Critical ranges for soil tests are based on results that inevitably involve some broad variance around the fitted relationship. Some of the variation is related to field-based factors affecting crop response to nutrients in the soil and some to the efficiency of the soil-test extractant itself. Most attempts to improve soil tests focus on the extractant, whereas here, we explore the variation that could be accounted for by field-based factors in the soil-test calibration relationship between Colwell phosphorus (P) and wheat yield, using the Australian Better Fertiliser Decisions for Crops database—the biggest dataset available for this relationship. Calibrations developed from this dataset have been criticised, and so we aimed to explore factors accounting for more of the variation in the relationships for the dryland, winter-dominant rainfall region of southern New South Wales.

As reported previously, soil type was shown to influence the critical range and r-value for the Colwell P soil-test calibration for P responses by wheat. We also identified a tendency for dry conditions, at sowing or during the season, to lower relative yields for a given soil-test value, indicating increased reliance on fertiliser P over soil P. A similar trend was evident for later sowing date, again suggesting an increased probability of crop P requirements being met from the fertiliser P. However, additional records need to be generated to establish definitively that early sowing or subsurface P reserves minimise response to fertiliser P. In general, factors that influence crop access to soil P will have an impact on response to fertiliser P.

Although this analysis shows that it is possible to ‘tighten’ the response curve for Colwell P and wheat by restricting the data for a given soil type to ideal management and seasonal conditions, the ‘outliers’ that are excluded frequently reflect an important subset of environmental conditions encountered by wheat crops in dryland agriculture.