Development and evaluation of an improved soil test for phosphorus. 2. Comparison of the Olsen and mixed cation-anion exchange resin tests for predicting the yield of ryegrass grown in pots

A glasshouse experiment was conducted on four soils contrasting in P sorption capacity and exchangeable Ca content with perennial ryegrass using six phosphate rock (PR) sources and a soluble P source applied at four rates (including a control). After three harvests (11 weeks) replicate pots of each treatment were destructively sampled and Olsen P and mixed cation-anion exchange resin (Resin P) extractions carried out. The remaining replicated treatments were harvested another seven times (during 41 weeks). Yields (for the last seven harvests) were expressed as percentages of the maximum yield attainable with MCP.In general, the Resin P test extracted more than twice as much P as the Olsen test. There was a significant increase in Resin P with an increase in the amount of each P source in all four soils, but Olsen P values were not significantly different for soils treated with different rates of each phosphate rock. The abilities of the Olsen and mixed resin soil P tests to predict the cumulative dry matter yield from 7 harvests and the relative yield of ryegrass were compared. Correlations between measured yield (for the last 7 harvests) and soil test for each soil, and relative yield and soil test for all four soils were assessed by regression analysis using Mitscherlich-type models.When dry matter yields were regressed separately against soil test values for each soil, the Resin P consistently accounted for 18–28% more of the variation in yield than did Olsen P. For Resin P a single function was not significantly different from the separate functions fitted to MCP and PR treatments. However, for Olsen P the separate functions for the MCP and PR treatments varied significantly from the single fitted function. The Resin P test (R² = 0.84) was a better predictor of relative yields over this range of soils than the Olsen test (R² = 0.75). Two regression models based on the regression of relative yield for MCP treatments against either Olsen or Resin were developed. These models were then fitted to the relative yield data on soils fertilized with PRs only. The Olsen P model was found to be a poorer predictor (R² = 0.41) than the Resin P model (R² = 0.73) because it underestimated the observed yield of the PR treatments.