Regression model for predicting yield of hard red spring wheat grown on stubble in the semiarid prairie
Soil testing laboratories require predictive equations to make accurate fertilizer recommendations to cereal producers in the Canadian prairies. We used results from two 12-yr experiments (one studying snow management × fertilizer rates, and the other a tillage experiment), conducted on a medium-textured Orthic Brown Chernozem at Swift Current, Saskatchewan, to develop a regression model to estimate grain yield of hard red spring wheat (Triticum aestivum L.) grown on stubble. Stepwise regression, with backward elimination, was used to develop the relationship:Y = 1006 + 10.53 WU − 0.017 WU2 + 5.52 FN − 0.095 FN2 − 33.16 SN + 0.436 SN2 − (0.112 FN × SN) + (0.057 FN × WU) + (0.159 SN × WU) − 1.26 DD (R2 = 0.89, P = 0.001, n = 262)where Y = grain yield (kg ha−1), WU = estimated water use (mm), SN = soil test N (kg ha−1), FN = rate of fertilizer N (kg ha−1), and DD = degree days >5 °C (°C-days). Water use was available spring water in the 0- to 1.2-m depth plus 1 May to 31 July precipitation + irrigation, and SN was NO3-N in 0- to 0.6-m depth, measured in fall. We validated this model using data from two other experiments in the Brown soil zone and one in the Dark Brown soil zone in Saskatchewan, and an irrigation × N rate experiment in the Brown soil zone in southern Alberta. The results showed that this model will provide reasonable yield estimates for fine-, medium- and coarse-textured soils, when SN ≤ 55 kg ha−1, over a wide range of water use. We recommend that this equation be tested by colleagues who have appropriate data and be considered for use by soil testing laboratories in Saskatchewan, Alberta, Montana and the Dakotas. Key words: Multiple regression, soil test N, fertilizer N, water use, degree-days