Predicting spatial distribution of soil organic matter using regression approaches at the regional scale (Eastern Azerbaijan, Iran)

• Published in: Environmental monitoring and assessment 2021-09, Vol.193 (9), p.615-615, Article 615
• Main Authors: Ebrahimzadeh, Golnaz, Yaghmaeian Mahabadi, Nafiseh, Khosravi Aqdam, Kamal, Asadzadeh, Farrokh
• Format: Article
• Language: English
• Subjects: Arid regions; Atmospheric Protection/Air Quality Control/Air Pollution; Carbon cycle; Datasets; Distribution; Distribution patterns; Earth and Environmental Science; Ecology; Ecotoxicology; Elevation; Environment; Environment models; Environmental factors; Environmental Management; Environmental modeling; Environmental monitoring; Environmental science; Estimation errors; Methods; Monitoring/Environmental Analysis; Normalized difference vegetative index; Organic matter; Organic soils; Regression coefficients; Regression models; Root-mean-square errors; Semi arid areas; Soil investigations; Soil moisture; Soil organic matter; Soil properties; Soil quality; Soil surfaces; Soil testing; Spatial distribution; Storage; Support vector machines; Training; Vegetation index
• ISSN: 0167-6369; 1573-2959
• DOI: 10.1007/s10661-021-09416-0

Soil organic matter (SOM) is one of the important factors in arid and semiarid areas, which describes the soil quality. Spatial estimation of SOM is important to understand the SOM storage and the emphasis of the SOM in the global carbon cycle and environmental issues. Mapping of SOM content can have significant uses in environmental modeling. In the current study, various methods have been evaluated for estimating the SOM content through soil samples and using auxiliary variables in the west of Eastern Azerbaijan province, Iran. In this study, support vector machine (SVM), multi-factor regression (MFR), and multi-factor weighted regression model (MWRM) approaches have been suggested for predicting and investigating the spatial distribution of SOM. In total, 155 surface soil samples (from the depth of 0 to 30 cm) were obtained. These soil samples were randomly divided into training data set (105 soil samples) and testing data set (50 samples). According to the results, SOM is affected by soil properties as well as environmental factors (normalized difference vegetation index (NDVI)). In this study, clay, silt/sand, NDVI, and soil moisture were used as auxiliary variables in the estimation of SOM. Three methods were compared to determine a suitable method for spatial estimation of SOM, and results showed that SVM has the lowest estimation error ( RMSE  = 0.100, MAE  = 0.07, and MRE  = 3.32) and highest regression coefficient ( R 2  = 0.719) during SOM estimation. The present results show the indirect effect of elevation by controlling auxiliary variables and confirm the importance of auxiliary variables in spatial distribution patterns of SOM.