Premium
Mapping soil salinity in 3‐dimensions using an EM38 and EM4Soil inversion modelling at the reconnaissance scale in central Morocco
Author(s) -
Dakak H.,
Huang J.,
Zouahri A.,
Douaik A.,
Triantafilis J.
Publication year - 2017
Publication title -
soil use and management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.709
H-Index - 81
eISSN - 1475-2743
pISSN - 0266-0032
DOI - 10.1111/sum.12370
Subject(s) - soil science , mean squared error , environmental science , salinity , residual , inversion (geology) , soil salinity , irrigation , hydrology (agriculture) , remote sensing , mathematics , soil water , geology , algorithm , statistics , geotechnical engineering , geomorphology , ecology , oceanography , structural basin , biology
Large areas of Morocco require irrigation and although good quality water is available in dams, farmers augment river water with poorer quality ground water, resulting in salt build‐up without a sufficient leaching fraction. Implementation of management plans requires baseline reconnaissance maps of salinity. We developed a method to map the distribution of salinity profiles by establishing a linear regression ( LR ) between calculated true electrical conductivity (σ, mS /m) and electrical conductivity of the saturated soil‐paste extract ( EC e, dS /m). Estimates of σ were obtained by inverting the apparent electrical conductivity ( EC a, mS /m) collected from a 500‐m grid survey using an EM 38. Spherical variograms were developed to interpolate EC a data onto a 100 m grid using residual maximum likelihood. Inversion was carried out on kriged EC a data using a quasi‐3d model ( EM 4Soil software), selecting the cumulative function ( CF ) forward modelling and S2 inversion algorithm with a damping factor of 3.0. Using a ‘leave‐one‐out cross‐validation' ( LOOCV ), of one in 12 of the calibration sites, the use of the q‐3d model yielded a high accuracy ( RMSE = 0.42 dS /m), small bias ( ME = −0.02 dS /m) and Lin's concordance (0.91). Slightly worse results were obtained using individual LR established at each depth increment overall (i.e. RMSE = 0.45 dS /m; ME = 0.00 dS /m; Lin's = 0.89) with the raw EM 38 EC a. Inversion required a single LR ( EC e = 0.679 + 0.041 × σ), enabling efficiencies in estimating EC e at any depth across the irrigation district. Final maps of EC e, along with information on water used for irrigation ( EC w) and the characterization of properties of the two main soil types, enabled better understanding of causes of secondary soil salinity. The approach can be applied to problematic saline areas with saline water tables.