Abstract: The changing trend of soil available phosphorus (Olsen-P) content in soil and its relationship with soil phosphorus surplus and crop yield are fundamental when making appropriate phosphate fertilizer recommendations. In this paper, the influences of long-term fertilization on crops phosphorus uptake, soil phosphorus surplus, changing trend of soil available phosphorus content and relationships of soil available phosphorus content with soil phosphorus surplus and crop yield were investigated through 34 years (1981-2015) long-term trial in loessial soil region on the Loess Plateau. The experiment had a completely-randomized-block split-plot design in triplicate. Two main-plot treatments were no farmyard manure and farmyard manure (M), and four subplot treatments were CK (no fertilizer), N (application of chemical fertilizer N), NP (application of chemical fertilizer NP) and NPK (balanced application of chemical fertilizer NPK), respectively. The results showed that fertilization treatments and crop types significantly influenced uptake amount of phosphorus and soil phosphorus surplus. Averaged over time from 1981 to 2015, wheat mean phosphorus uptake amounts of CK, N, NP, NPK, M, MN, MNP and MNPK were 8.63, 10.64, 16.22, 16.21, 16.25, 17.83, 20.39 and 20.27 kg·hm^-2, while rape phosphorus uptakeamounts of eight treatments were 4.40, 8.38, 15.08, 15.71, 10.52, 11.23, 17.96 and 17.66 kg·hm^-2, respectively. The surplus amount of soil phosphorus significantly correlated with the amount of phosphorus applied to soil. When soil phosphorus surplus amount equal zero, wheat and rape phosphorus input amounts were 10.47 kg·hm^-2 and 6.97 kg·hm^-2, respectively. Soil phosphorus surplus amount significantly influenced the changing trend of available phosphorus content in soil. CK and N treatments had no phosphorus input, and soil available phosphorus content exhibited a declining trend, annually decreased by 0.16 mg·kg^-1 and 0.15 mg·kg^-1, respectively. In contrast, NP, NPK, M, MN, MNP and MNPK six treatments were applied with phosphate fertilizer every years, and available phosphorus content gradually increased along with the duration of trial, with annual increase by 0.02-0.33 mg·kg^-1. Soil available phosphorus content significantly correlated with phosphorus accumulative surplus amount, and the linear models were y=0.012x+9.33 and y=0.009x+11.72 in manure and no manure treatments, respectively. In no manure treatments, wheat yields significantly positively correlated with soil available phosphorus content, however, in manure treatments, their relationships did not reach a significant level. The relationship of wheat grain yield with available phosphorus content could be significantly fitted by piecewise linear model, and available phosphorus agronomy threshold of wheat was 14.99 mg·kg^-1. Rape grain yield also increased with increasing soil available phosphorus content, but the relationship was not significant. This indicated when soil available P content is higher than 14.99 mg·kg ^-1, application of phosphate fertili-zer should be reduced or even avoided for planting wheat in loessial soil region on the Loess Plateau.