Performance evaluation of Artificial Neural Networks to estimate, summer savory (Satureja hortensis L.) essential oil yield based on the easily available soil properties

Document Type : original paper

Authors

1 Master Graduated, Dept. of Horticultural Science, Gorgan University of Agricultural Sciences and Natural Resources

2 Associate Prof., Dept. of Horticultural Science, Gorgan University of Agricultural Sciences and Natural Resources.

3 Head of Faculty-Gorgan Agricultural University And Natural Resources

4 Associate Prof., Dept. of Water Engineering, Gorgan University of Agricultural Sciences and Natural Resources

5 Lecturer, Department of Plant Production Engineering, Faculty of Agriculture, Technical and Vocational University (TVU), Khorasan Razavi

Abstract

Background and aim: One of the most important requirements in planning production and processing of medicinal plants in order to obtain high yield and high-quality is the initial assessment of the soil physical and chemical properties, which can reduce the production cost by avoiding the use of unnecessary soil analysis. Summer Savory(Satureja hortensis L.)is one the most widely used medicinal plants that quality index of plant is related to the quantity and the constituent of its essential oil content.Understanding the relations between the quantity and quality of medicinal plants with the several physical and chemical properties of soil is very complex and the estimation of parameters changes of medicinal plants affect by soil quality characteristics is more difficult. Today, with the introduction of multivariable regression models and artificial network models in the research, many complex relationships found in nature is understandable.Hence the need for estimation of the essential oil yield of savory using fast, cheap and acceptable accuracy methods is necessary.Material and method: The present study was performed as pot experiment based on completely randomized design with 3 replications. Fifty three soil samples were collected from different parts of Nishabur, and easily available soil properties including sand, silt and clay percentage, organic matter, pH, salinity, phosphorus, potassium, nitrogen and carbon contents of the soil samples were measured at laboratory and the primary results were obtained. Approximately 90 days after seed planting in mentioned soil samples, the sampling of plants was done based on the treatments. Samples were placed for 24 hours in an oven at 40 °C, for drying. Finally, the relationship between the essential oil yield and easily available soil parameters was determined using artificial neural network by Matlab7.9 software. To obtain the most sensitive parameters, sensitivity analysis was calculated by using sensitivity coefficient without dimension method. So that, if the parameter value is more than 0.1, then that parameter is considered as the sensitive parameter of the model.
Results: An artificial neural network is simulated from a human neural network model, which, after training, estimates the output parameters by applying the input parameters. In this research, the perceptron neural network structure was used with Marcoat Levenberg training algorithm to estimate the essential oil yield from easily available soil parameters such as soil texture, organic matter, and macro elements. The high R2 values and the low RMSE values indicate that predictive data are close to the measurement data and high accuracy of the model in the estimation of summer savory essential oil yield. Based on this, soil texture parameters (sand, silt and clay percentages) and organic carbon, organic matter, salinity, potassium and soil acidity were selected as the most sensitive parameters, respectively. High values of R2 and low levels of RMSE mentioned the proximity of the forecast data with measurement data and high accuracy of the model in summer Savory essential oil yield estimation. Accordingly, the parameters of organic carbon, nitrogen, phosphorus, organic matter, potassium, pH, salinity, clay, silt and sand respectively were selected as the most sensitive parameters. Conclusion: The results showed that the created neural models were not able to estimate the essential oil yield of summer savory with a maximum accuracy (R2=0.50).Among the 8 fitted models, a model based on independent variables EC+texture+carbon+organic matter + potassium + pH was better than the other, but the high number of input factors of this model is considered to be a limitation. Since the present study is an initial assessment of the essential oil yield of medicinal plants, it is recommended to continue the research in this regard as well as to predict the performance of other medicinal herbs.

Keywords

Main Subjects

References

1.Akbarpour, A., Khorashadizadeh, O., Shahidi, A. and Ghochanian, E. 2013. Performance evaluation of artificial neural network models in estimate production of yield saffron based on climate parameters. J. Saff. Res. 1: 1. 27-35.
2.Bremner, J.S. and Mulvaney, C.S. 1982. Nitrogen-total. In: A.L. Page (Ed.), Methods of Soil Analysis, Part 2. American Society of Agronomy. Madison, Wisconsin, Pp: 595-624.
3.Hill, M. 1998. Methods and guidelinesfor effective model calibration. U.S. Geological survey Water- Resources Investigations Rep. 98-4005.
4.Menhaj, M.B. 2001. Computational intelligence, fundamentals of neural networks. 2nd d., Amir Kabir University of Technology, Tehran: Iran. (In Persian (
5.Moazenzadeh, R., Ghahraman, B., Fathalian, F. and Khoshnoodyazdi, A.A. 2009. Effect of type and number of input variables on moisture retention curve and saturated hydraulic conductivity prediction. J. Water. Soil. 23: 3. 57-70. (In Persian)
6.Movahedi Naiini, A. 2008. Soil physics (foundations and applications). Gorgan University of Agricultural Sciences and Natural Resources. Press, 304p. (In Persian)
7.Nakhaei, M. 2005. Estimating the saturated hydraulic conductivity of granular material, using Artificial Neural Network, based on grain size distribution curv. Sci. I. R. Iran. J. 16: 1. 55-62.
8.Omidbaigi, R. 2005. Production and processing of medicinal plants. Astane Quds Publ. Tehran, 438p.(In Persian)
9.Page, A., Miller, R. and Keeney, D. 1982. Methods of Soil Analysis. 2th ed. Part 2: Chemical and biological properties. Soil. Sci. Soc. Am. Inc. Publisher.
10.Rao, V. and Rao, H. 1996. C++ Neural networks and fuzzy logic. BPB, New Dehli, India, Pp: 380-381.
11.Schaap, M. and Leij, F. 1998. Using neural networks to predict soil water retention and soil hydraulic conductivity. Soil.Till. Res. 47: 37-42.
12.Schaap, M., Leij, F. and Van Genuchten, M. 1998. Neural network analysisfor hierarchical prediction of soil hydraulic properties. Soil Sci. Soc. Am. J. 62: 847-855.
Journal of Plant Production Research
Volume 26, Issue 2
September 2019
Pages 47-58

Files

Share

How to cite

Statistics