Prediction of the impact of ecological restoration technology on the restoration of heavy metal pollution in agricultural soil

ABSTRACT

Heavy metal in agricultural soils increases its contamination and infertility in a due course of time. The vast expansion of industrialization along the agricultural soils has led to such adverse effects impacting vegetation. Several existing learning techniques are utilized to analyze the contamination in agricultural soil; however, the learning process consumes a high restoration error rate. This article introduces a Concentration-specific Restoration Prediction Technique (CRPT) for heavy metal-exposed agricultural soils. This technique relies on the heavy metal level present in different soil types for assessing its adverse impact. The vegetation failures and infertility progressions are accounted for by the continuous recurrent learning for predicting its restoration age. The restoration procedures preferred in the earlier stages are used for training the recurrent neural network. Based on the more feasible restoration method, the failure and infertility are validated for achieving a maximum prediction value. The learning process accounts for soil type, crop resistance, and climatic factors for holding restoration procedures. Therefore, the learning process performs restoration prediction along with resistance analysis for progressive restoration prediction. The least possible prediction value is validated using different training inputs for matching outputs. This technique is reliable in detecting pollution possibility with maximum restoration prediction (9.41%) and less training error (14.53%).

KEYWORDS:

• Agricultural soil
• ecology restoration
• heavy metal pollution
• predictive analysis
• recurrent learning

1. Introduction

Heavy metal (HM) pollution detection is an important task to perform in agricultural soil. The actual role is to detect the exact factors and features of HM in agricultural soil. HM pollution reduces the soil fertility range which decreases the plant growth and productivity ratio in farming (Liao et al., 2023). Many methods are used as HM pollution detection process which provides relevant data for the remediation process. Elements such as arsenic (As), mercury (Hg), copper (Cu), lead (Pb), zinc (Zn), and cadmium (Cd) were detected in the given sample of the soil (Ali et al., 2023). A meta-analysis-based detection method is used for HM pollution identification. The meta-analysis first evaluates the characteristics and elements that are presented in agricultural soil (X. Wang et al., 2023). Meta-analysis is a statistical method employed to merge and examine data from numerous autonomous investigations on a specific subject. In the area of heavy metal pollution detection, meta-analysis detection involves the utilization of this approach to combine and consolidate findings from diverse research pertaining to the identification of heavy metal pollution. The meta-analysis produces feasible data for HM pollution that reduces the latency in the detection process. It predicts the exact characteristics of elements that provide useful information for the HM pollution control development process (Sellami et al., 2022). A machine learning (ML) based HM pollution detection framework is used for agricultural development systems. The ML algorithm is mainly used to increase the accuracy of the HM detection process. A factor identification technique is used in the framework to evaluate the spatial and temporal features of the soil. The ML-based framework improves the performance range of the soil pollution control process (Ge et al., 2021; Shammi et al., 2021).

Ecological restoration is a process that initiates or recreates the recovery of an ecosystem. Ecological restoration involves an intentional and purposeful series of activities and interventions with the goal of reviving or reconstructing a degraded, damaged, or altered ecosystem. Ecosystems are susceptible to a range of hazards, including human activities, climate change, pollution, and natural catastrophes. These hazards can result in a reduction in biodiversity, disturbance of ecological processes, and general dysfunction of the ecosystem. Ecological restoration is a methodical and scientifically guided approach that aims to imitate natural processes and expedite the recuperation of an ecosystem. These operations encompass reforestation, habitat restoration, soil protection, reintroduction of native species, and eradication of invasive species. The objective is to bolster ecological resilience, facilitate the resurgence of indigenous plant and animal species, and reinstate a harmonious and enduring ecosystem. Ecological restoration is used in agricultural soil to enhance the productivity range of the crops (C. Jiang et al., 2021). An integrated evaluation framework using a multi-scenario trade-off is used for ecological restoration. The actual carbon storage, nitrogen retention, water conservation, and habitat quality are evaluated in the framework (Tao et al., 2022). The integrated framework provides quantified services to improve the fertility level of agricultural soil. The integrated framework provides necessary information for ecological restoration and protection services (Ou et al., 2021). An improved distribution method is used to identify the rate of earth elements (REE) on agricultural soil. The detection method is mainly used to detect the heavy metal elements that damage the soil (Zhang et al., 2022). The REE contains both harmful and useful elements which are used for ecological restoration. The REE is mainly detected for risk assessment which improves the lifespan of plants and human beings (Hassan et al., 2021). A facile strategy is used for ecological restoration in agricultural soil. The facile strategy identifies the HMs that are presented in the soil for the restoration process. The facile strategy improves the feasibility and significance level of ecological restoration systems (R. Jiang et al., 2023).

Machine learning (ML) algorithms and techniques are commonly used for the detection and identification process. ML-based land restoration methods are used in agricultural soil (Ang & Seng, 2021). A big data-based method is used for the ecological land restoration process. Both multispectral and hyperspectral information for agricultural soil is analyzed from big data (Wen et al., 2022). ML algorithm detects the necessary information that is relevant for the restoration process. The big data analysis process provides optimal data for yield forecasting, which minimizes the complexity of the agricultural improvement process. ML algorithm detects promising information for restoration that reduces the latency in land restoration development systems (Mentese et al., 2021; Zhao et al., 2022). Deep neural network (DNN) algorithm-based land restoration methods are used for agricultural soil. The DNN algorithm predicts the characteristics and elements that are presented in the sample soil. The predicted elements provide effective data for the sustainability improvement range of the soil. The DNN-based method balances the ecological aspects and land restoration process for agricultural purposes (Cui et al., 2021; T. Shi et al., 2022). Various researchers utilize different machine learning and classification techniques to identify heavy metal pollution. Here, a few researchers' opinions are discussed to understand the contamination prediction process. S. Yang et al. (2021). proposed a synthesis framework based on machine learning (ML) for heavy metal pollution. The main aim of the framework is to identify the major cause of air pollution in agricultural soil. A factor identification technique is used here to analyze the traffic and industrial activities of heavy metals. The identification technique minimizes the latency in the computation process. The proposed framework improves the accuracy of the soil pollution control process.

Liu et al. (2023). introduced a spatial distribution-principal component analysis (SD-PCA) model for heavy metal pollution identification in soil. The introduced SD-PCA model is mainly used to assess the heavy metal pollution in soil. It also identifies the potential sources and factors of soil for the improvement process. The introduced SD-PCA model enhances the overall performance and feasibility range of the systems. The study focuses on tackling soil contamination caused by heavy metals by utilizing a Spatial Distribution – Principal Component Analysis (SD-PCA) model. The research seeks to address deficiencies in current literature by introducing a novel approach for evaluating soil contamination. The SD-PCA model is selected for its spatial analysis capabilities, providing benefits in comprehending the dispersion of heavy metals in the environment. The findings offer useful insights into the scope and patterns of soil pollution, helping to both methodological progress and environmental management techniques.

F. Wang et al. (2023). developed a hybrid framework for delineating the migration route of heavy metals in soil. It is commonly used for intermediate controlling processes that control the metal pollution level in soil. The developed framework identifies the heavy metal similarities and functions that cause damage to the soil. The exact main source of the metals is also detected for further processes. The developed framework improves the accuracy of route migration in heavy soil.

Deng et al. (2023). designed a stepwise multiple linear regression (SMLR) and partial least squares regression (PLS) (MLP-PLS) based model for heavy metal distribution in agricultural soil. Remote sensing data is used in the model which produces accurate information for the prediction process. It is commonly used in low-contaminated areas for the soil monitoring process. The SMLR model improves the feasibility range of statistical analysis. The designed model predicts the heavy soil distribution ratio in the soil. The study examines the variables that impact the dispersion of high levels of heavy metals and metalloids in agricultural soils through the utilization of diverse data sources in their research. The research focuses on filling a significant knowledge gap regarding the complex dynamics of soil contamination. Utilizing several data sources enables a thorough examination, offering valuable understanding of the intricate relationships influencing the occurrence of heavy metals and metalloids in agricultural settings. The results enhance our comprehension of the intricate relationship between environmental elements and their influence on soil quality, and they have significant implications for the implementation of sustainable farming methods.

Mahvi et al. (2022). proposed a new heavy metal pollution and status detection method for soil. The actual goal of the method is to analyze the metal pollution category in soil. The ecological risk assessment and contamination indices are used in the method which provides feasible data for the detection process. It provides effective services to the soil improvement process. The proposed method increases the accuracy of metal pollution assessment which enhances the performance level of agricultural soil.

Chen et al. (2023). developed soil multifunctionality (MF) detection during ecological restoration. The exact driving forces and sources of MF are identified for climate regulation and forecasting processes. Both bacterial and fungal diversity of ecological restoration are verified for soil improvement. It minimizes the significant changes that occur in agricultural soil. The developed method reduces the overall network complexity ratio in the computation process.

Li et al. (2023). introduced an ecological conservation and restoration area (ECRA) identification framework. The main aim of the framework is to enhance the carbon storage ratio in ecological systems. It also evaluates the ecological priority scenario (EPS) to provide relevant carbon sources that reduce the latency in the detection process. The introduced framework increases the accuracy of ECRA detection which improves the efficiency level of the systems.

He and Shi (2022). designed a remote sensing technology-based geo-detector model for ecological degradation. It evaluates the impacts of anthropogenic factors on the development process. It provides relevant data for climate change forecasting that reduces the energy consumption in the computation process. The key factors that cause damage to the environment are identified to improve ecological degradation and restoration. The designed model enhances the sustainability range of environment management systems.

Lu et al. (2022). proposed a long-term metal pollution identification method for denitrification in agricultural soil. The proposed method is mostly used in recycling areas and mining areas which detects the relationship between the nitrification and denitrification interaction processes. The microbial functional profiles of heavy metals are identified from agricultural soil samples. The proposed method increases the accuracy of denitrification which improves the quality of agricultural soil.

H. Shi et al. (2023). developed a comprehensive framework for factor identification in soil. A geo-detector is used to detect the heavy metal sources in the sample. It is used as a digital elevation model which produces feasible information for environmental management. The developed framework evaluates the pollution level of soil for agricultural development and improvement processes. The developed framework improves the accuracy of the pollution source detection process.

L. Yang et al. (2022). proposed a geographically weighted regression and Principal Component Analysis (GWR-PCA) based spatial determinants detection. GWR is used to detect the relationship among the soil variables. The proposed method also identifies the heavy metal pollution range of the soil for the agricultural development process. It significantly improves the performance range of farming areas. Experimental results show that the proposed increases the accuracy of the determinant detection process. The study examines the factors that influence the distribution of heavy metals contamination in various agricultural soils by using geographically weighted regression (GWR). The study tries to uncover the specific elements that influence heavy metal contamination by using a modeling technique that takes into account spatial information. The research enhances our comprehension of the spatial heterogeneity in soil contamination, providing valuable insights into the distinctive attributes of various agricultural areas. The results have consequences for focused environmental management tactics, emphasizing the significance of taking into account regional variability when dealing with heavy metals contamination in agricultural areas.

Sun et al. (2023). introduced a long-term heavy metal (HM) accumulation for agricultural soil. The introduced method identifies the cumulative HM pollution range in soil. Copper, arsenic, zinc, and lead content which are presented in the soil are identified. It also decreases the overall time-consuming ratio in the computation process. It is mostly used for the fertilization improvement process. The introduced method provides effective information for agricultural applications.

J. Yang et al. (2022). proposed a risk assessment approach for heavy metal pollution in soil. The proposed approach uses a positive matrix factorization (PMF) model and is used in the approach to detect the priority resources of the soil. The worst ecological risks are identified and estimated to reduce the severe damage in agricultural soil. The proposed approach improves the significance and feasibility range of the agricultural soil.

Guo et al. (2023). developed a machine learning (ML) based heavy metal (HM) immobilization in agricultural soil. The main aim of the model is to predict the HM ratio of soil which reduces the agricultural performance range. The developed method identifies the critical factors and patterns of immobilization features. The developed method increases the performance and reliability level of the soil.

Hao et al. (2023). introduced an ensemble learning-based HM concentration (HMC) prediction (EL-HMC) method. A random forest (RF) algorithm is used in the method to evaluate the HM factors of the soil. It also analyzes the error level which causes damage to the soil. The introduced EL-HMC method improves the accuracy of HMC prediction which enhances the efficiency of environmental management. Even though the existing methods effectively analyze soil contamination, traditional techniques consume a high error rate while examining the large volume of data. The error rate leads to a reduction in the prediction rate and complexity. The research difficulties are overcome by applying the Concentration-specific Restoration Prediction Technique (CRPT). The CRPT techniques successfully predict heavy metal contamination in agricultural soil. The overall objective of this study is listed as follows.

Objectives

• To design a novel concentration-specific restoration prediction technique for pollution detection and technique recommendation for agricultural soils exposed to heavy metals

• To incorporate a recurrent neural network-based feasible restoration technique detection based on vegetation and climatic characteristics with resistance features

• To perform a data and metric-based analysis for validating the proposed technique’s efficacy and verification based on the descriptions.

Then, the overall manuscript is organized as follows: Section 2 analyzes the problem definition; section 3 discusses the working process of CRPT. Section 4 describes the vegetation impact, and Section 5 discusses the restoration prediction process. The conclusion is described in section 6.

2. Description

2.1. Problem description

The problem of heavy metals in agricultural soil reduces fertility and productivity due to long-term deposits. Predicting such pollution is tedious due to data availability, concentration, and recommendations that impact the scenario. Considering such lagging features, the methods discussed above are validated using soil, climate, and vegetation impacts. The joint validation is reduced due to complex system and analytical requirements. Therefore, precise data handling feature is required for improving the validity and precision of predicted and restoration abilities. Thus, considering these adverse effects, the proposed technique performs a sustainable predictive analysis for handling multiple heavy metal impacts over agricultural soils.

2.2. Data description

The reference (Okereafor et al., 2019) is used in this article for evaluating the CRPT’s improvement over the prediction process. This reference provides heavy metals such as Manganese$\left(\mathit{Mn}\right)$, Chromium (Cr), lead (Pb), Cobalt (Co), Arsenic (As), Cadmium (Cd), Nickel (Ni), and Zinc (Zn) deposits over the farms surrounding gold mines. This information is extracted from the mine dumps in Ekurhuleni, South Africa. The common metal concentration presented in this environment is tabulated in Table 1.

Table 1. Summary statistics of metal concentrations.

Composition	Locations	Number of Metals	Min. Concentration	Max. Concentration
Sand	8	Cd, Cr, Zn	<6	>7.1
Silt	10	Cd, Pb, Mn	<25	>0.1
Clay	5	Cu, Co	<25	>50
Sand + Silt	14	Cu, Cr	<25	>60
Sand + Clay	9	Cd, Zn	<40	>75
Silt + Clay	11	Pb, As	<40	>200
Sand+ Silt + Clay	4	Zn, Cr, Pb	<90	>200

The above concentration table illustrates the order of heavy metals present in agricultural soils. Based on the concentration, the metals are listed in their descending order as Cr-Al-As-Fe-Pb-Co-Ni-Ti-Cd-Zn-Cu. From the considered factors (for vegetation): soil infertility and vegetation failures, the restoration procedures are tabulated in Table 2. These procedures are identified from the references (Yi et al., 2017) along with their effects.

Table 2. Restoration procedures and effects.

Restoration	Metals	Procedure	Effects
Sepiolite	Cd	Reduced installments from 55% to 14%	Positive
Sepiolite + Limestone	Cd, Pb, Zn, Cu	Reduced Installments from 19.5% to 2.2%	Positive for Pb and Cd
Phosphate	Cd	Different degrees	Effective for Cd at limited degrees
Palygorskite	Cd, Pb, Zn, Cu	NA	Positive for Cd compared to others
Attapulgite	Cu, Cd, Zn	NA	Controlled Cu
Bentonite	Pb, Zn, Cd	Immobilization Rate	Weak effect on Pb

The above tabulations are used for validating the predictive restoration ages. Therefore, the recurrent neural process is split under two layers for procedure analysis (given above) and vegetation resistance, respectively. The congruent process identifies the prediction for their reliability across various restoration ages.

3. Concentration-specific restoration prediction technique

The proposed restoration prediction technique is designed to control heavy metal pollution and impacts on the agricultural soils in a smart farm. In this smart farming scenario, the adverse effects impacting vegetation are detected that are predicted through the proposed CRP technique. The heavy metal level that occurs in different soil types is addressed to reduce its impact. The heavy metal exposure analysis is performed to identify the vegetation failures and infertility progressions by continuous recurrent neural network learning. Vegetation failures occur when plants or crops exhibit below-average growth or productivity, deviating from anticipated results. Within ecological contexts, this phenomenon may arise as a consequence of variables such as climate change or the introduction of exotic species. In the field of agriculture, factors such as substandard soil quality, pest infestations, or insufficient farming techniques might contribute to these failures. Comprehending the reasons for vegetation failures is essential in both natural ecosystems and cultivated landscapes. This knowledge is vital for applying efficient ways to improve plant health and foster sustainable growth. The evolution of infertility entails a systematic evaluation and comprehension of the elements that impede a couple’s capacity to achieve conception. Initial assessments frequently prioritize fundamental aspects such as ovulation, sperm quality, and reproductive architecture. As the process progresses, more sophisticated diagnostic tests may be utilized to investigate hormone imbalances, hereditary causes, or structural abnormalities. The knowledge acquired from these stages informs individualized treatment strategies, encompassing modifications in lifestyle and pharmaceutical interventions, as well as the utilization of assisted reproductive technology. This promotes a holistic approach to tackling infertility.

Figure 1 portrays the proposed technique with suitable representations.

Figure 1. Portray of the proposed technique.

The restoration method is used to compute the failure and infertility for satisfying a maximum prediction value. The recurrent learning process accounts for holding restoration procedures and relies on crop resistance, soil types, and climatic factors along with resistance analysis for precise progressive restoration prediction. Crop resistance pertains to the capacity of plants to endure and endure numerous stressors, like pests, diseases, or environmental circumstances. Enhancing agricultural output and sustainability, it is an essential characteristic in agriculture. Plant breeders frequently employ selective breeding or genetic modification techniques to cultivate resistant cultivars, with the objective of augmenting the plant’s inherent defense mechanisms. Resistant crops are crucial for maintaining food security and decreasing dependence on chemical inputs for pest and disease management. This data is observed from the agricultural soils for validating the restoration procedures and resistance. Restoration techniques involve a wide range of activities that are intended to rejuvenate, rehabilitate, or improve the natural or constructed environment. Restoration in ecosystems entails the reintroduction of indigenous species, management of exotic plants, and the rehabilitation of habitats. Within the realm of cultural heritage, restoration processes involve the preservation and repair of structures, artworks, or artifacts in order to return them to their original or intended condition. In every given situation, successful restoration operations generally involve comprehensive evaluations, meticulous strategizing, and the implementation of suitable methods to attain enduring and visually appealing results. In the proposed technique, precise impact detection and problem prediction are validated using the different training inputs for matching outputs with already existing methods. This heavy metal in agricultural soils increases its infertility and contamination in a due course of time, this impact leads to vegetation failure and infertility progressions that reduce the crop growth and augment heavy metal pollution. These adverse effects impacting vegetation are controlled through Layer 1 and Layer 2 in recurrent learning. The procedure for restoration prediction is performed using soil type, climatic factors, and crop resistance. The CRPT operates between agricultural soils and restoration procedures. The impact on the soils is identified through the heavy metal exposure analysis, where ecology restoration prediction is made. The restoration prediction for controlling heavy metal pollution in different soils is pursued using the procedures preferred in the earlier stages are aided for training the learning process. Heavy metal pollution increases due to high amounts of metallic substances in soils like cadmium, lead, arsenic, mercury, chromium, sulfuric acid, etc. The heavy metals present in the agricultural soil $\mathit{Hv}{y}_m$ analyzed for predicting pollution possibility with maximum restoration predictions and then appropriate functions are performed to prevent the impacts and failures. Hence, the proposed technique is modeled into two segments namely vegetation impact detection and restoration prediction.

4. Vegetation impact detection

The heavy metal present in agricultural soil is accounted for by addressing the impact prediction using the recurrent learning process. The input can be of any type related to soil type, crop resistance, weather conditions, etc. In this analysis, the information observed from the agricultural soil$\left(\mathit{Soi}{l}_{\mathit{data}}\right)$ is computed as

(1) $\mathit{Soi}{l}_{\mathit{data}}=\sum _{\mathit{i}=0}^n\frac{\left[Max\left(Hv{y}_m\right)-\mathit{Min}\left(\mathit{Hv}{y}_m\right)\right]}{\mathit{IN}{F}_l+\mathit{Ve}{g}_f}+r_{\mathit{age}}$(1)

And,

(2) $\mathrm{\exists }=\frac{1}{2\mathit{\pi }}\left[\frac{\left(\frac{Min\left(Hv{y}_m\right)}{\mathit{Max}\left(\mathit{Hv}{y}_m\right)}-\frac{P}{Q}\right)}{2\left(\mathit{Soi}{l}_{\mathit{data}}-\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}\right)}\right]$(2)

Where the variables $\mathit{P}$ and $\mathit{Q}$ used to represent the current restoration method and the already existing method. Therefore, $\mathit{P}\in \mathit{Q}$ is defined for identifying heavy metal-exposed agricultural soils. If $\mathit{Max}\left(\mathit{Hv}{y}_m\right)$ and $\mathit{Min}\left(\mathit{Hv}{y}_m\right)$ are the maximum and minimum heavy metal presence observed in different soil types. The variable $\mathrm{\exists }$ and $\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}$ are used to denote the reliability of impact detection and previous data observed. The reliable impact detection is validated as the presence of heavy metals observed in different $\mathit{Soi}{l}_{\mathit{data}}$ observations. The minimum and maximum metal pollution experienced in the referenced location is pictured in Figure 2.

Figure 2. Minimum and maximum metal pollution in the referenced location.

The above illustration presents the observed pollution values from 2017 to 2022 from the given data references. The classification of $\mathit{Max}\left(\mathit{HV}{y}_m\right)$ and $\mathit{Min}\left(\mathit{HV}{Y}_m\right)$ is distinguished using its impact. This impact is validated by the mg of heavy metal present per kilogram of the minerals in the soil. Based on the mg/Kg the metal concentration is divided as nil, moderate, heavy, and last known value. These are the complete $\mathit{Soi}{l}_{\mathit{data}}$ and $\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}$ acquired from the recent years (Figure 2). If $r_{\mathit{age}}$ means restoration age of crops and $\mathit{n}$ means different types of soils. There are some cases of infertility progressions and vegetation failures occurring in agricultural soils due to heavy metal pollution. Therefore, these pollutions impact the agricultural soils at any period, for which the normalization is performed as

(3) $\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)=\frac{Rl{b}^{im{p}_d}}{{\left(\frac{P}{Q}-\mathrm{\nabla }\right)}^2}$(3)

Where,

(4) $\mathrm{\nabla }=\frac{1}{Q}\sqrt{\frac{1}{\mathit{P}-1}\sum _{\mathit{i}=1}^Q{\left(\frac{\mathit{Soi}{\mathit{l}}_{\mathit{data}}-\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}{\mathit{IN}{F}_l+\mathit{Ve}{g}_f}\right)}^2}$(4)

Based on Equations (3)(3) $\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)=\frac{Rl{b}^{im{p}_d}}{{\left(\frac{P}{Q}-\mathrm{\nabla }\right)}^2}$(3) and (4(4) $\mathrm{\nabla }=\frac{1}{Q}\sqrt{\frac{1}{\mathit{P}-1}\sum _{\mathit{i}=1}^Q{\left(\frac{\mathit{Soi}{\mathit{l}}_{\mathit{data}}-\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}{\mathit{IN}{F}_l+\mathit{Ve}{g}_f}\right)}^2}$(4) ), the normalization of $\mathit{Soi}{l}_{\mathit{data}}$ follows the maximum impact prediction value$\left(\mathit{\alpha }\right)$ and the standard restoration prediction value$\left(\mathrm{\nabla }\right)$ for improving successful vegetation. Here$\mathrm{\nabla }$ is a normalized measure, whereas$\mathit{\alpha }$ is the reliable prediction value, for which the appropriate computation of is observed at different periods. Based on $\mathit{Soi}{l}_{\mathit{data}}$ and $\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)$, the continuous impact prediction $\mathrm{\exists }\left[\mathit{Soi}{l}_{\mathit{data}},\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\right]$ is validated as

(5) $\left.\begin{array}{c}\mathrm{\exists }\left[\mathit{Soi}{l}_{\mathit{data}},\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\right]=\left[\frac{\mathcal{N}\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}{\mathit{Soi}{l}_{\mathit{data}}-\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}\right]12-\mathit{IN}{F}_l+\mathit{Ve}{g}_f\\ +\left[\frac{\mathcal{N}\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}{\mathit{Soi}{l}_{\mathit{data}}-\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}\right]22-\mathit{IN}{F}_l+\mathit{Ve}{g}_f+⋮\\ +{\left[\left(1-\frac{\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}{\mathit{Soi}{l}_{\mathit{data}}}\right)\mathit{\alpha }\right]}_{\mathit{P}}^2-\mathit{IN}{F}_l+\mathit{Ve}{g}_f\end{array}\right\},\mathit{P\epsilon Q}$(5)

Equation (5)(5) $\left.\begin{array}{c}\mathrm{\exists }\left[\mathit{Soi}{l}_{\mathit{data}},\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\right]=\left[\frac{\mathcal{N}\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}{\mathit{Soi}{l}_{\mathit{data}}-\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}\right]12-\mathit{IN}{F}_l+\mathit{Ve}{g}_f\\ +\left[\frac{\mathcal{N}\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}{\mathit{Soi}{l}_{\mathit{data}}-\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}\right]22-\mathit{IN}{F}_l+\mathit{Ve}{g}_f+⋮\\ +{\left[\left(1-\frac{\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}{\mathit{Soi}{l}_{\mathit{data}}}\right)\mathit{\alpha }\right]}_{\mathit{P}}^2-\mathit{IN}{F}_l+\mathit{Ve}{g}_f\end{array}\right\},\mathit{P\epsilon Q}$(5) computes the continuous reliable impact prediction until the current method is used to satisfy the maximum impact prediction value and restoration prediction pursued using the different agricultural soils observation. The current restoration procedure handling the impact reduction depends on the course of time until the impact is predicted from the soils. The above sequence of feasible restoration methods is analyzed using recurrent neural network learning. In a smart farm scenario, the information observed from the agricultural soils is analyzed for a heavy metal occurrence that must be distributed in appropriate and accurate time instances to improve synchronized working without any adverse effects. Besides, the present restoration method is to be instantaneous to meet the vegetation resistance or crop requirements. Therefore, connected recurrent neural network learning (RNN) and the proposed technique are used for $\mathrm{\exists }$ assessment. The RNN output is used to identify and segregate the heavy metal-exposed agricultural soils through $\mathit{Soi}{l}_{\mathit{data}}$ validation and $\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}$ based training for achieving a maximum impact and restoration prediction value. The first step of RNN learning is to sample the continuous $\mathit{Soi}{l}_{\mathit{data}}$ analysis, if the reliable impact detection is made. The convergence of satisfying$\left(1-\frac{\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}}{\mathit{Soi}{l}_{\mathit{data}}}\right)\mathit{\alpha }$ with the already existing restoration method is the beneficial output for predicting the impact and resistance. The analysis of the different types of agricultural soil data with two layers namely Layer 1 (Restoration procedure prediction) and Layer 2 (Resistance analysis) for detecting pollution possibility. The restoration procedures preferred in the earlier stages are used for training the recurrent neural network learning and the heavy metal presence and $\mathrm{\exists }$ are the additional metrics for identifying the impacts. For handling restoration procedures, Layer 1 and Layer 2 are used to predict the precise impact and failures. The impact prediction between $\mathit{Soi}{l}_{\mathit{data}}$ and $\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)$ is pursued using the observation of its contamination and course of time. Based on the Equation (1)(1) $\mathit{Soi}{l}_{\mathit{data}}=\sum _{\mathit{i}=0}^n\frac{\left[Max\left(Hv{y}_m\right)-\mathit{Min}\left(\mathit{Hv}{y}_m\right)\right]}{\mathit{IN}{F}_l+\mathit{Ve}{g}_f}+r_{\mathit{age}}$(1) , the condition$\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)>\mathit{Soi}{l}_{\mathit{data}}$ generates less impact from the different types of agricultural soils. The time-mapping for the vast expansion of industrialization and the continuous $\mathrm{\exists }\left[\mathit{Soi}{l}_{\mathit{data}},\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\right]$ based on$\left(\mathit{n}\times \mathit{CT}\right)$ are the verifying conditions for restoration prediction

(6) $\mathit{CT}=\sum _{\mathit{i}=1}^n\frac{{\left(Rl{b}^{im{p}_d}\right)}_{\mathit{i}}}{P_i}$(6)

And,

(7) $\mathit{IN}{F}_l=\frac{\mathit{CT}}{\left(n-Ve{g}_f\right)}-\left(\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)-\mathit{IN}{F}_l\right)$(7)

In the above equation$\mathit{CT}$ and $\mathit{IN}{F}_l$ variables represent the contamination and continuous infertility in agricultural soils in a due course of time. From the Equations (6)(6) $\mathit{CT}=\sum _{\mathit{i}=1}^n\frac{{\left(Rl{b}^{im{p}_d}\right)}_{\mathit{i}}}{P_i}$(6) and (7(7) $\mathit{IN}{F}_l=\frac{\mathit{CT}}{\left(n-Ve{g}_f\right)}-\left(\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)-\mathit{IN}{F}_l\right)$(7) ), the more feasible restoration method$\left(\mathit{RS}{T}_M\right)$ is computed for training the recurrent neural network at different times$\mathit{t}$, This computation is pursued for detecting the conditions $\mathit{n}\ne 0$ and $\mathit{n}=0$ at different $\mathit{t}$ periods using recurrent analysis. The recurrent analysis relies on $\mathit{Soi}{l}_{\mathit{data}}$ and $\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)$ such that $\mathit{RS}{T}_M$ is determined in the Layer 1 output$\left(\mathit{L}{1}^O\right)$. The linear solution of $\mathit{\alpha }$ $\mathrm{\nabla }$ is the detection of heavy metal pollution possibility for maximizing$\left(\mathit{n}\times \mathit{CT}\right)$. The$\mathit{L}1$ process for restoration procedure assessment is portrayed in Figure 3.

Figure 3. Portray of $\mathit{L}1$ process for restoration assessment.

The alternate minerals and their composition with different concentrations are used for restoration. This restoration is analyzed for$\mathit{\alpha }$ under different $\mathrm{\exists }\left[\mathit{Soi}{l}_{\mathit{data}},\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\right]$. Depending on the procedures the$\mathit{\alpha }$ and$\mathrm{\nabla }$ balance are validated. If $\mathrm{\nabla }>\mathit{\alpha }$ then restoration age is estimated otherwise, the contamination rate is estimated. This contamination rate is validated as $\mathit{mg}/\mathit{Kg}$ for which the classifications portrayed in Figure 2 are performed. If any of the above (sample) procedures address $\mathrm{\nabla }>\mathit{\alpha }$ condition, then training from other methods is performed in $L^{\prime }$ (Refer to Figure 3). The $\mathit{L}{1}^O$ and Layer 2 output$\left(\mathit{L}{2}^O\right)$ are crucial in determining $\mathit{RS}{T}_M$ along with detecting pollution possibilities. The inputs for the resistance analysis for progressive restoration prediction using the learning process are $\mathit{IN}{F}_l$ for both $\mathit{Soi}{l}_{\mathit{data}}$ and $\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)$ that different mappings. The RNN learning process for both the restoration procedure and resistance analysis is performed based on the conditions $\mathit{n}\ne 0$,$\mathit{IN}{F}_l=\left(\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)-\mathit{Ve}{g}_f\right)$ and$\mathit{CT}$. If heavy metal is present in the different types of agricultural soils outputs in $1$ or else$0$. The output of Layer 1 and Layer 2 generates a linear output, whereas the accurate impact prediction is made. The Equations (8)(8) $\left.\begin{array}{c}\mathit{L}{1}^{O^1}=n_1(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1)\\ \mathit{L}{1}^{O^2}=n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2)-\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_1-\mathit{Ve}{g}_{f_1}\\ \mathit{L}{1}^{O^3}=n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3)-\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_2-\mathit{Ve}{g}_{f_2}\\ \begin{array}{c}⋮\\ \mathit{L}{1}^{O^t}=n_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t)-\mathit{N}(\mathit{Soi}{l}_{\mathit{data}}{)}_t-\mathit{Ve}{g}_{f_t}\end{array}\end{array}\right\}$(8) and (9(9) $\begin{array}{c}\mathit{L}{2}^{O^1}=\mathit{L}{1}^{O^1}\\ \mathit{L}{2}^{O^2}=\mathit{L}{1}^{O^2}-p_1{\alpha }_1\\ \mathit{L}{2}^{O^3}=\mathit{L}{1}^{O^3}-p_2{\alpha }_2\\ \begin{array}{c}⋮\\ \mathit{L}{2}^{O^t}=\mathit{L}{1}^{O^t}-p_t{\alpha }_t\end{array}\end{array}∥\left.\begin{array}{c}\mathit{L}{2}^{O^1}=n_1(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1{\alpha }_1)\\ \mathit{L}{2}^{O^2}=n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2{\alpha }_2)-\mathit{IN}{F}_{l_1}+\mathit{Ve}{g}_{f_1}\\ \mathit{L}{2}^{O^3}=n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3{\alpha }_3)-\mathit{IN}{F}_{l_2}+\mathit{Ve}{g}_{f_2}\\ ⋮\\ \mathit{L}{2}^{O^t}=n_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t{\alpha }_t)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}\end{array}\right\}$(9) ) compute the resistance analysis for progressive restoration prediction. The computations are performed for both the estimation of impact detection and prediction in any $\mathit{t}$ interval. Therefore, the outputs are required for the resistance analysis. In the Layer 1, the identified impact $\mathit{Im}{p}_d$ serves as an input, post the detection of heavy metal exposed in agricultural soil is computed as

(8) $\left.\begin{array}{c}\mathit{L}{1}^{O^1}=n_1(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1)\\ \mathit{L}{1}^{O^2}=n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2)-\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_1-\mathit{Ve}{g}_{f_1}\\ \mathit{L}{1}^{O^3}=n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3)-\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_2-\mathit{Ve}{g}_{f_2}\\ \begin{array}{c}⋮\\ \mathit{L}{1}^{O^t}=n_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t)-\mathit{N}(\mathit{Soi}{l}_{\mathit{data}}{)}_t-\mathit{Ve}{g}_{f_t}\end{array}\end{array}\right\}$(8)

(9) $\begin{array}{c}\mathit{L}{2}^{O^1}=\mathit{L}{1}^{O^1}\\ \mathit{L}{2}^{O^2}=\mathit{L}{1}^{O^2}-p_1{\alpha }_1\\ \mathit{L}{2}^{O^3}=\mathit{L}{1}^{O^3}-p_2{\alpha }_2\\ \begin{array}{c}⋮\\ \mathit{L}{2}^{O^t}=\mathit{L}{1}^{O^t}-p_t{\alpha }_t\end{array}\end{array}∥\left.\begin{array}{c}\mathit{L}{2}^{O^1}=n_1(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1{\alpha }_1)\\ \mathit{L}{2}^{O^2}=n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2{\alpha }_2)-\mathit{IN}{F}_{l_1}+\mathit{Ve}{g}_{f_1}\\ \mathit{L}{2}^{O^3}=n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3{\alpha }_3)-\mathit{IN}{F}_{l_2}+\mathit{Ve}{g}_{f_2}\\ ⋮\\ \mathit{L}{2}^{O^t}=n_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t{\alpha }_t)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}\end{array}\right\}$(9)

From the above Equations (8)(8) $\left.\begin{array}{c}\mathit{L}{1}^{O^1}=n_1(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1)\\ \mathit{L}{1}^{O^2}=n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2)-\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_1-\mathit{Ve}{g}_{f_1}\\ \mathit{L}{1}^{O^3}=n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3)-\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_2-\mathit{Ve}{g}_{f_2}\\ \begin{array}{c}⋮\\ \mathit{L}{1}^{O^t}=n_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t)-\mathit{N}(\mathit{Soi}{l}_{\mathit{data}}{)}_t-\mathit{Ve}{g}_{f_t}\end{array}\end{array}\right\}$(8) and (9(9) $\begin{array}{c}\mathit{L}{2}^{O^1}=\mathit{L}{1}^{O^1}\\ \mathit{L}{2}^{O^2}=\mathit{L}{1}^{O^2}-p_1{\alpha }_1\\ \mathit{L}{2}^{O^3}=\mathit{L}{1}^{O^3}-p_2{\alpha }_2\\ \begin{array}{c}⋮\\ \mathit{L}{2}^{O^t}=\mathit{L}{1}^{O^t}-p_t{\alpha }_t\end{array}\end{array}∥\left.\begin{array}{c}\mathit{L}{2}^{O^1}=n_1(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1{\alpha }_1)\\ \mathit{L}{2}^{O^2}=n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2{\alpha }_2)-\mathit{IN}{F}_{l_1}+\mathit{Ve}{g}_{f_1}\\ \mathit{L}{2}^{O^3}=n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3{\alpha }_3)-\mathit{IN}{F}_{l_2}+\mathit{Ve}{g}_{f_2}\\ ⋮\\ \mathit{L}{2}^{O^t}=n_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t{\alpha }_t)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}\end{array}\right\}$(9) ), the linear solution is made for detecting heavy metal pollution in different agricultural soils, and if $\mathit{Im}{p}_d=0$, then $\mathit{RS}{T}_M=1$. Hence, the condition $\mathit{L}{2}^{O^t}=n_t\left(\mathit{IN}{F}_{l_t}\text{break}+P_t{\mathrm{\nabla }}_t{\mathit{\alpha }}_t\right)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}$ is used for satisfying optimal output for predicting impacts from the instance. Therefore, the reliable restoration method aided for failure and infertility validation is retained at$1$. In Layer 1 store the outputs of $\mathit{RS}{T}_M$,$\mathit{Soi}{l}_{\mathit{data}}$ and $\mathit{P}$ at different periods. Based on the output, this restoration procedure is analyzed with the existing method by continuous recurrent learning. Instead, the Layer 1 and Layer 2 outputs are computed as in Equations (10)(10) $\left.\begin{array}{c}\mathit{L}{1}^{O^1}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_1\\ \mathit{L}{1}^{O^2}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_2+\mathit{Im}{p}_{d_1}\\ \begin{array}{c}\mathit{L}{1}^{O^3}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_3+\mathit{Im}{p}_{d_2}\\ ⋮\\ \mathit{L}{1}^{O^t}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_{\mathit{t}}+\mathit{Im}{p}_{d_t}\end{array}\end{array}\right\}$(10) and (11(11) $\left.\begin{array}{c}\mathit{L}{2}^{O^1}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_1-n_{l_1}+(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1{\alpha }_1)\\ \mathit{L}{2}^{O^2}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_2+Im{p}_{d_1}-n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2{\alpha }_2)-\mathit{IN}{F}_{l_1}+\mathit{Ve}{g}_{f_1}\\ \mathit{L}{2}^{O^3}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_3+Im{p}_{d_2}-n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3{\alpha }_3)-\mathit{IN}{F}_{l_2}+\mathit{Ve}{g}_{f_2}\\ ⋮\\ \mathit{L}{2}^{O^t}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_{\mathit{t}}+Im{p}_{d_t}{n}_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t{\alpha }_t)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}\end{array}\right\}$(11) ) respectively.

(10) $\left.\begin{array}{c}\mathit{L}{1}^{O^1}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_1\\ \mathit{L}{1}^{O^2}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_2+\mathit{Im}{p}_{d_1}\\ \begin{array}{c}\mathit{L}{1}^{O^3}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_3+\mathit{Im}{p}_{d_2}\\ ⋮\\ \mathit{L}{1}^{O^t}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_{\mathit{t}}+\mathit{Im}{p}_{d_t}\end{array}\end{array}\right\}$(10)

(11) $\left.\begin{array}{c}\mathit{L}{2}^{O^1}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_1-n_{l_1}+(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1{\alpha }_1)\\ \mathit{L}{2}^{O^2}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_2+Im{p}_{d_1}-n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2{\alpha }_2)-\mathit{IN}{F}_{l_1}+\mathit{Ve}{g}_{f_1}\\ \mathit{L}{2}^{O^3}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_3+Im{p}_{d_2}-n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3{\alpha }_3)-\mathit{IN}{F}_{l_2}+\mathit{Ve}{g}_{f_2}\\ ⋮\\ \mathit{L}{2}^{O^t}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_{\mathit{t}}+Im{p}_{d_t}{n}_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t{\alpha }_t)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}\end{array}\right\}$(11)

The outputs as in Equations (10)(10) $\left.\begin{array}{c}\mathit{L}{1}^{O^1}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_1\\ \mathit{L}{1}^{O^2}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_2+\mathit{Im}{p}_{d_1}\\ \begin{array}{c}\mathit{L}{1}^{O^3}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_3+\mathit{Im}{p}_{d_2}\\ ⋮\\ \mathit{L}{1}^{O^t}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_{\mathit{t}}+\mathit{Im}{p}_{d_t}\end{array}\end{array}\right\}$(10) and (11(11) $\left.\begin{array}{c}\mathit{L}{2}^{O^1}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_1-n_{l_1}+(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1{\alpha }_1)\\ \mathit{L}{2}^{O^2}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_2+Im{p}_{d_1}-n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2{\alpha }_2)-\mathit{IN}{F}_{l_1}+\mathit{Ve}{g}_{f_1}\\ \mathit{L}{2}^{O^3}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_3+Im{p}_{d_2}-n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3{\alpha }_3)-\mathit{IN}{F}_{l_2}+\mathit{Ve}{g}_{f_2}\\ ⋮\\ \mathit{L}{2}^{O^t}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_{\mathit{t}}+Im{p}_{d_t}{n}_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t{\alpha }_t)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}\end{array}\right\}$(11) ) are required by validating the condition failures and infertility progressions in a step-by-step manner. If $\mathit{Im}{p}_d=0$ and then $\mathit{L}{2}^{O^t}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_{\mathit{t}}+\mathit{Im}{p}_{d_t}$ is the final output for ecological restoration prediction along with resistance analysis. Hence, the final restoration prediction is performed to prevent heavy metal pollution in soils. For this computation,$\mathit{RS}{T}_M=\left(\frac{\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_{\mathit{t}}+\mathit{Im}{p}_d}{\mathit{n}}\right)$ are the final output and that particular restoration method is distributed with all the outputs of $\mathit{L}{1}^{O^t}$ and $\mathit{L}{2}^{O^t}$ in Equations (10)(10) $\left.\begin{array}{c}\mathit{L}{1}^{O^1}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_1\\ \mathit{L}{1}^{O^2}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_2+\mathit{Im}{p}_{d_1}\\ \begin{array}{c}\mathit{L}{1}^{O^3}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_3+\mathit{Im}{p}_{d_2}\\ ⋮\\ \mathit{L}{1}^{O^t}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_{\mathit{t}}+\mathit{Im}{p}_{d_t}\end{array}\end{array}\right\}$(10) and (11(11) $\left.\begin{array}{c}\mathit{L}{2}^{O^1}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_1-n_{l_1}+(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1{\alpha }_1)\\ \mathit{L}{2}^{O^2}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_2+Im{p}_{d_1}-n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2{\alpha }_2)-\mathit{IN}{F}_{l_1}+\mathit{Ve}{g}_{f_1}\\ \mathit{L}{2}^{O^3}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_3+Im{p}_{d_2}-n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3{\alpha }_3)-\mathit{IN}{F}_{l_2}+\mathit{Ve}{g}_{f_2}\\ ⋮\\ \mathit{L}{2}^{O^t}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_{\mathit{t}}+Im{p}_{d_t}{n}_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t{\alpha }_t)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}\end{array}\right\}$(11) ). The processing output for failures and fertility based on the vegetation resistance factor is represented in Figure 4.

Figure 4. Processing output for failures and fertility.

The $\mathit{\alpha }\mathrm{\forall }$ pollution and vegetation impact due to climate are cumulatively used for learning combinations. In this combination the $\mathit{IN}{F}_l,\mathit{RS}{T}_M$, and reliability are extracted independently using $\mathit{\alpha }$ and $\mathrm{\nabla }$ conditions (i.e.) $\mathit{\alpha }>\mathrm{\nabla },\mathit{\alpha }<\mathrm{\nabla }$ or$\mathit{\alpha }=\mathrm{\nabla }$ for the above combinations. Based on these extracted features the prediction is performed; the maximum learning combinations are exploited for $\mathit{RS}{T}_M$ under $\mathit{L}{1}^o$ and $\mathit{L}{2}^o\mathrm{\forall t}$ (Refer to Figure 4). The restoration prediction process is discussed in the further subsection below.

5. Restoration prediction

The RNN learning assessment initiates from the continuous agricultural soil analysis with the first training for achieving maximum restoration prediction for time intervals. This $\mathit{Soi}{l}_{\mathit{data}}$ is the uncertain measure in the following instance, if any impacts (later), the restoration process are achieved. Hence in RNN, the consecutive process impact detection and pollution possibility estimation are precisely identified. The different training inputs for matching outputs are pursued for predicting accurate restoration procedure is defined as per Equation (12)(12) $\mathit{Z}\left\{\mathrm{\exists }\left[\mathit{Soi}{l}_{\mathit{data}},\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\right]\right\}=-\mathit{\alpha }\mathrm{\nabla }-\mathit{n}\left(\mathit{Soi}{l}_{\mathit{data}}\right)-\mathit{P}$(12) and (13(13) $\left.\begin{array}{c}\left(\mathit{L}{1}^{O^t}|\mathit{Soi}{l}_{\mathit{data}}\right)=\mathit{Im}{p}_d\left(\mathit{Soi}{l}_{\mathit{data}}+\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\mathit{\alpha }\right)\\ \left(\mathit{Soi}{l}_{\mathit{data}}|\mathit{L}{2}^{O^t}\right)=\mathit{Im}{p}_d\left(\mathit{Soi}{l}_{\mathit{data}}+\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\mathrm{\nabla }\right)\end{array}\right\}$(13) )

(12) $\mathit{Z}\left\{\mathrm{\exists }\left[\mathit{Soi}{l}_{\mathit{data}},\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\right]\right\}=-\mathit{\alpha }\mathrm{\nabla }-\mathit{n}\left(\mathit{Soi}{l}_{\mathit{data}}\right)-\mathit{P}$(12)

Such that,

(13) $\left.\begin{array}{c}\left(\mathit{L}{1}^{O^t}|\mathit{Soi}{l}_{\mathit{data}}\right)=\mathit{Im}{p}_d\left(\mathit{Soi}{l}_{\mathit{data}}+\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\mathit{\alpha }\right)\\ \left(\mathit{Soi}{l}_{\mathit{data}}|\mathit{L}{2}^{O^t}\right)=\mathit{Im}{p}_d\left(\mathit{Soi}{l}_{\mathit{data}}+\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\mathrm{\nabla }\right)\end{array}\right\}$(13)

In the above equation, the$\left(\mathit{L}{1}^{O^t}|\mathit{Soi}{l}_{\mathit{data}}\right)$ and $\left(\mathit{Soi}{l}_{\mathit{data}}|\mathit{L}{2}^{O^t}\right)$ factors used for satisfying the condition of $\mathit{Z}\left\{\mathrm{\exists }\left[\mathit{Soi}{l}_{\mathit{data}},\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)\right]\right\}$. As defined in the above equation, the reliable restoration prediction either satisfies$\left(\mathit{L}{1}^{O^t}|\mathit{Soi}{l}_{\mathit{data}}\right)$ or $\left(\mathit{Soi}{l}_{\mathit{data}}|\mathit{L}{2}^{O^t}\right)$ for achieving maximum prediction value. This case is not applicable for the first computation as in Equations (10)(10) $\left.\begin{array}{c}\mathit{L}{1}^{O^1}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_1\\ \mathit{L}{1}^{O^2}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_2+\mathit{Im}{p}_{d_1}\\ \begin{array}{c}\mathit{L}{1}^{O^3}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_3+\mathit{Im}{p}_{d_2}\\ ⋮\\ \mathit{L}{1}^{O^t}=\mathcal{N}{\left(\mathit{Soi}{\mathit{l}}_{\mathit{data}}\right)}_{\mathit{t}}+\mathit{Im}{p}_{d_t}\end{array}\end{array}\right\}$(10) and (11(11) $\left.\begin{array}{c}\mathit{L}{2}^{O^1}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_1-n_{l_1}+(\mathit{\daleth }\mathit{IN}{F}_{l_1}+P_1{\mathrm{\nabla }}_1{\alpha }_1)\\ \mathit{L}{2}^{O^2}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_2+Im{p}_{d_1}-n_2(\mathit{\daleth }\mathit{IN}{F}_{l_2}+P_2{\mathrm{\nabla }}_2{\alpha }_2)-\mathit{IN}{F}_{l_1}+\mathit{Ve}{g}_{f_1}\\ \mathit{L}{2}^{O^3}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_3+Im{p}_{d_2}-n_3(\mathit{\daleth }\mathit{IN}{F}_{l_3}+P_3{\mathrm{\nabla }}_3{\alpha }_3)-\mathit{IN}{F}_{l_2}+\mathit{Ve}{g}_{f_2}\\ ⋮\\ \mathit{L}{2}^{O^t}=\mathit{N}{(\mathit{Soi}{\mathit{l}}_{\mathit{data}})}_{\mathit{t}}+Im{p}_{d_t}{n}_t(\mathit{\daleth }\mathit{IN}{F}_{l_t}+P_t{\mathrm{\nabla }}_t{\alpha }_t)-\mathit{IN}{F}_{l_t}+\mathit{Ve}{g}_{f_t}\end{array}\right\}$(11) ) because it relies on infertility and vegetation failure occurrence in agricultural soils. Therefore, the $\mathit{RS}{T}_M$ along with resistance analysis is observed by the proposed technique and if it leads to low values, then climatic factors are analyzed to achieve high resistance. In the following sequence of analysis, $\mathit{RS}{T}_M$ on its previous time, determines the pollution possibility. If the sequence is observed in $\mathcal{N}\left(\mathit{Soi}{l}_{\mathit{data}}\right)>\mathit{Soi}{l}_{\mathit{data}}$, then the heavy metal-exposed soils are avoided to prevent impacts and failures in the vegetation process. An overview of the restoration prediction process based on the alternatives is presented in Figure 5.

Figure 5. Overview of the restoration prediction process.

The prediction process combines $\mathrm{\nabla }$ and $\mathit{\alpha }$ variants in the $\mathit{L}1$ and $\mathit{L}2$ under different$\mathit{t}$. This is required for validating multiple procedures for their restoration efficiency. Considering the $\mathit{Z}\left\{\mathit{E}\right\}$ for the multiple $\mathit{Max}\left(\mathit{HV}{y}_m\right)$ and $\mathit{Min}\left(\mathit{HV}{y}_m\right)$ the $\overline{\mathit{Soi}{\mathit{l}}_{\mathit{data}}}$ is used for normalization. In coherence to the $\left(\mathrm{\nabla },\mathit{\alpha }\right)$ combination the failures are validated for existing procedure recommendations. Thus, the predicted output is modified with the external procedures for recommendation (Figure 5). The RNN generates an alert to the smart agricultural platform to ensure appropriate actions to address the impacts. This prevents infertility and vegetation failures whereas, the restoration procedure is high. Based on the above discussion the pollution rate of 14 different locations in the reference image is tabulated for its actual and predicted values.

The tabulation above relies on the observed data from the reference (Okereafor et al., 2019) and the $\left(\mathrm{\nabla },\mathit{\alpha }\right)$ conditions stated in $\mathit{L}{1}^o$ and $\mathit{L}{2}^o$ respectively. The variations between the actual and predicted are validated as per $\mathit{RS}{T}_M$ for CT and $\mathit{IN}{F}_l$. Depending on the available $\mathit{P},\mathit{Q}$ the most suitable features are validated for preventing further failures in predicting observations. Therefore, the flexible combinations of $\mathit{\alpha }$ and$\mathrm{\nabla }$ are used for deciding the $\mathit{RS}{T}_M$ for improving prediction (Table 3 Data). The impact of heavy metals over the 8 different restoration methods proposed in (Yi et al., 2017) is validated in Table 4. This validation is based on $\mathrm{\exists }$ and $\mathit{\alpha }$ from the above discussion.

Table 3. Pollution rate and its predicted value.

Location	Cr	As	Pb	Co	Ni	Cd	Zn	Actual	Predicted
1	9.56	15.638	6.239	1.454	0.491	28.469	0.044	62.59	0.01
2	9.57	15.542	6.231	1.557	0.519	29.854	0.043	59.73	0.75
3	9.565	15.590	6.278	1.447	0.523	23.914	0.056	61.07	1.21
4	9.571	15.556	6.175	1.548	0.506	28.651	0.059	58.31	−0.01
5	9.551	15.556	6.272	1.584	0.525	27.252	0.040	61.82	1.17
6	9.567	15.562	6.146	1.533	0.505	29.919	0.048	62.89	0.05
7	9.578	15.489	6.115	1.514	0.490	27.790	0.043	62.03	1.25
8	9.58	15.487	6.207	1.530	0.529	26.444	0.057	59.02	−0.33
9	9.574	15.506	6.143	1.481	0.515	25.415	0.052	59.79	−1.09
10	9.579	15.608	6.277	1.491	0.481	27.771	0.049	60.70	1.30
11	9.57	15.481	6.127	1.511	0.482	26.288	0.058	62.21	0.96
12	9.56	15.584	6.135	1.476	0.524	26.271	0.052	61.30	0.05
13	9.57	15.460	6.174	1.556	0.514	24.650	0.049	58.55	0.11
14	9.578	15.473	6.233	1.446	0.487	26.861	0.058	61.77	−0.86

Table 4. Impact of$\mathit{\hspace{0.17em}}\mathrm{\nabla }$ over the heavy metal pollution.

$\mathrm{\nabla }$	Cr	As	Pb	Co	Ni	Cd	Zn	$\mathrm{\exists }$	$\mathit{\alpha }$
1	✓	✗	✓	✓	✓	✓	✗	0.84	0.85
2	✗	✓	✗	✓	✗	✗	✓	0.5	0.68
3	✗	✓	✗	✗	✓	✓	✗	0.59	0.62
4	✓	✗	✓	✗	✓	✗	✓	0.62	0.71
5	✗	✓	✗	✓	✗	✓	✗	0.71	0.81
6	✓	✓	✗	✓	✓	✓	✓	0.87	0.9
7	✗	✗	✓	✗	✗	✗	✗	0.5	0.74
8	✓	✓	✗	✓	✓	✓	✓	0.87	0.9

The green ticks in the above illustrate the feasible adaptation of restoration based on the methods proposed in the reference. The feasibility and less impact identification from low to high restoration method impacts positively. Contrarily, with the change in different reliability factors, the prediction varies accordingly. Therefore, the less feasible restoration techniques are also marked as non-impact-causing features (Table 4).

6. Discussion

Apart from the above discussions, a comparative analysis is performed using pollution detection, prediction ratio, training error, matching rate, and matching time. The mean concentration and maximum contribution ratio are varied in this comparative analysis. The methods GWR-PCA (L. Yang et al., 2022), SD-PCA (Liu et al., 2023), and MLR+PLS (Deng et al., 2023) from the related works section are incorporated for this comparative analysis.

In this proposed technique, the heavy metal-exposed agricultural soils are identified using continuous recurrent learning for predicting its restoration age and procedure for achieving the maximum impact prediction value through RNN learning (Refer to Figure 6). The infertility progressions and vegetation failures occurrence in agricultural land are detected from the soil types, crop resistance, and climatic factors to achieve maximum restoration prediction and impact detection. The identified impacts are mitigated using the proposed technique and recurrent neural network learning for detecting pollution possibility. The recurrent neural network learning process has two types of layers for balancing restoration procedures and crop resistance for handling maximum restoration procedures in any time interval. Continuous problem prediction is pursued for the different types of soils to reduce infertility progressions and vegetation failures. Maximum pollution detection based on identifying the heavy metal exposure in agricultural soils is reduced through the proposed technique, this technique depends on Layer 1 and Layer 2 outputs. Therefore, high pollution detection is achieved using this proposed technique.

Figure 6. Pollution detection.

This proposed technique achieves a high restoration prediction ratio depending on identifying adverse impacts of heavy metal presence in different soil types and the vegetation failures and infertility progressions are addressed by the recurrent learning process at any time intervals. The minimum/maximum impacts detected using the proposed technique for improving industrialization are represented in (Refer to Figure 7). The restoration procedures preferred in the earlier stages are used for training the neural network learning and restoration age is also predicted for maximum prediction; the more feasible restoration method is achieved. In this article, the already existing method is used for handling the restoration procedures at random time intervals to improve the matching ratio. The CRPT was used to identify vegetation failures and infertility through recurrent analysis. Based on the Layer 1 and Layer 2 outputs, the restoration prediction is pursued using the proposed technique relies on crop resistance observed at different time intervals to prevent training error. Therefore, the two layers are validated by the proposed technique to satisfy maximum impact detection and restoration prediction. In this proposed technique, a high prediction ratio is achieved using the RNN learning process.

Figure 7. Prediction ratio.

In Figure 8, the failure and infertility are validated using recurrent learning for analyzing the appropriate restoration procedures and crop resistance to improve prediction value and normalized value. Therefore, based on the Layer 1 and Layer 2 outputs, the different training inputs are matching for identifying adverse effects impacting vegetation to improve successful vegetation. Based on the restoration method, the soil types, climatic factors, and crop resistance are analyzed using the proposed technique. The assessment of adverse impacts on agricultural lands is analyzed with observed soil data and its contamination. Using recurrent learning, the problems are predicted and heavy metal presence in soil is also identified at different periods. The proposed technique and RNN learning process are used to improve the prediction ratio and matching rate. The proposed technique is used to reduce infertility progressions and vegetation failures through appropriate restoration methods and crop resistance analysis. The verification of soil types, weather conditions, and crop resistance is to increase vegetation. Therefore, the pollution possibilities are identified for which the proposed technique satisfies less training error.

Figure 8. Training error.

This article continuously processes the recurrent learning for predicting its restoration age and heavy metal presence in agricultural soils using the proposed technique. Based on the impact detection, the restoration procedure used to detect the heavy metal present in the soil for assessing its impacts is to satisfy less training error as compared to other factors as represented in Figure 9. Using this proposed technique, the adverse effects impacting vegetation are controlled through Layer 1 and Layer 2 recurrent analysis. The procedure for restoration prediction is used for checking the soil type, climatic factors, and crop resistance for successful vegetation without failures. The CRPT mainly performs agricultural soil data assessment and its appropriate restoration procedures for satisfying maximum prediction value. In this article, the high matching rate of the current restoration method with the existing method is instantaneous to meet fewer infertility progressions and vegetation failures from the instances using the proposed technique. Therefore, the heavy metal present in agricultural soils increases problems. This issue impacts vegetation failure and infertility progressions thereby reducing crop resistance. Based on the problem prediction, the high matching rate is achieved by the proposed technique.

Figure 9. Matching rate.

This proposed technique identifies the failures and pollution possibilities in the agricultural soils for maximizing restoration prediction along with crop resistance depending on agricultural soils for continuous predictive analysis. The failures and infertility are identified at the time of crop vegetation through the proposed technique and recurrent learning is employed at any interval. The heavy metal present in the agricultural soils is identified through recurrent analysis, for which the ecology restoration prediction is made. The restoration procedure used for controlling heavy metal pollution in different soils is pursued using the learning process. The heavy metal exposure analysis is performed to identify the vegetation failures and infertility progressions through continuous recurrent neural network learning. The restoration method is used to compute the failure and infertility for satisfying a maximum prediction value. Hence, the proposed technique identifies training errors and failures using recurrent learning. The proposed technique improves the impact detection and restoration prediction and thereby decreases matching time as presented in Figure 10.

Figure 10. Matching time.

7. Conclusion

In this article, the concentration-specific restoration prediction technique is introduced to provide recommendations on heavy metal-polluted agricultural soils. This technique is designed for detecting soil pollution using metal concentration from ages to provide suitable restoration methods that are reliable over long ages. The metal exposures from different levels are identified using sensed and accumulated data for which the infertility and vegetation factors are computed. This prediction process is validated using a recurrent neural network process for restoration age prediction and recommendation implications. The apt restoration method is validated based on soil attributes and climatic impact during the prediction. The recurrent training is performed for resistance analysis and the positive impact of the restoration process over the predicted outcome. The lowest and highest possible prediction values are estimated from the previous matching instances for improving the restoration factors. This proposed technique is validated using real-time data from different data sources and internal metrics. From the metric-based analysis, this technique is found to achieve 9.41% high detection, 8.32% high prediction ratio, and 14.53% less training error for the various heavy metal concentration rates.

Disclosure statement

No potential conflict of interest was reported by the author(s).