Soil salinity and its mapping are of great importance in arid and semiarid regions, especially where continued monitoring of soil salinity changes is costly. The aim of the present paper was to compare measured error resulted from calibration equations in different sampling campaigns and their use in soil salinity prediction with soft data in several time instants. Study area was located in western part of Urmia Lake, northwest Iran. Soil samples were taken from 0-20 cm during seven sampling campaigns between 2009-2015. Samplings were done on a grid of 500 meters for the first sampling campaigns and 250 meters and less for other time intervals. Soil electrical conductivity was measured in the field for all times series and in the laboratory, only for first, second and fifth data sets. After calibrating field EC by laboratory EC, soil salinity histograms of residuals were calculated by observed lab data minus estimated ones for the first and second datasets. Then probability density functions were calculated and used in soil salinity prediction using kriging with measurement errors for the third dataset. After validation of the predicted values for the third dataset, the same error and developed model were used in soil salinity prediction in other four time instants. ME and MSE were calculated as quantitative criteria for assessing the accuracy of predictions. Results from error calculation and its use in prediction showed that there were no significant differences among predicted values using measured error from the first and second datasets and also combination of both. Validation of the produced salinity maps of the study area revealed that application of the general prediction model developed for the study area and calculated error for spring 2010 (second dataset) could successfully predict soil salinity during other time intervals with ME and MSE equal to -0.12 and 0.92 for autumn 2014.The reduced need for constant calibration of field measured data and number o
