ACCURACY OF FIELD SAMPLING FOR SOIL TESTS

Abstract

Three widely separated Alberta fields were sampled intensively in a grid pattern, to estimate the effect of number and depth of cores and of field size on the accuracy of composite samples used in soil test analyses. Estimates were obtained by both statistical and computer simulation techniques and both gave similar results. Accuracy of composites increased with the number of cores, but even with as many as 30 cores per field there was still considerable inaccuracy in values for N and P. In the most variable field, a composite of 20 cores for the 0- to 15-cm depth predicted the true field mean within ± 10% for N, P and K only 47, 56, and 82 times out of 100, respectively. Corresponding values for the least variable field were 77, 70, and 98. Conductivity and pH were more accurately estimated with the same number of cores. Samples taken at the 15- to 30-cm and 30- to 61-cm depths showed greater coefficients of variability in N and P than did samples from the 0- to 15-cm depth. However, samples composed of the 0- to 30-cm and 0- to 61-cm depths were similar to the 0- to 15-cm depth in variability. The number of cores needed for a given level of accuracy increased very little with size of field, up to a maximum of 40 hectares used in this study. Degree of variation changed considerably from field to field, but even more from one position to another within the same field.