Dataset characteristics influence the performance of different interpolation methods for soil salinity spatial mapping

This study compared the performance of different interpolation methods for mapping soil salinity of three different agricultural fields having the same land use but different dataset characteristics. Four common spatial interpolation methods including global polynomial interpolation (GPI), inverse distance weighted (IDW), ordinary kriging (OK), and radial basis functions (RBF) were employed for mapping soil EC. The performance of interpolation methods in predicting soil EC was evaluated based on mean bias error, root mean square error, mean absolute percentage error, and coefficient of determinations criteria. Results showed that dataset characteristics, including central tendency and distribution, were significantly different among the studied fields. Experimental semivariogram and fitted model parameters indicated that three studied fields were also different in their spatial dependence strength. Considering all of the performance assessment measures used, the best interpolation method for fields A and C was OK and IDW for field B. The performance of interpolation methods was found to be affected by data characteristics of the studied fields, which were mostly ascribed to management practices. This study suggests in order to obtain accurate mapping of soil salinity in agricultural fields, it is essential to first find the best spatial interpolation method compatible with the characteristics of the collected data from the selected agricultural land.