo, Zr, Sr, U, Rb, Th, Pb, Se, As, Tl, Hg, Zn, Cu, Ni, Co, Fe, Mn, Cr, V, Ti, Ca, K, Cl, S, and P. Dried and finely ground samples were measured by filling an XRF container (model SC-4331, 26 mm internal diameter, 24 mm height, Premier Lab Supply Inc., Port St. Lucie, FL, USA) capped with a 4 µ m propylene film (model 240–255, 63 mm diameter, Premier Lab Supply Inc., Port St. Lucie, FL, USA). The samples were scanned in triplicate in each of the M and S modes and the average elemental content for each mode used hereafter. The analysis time for each scan was 90 s for the S mode and 120 s for the M mode. 2.2. Samples The samples of organic amendments corresponded to the MARSEP program, which is part of the Wageningen Evaluating Programmes for Analytical Laboratories (WEPAL) [ 25 ]. Samples corresponding to the quarterly periods one and three of 2015 to 2018 (32 samples, 4 per period) were included. The materials analyzed were composts (14 samples), sludge (12 samples), organic fertilizers (4 samples), and manures (2 samples) from four different European countries, and they were received as a dried and finely crushed powder. The dry test samples have been shown to be stable over a number of years when stored at room temperature. In the MARSEP programme, elements were determined as the acid-extractable form, X (ae), being aqua regia the most common extractant [ 26 , 27 ], although other extractants such as HNO 3 or HNO 3 /H 2 O 2 were used by a few participating laboratories. The elements considered were As, Ba, Ca, Cd, Co, Cr, Cu, Fe, Hg, K, Mg, Mn, Mo, Na, Ni, P, Pb, S, Sb, Se, Si, Sn, Sr, Ti, V, and Zn. Loss on ignition (LOI, i.e., organic matter) was also included. It was assumed that the MARSEP samples can be considered as Certified Reference Materials and that the average value for each determined element [ 28 ] obtained from the set of individual values of the participating laboratories was the “true value” for the element. The number of individual data used to calculate the average varied to a maximum of approximately 30, and depended on the element. Average values using less than n=5 were not considered in this study. It is doubtful which of the two methods provided by the PXRF instrument, Soil or Mining, offers better results in relation to the determination of the majority elements (Fe, Al, Ca, K, P, S), the concentrations of which ranged in the order of percentage. There are very few cases of use of the same model of Int. J. Environ. Res. Public Health 2019,16, 4317 4 of 18 instrument for the analysis of waste samples in the literature. Havukainen et al. [ 23 ] used the Mining method, but their findings indicate that there is not an acceptable level of agreement between PXRF and aqua regia-ICPMS for the examined element measurements, since the differences are too large. Two additional reference materials were used to assess the performance of the PXRF instrument for the determination of the real total content of major elements (Fe, Al, Ca, K, P, S, and Si). These samples were a sediment sample from the ISE (International Soil-Analytical Exchange) Programme (sample no. 859, period 2017.2.3, collected at De Bilt, Netherlands) [ 29 ], and sample SdAR-M2, that corresponded to a metal-rich sediment produced by the US Geological Survey [ 30 ]. For the verification check to be acceptable, it was considered that the measured value for each target analyte in these samples should be within ± 20% RD of the true value [ 15 ]. The mentioned major elements could present a greater difference than trace elements between the real-total contents and the extractable contents in aqua regia, since they can be associated with silicate that is difficult to dissolve. 2.3. Statistical Data Analysis Data were explored and transformed in the following way: For each element, the relationships among the reported average values corresponding to MARSEP program (X (ae)) and the average reading from the SOIL (X (S)) or MINING (X (M)) methods of PXRF were assessed by plotting the data. Multiple linear regression analysis was used to predict X (ae) as the dependent variable from X (S) or X (M) content. This first adjustment consisted of a single linear regression equation. After this first adjustment, the residuals were calculated and their correlations with other elemental contents were determined. The parameters with higher and significant Pearson coefficient (usually Si (M) but also other major and trace elements) were then added one by one into the equation, and the process was repeated to look for additional significant independent variables. X (ae) data and the predicted values obtained from the regression equations, X (predicted), were compared using a paired samples t-test and 99% confidence probability. All statistical analyses were carried out with IBM © SPPS © Statistics version 25 (IBM Corp., Chicago, IL, USA). The correlation coefficient (r) for both series of results should be 0.7 or greater for the PXRF data to be considered screening level data. If ris greater than 0.9 and inferential statistics (the slope of the line =1 and the y-intercept =0) indicate the two data sets are statistically similar at the 95% confidence level, the data quality is definitive [ 15 ]. The empirical calibration (single or multiple linear regression equation) is considered quantitatively acceptable for a specific analyte if the correlation coefficient is greater than 0.98 [15]. The relative standard deviation, RSD, is the ratio as a percentage of the standard deviation, SD, and the mean concentration of an element. The relative difference, RD, is the difference, as percentage, between the predicted and actual values. When the average RD was calculated, the absolute values were used. 3. Results and Discussion A summary of the results obtained for the MARSEP samples using both methods, Mining and Soil, is shown in Table 1. The range of values for several elements extended over one order of magnitude, which indicated a high variability among the different materials analyzed. These extended ranges also allowed a sufficient set of data for calibration extending between the concentration ranges that are usually given in organic amendments. For most of the elements, the number of valid results in the MARSEP programme [31] was greater than 30. Int. J. Environ. Res. Public Health 2019,16, 4317 5 of 18 Table 1. Comparison of ranges of elemental concentrations (mg kg −1 unless otherwise stated) determined using portable X-ray spectrometry (PXRF). Samples Method Soil Method Mining N1Min Max Mean N1Min Max Mean LOI 232 24.1 90.0 52.0 Bal 332 70.6 92.5 77.7 As 20 2.6 18.5 9.8 14 4.7 10.0 7.2 Ba 23 45.4 406 177 24 79.5 489 244 Cr 21 30.4 441 94.0 25 92.6 326 170 Cu 31 12.0 426 150.6 27 29.2 768 294 Mn 28 106 631 418 27 105 680 312 Mo 23 3.2 22.7 5.2 10 2.9 16.2 5.0 Ni 18 23.0 78.6 36.8 <LOD <LOD <LOD Pb 25 14.3 236 69.8 24 12.9 240 66.3 Rb 32 8.7 59.9 26.8 32 4.3 38.0 17.2 Sr 32 44.2 292 170 32 44.6 296 172 Ti 32 68.5 4142 1664 32 97.7 2669 1200 V 16 31.0 88.8 56.0 24 42.5 157 82.7 Zn 32 55.5 1122 419 32 59.0 1364 502 Zr 32 4.46 128 59.3 32 3.8 120 51.7 Al 430 1.0 33.5 11.4 Ca 432 33.6 145 79.2 32 26.4 118 61.0 Cl 432 0.30 10.3 2.57 Fe 432 0.9 147 29.9 32 1.4 130 33.3 K432 3.74 159 28.5 32 2.56 94.9 15.9 Mg 412 3.10 7.19 4.33 P432 1.01 35.4 17.1 S432 2.22 77.6 14.6 32 3.64 92.8 18.0 Si 432 4.08 159 59.2 1 N, valid cases above limit of detection (LOD). 2 LOI: Loss on ignition values in %. Data from the MARSEP programme, not from PXRF measurements. 3 Bal: Difference to 100% of the sum of all measured elements. 4 Values in g kg−1. 3.1. Loss on Ignition (LOI) The LOI (i.e., organic matter content) is not directly determined by the PXRF instrument since the constituent elements of the organic matter (C, N, O, H) are beyond its capability. However, in the measurement method M, a result called “Bal” (balance) is given, corresponding to the difference of up to 100% that has not been assigned to other elements. Bal and LOI were correlated (Figure 1), but the adjustment was weak (r 2 =0.66 for all the samples), and as can be seen in Table 1, the values for Bal were much higher than those of LOI (average data from the MARSEP programme). These two parameters adjusted well (Figure 1, Equation (1)) if the Si (M) and Ca (M) contents were included in Linear Equation (1) (Table 2) between LOI and Bal. Equation (1) assumed that Bal can include organic matter (LOI) and elements such as oxygen associated with silica, clay and calcium carbonate, that are likely constituents of the compost and sewage sludge samples. The stoichiometric coefficients in Equation (1) would be − 0.114 for SiO 2 , − 0.171 for SiO 3 (silicates, clay), and − 0.120 for CaCO 3 . The cause of the highest absolute values for the actual coefficients in Equation (1) is likely to be because other O-binding compounds, such as phosphates, sulfates, and iron and aluminum oxides, are included in the composition of the samples. The correlation coefficient for the predicted-LOI and actual-values, r=0.982 >0.98 (r 2 =0.965), indicated the validity of the adjustment for quantification. Additionally, the slope and the y-intercept of the relationship LOI-LOI (pred) met the quality criteria (slope =1 and y-intercept =0) whether for Equation (1), (2) or (3). Int. J. Environ. Res. Public Health 2019,16, 4317 6 of 18 Int. J. Environ. Res. Public Health 2019, 16, x 6 of 18 Figure 1. Relationships of Bal readings (difference to 100% of the sum of all measured elements) using PXRF and the LOI (loss on ignition) content in training samples of organic amendments (A). Adjustment of LOI predicted values using Equation (1): [LOI(pred)] = 0.998[Bal] − 0.248[Si(M)] − 0.176[Ca(M)], Equation (2): [LOI(pred)] = 1.145[Bal] − 0.272[Si(M)] − 0.268[Ca(M)] − 0.435[P(M)], and Equation (3): [LOI(pred)] = 0.885[Bal] − 0.258[Si(M)] − 0.032[Fe (M)] (Table 2) (B) and relative deviations (RD) of the predicted values (C). (r 2 is the coefficient of determination). Table 2. Multiple linear equations for the prediction of LOI 1 from the Bal 2 readings obtained by the PXRF using the Mining (M) method. Equation 󰇟𝑿(𝒑𝒓𝒆𝒅)󰇠=𝑨 󰇟𝑾(𝑴)󰇠+𝑩 󰇟𝑿(𝑴)󰇠+𝑪 󰇟𝒀(𝑴)󰇠+𝑫 󰇟𝒁(𝑴)󰇠 p Values Paired t-test p(A) p(B) p(C) p(D) r Mean Sd 4 (1) 2,3 󰇟𝐿𝑂𝐼(𝑝𝑟𝑒𝑑)󰇠=0.998 󰇟𝐵𝑎𝑙󰇠− 0.248 󰇟𝑆𝑖(𝑀)󰇠 − 0.176󰇟𝐶𝑎(𝑀)󰇠 0.000 0.000 0.000 0.982 52.0 3.31 (2) 󰇟𝐿𝑂𝐼(𝑝𝑟𝑒𝑑)󰇠=1.145 󰇟𝐵𝑎𝑙󰇠− 0.272 󰇟𝑆𝑖(𝑀)󰇠−0.268 󰇟𝐶𝑎(𝑀)󰇠 − 0.435󰇟𝑃 ( 𝑀 ) 󰇠 0.000 0.000 0.000 0.000 0.993 49.8 2.53 (3) 󰇟𝐿𝑂𝐼(𝑝𝑟𝑒𝑑)󰇠=0.885 󰇟𝐵𝑎𝑙󰇠− 0.258 󰇟𝑆𝑖(𝑀)󰇠−0.032 󰇟𝐹𝑒(𝑀)󰇠 0.000 0.000 0.019 0.965 55.8 1.24 1 LOI: Loss on ignition, i.e., organic matter content. Values in %. 2 [Bal]: Difference to 100% of the sum of all measured elements. 3 Si (M), Ca(M) and P(M) results expressed in g kg −1 . 4 sd, standard deviation of the differences. If the samples were separated according to their type (compost or sewage sludge) an even better adjustment could be obtained for the group of compost samples using Equation (2), which included a significant P (M) term, reducing the relative error of the predictions below 15% for all the sets of samples and below 10% if the LOI value was greater than 40% (Figure 1). For the samples of sewage sludge, the obtained adjustment (Equation (3), Figure 1) included an Fe(M) term, but it did not improve on Equation (1), although since these samples had a LOI value >50% their relative errors were Figure 1. Relationships of Bal readings (difference to 100% of the sum of all measured elements) using PXRF and the LOI (loss on ignition) content in training samples of organic amendments ( A ). Adjustment of LOI predicted values using Equation (1): [LOI(pred)] =0.998[Bal] − 0.248[Si(M)] − 0.176[Ca(M)], Equation (2): [LOI(pred)] =1.145[Bal] − 0.272[Si(M)] − 0.268[Ca(M)] − 0.435[P(M)], and Equation (3): [LOI(pred)] =0.885[Bal] − 0.258[Si(M)] − 0.032[Fe (M)] (Table 2) ( B ) and relative deviations (RD) of the predicted values (C). (r2is the coefficient of determination). Table 2. Multiple linear equations for the prediction of LOI 1 from the Bal 2 readings obtained by the PXRF using the Mining (M) method. Equation [X(pred)]=A[W(M)]+B[X(M)]+C[Y(M)]+D[Z(M)] pValues Paired t-test p(A) p(B) p(C) p(D) rMean Sd 4 (1) 2,3 [LOI(pred)] =0.998 [Bal]−0.248 [Si(M)] −0.176[Ca(M)] 0.000 0.000 0.000 0.982 52.0 3.31 (2) [LOI(pred)] = 1.145 [Bal]−0.272 [Si(M)] −0.268[Ca(M)] −0.435[P(M)] 0.000 0.000 0.000 0.000 0.993 49.8 2.53 (3) [LOI(pred)] =0.885 [Bal]−0.258 [Si(M)] −0.032[Fe(M)] 0.000 0.000 0.019 0.965 55.8 1.24 1 LOI: Loss on ignition, i.e., organic matter content. Values in %. 2 [Bal]: Difference to 100% of the sum of all measured elements. 3 Si (M), Ca(M) and P(M) results expressed in g kg −1 . 4 sd, standard deviation of the differences. If the samples were separated according to their type (compost or sewage sludge) an even better adjustment could be obtained for the group of compost samples using Equation (2), which included a significant P (M) term, reducing the relative error of the predictions below 15% for all the sets of samples and below 10% if the LOI value was greater than 40% (Figure 1). For the samples of sewage sludge, the obtained adjustment (Equation (3), Figure 1) included an Fe(M) term, but it did not improve on Equation (1), although since these samples had a LOI value >50% their relative errors were kept Int. J. Environ. Res. Public Health 2019,16, 4317 7 of 18 below 4%. The Fe (M) term was due to the high content of this element in sewage sludge (see Fe in Figure 2). To the best of the authors’ knowledge, there is no previous study in which the organic matter content (LOI) of organic amendments could have been inferred directly from the PXRF data. It is worth mentioning here the works of Andrade et al. [ 32 ] and Silva et al. [ 33 ] in which they found that Ca, P, Si and Fe, in line with the results presented here, were also among the most important variables for modeling the organic matter content in Brazilian soils. However, the r 2 values obtained in these studies were generally <0.8, possibly due the low organic content of the Brazilian soils studied. Int. J. Environ. Res. Public Health 2019, 16, x 7 of 18 kept below 4%. The Fe (M) term was due to the high content of this element in sewage sludge (see Fe in Figure 2). To the best of the authors’ knowledge, there is no previous study in which the organic matter content (LOI) of organic amendments could have been inferred directly from the PXRF data. It is worth mentioning here the works of Andrade et al. [32] and Silva et al. [33] in which they found that Ca, P, Si and Fe, in line with the results presented here, were also among the most important variables for modeling the organic matter content in Brazilian soils. However, the r2 values obtained in these studies were generally <0.8, possibly due the low organic content of the Brazilian soils studied. Figure 2. Relationships of PXRF readings of the major elements Fe, Al, Ca, K, P, and S and their aqua regia-extractable contents in training samples of organic amendments (A), and the adjustments of the predictions (B) of their aqua regia-extractable contents using the following multiple linear equations (Table 4): [Fe(pred)] = −0.242 + 0.804[Fe(M)] − 0.010[Si(M)]; [Al(pred)] = 2.077 + 1.622[Al(M)] − 0.138 [Si(M)]; [Ca(pred)] = −15.049 + 1.039[Ca(M)] + 0.145[K(S)]; [K(pred)] = −1.711 + 0.402[K(S)] + 0.117[S(S)]; [P(pred)] = −0.926 + 0.959[P(M)] + 0.033[Fe (M)]; [S(pred)] = 11.452 + 0.641[S(S)] − 0.162[Bal] + 0.004[Cu(S)]. 3.2. Macroelements (Fe, Al, Ca, K, P, S) A comparison of the contents of the certified samples SdAR-M2 and ISE 859, as determined by the M and S methods of PXRF, is shown in the Table 3. The EPA Method 6200 [15] indicated that the measured value for each target analyte should be within ±20 percent (% RD) of the true value for the verification check to be acceptable. The contents of Fe (M) were satisfactory for both samples, while Fe (S) readings underestimated Fe content in the SdAR-M2 sample. The Al (M) readings underestimated its content in both samples. The Ca (M) contents were satisfactory for both samples. The Ca (S) readings were also satisfactory for both samples, although the RDs were greater than those corresponding to the M method. The K (M) content was satisfactory for the ISE 859 sample but slightly exceeded the 20% error limit for the SdAR-M2 sample. On the contrary, K (S) content was satisfactory for the SdAR-M2 sample but slightly exceeded the 20% error limit for the ISE 859 sample. Considering both methods, the average absolute RD was lower for the S method. The P (M) measurement was adequate for the ISE 859 sample, but the RD was quite high for the SdAR-M2 sample. Nevertheless, the P content of this sample (0.35 g·kg−1) was lower than the minimum P content in the organic amendments (1.01 g·kg−1, Table 1). The S (M) readings were imprecise. The Figure 2. Relationships of PXRF readings of the major elements Fe, Al, Ca, K, P, and S and their aqua regia-extractable contents in training samples of organic amendments ( A ), and the adjustments of the predictions ( B ) of their aqua regia-extractable contents using the following multiple linear equations (Table 4): [Fe(pred)] = − 0.242 +0.804[Fe(M)] − 0.010[Si(M)]; [Al(pred)] =2.077 +1.622[Al(M)] − 0.138 [Si(M)]; [Ca(pred)] = − 15.049 +1.039[Ca(M)] +0.145[K(S)]; [K(pred)] = − 1.711 +0.402[K(S)] + 0.117[S(S)]; [P(pred)] = − 0.926 +0.959[P(M)] +0.033[Fe (M)]; [S(pred)] =11.452 +0.641[S(S)] − 0.162[Bal] +0.004[Cu(S)]. 3.2. Macroelements (Fe, Al, Ca, K, P, S) A comparison of the contents of the certified samples SdAR-M2 and ISE 859, as determined by the M and S methods of PXRF, is shown in the Table 3. The EPA Method 6200 [ 15 ] indicated that the measured value for each target analyte should be within ± 20 percent (% RD) of the true value for the verification check to be acceptable. The contents of Fe (M) were satisfactory for both samples, while Fe (S) readings underestimated Fe content in the SdAR-M2 sample. The Al (M) readings underestimated its content in both samples. The Ca (M) contents were satisfactory for both samples. The Ca (S) readings were also satisfactory for both samples, although the RDs were greater than those corresponding to the M method. The K (M) content was satisfactory for the ISE 859 sample but slightly exceeded the 20% error limit for the SdAR-M2 sample. On the contrary, K (S) content was satisfactory for the SdAR-M2 sample but slightly exceeded the 20% error limit for the ISE 859 sample. Considering both methods, the average absolute RD was lower for the S method. The P (M) measurement was adequate for the Int. J. Environ. Res. Public Health 2019,16, 4317 8 of 18 ISE 859 sample, but the RD was quite high for the SdAR-M2 sample. Nevertheless, the P content of this sample (0.35 g · kg −1 ) was lower than the minimum P content in the organic amendments (1.01 g · kg −1 , Table 1). The S (M) readings were imprecise. The content of S as read by the S method was satisfactory for the ISE 859 sample and slightly exceeded the 20% limit for the SdAR-M2 sample, but in this sample, the S content was lower than the minimum in the amendments (Table 1). The content of Si (M) was satisfactory in SdAR-M2 and slightly exceeded the 20% error limit in the ISE 859 sample. The Mg (M) content in ISE 859 was underestimated due to this element requiring air to be purged from the measurement window. In any case, the Mg element was not included in the following comparisons since a small number of data from PXRF were obtained, offering inconsistent results, possibly because they were close to the detection limit of the instrument. Additionally, in the case of Cl, there was not reported data in the MARSEP reports. Table 3. Real total contents (g kg −1 ) of the sediment samples ISE 859 and SdAR-M2 as determined using the Mining and Soil methods of the portable X-ray spectrometer (PXRF). Cert 1Mining Soil Value Value sd 2RD(%) 3Value sd 2RD(%) 3 Fe ISE 859 42.0 46.4 1.9 10.6 37.7 0.1 −10.2 SdAR_M2 18.4 18.5 0.1 0.3 13.9 0.1 −24.3 Al ISE 859 55.3 37.5 0.1 −32.2 - - - SdAR_M2 66.0 46.8 0.5 −29.0 - - - Ca ISE 859 33.9 35.0 0.0 3.1 39.5 0.2 16.6 SdAR_M2 6.00 5.72 0.1 −4.7 5.45 0.0 -8.7 KISE 859 15.3 12.3 0.0 −19.2 18.5 0.0 21.5 SdAR_M2 41.5 32.7 0.3 −21.2 40.1 0.2 -8.7 PISE 859 4.08 4.09 0.03 0.3 - - - SdAR_M2 0.35 0.58 0.11 68.3 - - - SISE 859 11.7 17.9 0.1 53.4 11.5 0.5 −1.1 SdAR_M2 0.97 1.56 0.04 60.6 1.19 0.08 22.6 Si ISE 859 213.0 167.5 0.3 −21.3 - - - SdAR_M2 343.3 304.5 1.1 −11.3 - - - Mg ISE 859 8.42 5.42 0.45 −35.7 - - - SdAR_M2 2.95 - - - - - - 1Certified value. 2sd, standard deviation. 3RD, percentage of difference to the certified value. It was decided to use the readings Fe (M), Ca (M), P (M), Si (M), K (S), and S (S). The manufacturer’s recommendations indicate that the M method is especially recommended for contents of metallic elements above 1%. Linear relationships among the average MARSEP values (X (ae)) and the selected X (M) or X (S) readings can be observed in Figure 2for the elements Fe, Al, Ca, K, P, and S. In the case of Al, two different trends can be observed depending on the type of sample considered (compost or sewage sludge). The relationships between the average values of the MARSEP program (X (ae), the vertical axis in each plot) and the calculated values from the best equations found (X (pred), the right horizontal axis in each plot) for the majority elements are shown in Figure 2. The obtained chemical linear equation for each element is shown in Table 4, and the results of the t-test among actual and predicted aqua regia-extractable contents are shown in Table 5. Int. J. Environ. Res. Public Health 2019,16, 4317 9 of 18 Table 4. Multiple linear equations for the prediction of aqua regia extractable contents of majority metal (X (pred)) from the readings obtained by the PXRF. Equation 1[X(pred)]=A B [W(M,S)]+C[X(M,S)]+D[Y(M,S)]+E[Z(M,S)] pValues p(B) p(C) p(D) rMean 2N4 (4) [Fe(pred)] =−0.242 +0.804 [Fe(M)] −0.010 [Si(M)] 0.000 0.005 1.000 25.9 32 (5) [Al(pred)] =2.077 +1.622 [Al(M)] −0.138 [Si(M)] 0.000 0.000 0.974 12.0 30 (6) [Ca(pred)] =−15.049 +1.039 [Ca(M)] +0.145 [K(S)] 0.000 0.017 0.956 52.1 32 (7) [K(pred)] =−1.711 +0.402 [K(S)] +0.117 [S(S)] 0.000 0.000 0.000 0.995 9.53 31 (8) [P(pred)] =−0.926 +0.959 [P(M)] +0.033[Fe(M)] 0.000 0.001 0.995 16.6 32 (9) 3[S(pred)] =11.452 +0.641[S(S)] −0.162[Bal]+0.004[Cu(S)] 0.000 0.001 0.011 0.994 8.30 31 1 Readings from M or S method. 2 Values in g kg −1 except Cu(S) in mg kg −1 and Bal in %. 3 Bal: Difference to 100% of the sum of all measured elements. 4N, number of data points. Table 5. Paired samples t-test between actual and predicted aqua regia-extractable elemental contents using the multiple regression equations. Paired Differences 99% CID 3 Mean Sd 1sEM 2Lower Upper t Df 4Sig 5 Fe 0.000 0.77 0.14 −0.37 0.37 0.000 31 1.000 Al 0.000 2.17 0.40 −1.09 1.09 0.000 29 1.000 Ca 0.000 6.52 1.15 −3.16 3.16 0.000 31 1.000 K 0.000 1.34 0.24 −0.66 0.66 0.000 30 1.000 P 0.000 1.37 0.24 −0.67 0.67 0.000 31 1.000 S 0.000 1.13 0.20 −0.56 0.56 0.000 30 1.000 Zn 0.000 0.77 0.14 −0.37 0.37 0.000 31 1.000 Cu 0.000 7.53 1.30 −3.58 3.58 0.000 30 1.000 Pb 0.000 3.72 0.74 −2.08 2.08 0.000 24 1.000 Sr 0.000 7.74 1.37 −3.76 3.76 0.000 31 1.000 V 0.000 3.32 0.83 −2.45 2.45 0.000 15 1.000 Cr 0.000 6.82 1.49 −4.24 4.24 0.000 20 1.000 Mn 0.000 17.7 3.35 −9.27 9.27 0.000 27 1.000 Ba 0.000 39.0 8.13 −22.9 22.9 0.000 22 1.000 Ni 0.000 6.14 1.44 −4.20 4.20 0.000 17 1.000 As 0.000 1.08 0.24 −0.69 0.69 0.000 19 1.000 1 sd, standard deviation of the differences. 2 sEM, standard error of the mean of the differences. 3 CID, 99% Confidence Interval of the Difference. 4df, degrees of freedom. 5sig, two-tailed significance level. The linearity was excellent for Fe. However, the slope of the regression was less than 1 and the Fe (ae) results were generally 20% lower than the Fe (M) readings. Caporale et al [ 17 ] found the soil aqua regia-extractable content of Fe was about 10–20% lower than the measured content using a similar PXRF instrument. This indicated a fraction of Fe was highly insoluble, likely due to the crystalline structures of the clay or other silicate compounds [ 17 ], which are probably also present in the compost and sludge samples. In the case that the Si (M) content was used to adjust the linear regression, Equation (4) in Table 4was obtained, with the coefficient for the Si(M) term being statistically significant (p(C) =0.005). The correlation coefficient (r=0.9999) and the results of the t-test (Table 5) showed a very good fitting between the actual Fe (ae) and predicted-Fe values, indicating the validity of the adjustment for quantification. The plot of Al (M) against the corresponding MARSEP results (Al (ae)) showed two different patterns (Figure 2). Sewage sludge samples tended to have higher Al (ae) values than Mining PXRF readings, while compost and biofertilizer samples tended to give lower Al (ae) values than Mining PXRF readings. It is likely that in both amendment types, a part of their Al was associated with Si in clay structures, but the effect was different depending on the Si content of the sample (average Si (M) content was 73.1 g · kg −1 for the compost samples and 36.0 g kg −1 for the sewage sludge samples). If this factor was taken into account (Equation (5), Table 4), the coefficient obtained for the Si (M) effect was significant and only one regression line was obtained. The paired sample t-test (Table 5) showed Int. J. Environ. Res. Public Health 2019,16, 4317 16 of 18 Abbreviations ICPMS Inductively coupled plasma—Mass Spectrometry ICP-OES Inductively coupled plasma—Optical emission spectrometry ISE International Soil Exchange LOI Loss on ignition MARSEP Manure and Refuse Sample Exchange Programme PXRF Portable X-ray fluorescence RD Relative difference: difference, as percentage, between the predicted and actual values RSD Relative standard deviation References 1. Martellozzo, F.; Landry, J.S.; Plouffe, D.; Seufert, V.; Rowhani, P.; Ramankutty, N. Urban agriculture: A global analysis of the space constraint to meet urban vegetable demand. Environ. Res. Lett. 2014 ,9, 064025. [CrossRef] 2. Meharg, A.A. Perspective: City farming needs monitoring. Nature 2016,531, S60. [CrossRef] [PubMed] 3. FAO. HABITAT III. Revised Zero Draft of the New Urban Agenda. In Proceedings of the United Nations Conference on Housing and Sustainable Urban Development, Quito, Ecuador, 18 June 2016. 4. Pfeiffer, A.; Silva, E.; Colquhoun, J. Innovation in urban agricultural practices: Responding to diverse production environments. Renew. Agric. Food Syst. 2015,30, 79–91. [CrossRef] 5. Rogus, S.; Dimitri, C. Agriculture in urban and peri-urban areas in the United States: Highlights from the census of agriculture. Renew. Agric. Food Syst. 2015,30, 64–78. [CrossRef] 6. Pearson, L.J.; Pearson, L.; Pearson, C.J. Sustainable urban agriculture: Stocktake and opportunities. Int. J. Agric. Sustain. 2010,8, 7–19. [CrossRef] 7. Johnson, M.S.; Lathuilli è re, M.J.; Tooke, T.R.; Coops, N.C. Attenuation of urban agricultural production potential and crop water footprint due to shading from buildings and trees. Environ. Res. Lett. 2015 ,10, 064007. [CrossRef] 8. Mok, H.F.; Williamson, V.G.; Grove, J.R.; Burry, K.; Barker, S.F.; Hamilton, A.J. Strawberry fields forever? Urban agriculture in developed countries: A review. Agron. Sustain. Dev. 2014,34, 21–43. [CrossRef] 9. Alloway, B.J. Contamination of soils in domestic gardens and allotments: A brief overview. Land Contam. Reclam. 2004,12, 179–187. [CrossRef] 10. Wei, B.; Yang, L. A review of heavy metal contaminations in urban soils, urban road dusts and agricultural soils from China. Microchem. J. 2010,94, 99–107. [CrossRef] 11. Young, K.E.; Evans, C.A.; Hodges, K.V.; Bleacher, J.E.; Graff, T.G. A review of the handheld X-ray fluorescence spectrometer as a tool for field geologic investigations on Earth and in planetary surface exploration. Appl. Geochem. 2016,72, 77–87. [CrossRef] 12. Weindorf, D.C.; Bakr, N.; Zhu, Y. Advances in Portable X-ray Fluorescence (PXRF) for Environmental, Pedological, and Agronomic Applications. Adv. Agron. 2014,128, 1–45. 13. Shand, C.A.; Wendler, R. Portable X-ray fluorescence analysis of mineral and organic soils and the influence of organic matter. J. Geochem. Explor. 2014,143, 31–42. [CrossRef] 14. Li, F.; Lu, A.; Wang, J. Modeling of Chromium, Copper, Zinc, Arsenic and Lead Using Portable X-ray Fluorescence Spectrometer Based on Discrete Wavelet Transform. Intern. J. Environ. Res. Public Health 2017 , 14, 1163. [CrossRef] [PubMed] 15. EPA. Method 6200: Field Portable X-ray Fluorescence Spectrometry for the Determination of Elemental Concentrations in Soil and Sediment. Available online: https://www.epa.gov/sites/production/files/2015-12/ documents/6200.pdf (accessed on 15 February 2019). 16. ISO. Soil quality—Screening Soils for Selected Elements by Energy-Dispersive X-ray Fluorescence Spectrometry Using a Handheld or Portable Instrument. Available online: https://www.iso.org/standard/ 53490.html (accessed on 12 February 2019). 17. Caporale, A.G.; Adamo, P.; Capozzi, F.; Langella, G.; Terribile, F.; Vingiani, S. Monitoring metal pollution in soils using portable-XRF and conventional laboratory-based techniques: Evaluation of the performance and limitations according to metal properties and sources. Sci. Total Environ. 2018,643, 516–526. [CrossRef] Int. J. Environ. Res. Public Health 2019,16, 4317 17 of 18 18. Sharma, A.; Weindorf, D.C.; Wang, D.; Chakraborty, S. Characterizing soils via portable X-ray fluorescence spectrometer: 4. Cation exchange capacity (CEC). Geoderma 2015,239–240, 130–134. [CrossRef] 19. Manohara, B.; Belagali, S.L. Evaluation of energy dispersive scanning electron microscopy and X-ray fluorescence techniques for analysis of compost quality. Anal. Methods 2017,9, 253–258. [CrossRef] 20. Weindorf, D.C.; Sarkar, R.; Dia, M.; Haggard, B.; McWhirt, A.; Wooten, A.; Wang, H.; Chang, Q. Correlation of X-ray Fluorescence Spectrometry and Inductively Coupled Plasma Atomic Emission Spectroscopy for Elemental Determination in Composted Products. Compost Sci. Util. 2008,16, 79–82. [CrossRef] 21. Weindorf, D.C.; Chakraborty, S.; Li, B.; Deb, S.; Singh, A.; Kusi, N.Y. Compost salinity assessment via portable X-ray fluorescence (PXRF) spectrometry. Waste Manag. 2018,78, 158–163. [CrossRef] 22. McWhirt, A.; Weindorf, D.C.; Zhu, Y. Rapid analysis of elemental concentrations in compost via portable x-ray fluorescence spectrometry. Compost Sci. Util. 2012,20, 185–193. [CrossRef] 23. Havukainen, J.; Hiltunen, J.; Puro, L.; Horttanainen, M. Applicability of a field portable X-ray fluorescence for analyzing elemental concentration of waste samples. Waste Manag. 2019,83, 6–13. [CrossRef] 24. Healy, M.G.; Fenton, O.; Forrestal, P.J.; Danaher, M.; Brennan, R.B.; Morrison, L. Metal concentrations in lime stabilised, thermally dried and anaerobically digested sewage sludges. Waste Manag. 2016 ,48, 404–408. [CrossRef] [PubMed] 25. Houba, V.J.G.; Uittenbogaard, J.; Pellen, P. Wageningen Evaluating Programmes for Analytical Laboratories (WEPAL), organization and purpose. Commun. Soil Sci. Plant Anal. 1996,27, 421–431. [CrossRef] 26. European Committee for Standardization (CEN). Soil Improvers and Growing Media—Extraction of Aqua Regia Soluble Elements, EN 13650; British Standards Institution: London, UK, 2001. 27. European Committee for Standardization (CEN). Sludge, Treated Biowaste and Soil-Digestion of Aqua Regia Soluble Fractions of Elements, EN 16174; British Standards Institution: London, UK, 2012. 28. Cofino, W.P.; Van Stokkum, I.H.M.; Wells, D.E.; Ariese, F.; Wegener, J.W.M.; Peerboom, R.A.L. A new model for the inference of population characteristics from experimental data using uncertainties. Application to interlaboratory studies. Chemom. Intell. Lab. Syst. 2000,53, 37–55. [CrossRef] 29. International Soil Analytical Exchange (ISE). Quarterlyy Report 2017.2; International Soil Analytical Exchange: Wageningen, The Netherlands, 2017. 30. International Association of Geoanalysts. Reference Material Data Sheet SdAR-M2 Metal-Rich Sediment; International Association of Geoanalysts: Nottingham, UK, 2015. 31. Wageningen Evaluating Programmes for Analytical Laboratories (WEPAL). MARSEP. Available online: http://www.wepal.nl/website/products/MARSEP.htm (accessed on 6 May 2019). 32. Andrade, R.; Silva, S.H.G.; Weindorf, D.C.; Chakraborty, S.; Faria, W.M.; Mesquita, L.F.; Guilherme, L.R.G.; Curi, N. Assessing models for prediction of some soil chemical properties from portable X-ray fluorescence (pXRF) spectrometry data in Brazilian Coastal Plains. Geoderma 2020,357, 113957. [CrossRef] 33. Silva, S.H.G.; dos Santos Teixeira, A.F.; de Menezes, M.D.; Guilherme, L.R.G.; de Souza Moreira, F.M.; Curi, N. Multiple linear regression and random forest to predict and map soil properties using data from portable X-ray fluorescence spectrometer (pXRF). Ciênc. Agrotec. 2017,41, 648–664. [CrossRef] 34. MetcalfandEddy, I. Wastewater Engineering: Treatment, Disposal and Reuse, 3rd ed.; McGraw-Hill Ltd.: New Delhi, India, 1991. 35. Nawar, S.; Delbecque, N.; Declercq, Y.; De Smedt,P.; Finke, P.; Verdoodt,A.; Van Meirvenne, M.; Mouazen, A.M. Can spectral analyses improve measurement of key soil fertility parameters with X-ray fluorescence spectrometry? Geoderma 2019,350, 29–39. [CrossRef] 36. Kabata-Pendias, A.; Pendias, H. Trace Elements in Soils ad Plants, 4th ed.; CRC Press: Boca Raton, FL, USA, 2011; p. 432. 37. Thermo Scientific. Thermo Scientific. Thermo Scientific Niton XRF Analyzers, Certificate of Analysis. In Part Number 143-00131, Rev. D.; Thermo Scientific: Boston, MA, USA, 2016. Int. J. Environ. Res. Public Health 2019,16, 4317 18 of 18 38. European Union. Regulation (EU) 2019/1009 of the European Parliament and of the Council of 5 June 2019 Laying down Rules on the Making Available on the Market of EU Fertilising Products and Amending Regulations (EC) No 1069/2009 and (EC) No 1107/2009 and Repealing Regulation (EC) No 2003/2003. Available online: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32019R1009 (accessed on 17 July 2019). 39. T ó th, T.; Kov á cs, Z.A.; R é k á si, M. XRF-measured rubidium concentration is the best predictor variable for estimating the soil clay content and salinity of semi-humid soils in two catenas. Geoderma 2019 ,342, 106–108. [CrossRef] © 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).