Neural Network Modeling of Physicochemical Processes in Fluvisols with the Application of Sewage Sludge as Fertilizer for Energy Crops

Abstract

This study investigates the impact of sewage sludge-based composts on the quality of sod-podzolic soils (Fluvisols, WRB), which are naturally acidic and low in fertility. A field experiment with eight variants was conducted, applying different doses of sewage sludge and composts mixed with organic materials to enhance soil properties. Energy crops, including Salix viminalis L., Miscanthus × giganteus, and Panicum virgatum L., along with Helianthus tuberosus, were cultivated in three replications to ensure research reliability. The study assessed changes in physicochemical soil properties, nutrient availability, and heavy metal accumulation within the soil-plant-ash continuum. Results indicated that compost application improved soil fertility, increased biomass yields, and influenced heavy metal dynamics, with variations depending on the applied compost type and dosage. The findings highlight the potential of sewage sludge composts to enhance soil productivity while maintaining environmental safety.

1. Introduction

Modern agriculture faces a double challenge: preserving soil fertility and simultaneously meeting the growing need for sustainable waste management solutions. This problem is particularly acute in acidic, nutrient-poor soils, such as alluvial soils (Fluvisols, WRB), which, despite a certain natural fertility, require additional enrichment with organic matter and macronutrients. Existing agrotechnical approaches are not always effective, environmentally safe, or economically feasible, especially in conditions of rapid growth in the amount of wastewater and accompanying sludge [1].

One of the effective methods of utilizing sewage sludge is the possibility of its transformation into a complex fertilizer for energy and agricultural crops. Taking into account the chemical composition of SS [2], it is important to develop optimally efficient and environmentally friendly technologies for its use in phytopower.

Wastewater sediment (WS) is considered a promising alternative to traditional fertilizers, as it contains a significant amount of organic matter, macro-, and microelements that can improve the physicochemical properties of the soil. Given this, WS is potentially able to solve two key problems—waste disposal and improving the condition of degraded soils. However, due to the complex and variable chemical composition of the sediment, as well as the risks of accumulation of heavy metals and toxic compounds, its use requires careful scientific justification [1,3,4,5,6].

Recent studies show significant progress in the study of the impact of WS on crops, in particular energy crops, as well as on soil properties. However, there is still insufficient data on the prediction of the long-term effects of sediment on different types of soils, including Fluvisols, and on the optimization of the dosage and conditions of fertilization based on it [3,7,8,9]. The use of composted WS, which has lower toxicity risks, but also requires additional research, becomes particularly relevant.

Since traditional assessment methods are time-consuming and do not always provide accuracy in predicting changes in soils under the influence of WS, more and more attention is paid to the use of modeling technologies. Modeling using neural networks (AI) allows us to take into account the nonlinear relationships between numerous variables—such as acidity, saturation with macronutrients, and concentration of heavy metals—and predict the results of fertilizer application with high accuracy. This approach not only increases the effectiveness of agrotechnical measures, but also contributes to the formation of environmentally safe recommendations for the introduction of WS [2,7,10].

In view of the above, the use of WS for the cultivation of energy crops such as Salix viminalis L., Helianthus tuberosus L., Silphium perfoliatum L., Miscanthus × giganteus, Panicum virgatum L., and others, is of particular interest. These crops have the ability to absorb nutrients from poor soils, withstand high concentrations of some elements, and can be used as a source of bioenergy. However, when assessing the effectiveness of their cultivation, it is necessary to take into account not only the yield, but also the composition of energy biomass, since it determines the possibilities of further use of raw materials, in particular, after thermal processing [11,12,13,14,15]. Here we are talking about the problem of the formation of aggressive ash during the burning of biomass due to its corrosive effect on the elements of heating equipment. It is also emphasized by the authors of the study [15], who focused on the properties of biomass residues after combustion, in particular ash. They note that ash can exhibit different corrosive properties, which is a significant limitation for the long-term operation of thermal installations. Similar results were obtained by Vitázek et al. [16], who investigated the kinetics of thermal decomposition of wood chips Malus domestica Borkh, which allowed a better understanding of the behavior of biomass during combustion and the potential formation of harmful ash residues.

Turf-podzolic soils (Fluvisols, WRB) of the Carpathians are characterized by an acidic reaction, low humus content, and weak cationic capacity, which limits their fertility. The use of sewage sludge enriched with organic matter and nutrients is considered a promising way to increase it, but needs control to prevent potential environmental risks.

The purpose of this study is to study the physicochemical processes occurring in alluvial soils under the influence of sewage sludge, with an emphasis on increasing the efficiency of growing energy crops, minimizing environmental risks, and substantiating optimal agrotechnical solutions. In this context, artificial neural networks are used as a tool for analyzing and predicting changes in soil indicators.

2. Materials and Methods

Taking into account the chemical composition of sewage sludge, it is important to develop optimally effective and environmentally safe technologies for its use in phytoenergy, which led to the study of the effectiveness of its application for the cultivation of energy crops:energy willow (Salix viminalis L.), giant miscanthus “Autumn Starflower” (Miscanthus × giganteus), pierced-leaved silybum (Silphium perfoliatum L.), switchgrass “Morozko” (Panicum virgatum L.), and Helianthus tuberosus “Lviv” (Helianthus tuberosus L.) [7,14].

Field studies were conducted on the territory of Tsenzhiv village of Yamnytsia rural amalgamated territorial community of Ivano-Frankivsk region (Precarpathian zone of western Ukraine) and on the collection and research field of Ivano-Frankivsk Professional College of Lviv National University of Environmental Management during 2016–2022 [6].

The analysis of the chemical composition of wastewater sediments from the Municipal Enterprise Ivano-Frankivsk Aeration Station “Ecotechprom” was carried out in accordance with generally accepted methods [6,7]. All laboratory studies were carried out in accordance with standardized methods: sampling for chemical analyses was carried out in accordance with DSTU 4287:2004 standard [17]; for physicochemical analyses—in accordance with DSTU ISO 11464:2007 standard [7,18]. Determination of the humus content in the soil—by the Tyurin method (oxidimetric method) in the modification of M. M. Kononova and N. P. Belchikova; pH—according to the DSTU ISO 10390:2022 standard [19]; hydrolytic acidity—according to the DSTU 7537:2014 standard [20]; content of exchangeable cations—according to the MVV 31-497058-007 [21]; and determination of trace elements and heavy metals was carried out according to the methods of DSTU 4770.1:2007—DSTU 4770.9:2007 standards [7,22,23,24,25,26,27,28,29]. Determination of nitrification activity according to the DSTU 8578:2015 standard [30]; determination of total biological activity by the gas chromatic method according to the DSTU 8644:2016 standard [31]; determination of the number of microorganisms in the soil by sowing on solid (agarized) nutrient medium according to the DSTU 7847:2015 standard [32]; and determination of total biological activity according to the DSTU 8644:2016 standard [7,31].

Field research was carried out on sod-pidzolic soils typical of the foothills of the Carpathian region. According to the Ukrainian classification of soils, these are poorly cultured acidic soils on alluvial sediments. According to the international classification WRB (World Reference Base for Soil Resources), these soils correspond to the Fluvisols type with features of the initial podzolic process. In the subsequent text, both terms are used to preserve the correspondence—sod-podzolic soils (Fluvisols, WRB).

The chemical composition was studied by X-ray fluorescence analysis. The method is based on the analysis of the fluorescence spectra of elements emitted during the adsorption of high-energy radiation. The method allows obtaining data on the chemical composition of a substance in a wide range with an accuracy of 1–10 ppm. The experiments were performed on the EXPERT 3L precision analyzer. The volume of the test sample was about 0.5 cm³. To average the composition, samples were taken from three different points of the sample. The test sample was placed in a special cylinder with a thin light-permeable film on the bottom. The cylinder with the sample was placed on an optical sensor placed inside the analyzer. After hermetically closing the analyzer, the atmosphere around the sample is automatically replaced with helium and the analysis process begins. If an element is not in the list of results, it means that its proportion in the sample is too small (approximately less than 1/1,000,000 by weight). The results are presented in ascending order of atomic mass of chemical elements in the periodic table [7]. The results of the study were processed by the methods of mathematical statistics using STATISTICA 12.0 for Windows.

Two separate field experiments were conducted to evaluate the effectiveness of sewage sludge and wastewater-based composts for growing energy crops: Experiment 1, with energy willow (Salix viminalis L.), included 10 fertilization options (pure sediment and compost based on it). Experiment 2, with Miscanthus × giganteus, Silphium perfoliatum L., Panicum virgatum L., and Helianthus tuberosus L., included eight options (sediment in different doses with compensation with mineral fertilizers and compost). This approach made it possible to study the dose-response relationship and compare the effectiveness of different forms of fertilizers.

Field studies to determine the ecological and agrochemical assessment of the use of sewage sludge during its repeated application during the cultivation of Salix viminalis L. on sod-podzolic soils of Ivano-Frankivsk region, namely, on the collection and research field of the Ivano-Frankivsk Professional College of the Lviv National University of Environmental Management were conducted according to the following scheme (the concept of “scheme” refers to the spatial placement of options on the field). Experimental variants (specific fertilizer application regimes that include control, mineral fertilizers, sewage sediment, and compost based on it):

• Control variant—without fertilization;
• Mineral fertilizers—N₁₀₀P₁₀₀K₁₀₀;
• SS—40 t.ha⁻¹;
• SS—60 t.ha⁻¹;
• SS—80 t.ha⁻¹;
• Compost (SS + sawdust (3:1)) —60 t.ha⁻¹;
• Compost (SS + straw (3:1))—20 t.ha⁻¹;
• Compost (SS + straw (3:1))—40 t.ha⁻¹;
• Compost (SS + straw (3:1))—60 t.ha⁻¹;
• Compost (SS + straw (3:1) + cement dust 10%)—40 t.ha⁻¹ (Figure 1), [6].

The planting scheme for Salix Viminalis L. was 0.33 × 0.70 m. Each variant was replicated three times, and the plots were placed systematically. Each experimental plot consisted of 10 rows with 10 rhizomes in each row [6].

Field studies with Miscanthus × giganteus (Osinnyi Zoretsvit variety), pierced-leaved sylph (Kanadchanka variety), Panicum virgatum L. (Morozko variety), and Helianthus tuberosus (Lvivskyi variety) were conducted on sod-podzolic soils of the Precarpathian region in the territory of Tsenzhiv village of Yamnytska rural united territorial community of Ivano-Frankivsk region.

Experimental variants:

• Control variant—without fertilization;
• N₆₀P₆₀K₆₀;
• N₉₀P₉₀K₉₀;
• SS —20 t.ha⁻¹+ N₅₀P₅₂K₇₄;
• SS—30 t.ha⁻¹ + N₃₀P₃₃K₆₆;
• SS—40 t.ha⁻¹ + N₁₀P₁₄K₅₈;
• Compost (SS + straw (3:1)) —20 t.ha⁻¹ + N₅₀P₁₆K₆₇;
• Compost (SS + straw (3:1)) —30 t.ha⁻¹ + N₃₀K₅₅.

All variants of the experiments, except for 1 and 2, were corrected with a compensatory dose of mineral fertilizers at the rate of N₉₀P₉₀K₉₀, taking into account the chemical composition of the sewage sludge. Thus, the amount of mineral fertilizers applied in variants 3–8 was the same. Helianthus tuberosus planting scheme—0.50 × 0.70 m (Figure 2) [6].

All experimental variants were laid out in a systematic design with three replications.

The experimental plots were 5.0 m wide and 7.0 m long, with a total area of 63.0 m² and an accounting area of 35.0 m² [6]. The planting material was manually placed into the soil using selected large Helianthus tuberosus tubers, which were embedded at a depth of 8–10 cm (Figure 3).

The planting scheme for Miscanthus × giganteus was 0.50 × 0.70 m. Rhizomes were planted manually by embedding them into the soil at a depth of 10–15 cm. Each experimental plot was 5.0 m wide and 7.0 m long, with a total area of 63.0 m² and an accounting area of 35.0 m². Each plot contained 10 rows, with 10 rhizomes planted in each row (Figure 3).

The study used the method of neural network modeling using the STATISTICA 12.0 program for Windows in order to predict changes in the justified physical and chemical parameters of the soil under the influence of sewage sludge. The neural network used in the study is a multilayer perceptron with two hidden layers of 15 neurons. The input parameters of the model were options for the introduction of sewage sludge and composts based on them, and the output parameters were the physicochemical parameters of the soil (pH, Ca²⁺, and Mg²⁺ content, degree of saturation with bases).

3. Results and Discussion

Dynamic changes in the physical and chemical properties of the soil under the influence of fertilizers directly affect the availability of nutrients to plants, and the nature and intensity of their absorption. The results of our research indicate a significant influence of the fertilizer application system on the dynamics of physicochemical parameters of sod-podzolic soil. In our studies, the pH value in the control variant was 4.8 in the arable (0–20 cm) and 4.5 in the subsoil (20–40 cm) layers of sod-podzolic soil. In the study, the upper soil layer refers to the arable humus horizon (A), which covers a depth of 0–20 cm. In this study, the soil layer at a depth of 20–40 cm was considered a poor horizon (under the upper humus), which allows us to assess the vertical effect of fertilizers on the soil profile.

The introduction of mineral fertilizers contributed to an increase in pH by 0.3 in the tilth and 0.8 in the subsoil layers. The use of the mineral fertilizer system contributed to the acidification of the soil solution. Compared to the control, fertilization with mineral fertilizers N₁₀₀P₁₀₀K₁₀₀ of variant 2 provided a change in the pH_KCl index by 0.6 in the tilth and 0.7 in the subsoil layers.

Under the influence of re-application of the sewage sludge, the pH gradually increased from 6.4 to 6.8 in the tilth and from 6.1 to 6.2 in the subsoil depending on the rate of the sewage sludge application (

Table 1. Changes in the physical and chemical parameters of sod-podzolic soil (Fluvisol, WRB) under the influence of repeated application of the sewage sludge (SS) for growing Salix viminalis L. (Hydrolytic acidity (Hr); Amount of absorbed bases, (S); Absorption capacity, (T); Degree of saturation with bases, (V, %)).

Experimental Design	Soil Layer (Depth)	pHKCl	Hr	S	T	V
	Cm	-	mmol per 100 g of Soil			%
1. Control variant—without fertilization	0–20	4.8	3.27	8.49	11.70	72.6
	20–40	4.5	3.30	8.42	11.75	71.6
2. Mineral fertilizers—N100P100K100	0–20	5.4	3.24	8.67	11.76	73.7
	20–40	5.2	3.26	8.64	11.75	73.5
3. SS—40 t.ha−1	0–20	6.4	2.96	8.92	11.78	75.7
	20–40	6.1	2.97	8.90	11.81	75.3
4. SS—60 t.ha−1	0–20	6.6	2.95	8.94	11.82	75.6
	20–40	6.2	2.93	8.86	11.84	74.8
5. SS—80 t.ha−1	0–20	6.8	2.90	9.10	11.87	76.7
	20–40	6.2	2.88	9.01	11.81	76.3
6. Compost (SS + sawdust (3:1))—60 t.ha−1	0–20	6.6	2.86	9.16	12.01	76.3
	20–40	6.3	2.88	9.11	11.96	76.2
7. Compost (SS + straw (3:1))—20 t.ha−1	0–20	6.4	2.80	9.06	11.98	75.6
	20–40	6.2	2.85	9.01	11.94	75.5
8. Compost (SS + straw (3:1))—40 t.ha−1	0–20	6.6	2.82	9.10	11.88	76.6
	20–40	6.3	2.89	9.06	11.86	76.4
9. Compost (SS + straw (3:1))—60 t.ha−1	0–20	6.7	2.86	9.13	11.99	76.1
	20–40	6.3	2.90	9.07	11.96	75.8
10. Compost (SS + straw (3:1) + cement dust 10%)—40 t.ha−1	0–20	6.4	2.90	8.96	11.96	74.9
	20–40	6.2	2.92	8.91	11.98	74.4
The least significant difference 0.05 (LSD0.05)	0–20	0.28	0.21	0.28	0.14	0.33
	20–40	0.40	0.18	0.03	0.02	0.34

).

This trend was also observed in the analysis of hydrolytic acidity. The use of the sewage sludge and composts based on it significantly influenced the dynamics of this indicator in the arable and subsoil layers of sod-podzolic soil. In the variant without fertilizers, the hydrolytic acidity content decreased from 3.27 to 2.80 mmol per 100 g of soil in the upper layer and from 3.30 to 2.85 mmol per 100 g of soil in the lower layer.

The use of a mineral fertilizer system contributed to a decrease in the content of hydrolytic acidity and amounted to 3.24–3.26 mmol per 100 g of soil. A decrease in the level of hydrolytic acidity was observed when compost based on SS + straw (3:1) was applied at a dose of 20 t.ha⁻¹. This indicator was 2.80 mmol per 100 g of soil in the upper (0–20 cm) layer and 2.85 mmol per 100 g of soil in the lower (20–40 cm) layer, which is 0.47–0.45 mmol per 100 g of soil compared to the control without fertilizers.

Increasing the doses of sewage sludge application contributed to a decrease in hydrolytic acidity in the upper and lower layers of sod-podzolic soil. In the control variant without fertilization, the amount of absorbed bases in the upper (arable) and lower (subsoil) layers of the soil was 8.49 and 8.42 mmol per 100 g of soil, respectively. The introduction of mineral fertilizers contributed to an increase in the indicator by 0.080–0.022 mmol per 100 g compared to the control without fertilizers. After the application of fresh sewage sludge, the indicator of the sum of absorbed bases increased to 9.10 mmol per 100 g in the tilth layer (0–20 cm) of the soil.

The cation absorption capacity of the sod-podzolic soil also depended on the fertilizers applied. In particular, in the control variant, it was 11.70 mmol per 100 g of soil. The use of mineral fertilizers had a positive effect on the cation absorption capacity and ensured its increase to 11.76 mmol per 100 g of soil in the upper (0–20 cm) soil layer. The introduction of fresh sewage sludge contributed to an increase in this indicator to 11.87 mmol per 100 g of soil in the tilth (0–20 cm) soil layer. It is under the influence of organic fertilizers of the sewage sludge and composts based on them that the microaggregate composition of the soil improves and the amount of calcium and potassium absorbed by the soil absorption complex increases [6]. The degree of saturation with bases in the control without fertilizers was 72.6% in the tilth layer and 71.6% in the subsoil layer. After the application of mineral fertilizers, this indicator increased by 1.1% at a depth of 0–20 cm and by 1.9% at a depth of 20–40 cm compared to the control variant. The highest indicators of the degree of saturation with bases (76.7%) were observed in the variant where fresh sewage sludge was applied at a dose of 80 t.ha⁻¹.

Under the conditions of the experiment, there is a clear tendency to increase the degree of saturation with bases under the influence of sewage sludge and composts based on them. This change in the experimental variants is observed within 0.9–1.2% of the absolute values.

The model in Figure 4 shows high accuracy in predicting the base saturation values with very small differences between the actual and predicted values. The average absolute error is usually less than 0.2%, indicating excellent prediction accuracy.

The key results of the analysis show that: The model accurately captures the increase in base saturation (V) from control treatments (72.6%) to organic treatments (75–76%). The highest values of base saturation are observed in the sewage sludge—80 treatment (76.7%), which the model correctly predicts. The model successfully recognizes the patterns of how different treatments affect soil properties, in particular the relationship between pH and base saturation. Visualization clearly demonstrates the close relationship between actual and predicted values for all treatments. The model can be used to predict the base saturation for new treatment combinations within the 0–20 cm layer, making it a valuable tool for soil management decision making. Based on the data presented in

Table 2. Changes in soil physicochemical parameters under the influence of the sewage sludge (SS) application for the case of growing grass crops.

Experimental Design	Soil Layer (Depth)	pHKCl	Hr	S	T	V
	Cm	-	mmol per 100 g of Soil			%
1. Control variant—without fertilization	0–20	4.8	3.10	8.49	11.70	72.5
	20–40	4.9	3.30	8.42	11.75	71.6
2. Mineral fertilizers—N60P60K60	0–20	5.2	3.08	8.53	11.71	72.8
	20–40	5.2	3.15	8.50	11.69	72.6
3. Mineral fertilizers—N90P90K90	0–20	5.2	3.06	8.56	11.72	73.0
	20–40	5.1	3.17	8.54	11.70	72.9
4. SS—20 t.ha−1+ N50P52K74	0–20	5.3	2.37	8.70	11.74	74.1
	20–40	5.1	2.99	8.66	11.76	73.6
5. SS—30 t.ha−1+ N30P33K66	0–20	5.4	2.94	8.74	11.79	74.1
	20–40	5.3	2.97	8.70	11.80	73.7
6. SS—40 t.ha−1+ N10P14K58	0–20	5.7	2.90	8.76	11.82	74.1
	20–40	5.4	2.92	8.74	11.83	73.8
7. Compost (SS + straw (3:1))—20 t.ha−1 + N50P16K67	0–20	5.2	2.88	8.70	11.79	73.7
	20–40	5.1	2.90	8.62	11.81	72.9
8. Compost (SS + straw (3:1))—40 t.ha−1 +N30K55	0–20	5.6	2.84	8.78	11.80	74.4
	20–40	5.3	2.90	8.72	11.81	73.8
The least significant difference 0.05 (LSD0.05)	0–20	0.1	0.02	0.02	0.01	0.2
	20–40	0.1	0.01	0.02	0.01	0.2

, the following analysis of the results of changes in the physicochemical parameters of the subsoil during the cultivation of cereals crops (Miscanthus × giganteus, Panicum virgatum L., and Silphium perfoliatum L.) can be made.

Analyzing the obtained results, we observe regularities in the profile of justification: Most indicators have a higher value in the upper layer (0–20 cm) of the soil. The difference between soil layers decreases with increasing doses of the sewage sludge and doses of composts based on them. The most pronounced effect is improving quality indicators in the upper soil layer. The optimal application dose is 30–40 t.ha⁻¹, but the application of composts based on the sewage sludge with straw (3:1) shows the best results compared to pure sewage sludge. The statistical reliability of the results is confirmed by the value of LSD_0.05, which is 0.01–0.02 for all studied indicators. Using STATISTICA 12.0 for Windows, we calculated neural network modeling to predict changes in the substantiated indicators when applying sewage sludge.

The analysis shows a clear upward trend, the dependence is nonlinear, and the predictive curve shows a gradual increase in growth at higher doses. The model demonstrates good prediction accuracy in the range of observed values. The most intensive growth of indicators is observed in the range of 0–30 t.ha⁻¹, but after the application of 30 t.ha⁻¹ of the sewage sludge, the growth rate of indicators slows down. At doses of the sewage sludge above 40 t.ha⁻¹, a slight increase in efficiency is predicted. Further increase in the dose of the sewage sludge is ineffective. This model can be used to plan the application of the norms of the sewage sludge. This analysis demonstrates the effectiveness of using neural network modeling to predict changes in reasonable indicators when applied (Figure 5).

The most intense changes are observed in the range of 0–20 t.ha⁻¹. Further improvement is predicted with an increase in the dose of compost application based on SS up to 40 t.ha⁻¹. The model demonstrates high reliability of predicting the degree of saturation of the basic parameters (Figure 6).

In the variant without fertilization, the hydrolytic acidity content decreased from 3.10 to 2.37 mmol per 100 g of soil in the upper layer and from 3.30 to 2.90 mmol per 100 g of soil in the lower layer. The use of a mineral fertilizer system contributed to a decrease in the content of hydrolytic acidity and amounted to 3.06–3.08 mmol per 100 g of soil (Figure 7).

This relationship can be described by the following linear regression equation:

$$y=0.319x+8.4525$$

(1)

where y is the cation absorption capacity (T), mmol per 100 g of soil;

x—the sum of absorbed bases (S), mmol per 100 g of soil.

The multiple coefficient of determination (R²) was 0.87, which indicates a close relationship between these indicators. The highest rates were provided by the variants of applying the sewage sludge at a dose of 40 t.ha⁻¹, and close in terms of calcium and magnesium content for variants with the introduction of composts based on sewage sludge with straw.

The predicted values for Ca²⁺ at 20 t.ha⁻¹ are 0.30 mmol per 100 g. The modeling characteristic has a linear upward trend. The predicted increase of 104% at the maximum dose of SS 50 t.ha⁻¹ is 0.34 mmol per 100 g (Figure 8). The predicted values for Mg²⁺ have a moderate non-linear increase. The predicted increase is 23.6% at the maximum dose of 50 t.ha⁻¹ of SS. Thus, our neural network model is effective for predicting fertilizer planning. Based on the presented graph of model accuracy (R²) for different parameters, the following analysis can be made. The neural network model demonstrates high forecasting accuracy for all studied indicators, as evidenced by the value of the coefficient of determination R² above 0.85 for all parameters. The highest accuracy of the prediction model is shown by the Ca²⁺ and Mg²⁺ content with a determination coefficient of about 0.95, which indicates a very high reliability of predictions about the dynamics of these elements based on the data. The pH prediction is also characterized by high accuracy with an R² of about 0.93, which ensures that the model is reliable for the predicted changes in the acidity of the substrate when applying the sewage sludge composts. A slightly lower but still high model accuracy is demonstrated by the degree of base saturation (R² ≈ 0.88) and hydrolytic acidity (R² ≈ 0.89). These values indicate a good predictive ability of the model for these parameters (Figure 9).

Disproportionality in the growth of calcium and magnesium content under the influence of different fertilization systems for sylphia pierced-leaved provided a change in the ratio between these cations. In the studies with the cultivation of cereals crops, the content of Ca²⁺ and Mg²⁺ in the soil absorption complex depended on the use of different fertilizer systems. In the control without fertilizers, the calcium content was 2.5 mmol per 100 g of soil in the 0–20 cm soil layer and increased with depth to 3.42 mmol per 100 g of soil. The use of mineral fertilizers increased the calcium content to 3.16–3.82 and 4.09–4.30 mmol per 100 g of soil, respectively, in the upper (0–20 cm) and lower (20–40 cm) soil layers. Under the conditions of the experiment, the ratio changed from 8.93 in the control without fertilizers to 14.38 in the variant where the sewage sludge was applied at a dose of 40 t.ha⁻¹ (

Table 3. Changes in the cation content in sod-podzolic soil (Fluvisols, WRB) after application of the sewage sludge (SS) for grass crops.

Experimental Design	Soil Layer (Depth)	Number of Cation Equivalents					Ca2+ + Mg2	Ca2+ Mg2+
		Ca2+	Mg2+	Na+	K+	∑
	cm	mmol per 100 g of Soil
1. Control variant—without fertilization	0–20	2.50	0.28	0.12	72	74.90	2.78	8.93
	20–40	3.42	0.29	0.10	57	60.81	3.71	11.80
2. Mineral fertilizers—N60P60K60	0–20	3.16	0.30	0.13	59	62.59	3.46	10.53
	20–40	4.09	0.32	0.12	81	85.53	4.41	12.78
3. Mineral fertilizers—N90P90K90	0–20	3.80	0.31	0.14	53	57.25	4.11	12.26
	20–40	4.30	0.33	0.12	90	94.75	4.63	13.03
4. SS—20 t.ha−1+ N50P52K74	0–20	4.17	0.29	0.12	59	63.58	4.46	14.38
	20–40	4.38	0.31	0.12	56	60.81	4.69	14.12
5. SS—30 t.ha−1+ N30P33K66	0–20	4.28	0.32	0.12	68	72.72	4.60	13.38
	20–40	4.45	0.33	0.12	61	65.90	4.78	13.48
6. SS—40 t.ha−1+ N10P14K58	0–20	4.57	0.33	0.13	76	81.03	4.90	13.85
	20–40	4.76	0.36	0.12	72	77.26	5.14	13.28
7. Compost (SS + straw (3:1)) —20 t.ha−1 + N50P16K67	0–20	4.16	0.28	0.12	60	64.56	4.44	14.86
	20–40	4.29	0.28	0.11	57	61.74	4.57	15.32
8. Compost (SS + straw (3:1))—40 t.ha−1 +N30K55	0–20	4.30	0.29	0.12	68	72.71	4.59	14.83
	20–40	4.57	0.29	0.12	60	64.98	4.86	15.76
The least significant difference 0.05 (LSD0.05)	0–20	0.2	0.1	0.1	1	2.1	1.02	1.01
	20–40	0.3	0.1	0.1	2	2.4	1.03	1.01

).

After the application of the sewage sludge in a dose of 40 t.ha⁻¹, the Ca²⁺ content in the soil was 4.57 mmol per 100 g of soil, and the use of composts based on the sewage sludge provided a wide range of Ca²⁺ content at the level of 4.16–4.30 mmol per 100 g of soil (Figure 10).

The Mg²⁺ content was in a much smaller range of values from 0.28 to 0.33 mmol per 100 g of soil and increased with increasing doses of the sewage sludge application. Neural network modeling showed the following characteristics of Ca²⁺ distribution in soil layers when using composts based on sewage sludge. In the upper soil layer (0–20 cm), calcium accumulation is more intense than in the lower layer. The initial values of Ca²⁺ content were 2.5 mmol per 100 g, with a gradual increase to 5.8 mmol per 100g at the maximum dose of compost application of 50 t.ha⁻¹. The predictive model demonstrates high accuracy with a determination coefficient of R² > 0.95, which indicates the reliability of the forecast (Figure 11).

In the lower layer (20–40 cm), the initial calcium content was higher—3.4 mmol per 100g—but the intensity of detecting accumulation was lower. At the maximum dose of compost, the level of 5.5 mmol per 100 g is predicted to be reached. The pattern of accumulation is more stable, with a lower amplitude of fluctuation with the top layer. The analysis of the vertical distribution of calcium shows a gradual equalization of its concentration between the layers with an increase in the dose of compost application. This indicates active migration of the element along the subsoil profile. The optimal application dose is 30–40 t.ha⁻¹, which achieves the most balanced distribution of calcium in both layers.

The predictive model demonstrates high reliability for both subsoil layers, which confirms the close correlation between predicted and actual data. This allows it to be used for fertilizer system planning and predicting changes in calcium content at different doses of sediment-based composts. Neural network modeling demonstrates the peculiarities of Mg²⁺ distribution in different layers of the subsoil when using composts based on sewage sludge.

In the upper soil layer (0–20 cm), each step increases the magnesium content. The initial value is 0.278 mmol per 100g with a predicted increase to 0.293 mmol per 100 g at a maximum compost dose of 50 t.ha⁻¹. The accumulation pattern has a linear trend with a moderate growth rate. The actual values are in good agreement with the predicted ones, which confirms the reliability of the model. In the lower layer (20–40 cm), the pattern is different. The initial magnesium content is 0.288 mmol per 100 g, but with an increase in the dose of compost, a slight decrease is predicted to 0.284 mmol per 100 g at a dose of 50 t.ha⁻¹. The actual data show some variability in the value, especially in the 20–30 t.ha⁻¹ dose range, where there is a temporary increase in magnesium content. Comparative analysis of the two layers reveals opposite trends: while the upper layer shows a gradual accumulation of magnesium, the lower layer tends to slightly decrease its content. This pattern may be related to the peculiarities of magnesium migration along the profile of the base and its interaction with other elements, especially calcium (Figure 12).

The optimal dose of compost application is 20–30 t.ha⁻¹, which ensures the most balanced distribution of magnesium between the layers of the substrate. At higher doses, there is a risk of disruption of the optimal Ca²⁺/Mg²⁺ content, which can negatively affect the justified processes and plant nutrition.

4. Conclusions

The study allowed us to thoroughly evaluate the impact of sewage sludge (SS) and composts based on it on the physicochemical properties of Fluvisols for growing energy crops. The results of the study demonstrate the multifaceted effect of applying different doses of sewage sludge and composts, which is important for sustainable soil management and waste disposal.

It has been experimentally proven that the application of SS and composts based on them significantly improves soil fertility. In particular, there was an increase in pH from 4.8–4.9 to 5.4–6.8 in the upper soil layer and from 4.5–4.9 to 5.2–6.3 in the lower layer, which creates more favorable conditions for the growth of energy crops. The hydrolytic acidity decreased from 3.10–3.30 to 2.80–2.90 mmol per 100 g of soil, which indicates the neutralizing effect of organic fertilizers. A significant result is the positive effect of the SS on the cation exchange capacity, which increased from 11.70–11.75 to 11.80–12.01 mmol per 100 g of soil, and the degree of saturation with bases, which increased from 71.6–72.6% to 74.9–76.7%. The increase in calcium and magnesium content in the soil absorption complex is particularly valuable, as it improves soil structure and provides plants with the necessary nutrients. The effectiveness of the application of SS shows a nonlinear dependence on the doses of application. The most intensive improvement of indicators is observed at doses of 20–30 t.ha⁻¹, and further increase to 40–80 t.ha⁻¹ gives a less pronounced increase. For most physicochemical parameters, the optimal doses are 30–40 t.ha⁻¹, which ensures a balanced distribution of nutrients in the soil profile and the best Ca²⁺/Mg²⁺ ratio.

The use of neural network modeling methods allowed us to predict with high accuracy changes in soil physicochemical parameters under the influence of different doses of SS. The coefficients of determination R² for the developed models are 0.85–0.95, which confirms their reliability and practical value for planning fertilizer systems. The model demonstrated particularly high prediction accuracy for Ca²⁺ and Mg²⁺ content (R² ≈ 0.95) and pH (R² ≈ 0.93).

A comparative analysis of the effectiveness of different forms of organic fertilizers showed that composts based on SS with the addition of plant residues (straw, sawdust) in a ratio of 3:1 provide a more balanced effect on soil properties compared to pure SS. Such composts contribute to the formation of a more stable soil structure, a more even distribution of nutrients across the profile, and reduce the risk of salinization. Of particular value is the established nature of the vertical migration of macronutrients when using the SS. It was found that calcium tends to gradually equalize the concentration between soil layers with an increase in the dose of compost, while magnesium shows opposite trends—accumulation in the upper layer and a slight decrease in the lower layer.

Further research should focus on long-term monitoring of the delayed effects of SS application, assessment of its impact on soil biological indicators, and the development of technologies aimed at minimizing potential risks of heavy metal accumulation and other pollutants. Improving neural network modeling methods to predict the complex impact of SS on the “soil–plant–atmosphere” system, taking into account climate change, is also considered a promising direction.