Soil and Water Research (Jan 2022)

Assessment and mapping of soil salinity using electromagnetic induction and Landsat 8 OLI remote sensing data in an irrigated olive orchard under semi-arid conditions

  • Mohamed Elhedi Gharsallah,
  • Hamouda Aichi,
  • Talel Stambouli,
  • Zouhair Ben Rabah,
  • Habib Ben Hassine

DOI
https://doi.org/10.17221/178/2020-SWR
Journal volume & issue
Vol. 17, no. 1
pp. 15 – 28

Abstract

Read online

Salinisation threatens the sustainability of irrigated olive orchards in Tunisia. Electromagnetic induction measurements and soil spectral index calculations could help to survey the soil salinity. This study aimed to map changes in the soil salinity spatial pattern using geostatistical techniques and soil spectral index regression. The study area is located in Sminja, Tunisia. It is a 665 ha olive orchard, landscaped in ridges and furrows and managed following a very high-density planting system (1.5 × 4 m2). Electromagnetic readings measured in situ with an electromagnetic device (EM38) that was fitted, in turn, to the electrical conductivity of the saturated paste of five soil depths namely: 0-20, 20-40, 40-60, 60-80 and 80-100 cm and to the average electrical conductivity of the saturated paste of the 0-100 cm soil depth. Both the ordinary kriging and universal kriging performed similarly and well in mapping the soil salinity. (R2= 0.86 and 0.89 for the 0-20 cm and the 0-100 cm depths, respectively). Our results prove that mapping the soil salinity based on electromagnetic induction and kriging methods is an effective approach, which allows one to monitor the soil salinity within permanent croplands in semi-arid conditions. Salinisation that reaches intolerable values by olive trees, is especially accumulated on the top of the ridges, where the drippers are installed. Furthermore, based on two Landsat 8 images acquired on April 30, 2019 and May 16, 2019, respectively, we calculated seven soil spectral indices. Nevertheless, multiple regression models between the electromagnetic readings and various combinations of soil spectral indices were poor. In the coming investigations, under permanent land cover, spectral index regression models should integrate not only the soil, but also vegetation indices.

Keywords