Investigating the effect of Wetness index and spectral data on the estimation of soil particles percentage using neuro-fuzzy, artificial neural network and regression tree models

Mehrabi Gohari, Elham; Matinfar, Hamidreza; Taghizade Mehrjardi, Rohollah; JAFARI, AZAM

doi:10.22125/iwe.2019.100742

Investigating the effect of Wetness index and spectral data on the estimation of soil particles percentage using neuro-fuzzy, artificial neural network and regression tree models

Document Type : Original Article

Authors

elham mehrabi gohari ¹

Hamidreza Matinfar ²

Rohollah Taghizade Mehrjardi ³

AZAM JAFARI ⁴

¹ lorestan

² Department of Soil Science, College of Agriculture, Lorestan University, Lorestan, Iran

³ Faculty of Agriculture and Natural Resources, Ardakan University, P.O. Box 89516-56767, Yazd,

⁴ Department of Soil Science, Agriculture Faculty, Shahid Bahonar University of Kerman, Kerman, 76169-14111, Iran

10.22125/iwe.2019.100742

Abstract

Direct estimation of some soil characteristics is time consuming, costly and sometimes not possible. In recent years, indirect methods have been used to estimate these properties. In the present study, to predict the soil texture fractions, 115 profiles were identified based on the Hypercube technique, and the horizons were sampled and the percentage of sand, clay and silt of soil samples were measured. Environmental variables used in this study include the terrain attributes (derived from a digital elevation model), Landsat 8 image data (acquired in 2015), geomorphological map, and spectrometric data (laboratory data). Artificial neural network, regression tree and neuro-fuzzy models were used to make a correlation between soil data (clay, sand and silt) and environmental variables. The results of this study showed that the neuro-fuzzy model was more accurate in prediction of the three parameters of clay, sand and silt than artificial neural network and tree regression. The RMSE value in the neuro fuzzy model was compared to regression tree model. The neuro fuzzy model results were, for clay surface 1.43 %, for sand surface 1.98% and for silt surface 2.1% that reduced by 6.71%, 8.49% and 5.42% for clay, sand and silt respectively, compared to regression tree model. The results also showed that the most important auxiliary variables are spectrometric data followed by MrVBF and wetness index.

Keywords

Wetness Index

VIS_NIR

Artificial Neural Network

Tree Regression

Neuro-Fuzzy