WATER-CONTENT EFFECT ON SOIL-SALINITY PREDICTION - A GEOSTATISTICAL STUDY USING COKRIGING

A geostatistical analysis of soil salinity in an agricultural area in the San Joaquin Valley included measurements of electrical conductivity of soil paste extract (EC(e)) and water content of soil samples supplemented by surface measurements of apparent electrical conductivity (EM(H)). Prediction of soil salinity at unsampled points by cokriging log(e)(EC(e)) and EM(H) is worthwhile because EM(H) measurements are quicker than soil sampling. This work studies how patterns of log(e)(EC(e)) predicted by cokriging with EM(H) are influenced by variation in gravimetric water content (W). The data are mean EM(H) = 1.00 +/- 0.13 dS m(-1) for 2378 locations, mean log(e)(EC(e)) = 1.40 +/- 0.29 dS m(-1), and mean gravimetric W = 0.260 +/- 0.003, both averaged for four samples from 0.3-m intervals to 1.2-m depth for 315 locations. The coefficient of determination (R(2)) for EM(H) vs. log(e)(EC(e)) increased with depth from 0.05 to 0.54 whereas the R(2) for EM(H) vs. W decreased from 0.48 to 0.28. A gray-scale EM(H) map contained nine out of 56 quarter-section boundaries coinciding with step variations in EM(H). The t-statistics for differences in mean W were six of nine significant at 0.001 and nine of nine at 0.05, but mean log(e)(EC(e)) had only two of nine at 0.05, implying that W caused EM(H) steps. Water-affected EM(H) impaired prediction of EC(e) at depth by cokriging, because near-surface variations in W masked EC(e). Two subareas were defined, one where management factors, such as irrigation, controlled EM(H), causing steps, and one where near-surface W varied less, making cokriging predictions more reliable.