An assessment of soil tests for phosphate for the prediction of cereal yields on a sandy soil in Western Australia

Nine soil tests for phosphate were evaluated for predicting the yield and P content of wheat, barley and oats grown on a sandy soil in Western Australia: Olsen, modified Olsen 1 (soil:solution ratio 1:5), modified Olsen 2 (soil:solution ratio 1:50), Colwell, Bray 1, Bray 2, modified Bray 2T (shaking time 10 minutes), modified Bray 2C (pH 3.7) and lactate. The soil had been fertilized 5 years previously with 20 levels each of superphosphate (OSP, range 0 to 400 kg P ha^–1) and Queensland rock phosphate (QRP, range 0 to 20 000 kg P ha^–1). For each species and fertilizer taken separately, all the tests, except for lactate, gave a good prediction of yield. When data for OSP and QRP were pooled, Bray 2 and modified Bray 2T tests were unsatisfactory predictors of both yield and P content.A linear relationship (P < 0.05) between mean soil tests value () and the standard deviation ( ) of the test value was observed for each soil test. For QRP, the results for lactate were the most variable (i.e./ was greatest) followed by modified Olsen 2 > Bray 1 > Bray 2 > Olsen > modified Bray 2C > modified Olsen 1 > modified Bray 2T > Colwell. The order for OSP fertilized soil was Bray 1 > modified Bray 2T > Bray 2 > Olsen > Colwell > modified Bray 2C > modified Olsen 1 > lactate > modified Olsen 2. For combined OSP and QRP data, the results of the Olsen 1 and Colwell extractions were the least variable.Errors in the prediction of yield ( _Y ) for all crops resulting from an error in soil test values () were calculated. For OSP-fertilized soil variability in values for the Bray-1 test provided the highest error (about 16%) in the prediction of the yield, followed by Bray 2 (12%) > Bray 2T (10%) > Olsen (8%) > Colwell (7%) > modified Bray 2C (6%) > lactate (4%). Maximum error was at yields of about 65% of maximum yield. For soil fertilized with QRP, lactate provided the highest error (about 10%) in the prediction of yield, followed by the other tests (< 6%). Maximum error was at yields of about 35% of maximum yield.The Colwell soil test gave the most accurate overall prediction of yield for both fertilizers.