Mapping Soil Salinity Using Calibrated Electromagnetic Measurements

A statistical modeling approach is presented that predicts spatial soil salinity patterns from aboveground electromagnetic induction (EM) readings. In this approach, EM readings are obtained from a field sampled on a uniform (centric systematic) grid. A small number of these sample sites are chosen for soil sampling, based on the observed EM field pattern. The salinity levels for these soil samples are determined and then the remaining nonsampled salinity values are predicted from the corresponding EM readings through a multiple linear regression equation. Experimental results suggest that this approach will work well in fields having low to moderate levels of soil textural variability. For example, 95% of the spatial variability in soil salinity within typical 16.2-ha (40-acre) cotton (Gossypium hirsutum L.) fields could be accounted for with only 36 soil samples, as opposed to the 200 to 300 soil samples typically required if no EM readings were available. This approach makes EM readings a more practical and cost-effective tool by substantially reducing the number of soil samples needed for accurate mapping of spatial salinity patterns at the field scale. A SOIL SALINITY ASSESSMENT is neCCSsary for agriculture management. Excessive soil salinity affects crop production and may cause crop loss and, eventually, land degradation. Cost-effective appraisal techniques designed to assess and monitor the salinity levels can help minimize such losses. Field-scale soil salinity conditions can be characterized using ECa measurements. Soil ECa is influenced by chemical and physical properties of the soil liquid and solid phases. Soil salinity, as represented by ECe, can be determined from field ECa measurements (Rhoades et al., 1989b). On farmland, practical measurements of ECa can be made with either in situ or remote devices. Three kinds of portable sensors are available: (i) four-electrode sensors, including either surface-array or insertion probes; (ii) EM induction sensors, and (iii) time domain reflectometric sensors (Rhoades and Oster, 1986; Rhoades and Miyamoto, 1990; Rhoades, 1990). The surface-array, four-electrode, or EM techniques all give depth-weighted ECa measurements. The weighting functions vary with the configuration of the electrodes or electromagnetic coils, frequency of electrical current used in the measurement, distribution of ECa within the various depths of the soil profile, and other factors. All of these factors must be compensated for when interpreting soil salinity with these devices. Two approaches have been used to determine U.S. Salinity Lab., USDA-ARS, 4500 Glenwood Dr., Riverside, CA 92501. Contribution from the U.S. Salinity Lab. Received 4 Dec. 1990. *Corresponding author. Published in Soil Sci. Soc. Am. J. 56:540-548 (1992). salinity by depth in the soil from EM measurements. Rhoades et al. (1989a) developed a salinity prediction model to estimate ECe using either EM or four-electrode measurements, provided that soil water and clay content were known. Empirical equations that predict soil ECe from EM measurements have also been developed for specific soil and water-content situations (Rhoades and Corwin, 1981; Corwin and Rhoades, 1982; Williams and Baker, 1982; Williams and Hoey, 1987; Slavich and Petterson, 1990; McKenzie et al., 1989). For some purposes, establishing a direct ECe = f(EM) prediction equation is advantageous, such as in a single field under uniform management where water content, bulk density, and other soil properties are reasonably homogeneous. Under such conditions, it is possible to establish an accurate ECe-EM relationship using a limited number of soil samples. Geostatistical procedures have traditionally been used for salinity mapping when soil samples are available (Webster, 1989). However, such procedures generally require intensive sampling to obtain accurate variogram estimates. We have developed a more practical, cost-effective predictive technique—field-specific models that use limited ECe ground-truth data and extensive EM measurements for spatial salinity predicion and mapping.