Agricultural Engineering

Abstract Introduction: Wheat (Triticum aestivum L.) is one of the world's most important cereal crops and plays a vital role in global food security. Efficient potassium management is essential for sustaining wheat productivity and improving potassium use efficiency, particularly in calcareous soils where potassium availability is often restricted despite adequate total soil K reserves. Potassium sulfate (SOP) and potassium chloride (MOP) are the two most commonly used potassium fertilizers; however, comparative information on their agronomic performance under irrigated calcareous soil conditions remains limited. Therefore, this study was conducted to evaluate the effects of these two potassium sources on wheat productivity, nutrient uptake, potassium use efficiency, and economic profitability.

Materials and Methods: A field experiment was conducted during the 2023–2024 growing season at the Aibolu experimental farm (Urmia) using a randomized complete block design with three treatments (NP control, SOP at 100 kg ha^-1, and MOP at 80 kg ha^-1) and five replications. Application rates were adjusted to provide equivalent amounts of K₂O, and potassium was applied in split applications at tillering and stem elongation. At physiological maturity, grain yield, yield components, and leaf nutrient concentrations were determined. Agronomic efficiency of potassium (AEK), partial factor productivity of potassium (PFP_K), and partial budget analysis were calculated. Data were subjected to analysis of variance (ANOVA), and treatment means were separated using the LSD test at P ≤ 0.05.

Results and Discussion: Potassium fertilization significantly increased grain yield and improved all measured agronomic traits compared with control. Grain yield increased from 5620 kg ha^-1 in the control treatment to 6500 kg ha^-1 with potassium sulfate (SOP) and 6230 kg ha^-1 with potassium chloride (MOP). Straw yield, biological yield, thousand-kernel weight, spike density, and grains per spike also increased significantly following potassium application. Among the two potassium sources, SOP consistently produced superior agronomic performance, suggesting that the additional sulfur supplied by SOP contributed to improved crop performance. Leaf nutrient analysis demonstrated that potassium fertilization significantly improved leaf potassium concentration and enhanced the accumulation of iron and manganese. The greater micronutrient concentrations recorded in the SOP treatment were likely associated with the additional sulfur supplied by potassium sulfate, which improved nutrient availability and uptake under alkaline calcareous soil conditions. This improved nutritional balance likely contributed to enhanced physiological activity and, consequently, greater grain production.

Potassium use efficiency indices also favored the SOP treatment. Agronomic efficiency of potassium (AEK) reached 17.6 kg grain kg^-1 K₂O for SOP, compared with 12.71 kg grain kg^-1 K₂O for MOP, indicating a substantially greater yield response per unit of applied potassium. In contrast, partial factor productivity of potassium (PFP_K) was nearly identical for the two fertilizer sources, with values of 130.0 and 129.8 kg grain kg^-1 K₂O for SOP and MOP, respectively. These findings suggest that although both fertilizers produced comparable grain yield per unit of potassium applied, potassium sulfate produced a greater agronomic response than MOP, reflecting the synergistic contribution of sulfur to nutrient utilization and crop productivity. The economic evaluation corroborated the agronomic results. Potassium sulfate produced the highest net benefit (164,150 × 10³ Rials ha^-1) and benefit–cost ratio (1.13), compared with potassium chloride, which resulted in a net benefit of 113,090 × 10³ Rials ha^-1 and a benefit–cost ratio of 1.09. Despite its relatively higher purchase cost, the superior yield and agronomic efficiency of SOP resulted in greater economic profitability.

Conclusion: potassium fertilization substantially enhanced wheat productivity, nutrient uptake, fertilizer use efficiency, and economic profitability under irrigated calcareous soil conditions. Although both potassium sources exhibited similar partial factor productivity, potassium sulfate consistently outperformed potassium chloride in grain yield, agronomic efficiency, and economic return. The superior performance of potassium sulfate was likely associated with its combined supply of potassium and sulfur, which improved nutrient balance and crop productivity. Therefore, potassium sulfate appears to be a suitable potassium fertilizer source for irrigated wheat cultivated on calcareous soils in northwestern Iran.

Abstract Introduction:One of the problems of soil pollution is heavy metals.Heavy metals cause significant and serious pollution in many large cities.They may cause permanent damage to the ecosystem and humans.Excessive accumulation of heavy metals in soil and plants poses risks to human health.Therefore, accurate knowledge of heavy metals such as their levels and sources of pollution is essential.The aim of this study is to investigate the pollution of soils in some agricultural lands of Isfahan province with heavy metals such as cadmium, lead, copper, nickel, zinc, arsenic and iron.




 Materials and Methods: For this purpose, a total of 90 samples of agricultural topsoil were collected from areas of Khomeini Shahr, Isfahan, Falavarjan, Lenjan and Ardestan counties, and all samples were placed in polyethylene bags and transported to the laboratory. First, the soil samples were air-dried, then passed through a 2 mm sieve. Then, they were extracted with 4 M nitric acid. The concentration of heavy metals were determined by ICP-MASS. Descriptive statistics parameters including mean, maximum, minimum, skewness and kurtosis of the studied heavy metal concentrations were determined by SPSS software. Kolmogorov-Smirnov test was used to test normality and since the data were not normally distributed, Spearman correlation analysis was performed to determine the relationship between trace elements. A significance level of pZn>Ni>Pb>As>Cu>Cd. There was a significant positive correlation coefficient between all elements except iron, which may be due to the common source of these elements. The average of lead, zinc and cadmium in Isfahan city with the other three cities was significantly higher than in other study areas, which could be due to the presence of the Bama lead and zinc mine and the cement factory, as well as more vehicle traffic. Also, Ardestan city showed higher arsenic pollution than other cities, which is likely to have a geological origin in this area. The enrichment factor values ​​showed that the soils of the region have significant to high enrichment for cadmium, lead and zinc elements. According to the geoaccumulation results, it was determined that cadmium and arsenic elements have moderate to severe pollution in the entire region. Also, in Isfahan city, the geoaccumulation index results showed that lead and zinc elements have moderate to severe pollution. The ecological risk index was higher for cadmium in the entire region and cities, which indicates that this element is affected by human activities.




 Conclusion: The entry of heavy metals into the human body through consumption of plants, livestock, inhalation and skin contact has harmful toxic effects, and these effects are intensified as these elements accumulate in the human body over time. Therefore, it is suggested that the amount of these metals (especially arsenic) in dust, plants, livestock and humans be investigated.




 

Abstract Introduction



Tomato is one of the most widely consumed and nutritious crops, and its rich antioxidant content plays an important role in the prevention and treatment of various diseases and cancers. On the other hand, pesticide residues in agricultural products pose a serious threat to human health and may cause significant short- and long-term effects; therefore, accurate monitoring and control of these residues are highly important. In recent years, spectroscopic methods, especially NIRS and VIS/NIR, have emerged as fast, low-cost, non-destructive, and environmentally friendly techniques for detecting pesticide residues. These methods allow qualitative and quantitative analysis without complex sample preparation, and their accuracy can be improved using statistical and machine-learning models (such as artificial neural networks, deep learning, multiple regression, SVM, and optimization algorithms). Previous studies have shown promising results for these technologies in detecting pesticide residues in agricultural products like apples, oranges, pumpkin, and okra. Specifically, in the case of tomatoes, which are commonly found in daily meals, detecting pesticide residues is particularly important. Therefore, this research aims to investigate the capability of the VIS/NIR method for the non-destructive detection of indoxacarb pesticide residue in tomatoes



Materials and Methods



Experimental Design: Twenty-five tomato seedlings (variety 8320) were grown under uniform conditions. After fruiting, plants were sprayed with Indoxacarb 15% SC at volumetric concentrations of 0.625%, 1.25%, 2.5%, and 5%, plus an untreated control group. A total of 75 samples were harvested. Spectral Measurements: A NIRS spectrometer (wavelength range 400-2500 nm) was used on days 1, 3, 5, 7, and 8 after spraying, with three replicates. Data Preprocessing: Five methods were applied: Multiplicative Scatter Correction (MSC), Standard Normal Variate (SNV), Vector Normalization (VN), First Derivative (FD), and Second Derivative (SD). Synthetic Minority Over-sampling Technique (SMOTE) was used for data balancing, and Principal Component Analysis (PCA) for dimensionality reduction. Classification & Analysis: Support Vector Machines (SVM) were employed to classify different concentrations (70% training, 30% testing). Model performance was evaluated using confusion matrix, accuracy, sensitivity, detectability, and F1-score. Validation: HPLC (AZURA model, Germany) validated pesticide residue presence and established a linear equation correlating chromatogram peak area with concentration.



Results and Discussion In this section, the spectral signature of tomato samples treated with indoxacarb pesticide was compared with that of the control samples. The results showed that on the first day after spraying, the absorption pattern of the contaminated samples was generally similar to that of the control, although their absorption amplitude was reduced. These features were mainly associated with internal fruit constituents such as sugars, acids, and pigments including chlorophyll, beta-carotene, and lycopene. Moreover, the key spectral regions in the visible and near-infrared ranges were linked to chemical bonds in the indoxacarb structure, particularly C-H, N-H, C=O, and aromatic groups. Over time, especially by the eighth day after spraying, the spectral differences between treated and control samples decreased markedly, particularly at lower pesticide concentrations. This indicates that the residual effect of the pesticide gradually diminished and the fruit characteristics became closer to those of untreated tomatoes. The greatest spectral differences were observed in the 900–2500 nm range, especially within 900–1350 nm and 1550–1900 nm. Finally, support vector machine classification was used to distinguish contaminated samples from the control group, and raw data were compared with preprocessed data to identify the best classification performance. The classification results showed that among the preprocessing methods, vector normalization (VN) achieved the best performance, followed by SNV. This superiority is likely due to VN’s ability to minimize the effects of illumination differences in hyperspectral data, thereby removing intensity-related variation while preserving the spectral shape. It was also found that the 400–900 nm and 900–1900 nm spectral ranges were effective for distinguishing samples, especially on the eighth day after spraying. Although pre-processing improved accuracy on the first day, methods such as SNV and VN may have reduced performance on the eighth day by removing some useful signals. Nevertheless, the results showed that even at lower pesticide concentrations, classification accuracy remained acceptable, exceeding 80%.This suggests that indoxacarb can also influence the visible range, making it possible to detect pesticide residues using simpler and lower-cost instruments without the need to analyze higher spectral bands. In the residue estimation section, partial least squares regression (PLSR) was used. Five principal components explained more than 97% of the output variation, and the model was compared with HPLC measurements. The results indicated that pesticide residue decreased over time, especially on days seven and eight. Overall, visible/near-infrared spectroscopy was found to be an effective non-destructive, fast, and low-cost method for monitoring pesticide residues in tomatoes. Previous studies show that spectroscopy in different wavelength ranges can be used to detect pesticide residues. It also cites a study on eggplant, where indoxacarb residues were measured using HPLC. The initial residue levels on eggplant fruits decreased over time and became undetectable after about 15 to 20 days. Overall, the results of this study suggest that the developed regression model for estimating indoxacarb residues can be extended in the future to detect other pesticides as well, offering a fast, accurate, and non-destructive approach for monitoring pesticide residues in agricultural products.



Conclusion The results showed that visible/near-infrared spectroscopy in the 400–2500 nm range, especially when applied to pre-processed data, was able to distinguish sprayed samples from unsprayed ones. The method achieved higher accuracy on days close to spraying and still maintained about 77% accuracy even eight days after application. In addition, the partial least squares regression model in the 400–900 nm range was able to estimate pesticide residue effectively, with a coefficient of determination of 90%.

Abstract Hyperspectral sensors record surface reflectance in hundreds of narrow, contiguous spectral bands and therefore allow crops that appear almost identical in broadband imagery to be discriminated. This spectral richness, however, introduces the Hughes phenomenon: as the number of bands grows relative to the number of labelled training samples, the accuracy of conventional classifiers first improves and then deteriorates, while the strong correlation between neighbouring bands adds redundancy and computational cost. Dimensionality reduction, either by feature extraction such as principal component analysis (PCA) or by band selection, is therefore an essential pre-processing step. Classical learners such as artificial neural networks (ANN) and support vector machines (SVM) remain standard baselines and recent deep architectures exceed 99% accuracy on benchmark scenes, although they require large training sets, heavy computation and complex tuning and offer limited interpretability. Two issues have received comparatively little attention in the benchmark literature: the strong sensitivity of shallow-network accuracy to training hyperparameters in high-dimensional spaces, and the use of unsupervised clustering to reveal the intrinsic spectral-affinity structure among crops without any labelled data.


The standard Salinas Valley hyperspectral image acquired by the AVIRIS sensor was used. The scene covers 512 by 217 pixels at a spatial resolution of 3.7 m and, after the removal of the water-absorption bands 108-112, 154-167 and 224, retains 204 effective spectral bands spanning 400-2500 nm; it contains 54,129 labelled pixels distributed very unevenly among sixteen crop classes. Band collinearity was quantified before any reduction: the mean absolute correlation over all band pairs was 0.65 and that between adjacent bands 0.99, with 33% of all pairs exceeding 0.90, and the participation ratio of the covariance eigenvalues was 1.6, so the effective dimensionality is close to two. Three input scenarios were then defined: all 204 bands, all principal components, and the first two components, which are the only ones above the 2% variance threshold and together account for 98.3% of the total variance. The labelled pixels were split 70/30 by stratified random sampling, and a spatially disjoint split along a column cut served as a complementary test of spatial information leakage. An ANN and an RBF support vector machine were trained on identical training pixels and scored on identical test pixels in every scenario, using overall accuracy, the kappa coefficient, producer's and user's accuracy and the F1 score. The ANN sensitivity analysis covered the number of hidden nodes (2 to 50), the number of layers (one to three), the early-stopping patience (1 to 12) and five training algorithms, searched one factor at a time with ten repetitions per configuration. In the unsupervised stage, K-means and self-organizing maps were applied for cluster numbers from 2 to 16 and assessed with the silhouette, Davies-Bouldin and Calinski-Harabasz indices, while the spectral angle between class centroids provided an independent measure of spectral affinity.


Training hyperparameters proved to be as decisive as the choice of classifier. Increasing the early-stopping patience raised the ANN test accuracy on PCA features from about 66% to 90.3%, a change exceeding 24 percentage points; across the three scenarios the range of accuracy over the patience sweep was 20.1, 24.4 and 23.0 percentage points, with Spearman correlations of 0.62 (p = 0.031) for all bands and 0.64 (p = 0.026) for two components, so the effect is large but not monotonic. In the high-dimensional space the scaled conjugate gradient and one-step secant algorithms performed best, whereas the Levenberg-Marquardt algorithm failed completely, which is attributable to its heavy memory demand when the number of parameters is large. In the like-for-like comparison the support vector machine ran without difficulty on the full 204-band space, contrary to the initial analysis, and reached 93.9% overall accuracy with a kappa of 0.932, against 91.1% and 0.901 for the ANN; on all principal components the two reached 92.7% and 90.0%, and on the first two components only 83.4% and 81.8%. The earlier finding that the support vector machine was ineffective therefore reflected the restriction of its input to two components and untuned kernel parameters rather than any property of the method. Dimensionality reduction brought no gain in accuracy on this scene: retaining all components, an invertible rotation of the original space, left accuracy almost unchanged, while discarding components on a variance criterion cost about ten percentage points. Under the spatially disjoint split the mean F1 fell from 97.2% to 83.4%, showing that spatial autocorrelation does inflate the conventional random-split estimate. The clustering analysis showed that the intrinsic spectral structure is coarser than the nominal sixteen classes: the silhouette peaked and the Davies-Bouldin index reached its minimum at the same seven clusters. The stubble class formed the purest cluster with more than 99% of its pixels in a single cluster, while the spectral angle confirmed the overlaps independently, the closest pairs being fallow and developing vineyard soil at 1.1 degrees and untrained grapes and untrained vineyard at 1.4 degrees.


In the supervised recognition of crop patterns from hyperspectral imagery with classical methods, the tuning of the training parameters is at least as important as the choice of the learning method: the early-stopping rule alone moved accuracy by more than 24 percentage points, whereas the gap between the two classifiers never exceeded three. Dimensionality reduction on a variance criterion did not improve accuracy on this scene; the value of principal component analysis here is computational, and the number of retained components should be tied to classification performance rather than to a fixed variance threshold. The intrinsic spectral structure of the scene is described by seven clusters rather than sixteen nominal classes, and the classes prone to confusion can be identified before any classification, without labelled data, from the spectral angle between class centroids. These results give three practical recommendations for crop-recognition systems: budget and document the parameter tuning before selecting a classifier,, choose the number of components by classification performance, and concentrate ground sampling on the class pairs that the spectral angle identifies as close.

Abstract Introduction


Accurate land suitability evaluation is an important basis for sustainable agricultural land management and improved crop productivity. Conventional parametric land evaluation methods commonly represent the effects of soil limitations using predefined, piecewise-linear relationships between soil properties and suitability scores. Although these relationships provide a continuous representation of suitability, their fixed linear form may not adequately capture nonlinear crop responses to variation in soil properties. Fuzzy logic provides an alternative framework by expressing the degree of suitability through continuous membership functions whose shape and parameters can be adjusted to represent nonlinear responses. However, the parameters of these functions are commonly defined using expert knowledge or predefined threshold values, which may not fully reflect crop responses under specific field conditions. A data-driven calibration approach can reduce reliance on manual parameter setting by optimizing membership-function parameters against observed crop performance. Therefore, the objective of this study was to develop a data-driven fuzzy soil suitability model for irrigated grain maize by optimizing the parameters of fuzzy membership functions using the Differential Evolution (DE) algorithm, and to examine the contribution of individual soil properties to the relationship between soil suitability and observed maize yield.


Materials and Methods


The study was conducted using observations from 100 irrigated grain-maize farms distributed across four agricultural regions of Iran, located in Khuzestan, Fars, Qazvin, and Kerman provinces. Climatic suitability was considered similar and suitable for maize across the study areas; therefore, the analysis focused on soil-related limitations. The soil and land characteristics considered as model inputs were soil pH, electrical conductivity (EC), exchangeable sodium percentage (ESP), calcium carbonate equivalent (CCE), gypsum content, soil texture, drainage condition, and slope. Actual maize yield, calculated as the average yield of each farm over five years, was used as an independent criterion for evaluating the relationship between the soil suitability index and crop performance and as the objective variable for data-driven optimization; it was not used as a direct input for calculating the suitability index. A classical parametric suitability model based on the framework of Sys et al. (1991) was first implemented as the reference approach. Subsequently, a fuzzy model was developed in which continuous membership functions were defined for the quantitative soil properties according to the agronomic response of maize, while qualitative variables were represented using expert-defined suitability functions. The parameters of the membership functions were then optimized using the DE algorithm by maximizing the coefficient of determination (R²) between the resulting fuzzy soil suitability index (FSI) and observed maize yield. The initial fuzzy model was also evaluated before optimization to quantify the effect of parameter optimization. Sensitivity analysis was conducted by removing each soil variable individually from the optimized model and assessing the resulting change in R². Permutation-based importance analysis was additionally used to examine the independent contribution of each predictor by randomly permuting its values while keeping the other predictors unchanged. Alternative aggregation operators were examined using the generalized power mean, and the effect of assigning variable-specific weights on the resulting suitability index was also evaluated.


Results and Discussion


The results showed that data-driven optimization of the membership-function parameters improved the relationship between the fuzzy soil suitability index and observed maize yield. The coefficient of determination increased from 0.45 for the initial fuzzy model to 0.69 after optimization, whereas the classical parametric suitability approach produced an R² of 0.59. Accordingly, the optimized fuzzy model provided a stronger explanation of the observed variation in maize yield than both the initial fuzzy formulation and the classical parametric approach. These results indicate that, in addition to adopting continuous fuzzy membership functions, appropriate calibration of their parameters is important for representing the field-level response of maize to soil limitations. Sensitivity analysis and permutation-based importance analysis did not produce identical rankings of soil properties, reflecting differences in the information captured by the two approaches; nevertheless, both analyses highlighted the greater contribution of several chemical soil properties to the observed relationship between soil suitability and maize yield. The complementary nature of the two analyses provides additional information on the influence of individual soil properties without implying that either analysis alone represents a unique causal ranking of the variables. Evaluation of alternative aggregation operators showed that the generalized power mean approached the geometric mean, with the optimized power parameter close to zero. Consequently, the geometric mean was retained for the final suitability index. Furthermore, applying variable-specific weights did not increase the coefficient of determination of the fuzzy model, indicating that weighting the membership functions did not provide an improvement over their unweighted geometric aggregation under the conditions of this study.


Conclusion


This study developed a data-driven framework for fuzzy soil suitability assessment of irrigated grain maize by optimizing membership-function parameters using the Differential Evolution algorithm and observed field yield. The results showed that optimization increased the coefficient of determination between the soil suitability index and observed maize yield from 0.45 to 0.69, compared with 0.59 obtained using the classical parametric approach. The findings demonstrate that data-driven adjustment of membership-function parameters can improve the ability of a fuzzy soil suitability index to explain variation in observed maize yield under field conditions. Sensitivity and permutation-based importance analyses provided complementary information on the contribution of individual soil properties, with several chemical properties showing greater influence than the evaluated topographic factors. The geometric mean remained the appropriate aggregation operator after evaluation of alternative power-mean formulations, while variable-specific weighting did not improve the model’s R². Overall, the proposed framework provides a data-driven approach for reducing reliance on manual specification of membership-function parameters and may be considered in future soil suitability studies for other crops and regions where reliable field-performance data are available.

Abstract Introduction


This investigation presents a rigorous experimental evaluation of a low-cost electronic nose (E-nose) framework, coupled with multivariate chemometric and supervised machine learning algorithms specifically Principal Component Analysis (PCA), Random Forest (RF), and Artificial Neural Networks (ANN) for the identification and discrimination of agricultural biofuel blends. The primary objective is to critically appraise the operational validity and physical limitations of static metal-oxide semiconductor sensing for fuel quality assurance, thereby establishing an empirical baseline to prevent publication bias and guide the robust deployment of bio-based renewable fuels in compression ignition engines.


Materials and Methods


An electronic nose system comprising an array of eight commercial metal-oxide semiconductor (MOS) gas sensors (MQ series: MQ2, MQ3, MQ5, MQ8, MQ9, MQ135, MQ137, and MQ138) was developed and evaluated across 15 distinct fuel classes, including neat fossil diesel (D100), waste cooking oil biodiesel (B100), fusel oil (F100), and their corresponding binary and ternary blends. A strictly balanced dataset of 225 experimental trials was established and partitioned into a training set (n=150) and an independent test set (n=75, with 5 replicates per class). Static steady state voltage equilibrium responses were acquired and standardized. Data structure and variance distribution were explored using unsupervised Principal Component Analysis (PCA), while supervised multi class classification and hyperparameter tuning were conducted using Random Forest (RF) and a Multilayer Perceptron Artificial Neural Network (ANN/MLP, 8-4-15 architecture). Sensor contribution was quantitatively assessed via Mean Decrease in Impurity (MDI), and dimensional sensitivity was examined by contrasting the full eight sensor array against an optimized four sensor subset.


Results and Discussion


Unsupervised exploratory analysis demonstrated that although the first two principal components captured 74.96% of the cumulative variance (PC1: 46.10% and PC2: 28.86%), extensive spatial cluster overlap persisted across all 15 fuel treatments without evident class separation boundaries. Supervised classification on the independent test set yielded poor discrimination, with RF and ANN attaining overall test accuracies of only 6.67% (95% CI: [2.88%, 14.68%]) and 8.00% (95% CI: [3.72%, 16.37%]), and Cohen’s kappa coefficients of κ=0.0000 and κ=0.0143, respectively. Both models converged near the theoretical random chance baseline (1/15≈6.67%), with McNemar’s test confirming no statistically significant performance difference (p=1.0000). Random Forest feature importance revealed a homogenous impurity reduction across the array (MDI spanning narrowly between 0.1008 for MQ9 and 0.1465 for MQ8), evidencing pronounced sensor cross sensitivity and strong collinearity. Furthermore, pruning the array to four sensors marginally altered RF and ANN test accuracies to 10.67% and 5.33%, respectively, demonstrating the adverse impact of collinear dimensions (curse of dimensionality) under small sample constraints and confirming that scalar equilibrium signals lack the kinetic resolution required to decouple highly overlapping volatile organic compound profiles.


Conclusion


The findings demonstrate that low cost MOS electronic nose systems relying solely on steady state equilibrium readings cannot achieve granular 15 class discrimination among chemically proximate diesel, biodiesel, and fusel oil blends, with machine learning complexity failing to compensate for baseline informational deficits. Consequently, future developments must transition from scalar static signals to dynamic transient response profiling, sensor temperature modulation, and continuous chemometric regression modeling to establish dependable, real-time quality control protocols in biofuel refining.