Effect of long-term irrigation application on the variation of soil electrical conductivity in vineyards

The soil of a section of the farm Broodkraal, in the South Western Cape province of South Africa, was sampled on a 45 by 50.5 m grid to model spatial variation of salinity and predicts the amount of salts lost through irrigation. Sampling was done to a depth of 45 cm. The sampling included vineyards of different age that varied between 2 and 5 years since the first irrigation application. Salts mobilized by irrigation have a profound effect on the water quality in the Berg River catchment. We propose inverse distribution functions as a method to model the temporal change in salt content of the soil. We assume that the parameters of the distribution function for salt content depend on time since land was taken into irrigation. We may therefore take observations of salt content from land that has been irrigated for some known time, and predict the corresponding values in past or future by a simple transformation of the data. To test this hypothesis we undertook a second sampling campaign on sections of the farm irrigated for the same length of time as two of our original data sets. We showed that temporal variability of the distribution functions of all data was more substantial than the spatial variability between data sets on land irrigated for the same period of time. It is clear that the general lowering of ECe over time caused the variance to reduce. The comparison of variograms showed that the use of only one variogram model to krige the total landscape might result in an over simplification of the young irrigated surfaces while it complicated the older ones. Any mapping of ECe aimed over larger areas should therefore take cognisance of the irrigation age and the sampling frequency should be increased for younger or non-irrigated surfaces. Finally, the results were used to estimate the yearly salt return to the Berg River.