Orthogonal polynomial models to describe yield response of rice to nitrogen and phosphorus at different levels of soil fertility

Yield responses of rice (Oryza sativa L. cv JD1187) to variable rates and combinations of fertilizer nitrogen (N) and phosphorus (P), to available soil N and P, and to cation exchange capacity (CEC) were examined in 41 field trials. The trial sites were located in the rice-growing region of Hebei Province, Northern China, covering a diversity of soil fertility levels. Relationships between crop yields, fertilizer rates and soil nutrient supply were established using an orthogonal polynomial model. The optimum economic fertilizer rates at different levels of soil N and P supply were estimated by using fertilizer-yield response functions and soil properties for a particular site. A difference in trend coefficients, reflecting yield response to fertilizer, was observed, which was mainly dependent on variance of soil fertility across 41 experimental sites under the relatively consistent climate and farming practice. The results of correlation analysis showed that there was a positive correlation between trend coefficient T-0 and yield without fertilizer (Y-0) (r = 0.80**). Trend coefficient T-1, the major effect of N fertilizer on yield, showed negative relationships with the quadratic transformation of alkali-hydrolysable N (N-s) and CEC (r = -0.60**). Trend coefficient T-2 the major effect of P fertilizer on yield, had negative relationships with the logarithmically transformed data of soil NaHCO3-extractable P (P-s) (r = -0.57**). Trend coefficients T-3, T-4 and T-5, representing linear interaction of N and P, and quadratic trend of N and P, respectively, showed no significant correlation with soil fertility variables. Yield responses to N and P fertilizers, optimum economic fertilizer rates and gross profit from fertilization decreased with increase in available soil N and P, and CEC. The model was validated in 19 field trials. The results suggest that the model developed in this study can reasonably predict optimum economic fertilizer rates through routine soil tests in the rice production region. The model can also be extrapolated to other regions and include other variable factors, provided that new relationships between yield response to fertilizer and site variables are established and incorporated into the model.