مدل سازی رابطه غلظت آهن در برگ مرکبات با برخی خصوصیات خاک با استفاده از شبکه عصبی مصنوعی (مطالعه موردی جنوب استان کرمان)

نوع مقاله : مقاله پژوهشی

نویسندگان

• صابر حیدری ¹
• سید علی غفاری نژاد ²
• جواد سرحدی ³
• مهری شریف ⁴

¹ عضو هیات علمی، بخش تحقیقات خاک و آب، مرکز تحقیقات و آموزش کشاورزی و منابع طبیعی جنوب استان کرمان، سازمان تحقیقات، آموزش و ترویج

² عضو هیات علمی موسسه تحقیقات خاک و آب، سازمان تحقیقات، آموزش و ترویج کشاورزی، کرج، ایران.

³ عضو هیات علمی بخش تحقیقات خاک و آب، مرکز تحقیقات و آموزش کشاورزی و منابع طبیعی جنوب استان کرمان، سازمان تحقیقات، آموزش و ترویج کشاورزی،

⁴ بخش تحقیقات خاک و آب، مرکز تحقیقات و آموزش کشاورزی و منابع طبیعی جنوب استان کرمان، سازمان تحقیقات، آموزش و ترویج کشاورزی، جیرفت، ایران

چکیده

کلیدواژه‌ها

• پرسپترون چند لایه
• رگرسیون گام به گام
• ماده آلی خاک
• مرکبات

موضوعات

• حاصلخیزی خاک و تغذیه گیاه

عنوان مقاله [English]

Modeling the relationship between iron concentration in citrus leaves and some soil properties using artificial neural network (case study of southern Kerman province)

نویسندگان [English]

• saber heidari ¹
• Seyed Ali Ghaffari Nejad ²
• Javad Sarhadi ³
• Mehri Sharif ⁴

کلیدواژه‌ها [English]

• Citrus
• Multilayer Perceptron
• Soil Organic Matter
• Stepwise Regression

Modeling the Relationship between Iron Concentration in Citrus Leaves and Some Soil Properties Using Artificial Neural Network (Case Study of Southern Kerman Province)

EXTENDED ABSTRACT

Introduction

Studying the relationships between the chemical and physical characteristics of a plant's growing environment and the uptake of required elements by the plant can lead to better knowledge and design of the growing environment in that plant. Therefore, statistical and mathematical tools such as linear regression, logistic regression, and their combined models can be used in this direction. In recent years, the use of the artificial neural network (ANN) in predicting and modeling the nonlinear relationships of various phenomena that have great complexity, and the usual linear models and statistical analyzes that are unable to explain the relationships has expanded. The purpose of this study was conducted to evaluate the relationship between leaf iron and some easily-available soil properties in citrus orchards in the southern region of Kerman province by artificial neural network modeling and stepwise regression.

Materials and methods

The studied area is the south of Kerman province and includes seven cities: Jiroft, Kohnouj, Anbarabad, Manojan, Rudbar Janub, Ghaleganj and Faryab. To conduct this research, 40 orchards were selected from the study area and the physical and chemical properties of soil and iron in the plant leaves were measured. To model the relationship between iron concentration in leaves and some soil characteristics, stepwise regression and artificial neural network methods were used. To model the artificial neural network, MATLAB software, multilayer perceptron neural network, and backpropagation learning algorithm were used. using artificial neural networks in different models with different data from soil properties as input and leaf iron as output, the ability of these models to predict leaf iron concentration was evaluated and compared with the stepwise regression model.

Results and Discussion

The results of the regression model showed that soil organic carbon with a coefficient of 116.79 was the most and the only effective factor in the estimation of leaf iron, and the rest of the independent variables in the model did not affect the estimation of leaf iron. Most of the soils of citrus orchards in the region had low organic matter. The modeling results showed that in the ANN 1 model, with the selection of organic carbon as the only model input and the number of 10 neurons in the middle layer, the value of the regression coefficients was 0.33 and 0.32 for the training and test data, which had a better performance than the same model in stepwise regression. In the ANN 2 model, by adding the pH variable as an input, the prediction accuracy (regression coefficient) increased and reached 0.50 and 0.47 for the training and test data. Further, by adding the number of input variables in the ANN 3, ANN 4, and ANN 5 models, the relative improvement of the regression coefficients and error in the training and test data was achieved. The best result in the estimation of leaf iron was obtained in the ANN 6 model. The artificial neural network with variables of organic carbon, pH, clay, phosphorus, TNV, and electrical conductivity with regression coefficients of 0.86 and 0.81 and root mean square error (RMSE) of 14.60 and 20.13 mg/kg for data Training and testing were the best models in estimating leaf iron. The results of the neural network showed that the most important soil characteristic affecting iron absorption and concentration in plant leaves was organic carbon, followed by pH, and after these two variables, the most important characteristics in order of importance were phosphorus, and electrical conductivity. Comparison of regression and neural network models in the test data showed that the neural network had a higher accuracy with a regression coefficient of 0.81 than stepwise regression with a regression coefficient of 0.2. The amount of RMSE in the neural network also improved and increased from 27.72 mg.kg^-1 in the stepwise regression model to 20.13 mg.kg^-1 in the neural network.

Conclusion

The results of modeling using two methods of regression and an artificial neural network showed that the iron in plant leaves was most closely related to the amount of organic matter and soil pH. Artificial neural networks have been able to predict the iron in plant leaves based on the easily-available properties of the soil so by choosing organic carbon as the input of the first model to the best model by selecting organic carbon, pH, clay, phosphorus, TNV, and electrical conductivity, model accuracy increased.