Predicting Soil Particulate Organic Mattter Using Artificial Neural Network with Wavelet Function

ABSTRACT

Particulate organic matter is a part of the labile, easily decomposable, pool of soil organic matter; therefore its study in soil is very important. Estimating particulate organic matter through some related variables at an acceptable level of accuracy is very importance. In this study three model were utilized and compared for predicting particulate organic matter. (i) wavelet function of artificial neural network, (ii) perceptron function of artificial neural network and III) multiple linear regression. The highest R² value and lowest NRMSE in the regression models were 0.94 and 0.071, respectively, in the wavelet function of artificial neural network were 0.97 and 2.10 E-05, respectively. These values in perceptron function of artificial neural network were intermediate. The results indicated that although wavelet function of artificial neural network was the most reliable model for predicting POM, but multiple linear regression was also an appropriate predictor due to its acceptable R², especially when the logarithmic forms of the data were used in the models. The results revealed that the best inputs for modelings are soil physicochemical properties only. Wavelet function of the ANN was utilized in the POM prediction for the first time.

KEYWORDS:

• Modeling
• multiple linear regression
• perceptron function of artificial neural network
• wavelet function of artificial neural network
• particulate organic matter

Disclosure statement

The authors declare that they have no conflict of interest.

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.