Long-Term Manure Application in Urban Gardens: Impacts on Soil Fertility, Mineral Composition, and Variability

Abstract

Urban and peri-urban agriculture (UA) plays an increasingly important role in promoting sustainable urban development, providing socioeconomic, environmental, and educational benefits. However, UA is often linked to nutrient accumulation in soils since vegetable-growing areas typically receive substantial inputs of both organic and inorganic fertilizers. This study examines soil variability in two sections of an urban allotment garden subjected to long-term manure fertilization for 12 or 16 years, with application rates up to 10–12 kg m⁻² yr⁻¹. Surface soils were analyzed for organic and inorganic carbon, total-N, available-P and -K, pH, and elemental composition using portable X-ray fluorescence (pXRF). Prolonged manure incorporation substantially enhanced soil fertility, as evidenced by increases in soil organic carbon (up to 3.78%), total-N (up to 0.38%), available-K (up to 412 mg kg⁻¹), and both total- and available-P (up to 2485 and 276 mg kg⁻¹, respectively). Marked shifts in mineral composition were also detected, including significant increases in total Ca, inorganic C (as calcium carbonate), Sr, and S. Despite the high manure inputs, no accumulation of potentially toxic elements (PTEs) was observed. However, pronounced spatial heterogeneity emerged among individual plots, with coefficients of variation reaching 58% for S and 47% for Zn, reflecting differences in fertilization intensity and management practices. Portable X-ray fluorescence (pXRF) analysis proved highly effective for detecting soil compositional changes and adequate for predicting K and P availability, highlighting its value as a rapid diagnostic tool for precision agriculture. Overall, these findings demonstrate the agronomic benefits of long-term organic fertilization while emphasizing the need for careful management to avoid nutrient imbalances and ensure sustainable practices that minimize environmental risks.

1. Introduction

Urban and peri-urban agriculture (UA) plays an increasingly crucial role in urban development, not only by enhancing food availability and accessibility but also through its multidimensional contributions to sustainable urban growth, including socioeconomic, environmental, and educational benefits [1]. There are compelling reasons for promoting the expansion of UA, such as responding to rapid urbanization, concerns about mitigating the impacts of conventional agriculture on climate change, and tackling challenges related to the need to address food security and accessibility issues [2].

As a localized food production system, UA brings food production closer to consumers, providing access to affordable and nutrient-rich foods and helping to mitigate the adverse effects associated with prevailing long-distance food supply chains [2]. Urban agriculture (UA) is a rapidly expanding global practice—driven both by the pace of urban growth and by supportive policies in more developed countries [3]. The Food and Agriculture Organization of the United Nations, on its page dedicated to UA, indicates that 800 million people worldwide were involved in UA in 1996 and that it was practiced in 266 million households in developing countries [4]. Algert [5] indicates that in 2013, 42 million American households grew their own food, either at home or in a community garden. In Andalusia, the southern region of Spain, the regional government recorded in 2021 that there were 270 social gardens covering 102 hectares, with 14,571 family plots averaging 70 m² each, along with 2000 school gardens [6]. Municipal and regional governments often promote UA by providing essential resources such as land, water, irrigation systems, and fencing, while requiring only minimal financial contributions from participating gardeners. In less developed countries, UA is frequently associated with low-input agriculture and limited productive capacity. The multiple benefits of urban agriculture (UA) make it an effective pathway toward advancing several of the United Nations’ 2030 Agenda Sustainable Development Goals (SDGs) [7], particularly Goals 2 (Zero Hunger), 3 (Good Health and Well Being), 5 (Gender Equality), 8 (Decent Work and Economic Growth), 11 (Sustainable Cities and Communities), 12 (Responsible Consumption and Production), 13 (Climate Action), and 15 (Life on Land). Through improved access to fresh and nutritious food, the promotion of healthier lifestyles, the empowerment of women in food-related activities, the creation of local green jobs, the enhancement of urban resilience, the reduction in food waste, and the contribution to biodiversity and soil restoration, UA offers a multidimensional set of contributions that directly support the achievement of these global targets. Several recent review studies [8,9,10,11] show the interest that AU arouses in the scientific community at a global level. However, in developed or recently developed countries, UA is frequently linked to the accumulation of soil nutrients, as vegetable-growing soils receive large amounts of organic and inorganic fertilizers [12]. Elevated nutrient levels typically result from intensive fertilization practices, particularly the use of animal manures [12].

Organic farming (OF) practices are commonly adopted within UA systems [3]. In organic vegetable production, due to the high nitrogen and potassium demands of crops, fertilization management requires external sources of these nutrients such as manures, compost, or organic amendments—preferably derived from OF systems—which are applied as base fertilizers [13,14]. Although organic fertilizers can contribute to greenhouse gas (GHG) emissions, their overall effect is a negative carbon balance, as they enhance soil carbon sequestration and reduce fossil fuel use associated with synthetic fertilizer production. Overall, while organic fertilizers present challenges regarding GHG emissions, their multiple benefits justify careful consideration and strategic implementation in agricultural systems [15].

Manures not only recycle macro- and micronutrients back into the soil but also contribute substantial amounts of organic matter and microorganisms [16]. Soil organic matter exerts profound physical (e.g., structure, aeration, water retention), chemical (e.g., nutrient retention), and biological (e.g., biodiversity, disease suppression, biomass) effects, thereby maintaining soil health and enhancing crop productivity [16,17].

While the advantages of manure use are well known, its mismanagement can lead to serious problems. This has prompted the European Union and other countries to implement restrictive regulations on its agricultural use, setting limits on application rates, conditions, and methods, primarily to prevent nitrate contamination of water resources, as outlined in the Nitrates Directive [16].

Despite the long-standing tradition and extensive knowledge of manure use in agriculture, a recent European Union report [16] highlights emerging research priorities that extend beyond the well-documented nitrogen-related adverse effects. These priorities include additional impacts of manures and their derivatives on human health and the environment, such as soil fertility outcomes linked to manure management, the presence of microcontaminants (e.g., metals and pesticides), and the influence of manure processing on the biogeochemical cycle of phosphorus [16] (pp. 28–29).

Family or allotment gardens are a common UA typology in developed countries [3] and are expanding globally. In these systems, non-professional gardeners cultivate vegetables on small plots, following their own management practices, albeit often within a set of shared regulations. These gardens are non-commercial in nature, and although organic practices are frequently adopted, certification processes are typically avoided to minimize costs. Organic amendments are commonly incorporated with limited oversight, typically without nutrient content analyses of the applied materials applied or the receiving soil. The relevance of both organic and urban agriculture in developing countries is also highlighted in numerous current studies [18,19].

A global concern in urban gardens is the presence of contaminants such as potentially toxic trace elements (PTEs) [20,21], which may eventually reach crops and pose health risks to humans [22]. Soil contamination in these gardens can result from multiple sources, including nearby housing, road traffic, urban or industrial activities, as well as from the gardeners’ own practices, such as the use of fertilizers, waste materials, or pesticides [23,24].

A recent study indicates that organic residues and tillage practices often induce spatial heterogeneity in soil properties [25]. Understanding soil spatial variability is essential across disciplines such as ecological modelling, natural resource management, precision agriculture, and environmental prediction [26]. Despite its influence on many soil properties, few studies have examined this variability within urban gardens, and those that have mostly focused on differences between gardens within or across cities [27]. Some intensive sampling studies have focused on lead (Pb) contamination [28,29]. Portable X-ray fluorescence (pXRF) is particularly suitable for such detailed studies [27,30,31]. As Stevenson and Hartemink [31] noted, “many studies stated that urban soils are heterogeneous yet did not quantify small-scale spatial patterns or variation. The use of proximal soil sensors such as pXRF can aid in increasing sample density”; they further indicated that evaluating and summarizing spatial variation across different urban land uses remains a key area for future research.

The present study hypothesizes that individually managed plots within urban vegetable gardens may develop significant soil heterogeneity due to divergent management practices, particularly the uneven and long-term application of manure by gardeners. To test this hypothesis, the study pursued four specific objectives:

• Quantify elemental nutrients and PTEs in multiple plots managed under different practices, using both pXRF and conventional wet-chemistry analysis;
• Evaluate the analytical performance of pXRF for determining soil elemental composition of soils;
• Characterize spatial variations in elemental composition and soil properties and analyze the relationships among these variables;
• Identify key variables driving the soil variability and develop models capable of predicting plant available-P and -K.

The use of pXRF—along with visible–near-infrared and mid-infrared spectroscopy—to predict soil properties is increasing exponentially [32]. This study may contribute to a deeper understanding of small-scale horticultural production systems by applying an analytical technique that is rapid, low-cost, and highly suitable for use in these contexts.

2. Materials and Methods

2.1. Study Site and Soil Sampling

The study was conducted in the urban community garden of Utrera, in the province of Seville, southwest of Spain, located within the Quinto Centenario Park. The site is municipally owned and consists of individual plots allocated for long-term use by retired citizens. Each gardener independently manages their agricultural practices but is required to follow organic farming guidelines, which prohibit the use of most synthetic fertilizers and pesticides. However, the organic farming system is not formally certified.

All plots were equipped with a common drip irrigation system that provides uniform watering frequency. Horticultural practitioners exercise autonomy in the selection of crops. Typically, two cultivation cycles are undertaken annually, corresponding to autumn–winter and spring–summer rotation. Within each cycle, between five and ten species may be established, arranged in parallel rows, as illustrated in Figure 1. Leguminous species (e.g., Vicia faba, Pisum sativum) were frequently incorporated into the rotation, at least within certain sections of the plot. The practice of intercropping small, fast-growing species—such as Lactuca sativa—during the early developmental stages of larger-statured crops was likewise common. This resulted in a highly fragmented appearance of the garden (Figure 1a,b).

The park contains two adjacent garden areas. Zone 1 (centre at 37°11′41.3″ N, 5°45′58.9″ W) comprises 53 plots of approximately 85 m² each and has been cultivated since 2008. Zone 2 (centre at 37°11′37.1″ N, 5°46′03.5″ W) includes 34 plots of approximately 55 m² each and was established in 2012.

Geologically, the garden is located within the Upper Cenozoic Guadalquivir River Basin, composed of Quaternary fluvial terrace deposits [33] overlying Upper Miocene (Messinian, ~5 Ma) sediments [34]. From a previous study [35], it was known that the soil texture was loamy sand, consisting of 70.0 ± 10.7% coarse sand, 10.8 ± 1.7% fine sand, 7.0 ± 5.4% silt, and 12.2 ± 5.2% clay.

Samples from Zone 1 were collected during three sampling campaigns: in December 2023 (15 samples), four months after the last amendment; in June 2024 (13 samples), eight months after the amendment; and in September 2024 (11 samples), five months after the amendment. Control soil samples and samples from Zone 2 were collected in January 2025, five months after the last amendment. Composite topsoil samples (0–15 cm) were obtained by collecting 3 to 5 subsamples at intervals of 1–1.5 m along a 3–6 m transect corresponding to a single crop row. The subsamples were thoroughly mixed to produce a composite sample of approximately 1 kg. Sampling was carried out at the end of the cropping cycle, approximately 4 to 6 months after the last application of organic amendments. When plants were still present, subsamples were collected between plants, avoiding areas with high root density. Within the framework of a science outreach project [36], the authors were conducting trials applying biowaste compost, including separately collected household waste and waste from parks and gardens, to several plots in Zone 1 of the orchard. It was observed that garden plots varied considerably in their soil organic carbon (SOC) content, and that soil colour was related to their SOC content. Therefore, the entire orchard was examined, and sampling was extended to Zone 2, including plots that visually displayed a colour gradient. In Zone 1, a total of 39 samples were collected from 19 plots, with up to six samples taken from several plots to evaluate both intra- and inter-plot variability. In Zone 2, eight soil samples were collected from seven plots. The sampled plots are shown in Figure 2.

Additionally, three samples were collected from an adjacent area outside the cultivated plots (Figure 2). For these soil samples, four subsamples were taken from the vertices of a 1 m square. This area corresponded to a former Citrus aurantium (sour orange) orchard that predated both the gardens and the park. It has remained untilled for over 25 years, characterized by spontaneous weed vegetation and biannual mowing. These samples were identified as Control soil, as they had not been subjected to any agricultural management during this period.

Information regarding organic fertilization practices was obtained through personal interviews with gardeners. Sheep, horse, and cattle manure were identified as the primary organic fertilizers. Application rates ranged from 4 to 10 kg m⁻² year⁻¹, typically applied prior to the autumn–winter cropping season (late August–early September). Many gardeners performed a second, smaller application (1–2 kg m⁻² year⁻¹) before the spring–summer crops. In several plots associated with experiments from the same research team, compost derived from biowaste or commercial manure compost was incorporated during the season immediately preceding soil sampling at rates of approximately 2–4 kg m⁻². A few gardeners also applied small quantities, up to 0.5 kg m⁻², of vermicompost and concentrated commercial organo-mineral fertilizers. Several samples of the organic amendments applied in 2024 were collected and analyzed.

2.2. Soil and Organic Amendment Analyses

Soil samples were oven-dried at 40 °C and sieved through a 2 mm mesh. The oven-dried fraction was used for standard soil characterization, including pH, electrical conductivity (EC), inorganic carbon, soil organic carbon (SOC), total nitrogen (total-N), available phosphorus (available-P), and available potassium (available-K). Soil organic carbon (SOC) content was determined on a PRIMACS SNC 100 IC-E elemental analyser (Skalar Analytical B.V., Breda, The Netherlands): soil total carbon (STC) is converted to CO₂ after sample combustion and determined by non-dispersive infrared (NDIR) detection. Soil inorganic carbon (SIC) was determined on the same apparatus by introducing the sample into a reactor at 150 °C, where the SIC is automatically acidified with phosphoric acid to convert carbonates into CO₂, the extent of which is determined by NDIR detection. SOC was calculated as the difference between the STC and the SIC (SOC = STC − SIC). SIC is expressed as CaCO₃, the most abundant form in these semiarid Mediterranean soils, and reported in this form in the dataset. Total-N was determined in the same instrument. A 1:2.5 aqueous extract was used to measure pH and a 1:5 ratio for EC. Available-K was determined after extraction with ammonium acetate and available-P by extraction with sodium bicarbonate (Olsen method).

An aliquot of each sample was further oven-dried at 105 °C and analyzed for elemental concentrations using a portable X-ray fluorescence spectrometer (pXRF), (Niton XL3t 950s GOLDD+, Thermo Scientific Inc., Billerica, MA, USA) mounted on a shielded laboratory stand. Analyses followed the USEPA Method 6200 [37], with experimental details as described by [38]. Each soil sample was scanned twice using two precalibrated measurement modes, repositioning the sample between scans to cover different portions of the material. The mean value of both measurements was used after verifying their consistency.

The Soil mode is recommended for elements present at low concentrations, typically within the range of 10 to 1000 mg kg⁻¹. In this mode, the elements As, Ba, Cu, K, Mn, Ni, Pb, Rb, Sb, Sr, Ti, V, Zn, and Zr were quantified. Other elements—including Cr, V, Sc, U, Th, Au, Se, Co, Hg, W, Cs, Te, Sn, Cd, Ag, and Pd—can also be analyzed using this mode; however, their concentrations were below the limit of detection (LOD) of the technique (10–50 mg kg⁻¹, depending on the element). The Mining mode, recommended for concentrations on the order of percent levels (≈10,000 mg kg⁻¹), was used to measure Al, Mg, P, and Si, which are not included in the Soil mode, as well as Ca and Fe due to their higher concentrations.

The reference sediment material SdAR-M2, produced by the U.S. Geological Survey [39], was analyzed to verify instrument performance. It should be noted that XRF analysis provides total elemental concentrations, in contrast to the quasi- or pseudo-total values obtained through acid digestion methods (e.g., aqua regia or nitric acid).

The chemical characterization of organic amendments (pH, EC, total organic carbon, total-N, and C/N ratio) was conducted following standardized European procedures for soil improvers and growing media. Total C and N were determined by dry combustion according to ISO 10694:1995 [40] using a PRIMACS SNC 100 IC-E elemental analyzer (Skalar Analytical B.V., Breda, The Netherlands). Extractable elements in aqua regia were determined following EN 13650:2001 [41], and their concentrations were measured by inductively coupled plasma–optical emission spectrometry (ICP-OES; VARIAN 720-ES, Agilent Technologies, Santa Clara, CA, USA).

This method provides total elemental concentrations, in contrast to the quasi- or pseudo-total values obtained through acid digestion methods (e.g., aqua regia or nitric acid).

All results are expressed on a dry matter basis.

2.3. Pollution Indices

Soil contamination factors (CFᵢ) for Ni, Cu, Zn, As, and Pb were calculated as the ratio between the measured concentration (Cᵢ) of each metal i and its corresponding background concentration (CBᵢ) in regional soils [42]:

CFᵢ = Cᵢ/CBᵢ

(1)

Background values (CBᵢ) were obtained from geochemical reference data for trace elements reported by [43] for the Guadalquivir River Basin. The study area was undoubtedly included in this geological zone, as can be verified in the geological maps of the cartographic portal of the regional administration [44]. The background levels used were those reported as the median (50th percentile) of the data collected in the study [43]. The background values were 6 mg kg⁻¹ for As, 66 mg kg⁻¹ for Cr, 24 mg kg⁻¹ for Cu, 30 mg kg⁻¹ for Ni, 17 mg kg⁻¹ for Pb, and 56 mg kg⁻¹ for Zn. These values should be considered quite restrictive for detecting contamination.

The overall pollution status was assessed using the Pollution Load Index (PLI) [42] calculated as the nth root of the product of n contamination factors:

PLI = (CF1 × CF2 × … × CFn)^1/n

(2)

The maximum value of n was five when concentrations above the LOD were available for all five metals (Ni, Cu, Zn, As, and Pb). The index was also computed when only four or three metals exceeded the detection limit.

2.4. Statistical Analysis

Descriptive statistical analyses were conducted, including the calculation of mean, minimum, maximum, standard deviation, and coefficient of variation. Differences among means were tested using one-way analysis of variance (ANOVA) followed by Scheffé’s or Games–Howell’s post hoc tests depending on whether or not homoscedasticity is met. Data normality was verified using the Shapiro–Wilk test, and homogeneity of variances was evaluated with Levene’s test.

To evaluate the relationships among the parameters determined in soil samples from the orchard plots (excluding the Control soil), a principal component analysis (PCA) was performed, incorporating general soil parameters (pH, EC, CaCO₃ content, SOC content, total-N, available-P, and available-K contents) and elemental determinations obtained by pXRF, excluding elements with a high number of measurements below the LOD (Ni, As, Mg). A linear model was fitted to predict available-K and available-P from the remaining parameters. To achieve this, the dataset was divided into a training subset (70% of the data) and a validation subset (30%). The models were then created using a Forward Stepwise procedure and the Corrected Akaike Information Criterion (AICC) in the automatic linear modelling option of SPSS.

As criteria for the accuracy of the predictions, R², RMSE, RMSE%, and RPD have been calculated for both datasets. RMSE was calculated as:

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{N}}{\left|\mathrm{y}\mathrm{i}-\mathrm{ŷ}\mathrm{i}\right|}^2}{\mathrm{N}}}}$$

(3)

where N is the number of data points, yi is the ith measurement, and ŷi is its corresponding prediction.

The relative error RMSE% was calculated by dividing the RMSE of the prediction by the range of the laboratory measured soil property.

The RPD was calculated as the ratio between the standard deviation (sd) of the measured soil property and the RMSE of its prediction.

All statistical analyses were conducted using SPSS v.29.0.0.0.

3. Results

3.1. Evaluation of the Performance of the X-Ray Fluorescence Instrument

The results of the performance assessment of the pXRF instrument are shown in

Table 1. Performance of the instrument: Variations in daily measurements (N = 7) of the certified reference material sediment SdAR-M2.

	Certified Value mg kg−1	Average Value mg kg−1	Standard Deviation	CV %	Recovery %	Minimum Recovery %	Maximum Recovery %
Ag	15	19.1	2.3	12.0	127.2	105.9	148.2
As	76	79.7	9.3	11.7	104.9	86.1	117.1
Ba	990	871	13.7	1.6	88.0	86.1	90.2
Cr	49.6	61.1	12.2	19.9	123.2	84.6	149.6
Cu	236	219	13.3	6.1	92.9	86.3	99.7
Mn	1038	844	28.8	3.4	81.3	76.0	84.8
Ni	48.8	76.4	6.1	8.0	156.6	134.7	177.2
Pb	808	796	17.9	2.2	98.5	93.6	99.9
Rb	149	133	3.2	2.4	89.5	85.5	91.6
S	970	1068	98.9	9.3	110.1	94.0	121.1
Sb	107	103	3.2	3.1	95.9	91.8	100.1
Sr	144	142	3.1	2.2	98.5	95.1	101.4
Th	14.2	16.9	3.1	18.1	119.0	87.5	145.7
Ti	1798	1407	97.1	6.9	78.3	67.2	83.1
Zn	760	716	17.5	2.4	94.2	90.6	98.2
K × 10−3	41.5	37.6	2.5	6.8	90.7	78.4	96.4
1 Fe × 10−3	18.4	18.5	0.2	1.0	100.7	99.2	101.7
1 Ca × 10−3	6.00	5.88	0.14	2.4	98.0	94.1	101.5
1 Mg × 10−3	2.96	4.63	1.24	26.9	156.4	92.9	205.7
1 Al × 10−3	66.0	37.5	2.43	6.5	56.8	53.5	63.3
1 Si × 10−3	343	292	13.0	4.4	85.1	79.9	90.7
1 P	345	549	91	16.8	159.1	129.9	183.8

¹ For these elements, measurements were performed in the precalibrated Mining mode. For the rest of the elements, the measurements were made with the precalibrated Soil mode.

. Measurement precision for each element is indicated by the percentage recovery obtained, while the coefficient of variation reflects the reproducibility of the results. As shown, the percentage recovery (ratio of the mean value to the certified value) ranged from 90 to 110% for As, Cu, Pb, S, Sb, Sr, Zn, K, Fe, and Ca. Among these elements, the coefficient of variation was below 10% for Cu, Pb, S, Sb, Sr, Zn, K, Fe, and Ca.

Percentage recovery was between 100 ± 10 and 100 ± 20% for Ba, Mn, Rb, Th, and Si, with coefficients of variation of 1.6, 3.4, 2.4, 18.1, and 4.4%, respectively. Recovery values ranged from 100 ± 20 to 100 ± 30% for Ag, Cr, and Ti, with coefficients of variation of 12, 19.9, and 6.9%, respectively. Recovery exceeded ±30% for Ni, Mg, Al, and P, with CVs of 8.0, 26.9, 6.5, and 16.8%, respectively.

3.2. Element Concentration and Inter-Plot Variability

The element concentrations in soils from the urban garden plots are shown in

Table 2. Statistics of elemental concentrations in urban garden soil samples and elemental concentrations in the Control soil (the elements are arranged in decreasing order of CV).

	Control Soil mg kg−1	Average Value mg kg−1	Standard Deviation mg kg−1	CV 1 %	Minimum mg kg−1	Maximum mg kg−1	N 2
S	396	1337	769	57.5	598	5146	46
Zn	22.0	66.2	31.1	47.0	36.0	244.3	46
3 Ca × 10−3	14.3	62.7	23.6	37.6	20.1	123.4	46
Pb	22.3	20.8	6.9	33.4	10.8	51.8	46
P	1390	2485	689	27.7	1009	3961	46
Sr	20.7	60.0	16.3	27.2	29.0	90.2	46
Cu	19.7	33.2	8.4	25.4	18.6	51.0	44
Zr	164	140	34.1	24.3	76.9	233	46
As	BLD 4	7.85	1.74	22.2	5.48	11.64	25
Ba	215	202	43.6	21.5	104	291	46
3 Fe × 10−3	10.7	12.6	2.4	18.7	8.85	19.4	46
Sb	23.2	18.6	3.4	18.5	12.7	26.4	38
V	35.7	37.6	6.9	18.3	28.7	57.2	44
Mn	166	169	29.1	17.2	110	257	46
3 Al × 10−3	19.1	15.3	2.55	16.6	10.9	21.3	46
Ni	29.7	32.8	4.9	15.0	25.6	44.2	27
3 Mg × 10−3	bd	5.13	0.76	14.7	4.47	6.30	5
Rb	19.3	23.9	3.5	14.6	18.0	31.1	46
3 K × 10−3	7.01	7.48	1.09	14.5	5.42	10.9	46
Ti	2020	1607	227	14.2	1198	2172	46
3 Si × 10−3	245.5	208.7	24.6	11.8	157.0	274.4	46

¹ Coefficient of variation. ² Number of cases above limit of detection. ³ For these elements, measurements were performed in the precalibrated Mining mode, and for the rest of the elements, the measurements were made with the precalibrated Soil method; ⁴ BLD: below limit of detection.

. For each element, the mean value, standard deviation (sd), coefficient of variation (CV = sd × 100/average value, expressed as a percentage), and minimum and maximum values were shown.

In decreasing order of CV, the elements with the highest variability were S > Zn > Ca, all with CV > 35%. The mean concentrations of these elements in the garden soils were higher than those in the Control soil: 3.4-fold for S, 3.0-fold for Zn, and 4.4-fold for Ca. For these same elements, the differences between the minimum and maximum measured values were very large, approaching nearly one order of magnitude for S and Zn. In decreasing order of CV, the elements Pb > P > Sr > Cu > Zr > As > Ba > Fe > Sb > V > Mn > Al and Ni showed intermediate variability. Among these, the mean concentration in the plots was higher than in the Control soil for P (1.8×), Sr (2.9×), and Cu (1.7×), while for the remaining elements the mean concentrations in the plots and the Control soil were relatively similar. Variability was low for Rb, Ti, and Si, with mean Ti and Si values being lower in the plot soils than in the Control.

The CF values for Ni, Cu, Zn, As, and Pb and the PLI for both garden sectors are shown in Figure 3. Sector 1 has been cultivated since 2008, and Sector 2 since 2012. The CFCu, CFZn, and PLI values were higher in both garden sectors compared to the Control soil, although statistically significant differences were observed only for CFZn and PLI. For CFCu, only one replicate in the Control soil was above the LOD, and consequently, post hoc tests on the means could not be performed. Nevertheless, the difference between the garden soils and the Control soil is clearly evident in the graph. The CFAs value may also be considered significant, as its concentration in the Control soil could not be determined because it was below the LOD of the analytical technique. Mean values of CFCu, CFZn, and PLI in the Control soil were below 1. In contrast, the CFCu, CFAs, and PLI values for both garden sectors were above 1. Meanwhile, the CFZn and CFNi values in both sectors were close to 1. Finally, CFPb values were similar across all three areas, with values >1. According to Figure 3, considerable dispersion of the concentration values was observed, particularly for Cu, Zn, and As in Sector 1 and for Pb in Sector 2. Furthermore, Ni and As concentrations were frequently undetectable, with numerous soil samples falling below the detection limit. In such cases, CFNi and CFAs values could not be calculated, and therefore the mean CF values for these two elements should be regarded as overestimates.

The fertility soil properties for the garden sectors (Sectors 1 and 2) and the Control soil is shown in

Table 3. Fertility soil properties in urban garden and Control soil samples.

		Control Soil mg kg−1	Average Value mg kg−1	Standard Deviation mg kg−1	CV 1 %	Minimum mg kg−1	Maximum mg kg−1	N 2
			Sector 1 and 2
pH		7.36	7.41	0.40	5.50	6.33	8.21	41
EC 3	dS m−1	0.05	0.59	0.49	83.0	0.12	2.60	41
CaCO3	%	1.93	9.86	5.48	55.6	1.50	25.80	41
SOC 4	%	1.23	3.57	1.18	33.2	1.86	5.83	41
total-N	%	0.073	0.355	0.116	32.7	0.203	0.605	41
C/N		16.6	10.1	1.2	12.2	8.3	13.2	41
avail-P 5	mg kg−1	26	264	108	41.0	97	523	46
avail-K 6	mg kg−1	113	375	294	78.4	116	1755	41
			Sector 1
pH		7.36	7.39	0.44	5.9	6.33	8.21	33
EC	dS m−1	0.05	0.57	0.39	67.7	0.12	1.67	33
CaCO3	%	1.93	10.64	5.8	54.3	1.5	25.8	33
SOC	%	1.23	3.78	1.20	31.7	1.99	5.83	33
total-N	%	0.073	0.377	0.118	31.3	0.203	0.605	33
C/N		16.6	10.1	1.18	11.7	8.3	13.2	33
avail-P	mg kg−1	26	276	110	40.0	97	523	38
avail-K	mg kg−1	113	412	315	76.4	142	1755	33
			Sector 2
pH		7.36	7.49	0.23	3.1	7.22	7.84	8
EC	dS m−1	0.05	0.66	0.82	125	0.16	2.60	8
CaCO3	%	1.93	6.64	2.09	31.5	3.9	9.5	8
SOC	%	1.23	2.72	0.66	24.3	1.86	3.70	8
total-N	%	0.073	0.264	0.045	17.1	0.208	0.344	8
C/N		16.6	10.2	1.52	14.9	8.6	12.4	8
avail-P	mg kg−1	26	208	83	39.8	113	370	8
avail-K	mg kg−1	113	221	81	36.7	116	339	8

¹ Coefficient of variation. ² Number of samples analyzed. ³ Electrical conductivity. ⁴ Soil organic carbon. ⁵ Available phosphorus. ⁶ Available potassium.

, and their mean differences are shown in

Table 4. Differences in soil properties between different areas of the orchard.

		Control Soil	Zone 1	Zone 2
pH		7.36	7.39	7.49
EC 1	dS m−1	0.05 a	0.57 b	0.66 ab
CaCO3	%	1.93 a	10.64 b	6.64 a
SOC 2	%	1.23 a	3.78 b	2.72 c
total-N	%	0.073 a	0.377 b	0.264 c
C/N		16.6	10.1	10.2
avail-P 3	mg kg−1	26 a	276 b	208 b
avail-K 4	mg kg−1	113 a	412 b	221 c

¹ Electrical conductivity. ² Soil organic carbon. ³ Available phosphorus. ⁴ Available potassium. For each parameter, different letters indicate significant differences between sampling areas according to Games–Howell post hoc test, and values followed by the same letter or without a letter are not significantly different (p < 0.05).

. In general, the Control soil showed SOC and nutrient levels corresponding to low–medium soil fertility. Mean values were higher in both sectors than in the Control soil for SOC, total-N, C/N, and the available-P (Table 4). Calcium carbonate content was also higher in Sector 1 than in the Control soil. Carbonate content, SOC, and N were higher in Sector 1 than in Sector 2. For EC and available-K, although statistical differences were not detected, concentrations in both sectors were higher than in the Control soil (Table 4). Finally, pH was nearly identical in the two sectors and the Control soil, and the C/N ratio was close to 10 in both sectors but higher (16.6) in the Control soil. Regarding variability among plots (Table 3), CV values were very high (>35%) for EC, available-P, and available-K in both sectors, as well as for calcium carbonate in Sector 1. Moderate variability (15–35%) was observed for calcium carbonate in Sector 2 and for SOC and total-N in both sectors. In contrast, pH and the C/N ratio exhibited low variability (CV < 15%) across both sectors.

3.3. Composition of Manures

The characterization of four manure samples and two composts collected from the gardens is presented in

Table 5. General properties and extractable aqua regia contents of manures and composts. (HM: horse manure; CD: cow dung; CC: commercial compost; BC: biowaste compost).

		HM1	HM2	HM3	CD	CC	BC
Moisture	%	53.0	26.0	72.0	20.9	23.5	11.7
pH		8.85	7.88	8.14	7.58	9.05	8.51
E.C. 1	dS m−1	1.40	1.84	1.86	4.08	1.52	3.45
Inorg-C	g kg−1	2.25	4.48	2.29	5.94	5.95	15.0
Org-C	g kg−1	404	99.6	365	356	121	328
OM 2	g kg−1	768	189	693	676	229	607
N	g kg−1	13.9	9.34	14.0	22.5	12.9	32.6
C/N ratio		29.1	10.7	26.0	15.8	9.0	10.1
Ca	g kg−1	19.1	24.3	12.6	23.8	19.4	83.2
K	g kg−1	12.0	5.68	19.4	15.1	11.3	11.4
Mg	g kg−1	2.83	2.33	3.66	4.42	6.69	7.79
Na	g kg−1	7.62	1.70	4.26	6.58	3.14	6.43
S	g kg−1	2.66	1.55	2.53	3.90	2.55	5.52
P	g kg−1	3.95	2.74	3.78	4.70	5.20	6.99
Al	g kg−1	1.55	7.19	2.96	2.22	8.57	5.07
As	mg kg−1	<1.0	<1.0	<1.0	<1.0	21.9	1.02
B	mg kg−1	11.9	10.7	12.2	13.8	20.7	34.4
Ba	mg kg−1	39.8	53.0	69.0	34.4	67.0	49.9
Cd	mg kg−1	0.24	0.32	0.30	0.16	0.23	0.19
Co	mg kg−1	0.55	2.39	0.87	1.44	3.80	1.31
Cr	mg kg−1	10.1	38.2	10.9	12.9	92.8	40.2
Cu	mg kg−1	10.8	23.6	8.1	22.1	46.6	56.7
Fe	mg kg−1	1697	8286	2687	2546	9297	7097
Li	mg kg−1	2.01	7.70	4.09	3.31	28.3	4.57
Mn	mg kg−1	95.0	123	90.0	159	343	207
Mo	mg kg−1	1.43	0.79	2.38	1.87	2.80	3.44
Ni	mg kg−1	5.70	8.66	5.39	7.30	44.7	26.1
Pb	mg kg−1	8.0	9.5	14.5	1.9	14.9	24.1
Sn	mg kg−1	0.91	0.84	0.58	0.48	1.9	13.8
Sr	mg kg−1	48.0	58.6	101.4	56.6	84.8	193
V	mg kg−1	4.2	14.6	6.3	5.0	16.3	10.9
Zn	mg kg−1	41.1	68.5	49.6	155.1	153.5	167

¹ E.C.: Electrical conductivity 1:5 v/v extract; ² OM: organic matter content.

. Beyond C and N, the elements at the highest concentrations in the samples were Ca and K, with values ranging approximately from 10 to 20 g kg⁻¹. In the case of biowaste compost, Ca concentrations were higher than in the manures, although this product was used only during a single season in a few plots. Excluding the biowaste compost, Na concentrations in the manures ranged from 1.7 to 7.6 g kg⁻¹, Mg from 2.3 to 6.7 g kg⁻¹, S from 1.5 to 3.9 g kg⁻¹, and P from 2.7 to 5.2 g kg⁻¹.

Regarding PTEs, the observed concentration ranges in the organic amendments were Mn 90–343 mg kg⁻¹, Zn 41–167 mg kg⁻¹, Cr 10–93 mg kg⁻¹, Cu 8–57 mg kg⁻¹, Ni 5–45 mg kg⁻¹, and Pb 2–24 mg kg⁻¹. Among the more potentially toxic elements, As concentrations were below 1 mg kg⁻¹, except in one sample that exceeded 20 mg kg⁻¹, and Cd was <0.3 mg kg⁻¹. Although the manures did not originate from organic farming operations, these values were below the limits established in the European Union Ecolabel [45] for Zn, Cr, Cu, Pb, and Cd in soil amendments. Only Ni slightly exceeded the regulatory limit in one sample. The organic matter content also surpassed the minimum value required for the Ecolabel.

3.4. Principal Component Analysis

Figure 4 shows the first two components or factors from the principal component analysis (PCA) performed on the dataset of the studied soils. The first factor explained 33.2% of the variance in the data. In the component matrix, the highest loadings in this factor corresponded to Sr (0.935), Ca (0.926), Si (–0.891), calcium carbonate (0.837), total-N (0.823), available-P (0.811), Rb (0.792), and SOC (0.757).

The second factor explained 18.7% of the variance, with the highest loadings observed for Al (0.897), Ba (0.812), Fe (0.684), and V (0.662), and in the opposite direction P (–0.659).

The third factor explained 11.2% of the variance, with the highest loadings corresponding to S (0.732), EC (0.725), Pb (0.558), and Zr (0.534), and in the opposite direction pH (–0.529).

As can be inferred from Figure 4, many of the analyzed parameters showed correlations with each other. These pairwise correlations are shown in Supplementary Material Table S1.

3.5. Prediction of Available-K and Available-P

The multiple linear regression analyses performed to estimate available-P and -K concentrations as a function of elemental composition and other soil properties are shown in

Table 6. Results of multiple linear regression for the prediction of available-K and available-P.

Predictor	t	Sig.	Importance
Available-K					Training Set	Validation Set
K	13.39	0.000	0.688	R2	0.919	0.091
Al	−6.10	0.000	0.143	RMSE	233	217
Mn	−5.47	0.000	0.115	RMSE %	14	101
pH	−3.29	0.003	0.042	RPD	1.26	1.36
Available-P					Training Set	Validation Set
Sr	7.39	0.000	0.640	R2	0.787	0.863
P	4.32	0.000	0.219	RMSE	192	40
Mn	−2.60	0.015	0.079	RMSE %	40	10
Al	2.29	0.030	0.062	RPD	0.56	2.72

. Only predictors with statistically significant effects are included.

The model for available-K achieved an R² of 0.919 in the training dataset, but it decreases to a very low value (0.09) in the validation dataset. Correspondingly, RMSE% increases from 14% in the training dataset to 101% in the validation dataset. The RPD value remains close to 1.3 in both datasets. Total K concentration was the most influential predictor, followed—acting with negative coefficients and in decreasing order of importance—by the concentrations of Al, Mn, and by soil pH.

The concentration of available-P was predicted with an adjusted R² of 0.787 in the training dataset, but it increases to 0.86 in the validation dataset. The predictor with the greatest importance was Sr concentration, followed by total-P. In contrast, Mn acted in the opposite direction, although with lower importance than the aforementioned predictors. The concentration of Al is also significant in this case with a positive sign.

The correspondence between the values predicted by the linear models and the observed values is shown in Figure 5.

3.6. Intra-Plot Variability

Intra-plot CVs in 3 plots at Zone 1, where several samples (from 3 to 6) were taken, and those for Zone 1 for comparison purposes are shown in

Table 7. Coefficient of variation (%) of soil properties in selected plots (number in brackets indicates the number of cases).

Plot	35	34	45	Zone 1
EC	56.3 (6)	16.5 (3)	57.3 (4)	67.7 (33)
CaCO3	21.0 (6)	32.3 (3)	15.0 (4)	54.3 (33)
SOC	14.2 (6)	16.0 (3)	32.6 (4)	31.7 (33)
Total-N	12.8 (6)	9.1 (3)	27.3 (4)	31.3 (33)
Available-P	7.7 (6)	8.9 (4)	14.3 (4)	40.0 (38)
Available-K	55.1 (6)	34.5 (3)	40.6 (4)	76.4 (33)
S	11.3 (6)	6.9 (4)	6.4 (4)	40.8 (38)
Zn	12.1 (6)	6.4 (4)	5.5 (4)	47.8 (38)
Ca	10.3 (6)	5.8 (4)	7.3 (4)	37.1 (38)
Pb	9.1 (6)	18 (4)	18.0 (4)	17.2 (38)
Cu	26.1 (6)	7.6 (4)	2.1 (4)	26.0 (37)
Ni	24.0 (4)	7.8 (3)	--	15.9 (23)

. In general, the intra-plot CVs were lower than those for Zone 1, although CVs > 30 were observed for available-K in all three plots, for EC in plots 35 and 45, for calcium carbonate in plot 34, and for SOC in plot 45. The lower intra-plot CVs for available-P, S, Zn, and Ca compared to the zone values were noteworthy.

4. Discussion

4.1. Evaluation of the Performance of the X-Ray Fluorescence Instrument

The assessment of the performance of the measuring instrument is crucial for determining how much of the variability in the measurements can be attributed to the soil itself and how much to deviations inherent to the analytical technique. In addition, it provides the basis for assigning a level of confidence to the various results presented throughout the study. However, the determination of precision and accuracy is not always considered in studies employing pXRF, despite the widespread use of this technique [46].

The USEPA protocol [37] states that, for calibration verification to be acceptable, the measured value for each target analyte must fall within ±20% of the true value, i.e., a recovery between 80% and 120%. As shown in Table 1, this requirement was fulfilled—considering the recovery range (minimum and maximum recovery)—for most of the analyzed elements. Specifically, for As, Cu, Pb, S, Sb, Sr, Zn, Fe, Ca, Ba, Mn, Rb, and Si, all verification measurements were acceptable. Among these, Cu, Pb, S, Sb, Sr, Zn, Fe, and Ca also exhibited low coefficients of variation (CV < 3%), except for Cu and S, which showed CV > 10%; therefore, the stability of their measurements can also be considered as high. For comparison, Qu et al. [47] reported a CV of 6.51% for Pb based on seven measurements, whereas the present study yielded a CV of 2.2%.

The elements Ni, Mg, Al, and P displayed recovery values with errors exceeding 50%, but this is attributable to their concentrations in the reference material being close to the LOD of the technique. Poor pXRF performance for light elements has been frequently reported in the literature [46]. The recovery of Ag, Cr, and Ti was intermediate (±20 to ±30%), and in the case of Ag and Cr, their concentrations in the reference material were also close to the LOD. In the case of Ti, the CV was low, indicating that the deviation must be due to another cause or to a specific interference.

Consistent with these findings, in a study limited to Ni, Cu, Zn, Pb, Cd, As, and Mn [30], high precision was reported for elements with higher atomic mass, such as Pb, while an opposite result was observed for elements with lower atomic mass, such as Ni, with decreasing data reliability in the order Pb > Cu > Zn > Cd > Ni. The same authors also noted that the error margins for Pb and Cu were similar to those obtained using ICP-OES.

4.2. Variability of Elements in the Soil of the Plots

The elements S, Zn, Ca, P, Sr, and Cu—which exhibited the greatest variability (CV) in the soils of the garden plots (Table 2)—also displayed the largest increases in mean concentrations relative to the Control soil: S (238%), Zn (201%), Ca (338%), P (79%), Sr (190%), and Cu (69%). Since their CVs (>25%) were far higher than those attributable to analytical deviations or fluctuations (CVs in Table 1, generally on the order of 10% or lower except for P), the variation in elemental content can be a consequence of the incorporation of these elements via the manures or organic amendments used by the gardeners. The lower intra-plot variability of the studied parameters (Table 7) also corroborates that the different fertilization strategies of each horticulturist are responsible for the inter-plot differences. In the case of P, its concentration in the reference material (345 mg kg⁻¹) was close to the detection limit of the technique, whereas concentrations in the garden soils were higher; therefore, its reproducibility can be assumed to be greater in the garden soils.

Indeed, the analyzed manures (Table 5) showed high concentrations of nutrients such as Ca (12.6–24.3 g kg⁻¹), P (2.7–5.2 g kg⁻¹), and S (1.5–3.9 g kg⁻¹). These concentrations were similar to those reported by other authors for horse manure [48], for farmyard manure [15], and for raw manures from four European countries [16]. Kinoshita et al. [12] also showed that the use of animal manure is an important factor contributing to the high levels of P and Ca observed in peri-urban agricultural fields in Kenya. In all cases, the manure concentrations exceeded those in the Control soil (14.3, 0.14, and 0.04 g kg⁻¹ for Ca, P, and S, respectively), indicating that continued manure application has resulted in the accumulation of these elements in the soil.

This increase, in addition to total-P concentration, is also reflected in the available-P fraction, which has increased by nearly one order of magnitude—from 26 mg kg⁻¹ to more than 200 mg kg⁻¹ in both garden areas (Table 3). Losses of P from manures to water bodies that cause eutrophication are a well-known environmental issue in Europe [16] and all over the world. However, in the studied soils, such losses are unlikely due to their high Ca content, which promotes the precipitation of phosphates as insoluble apatite. From an agronomic standpoint, the available-P (Olsen–P) levels reached in the garden plots greatly exceed the nutritional requirements of most crops [49].

Potassium was also supplied in significant quantities by the manures (5.7–19.4 g kg⁻¹), but it showed only a slight increase in its total soil concentration (from 7 to an average of 7.5 g kg⁻¹). The manure concentrations fall within the lower end of the range reported for European manures (15–157 g kg⁻¹) [16]. Even so, assuming manure application rates of around 100 Mg ha⁻¹, approximately 1000 kg K ha⁻¹ would be added—an amount exceeding the uptake and requirements of most horticultural crops [50]. Although the increase in total K content was small, the available-K fraction (Table 3 and Table 4) increased markedly in many garden plots (see maximum values in Table 2). Thus, a small fraction of the K supplied by manure has remained in the soil in available forms, while the excess has probably leached into deeper horizons due to the sandy nature of the soil. Similarly, Ref. [48] observed increases in available-K, although not statistically significant, following large applications of equine manure.

Three PTEs—Zn, Pb, and Cu—showed wide variability in the soils (Table 2). Between them, Cu and particularly Zn showed a concentration increase compared to the Control soil. Both Cu and Zn are added to animal feed, although regulatory limits exist [16]. The mean Zn contamination factor (CFZn) for both garden areas (Figure 3) was close to 1, indicating no contamination [35], despite the increase compared to the Control soil. This is because the Control soil concentration was below the reference value for this geological domain (56 mg kg⁻¹). Thus, although the amendments analyzed contained higher concentrations (41–167 mg kg⁻¹) than the original soil (22 mg kg⁻¹), risk Zn levels were not reached.

A similar situation was observed for Cu, although the CFCu values in both areas were greater than 1 (Figure 3). The Cu concentration in the original soil (~20 mg kg⁻¹) was close to the geological background (24 mg kg⁻¹). However, the Cu concentrations in the analyzed amendments (mean of 28 mg kg⁻¹), including manure+greenwaste and biowaste compost do not fully explain the 69% increase observed in the garden soils. The use of Cu-based fungicides authorized in organic farming can contribute to this increase. A similar situation was reported in a previous study [35].

The specific case of Pb, with CFPb values > 1 across all zones, may indicate the presence of an additional pollution source. The greater dispersion and higher values observed in Zone 2 (Figure 3) are consistent with pollution related to historical traffic, as this zone is closer to the urban centre and, in particular, to a highly frequented church (less than 200 m from its parking lot). In a study of several parks in the city of Seville, Madrid et al. [51] attributed high Pb levels to traffic.

Overall, the PLI values for the Control soil were <1, suggesting contamination levels of PTEs below the local background, likely due to the sandy nature of these soils [35]. In contrast, zones 1 and 2 presented mean PLI values slightly >1, indicating that inputs such as manures, fungicides, and traffic emissions have produced mild contamination.

As shown in Table 2, a group of elements—Zr, Sb, Al, Ti, and Si—exhibited lower mean concentrations in the cultivated plots compared with the Control soil. Aluminum, Ti, Zr, and Si are crustal elements commonly found in the Earth’s crust, and Si, Ti, and Zr often form a geochemical association [52] (p. 49). The observed concentration decrease in these elements can be due to the dilution effects associated with increases in Ca concentration or soil organic matter.

The principal component analysis (Figure 4) visually summarizes the associations among the variables previously discussed. One-third of the variability in soil properties (explained by Factor 1) was associated with calcium carbonate content, including CaCO₃, Ca, and Sr (the latter via isomorphic substitution in Ca-bearing minerals). SOC, total-N, Zn, and P also load positively on this factor, reflecting their shared origin from manure inputs. Figure 6 shows the statistically significant relationship between the CaCO₃ content of the soil and the SOC. These inputs diluted Si, leading to its negative loading. In a comparable study of urban soils in Ghent, Belgium, Delbecque et al. [53] reported that 17% of the variability was related to soil carbonate content.

Factor 2, accounting for nearly 20% of the variability, was primarily influenced by Al, Ba, V, and Fe concentrations (Figure 4, upper-right quadrant). These elements did not exhibit a clear trend in the soils (Table 2) and displayed moderate variability (16.6 < CV < 21.5), suggesting that their distribution is mainly regulated by geological or pedological rather than anthropogenic processes.

Factor 3 explained approximately 11% of the variance and groups variables linked to manure-derived inputs as well as soil salinity, as it includes EC, which may partly reflect sulfate content. As noted by Kenny et al. [54], increases in salinity due to manure application can become problematic for plants. The inclusion of Zr in this factor may be related to particle-size distribution, itself governed by lithological driving processes [52].

The factor analysis including the soil variables enabled the predictions of K and P availability using only a small set of parameters (Figure 5; accuracies in the complete dataset of 78% and 80%, respectively). The model for available-K shows that its availability is, as expected, proportional to the total K concentration, while Al, Mn, and pH exert negative effects. Kabata-Pendias and Pendias [52] report antagonistic interactions of Al and Mn with K in plant nutrition, although in this case, their negative influence on availability may instead reflect the geological characteristics of the parent materials. The increase in pH is expected to reflect a higher calcium (Ca) content, which may displace exchangeable potassium (K) from cation exchange sites. The displaced K could subsequently leach, leading to a reduction in the amount of available-K.

The model for available-P incorporated four elements measurable by pXRF. In addition to total-P, Sr and Al contributed positively to P availability. Strontium (Sr) played a positive effect because it was applied together with Ca in manure amendments, in a manner similar to phosphorus (P). In a soil study conducted in the Czech Republic, Janovský et al. [55] associated the enrichment of P, Sr, Zn, and Mn with historical agricultural management, possibly linked to manure application. Strontium content in the manures analyzed was higher than in the Control soil, and it is plausible that newly formed or added Ca and Sr carbonates in the soils possess a distinct surface adsorption capacity for phosphate thereby enhancing its availability. The model also indicated a positive effect of Al content, which might appear contradictory to the blocking effect on phosphate that is usually attributed to Al oxides. This effect is particularly relevant in acidic soils such as tropical soils, where P deficiencies often occur. However, in the case of these soils, their higher pH, close to 7.5, justifies the different behaviours. Hesse [56] indicated that in calcareous and non-calcareous soils among the most available forms of P are discrete precipitates of calcium phosphate and aluminum phosphate on the surfaces of calcium carbonate and aluminum oxides, respectively. Rubidium, by contrast, reduced P availability. Although its influence was relatively small, it was statistically significant given the low CV of this element (Table 2). Rubidium, as well as Mn, is considered an antagonist of P [52].

Although these models are grounded in the chemical principles discussed above, the evaluation of the accuracy parameters in Table 6 indicates that the available-K model is weak. Chang et al. [57] and Tavares et al. [58] established four categories for prediction performance based on RPD values: poor (RPD < 1.40), reasonable (1.40 < RPD < 2.00), good (2.00 < RPD < 3.00), and excellent (RPD > 3.00). The RPD values (≈1.3) obtained for available-K prediction therefore classify the model as poor. In the case of available-P, the prediction was poor for the training set but good (2.7) for the validation set. The reason for the discrepancy between both sets lies in the fact that, as shown in Figure 5, the relative deviation of the predictions is large at low concentrations of available-K and -P (<500 mg kg⁻¹ for available-K), whereas the relative deviations decreased at higher concentrations. Considering the R² values, the prediction quality is similar to or better than that reported by Gozukara et al. [32] for various soil properties using pXRF (R² = 0.36–0.86). However, Tavares et al. [58] obtained higher RPD values for the prediction of exchangeable-K (a parameter analogous to available-K) using XRF in tropical soils, although in that case the concentration ranges were likely narrower.

Obviously, the addition of manures has supplied organic matter that has increased the soil SOC and N contents (Table 3 and Table 4). Since Zone 1 has been used as an urban garden for a longer period than Zone 2 (16 and 12 years, respectively), the mean SOC concentration reached in Zone 1 was significantly higher than in Zone 2. The SOC content varied widely—by up to a factor of 3 between the maximum and minimum measured values—reflecting the different rates and types of manure applied, with a larger range in Zone 1 than in Zone 2. The observed SOC levels clearly exceed the usual values in agricultural soils of the area, which rarely surpass 0.6–1.2% [59] due to the semiarid climatic conditions of this Mediterranean region. Despite the variability in SOC, the C/N ratio was relatively stable, with a mean value of 10, which is typical for well-developed agricultural soils [49] (p. 34). Only in 4 out of the 41 samples, the C/N ratio exceeded 11. This may indicate that the mineralization of organic materials with an initial C/N ratio greater than 10 occurred rapidly, conducing to a situation with stable organic matter. In a study carried out in a nearby soil under the same climatic regime, San Emeterio et al. [60] found that the addition of aged compost could be beneficial for the soil microbial community, producing a pronounced short-term priming effect that enhanced the decomposition of soil organic matter and reduced the mean residence time of the labile, rapidly mineralizable fraction. From an agronomic perspective, this would result in a soil capable of mineralising amounts of N enough to ensure the nutritional requirements of vegetables, as was indeed observed from crop growth and yield (not addressed in this study).

5. Conclusions

The continuous incorporation of manure over 12 or 16 years in an allotment garden increased fertility, with notable rises in SOC, N, available-K, and total- and available-P. It also affected its mineral composition, with substantial increases in total Ca, inorganic C (calcium carbonate), Sr, and S. Despite the use of considerable manure rates, no increases in PTEs, which could pose a concern, have been detected. Exposure pathways of PTE include direct ingestion, dermal contact, and inhalation [61,62]. It is well known that elements such as As, Cd, Cr, Ni, and Pb can seriously affect human health when they exceed safe limits and also cause cancers [63,64]. However, the individual plots have shown considerable variability among them, attributable to the differing fertilization strategies (amount and type of manure) adopted by each gardener.

Although this variability could be problematic from an agronomic management perspective, it provides urban gardens and green spaces with unique characteristics for modelling, dynamics, and the study of processes related to soil chemistry or microbiology, since gradients in various properties can be identified over a common and homogeneous pedological and geological substrate.

The pXRF technique, due to its speed, cost-effectiveness, ease of use, and robust performance, has proven satisfactory for studying soil composition and for predicting soil characteristic parameters traditionally determined through wet-chemistry laboratory procedures, such as levels of available nutrients. However, predictions could be improved by combining pXRF with other proximal sensors such as visible and near-infrared (vis-NIR) diffuse reflectance or by narrowing the prediction limits in the higher concentration ranges, where the accuracy of pXRF is adequate. Portable-XRF proved reliable for most elements, meeting recovery standards between 80 and 120% and showing low variability. However, accuracy dropped for light elements (Mg, Al, P) or for those with concentrations near the detection limit (Mg, Cr, Cd). These findings highlight the need for caution when interpreting results near the LOD and suggest complementary methods for critical analyses. The combination of the spatial variability detected in small areas such as urban gardens with proximal techniques such as pXRF (or other spectral techniques like NIR or MIR) can provide, at low cost, the necessary tools to support the calibration of fertilization strategies in precision agriculture for soils or agricultural systems of similar nature or characteristics. This study may contribute to a deeper understanding of small-scale horticultural production systems by applying an analytical technique that, due to its speed and low cost, can be highly suitable for use in these contexts.