Comparison of estimation of soil salinity using spectroscopy, electromagnetic induction, and remote sensing

Soil salinity is one of the most important soil properties and it's variability investigation is essential to crop management, land degradation and environmental studies. Soil salinity is measured using electrical conductivity (EC) and estimation of soil salinity contents using experimental methods is expensive and time consuming. Therefore, the collection of information on the spatial distribution of soil salinity in n vast areas requires new inexpensive techniques. Recently, new techniques such as electromagnetic induction, visible - near infrared spectroscopy and remote sensing were applied to measure soil salinity. The purpose of this study is the estimation of soil salinity using visible- near infrared spectroscopy, electromagnetic induction, and remote sensing methods. The study area is located 20 km northeast of Ghorveh city in Kurdistan Province and covers a surface of 26000 hectares. 100 soil samples (0–30 cm depth) were collected and. soil electrical conductivity was measured in a saturated extract. Applied auxiliary data in this study were spectral information of visible - near infrared spectroscopy method, reading of electromagnetic induction method, and ETM+ data of Landsat 8. In the 100 sampling sits, horizontal and vertical readings were read using EM38 and salinity index (SI) and normalized difference vegetative index (NDVI), bright index, and Bands 1, 2, 3, 4, 5, 6 , and 7were computed and extracted using Landsat 8 ETM+ data and Arc GIS software. Moreover, the 100 samples were scanned using spectrometer (model of FieldSpec®3, ASD, FR, USA) with a spectral range of 350 to 2500 nm. To make a relationship between soil salinity and auxiliary data of the three methods, artificial neural network (ANN) model were applied. Finally, soil salinity were estimated using ANN and were validated using cross validation method. Soil salinity contents were low to high (0.23 to 14.47 dSm-1). The highest contents of soil salinity were observed in central regions (low and bare land) and the lowest contents of soil salinity were located in high and range land. Based on sensitive analysis of artificial neural network model, in remote sensing methods salinity index, NDVI index, band 7, and band 3 were the most variables to predict soil salinity. In general, the results showed the most important auxiliary variables to predict soil salinity were spectral information of visible - near infrared range, vertical reading, and remote sensing data, respectively. Soil visible - near infrared spectroscopy method to predict soil salinity had 0.94, 0.27 and 0.64, respectively for determination of coefficient (R2), mean error (ME), and root mean square root (RMSE) and was better compared to the electromagnetic induction, remote sensing although combination of three methods together had the best results to estimate soil salinity. The most important auxiliary data to predict soil salinity in the study area was spectral information of visible - near infrared range. Electromagnetic induction method also is suitable auxiliary data to predict soil salinity and it can recommend as speed, accurate and cheap method to predict soil salinity. Combination of three methods together (electromagnetic induction, visible - near infrared spectroscopy and remote sensing) had the best results to estimate soil salinity. Therefore, it is suggested to predict soil salinity, ANN model and auxiliary data such as spectral information of visible - near infrared spectroscopy method and electromagnetic induction will be applied in the future studies.