Soil temperature modeling using machine learning techniques

Document Type: Research Paper

Authors

1 Department of Soil Science, Faculty of Agriculture, University of Zanjan,Zanjan, Iran and Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardebili, Ardabil, Iran

2 Department of Soil Science, Faculty of Agriculture, University of Zanjan,Zanjan, Iran

3 Department of Remote Sensing and GIS, University of Tehran, Tehran, Iran and Department of Geography, Humboldt University Berlin, Berlin, Germany

4 Forschungszentrum Jülich, Institute of Bio- and Geosciences: Agrosphere (IBG-3), 52428 Jülich, Germany

5 Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardebili, Ardabil, Iran

6 School of Environmental Sciences, University of Guelph, Canada

10.22059/jdesert.2020.79256

Abstract

Soil Temperature (ST) is critical for environmental applications. While its measurement is often difficult, estimation from environmental parameters has shown promise. The purpose of this study was to model ST in cold season  from soil properties and environmental parameters. This study was conducted as a pot experiment in Ardebil, Iran. Automatic thermal sensors were installed at 5 and 10 cm depths. Besides, soil properties and environmental parameters were determined based on field and laboratory works. Machine learning methods including Multiple Linear Regression (MLR), Artificial Neural Network (ANN), and Adaptive Neuro-Fuzzy Interface System (ANFIS) were used for modeling ST. The air temperature was observed as the most effective factor in ST modeling. The relationship between soil and air temperature was stronger at 5 cm depth compared to 10 cm. The R2 between soil and air temperature was higher in the absence of sunlight than in its presence. The prediction of ANFIS (R2= 0.96 and MAPE= 10.5) was closer to the observed ST values compared to the ANN (R2= 0.91 and MAPE= 35) and MLR (R2= 0.57 and MAPE= 41). The results revealed the advantage of ANFIS method for ST modeling. This approach can be applied for soil depths and locations with data gap.
 

Keywords