Published online by Cambridge University Press: 27 March 2009
The reliability of the soil series as a basis for crop yield prediction was examined by comparing the influence of soil and management (inter-farmer differences) on the variance of barley grain yield. Yields of barley were measured from farmers' fields for 2 years and in undisturbed soil cores for 5 years. Fields on different soil series within farms and on the same soil series across farms were used. Linear stepwise regressions of yield on a number of soil properties were also examined to assess the relative influence of soil and management on the properties that were significant to yield.
The results show that generally, soil classification has a strong influence on yield variance, but there is clear evidence that the influence of management, specifically the cumulative effect of P and K fertilizer applications, is considerable. Thus in an undisturbed core, for the first 3 years when N, P, K and Mg were applied, the influence of soil was stronger than that of management, but this was reversed in the 4th year when P and K were not applied. Similarly in the field, the influence of soil was stronger in the 1st year, but this was reversed in the 2nd year, although on different field-farm combinations. In all cases, the influence of neither soil nor management was significantly stronger than the other.
The results of the regression studies also confirmed those of yield variance in that in general, neither soil nor management has exclusive control of the yield variance. The soil was significantly stronger than management in the control of only coarse sand content at 0–15 cm out of the four soil properties (including Mg and pH (0–15 cm) and K (15–30 cm)) which were significant to yield in the cores. On the other hand, management was significantly stronger than soil in the control of only available P at O–15 cm out of the three soil properties (including Cu and Mg both at 0–15 cm) that were significant to field yield.
It is concluded that for soil classification to be a reliable basis for yield prediction and/or agrotechnology transfer, the effect of management must be emphasized.