Journal of Soil and Plant Interactions
Back to the articles list |
Back to browse issues page
Agricultural Research, Education and Extension Organization (AREEO)
Abstract: (90 Views)
Abstract
This study aims to predict the concentrations of copper (Cu) and manganese (Mn) in citrus leaves. It also seeks to identify the key soil properties that affect their availability in the citrus orchards of southern Kerman, Iran. Samples were collected from 40 orchards, with a focus on both soil and leaf samples taken during the spring growth season. We used standard methods to analyze soil properties and measure Cu and Mn levels in the leaves. The results revealed that in the topsoil, soil pH significantly hindered nutrient absorption, with Cu and Mn showing negative correlations of -0.41 and -0.33, respectively. However, Cu was positively correlated with both clay content (0.32) and organic carbon (OC) (0.39), while Mn exhibited a negative correlation with soil salinity (EC = 0.29). Stepwise regression analysis revealed that pH, clay, and phosphorus were the key predictors for Cu (R² = 0.37), while pH and EC were key predictors for Mn concentrations (R² = 0.18), explaining only a fraction of the variation. To overcome the limitations of linear models, we developed an artificial neural network (ANN) model, using the Levenberg-Marquardt algorithm, which outperformed the regression models. The best ANN model for Cu, incorporating pH, OC, clay, and phosphorus with ten hidden neurons, achieved an R² of 0.78 and an RMSE of 0.88 mg kg⁻¹. For Mn, an ANN model using pH, OC, clay, phosphorus, and EC with 11 hidden neurons resulted in an R² of 0.66 and an RMSE of 5.68 mg kg⁻¹. The AI models reduced prediction errors by 83% for Cu and 44% for Mn compared to regression models. These results suggest that ANNs are powerful tools for improving nutrient management and boosting productivity in citrus orchards.
Background and Objective: Citrus farming is a critical agricultural activity in southern Kerman, Iran. However, achieving optimal productivity is often challenged by the limited availability of essential micronutrients like copper (Cu) and manganese (Mn), which are vital for plant health and fruit quality. This study aims to predict the concentrations of Cu and Mn in citrus leaves by identifying the key soil properties that affect their availability. Furthermore, it compares the predictive accuracy of traditional stepwise regression with a more advanced artificial neural network (ANN) model to develop a robust tool for nutrient management.
Methods: Soil and leaf samples were collected from 40 citrus orchards during the spring growth season. Soil samples were analyzed for key physicochemical properties, including pH, electrical conductivity (EC), organic carbon (OC), texture (clay content), and available phosphorus (P), using standard laboratory methods. Leaf samples were also analyzed to determine their Cu and Mn concentrations. The collected data was used to develop two types of predictive models: a stepwise linear regression model and a multi-layer perceptron ANN trained with the Levenberg-Marquardt algorithm. The dataset was split into training (70%) and testing (30%) sets, and model performance was evaluated using the coefficient of determination (R2) and root mean square error (RMSE).
Results: Correlation analysis revealed that soil pH was the most significant factor hindering nutrient absorption, showing negative correlations with both Cu (-0.41) and Mn (-0.33). Conversely, Cu concentration was positively correlated with clay content (0.32) and organic carbon (0.39), while Mn showed a negative relationship with soil salinity (EC) (-0.29). The stepwise regression models showed limited predictive power; the model for Cu (inputs: pH, clay, P) yielded an R2 of only 0.36, and the model for Mn (inputs: pH, EC) yielded an R2 of just 0.28.
In contrast, the ANN models performed significantly better. The optimal model for Cu, which incorporated pH, OC, clay, and P as inputs with ten hidden neurons, achieved an R2 of 0.78 and an RMSE of 0.88 mg kg⁻¹ in the test phase. For Mn, the best model used pH, OC, clay, P, and EC as inputs with 11 hidden neurons, resulting in an R2 of 0.66 and an RMSE of 5.68 mg kg⁻¹. Compared to the regression models, the ANN models reduced prediction errors by a remarkable 42% for Cu and 38% for Mn.
Conclusion: This research highlights that soil properties such as pH, clay content, organic carbon, and salinity are critical drivers of Cu and Mn bioavailability in the calcareous soils of southern Kerman. The study definitively demonstrates that ANN models, by capturing the complex and non-linear interactions between these soil properties, are far more accurate and reliable for predicting nutrient concentrations than traditional linear regression methods. These findings underscore the potential of ANN-based models as powerful decision support tools for precision agriculture, enabling the early diagnosis of nutrient deficiencies and the development of targeted fertilization strategies to boost citrus orchard productivity and sustainability.
This study aims to predict the concentrations of copper (Cu) and manganese (Mn) in citrus leaves. It also seeks to identify the key soil properties that affect their availability in the citrus orchards of southern Kerman, Iran. Samples were collected from 40 orchards, with a focus on both soil and leaf samples taken during the spring growth season. We used standard methods to analyze soil properties and measure Cu and Mn levels in the leaves. The results revealed that in the topsoil, soil pH significantly hindered nutrient absorption, with Cu and Mn showing negative correlations of -0.41 and -0.33, respectively. However, Cu was positively correlated with both clay content (0.32) and organic carbon (OC) (0.39), while Mn exhibited a negative correlation with soil salinity (EC = 0.29). Stepwise regression analysis revealed that pH, clay, and phosphorus were the key predictors for Cu (R² = 0.37), while pH and EC were key predictors for Mn concentrations (R² = 0.18), explaining only a fraction of the variation. To overcome the limitations of linear models, we developed an artificial neural network (ANN) model, using the Levenberg-Marquardt algorithm, which outperformed the regression models. The best ANN model for Cu, incorporating pH, OC, clay, and phosphorus with ten hidden neurons, achieved an R² of 0.78 and an RMSE of 0.88 mg kg⁻¹. For Mn, an ANN model using pH, OC, clay, phosphorus, and EC with 11 hidden neurons resulted in an R² of 0.66 and an RMSE of 5.68 mg kg⁻¹. The AI models reduced prediction errors by 83% for Cu and 44% for Mn compared to regression models. These results suggest that ANNs are powerful tools for improving nutrient management and boosting productivity in citrus orchards.
Background and Objective: Citrus farming is a critical agricultural activity in southern Kerman, Iran. However, achieving optimal productivity is often challenged by the limited availability of essential micronutrients like copper (Cu) and manganese (Mn), which are vital for plant health and fruit quality. This study aims to predict the concentrations of Cu and Mn in citrus leaves by identifying the key soil properties that affect their availability. Furthermore, it compares the predictive accuracy of traditional stepwise regression with a more advanced artificial neural network (ANN) model to develop a robust tool for nutrient management.
Methods: Soil and leaf samples were collected from 40 citrus orchards during the spring growth season. Soil samples were analyzed for key physicochemical properties, including pH, electrical conductivity (EC), organic carbon (OC), texture (clay content), and available phosphorus (P), using standard laboratory methods. Leaf samples were also analyzed to determine their Cu and Mn concentrations. The collected data was used to develop two types of predictive models: a stepwise linear regression model and a multi-layer perceptron ANN trained with the Levenberg-Marquardt algorithm. The dataset was split into training (70%) and testing (30%) sets, and model performance was evaluated using the coefficient of determination (R2) and root mean square error (RMSE).
Results: Correlation analysis revealed that soil pH was the most significant factor hindering nutrient absorption, showing negative correlations with both Cu (-0.41) and Mn (-0.33). Conversely, Cu concentration was positively correlated with clay content (0.32) and organic carbon (0.39), while Mn showed a negative relationship with soil salinity (EC) (-0.29). The stepwise regression models showed limited predictive power; the model for Cu (inputs: pH, clay, P) yielded an R2 of only 0.36, and the model for Mn (inputs: pH, EC) yielded an R2 of just 0.28.
In contrast, the ANN models performed significantly better. The optimal model for Cu, which incorporated pH, OC, clay, and P as inputs with ten hidden neurons, achieved an R2 of 0.78 and an RMSE of 0.88 mg kg⁻¹ in the test phase. For Mn, the best model used pH, OC, clay, P, and EC as inputs with 11 hidden neurons, resulting in an R2 of 0.66 and an RMSE of 5.68 mg kg⁻¹. Compared to the regression models, the ANN models reduced prediction errors by a remarkable 42% for Cu and 38% for Mn.
Conclusion: This research highlights that soil properties such as pH, clay content, organic carbon, and salinity are critical drivers of Cu and Mn bioavailability in the calcareous soils of southern Kerman. The study definitively demonstrates that ANN models, by capturing the complex and non-linear interactions between these soil properties, are far more accurate and reliable for predicting nutrient concentrations than traditional linear regression methods. These findings underscore the potential of ANN-based models as powerful decision support tools for precision agriculture, enabling the early diagnosis of nutrient deficiencies and the development of targeted fertilization strategies to boost citrus orchard productivity and sustainability.
Type of Study: Research |
Subject:
Modeling of soil-water-plant relations and root water uptake
Received: 2025/04/27 | Accepted: 2025/08/3
Received: 2025/04/27 | Accepted: 2025/08/3
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
Journal of Soil and Plant Interactions by http://jspi.iut.ac.ir is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
This work is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported License which allows users to read, copy, distribute and make derivative works for non-commercial purposes from the material, as long as the author of the original work is cited properly.
This work is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported License which allows users to read, copy, distribute and make derivative works for non-commercial purposes from the material, as long as the author of the original work is cited properly.
