Capturing the in-field spatial-temporal dynamic of yield variation

Many researchers have predicted that within-field spatial variation in crop yield could be exploited for economic benefit. However, the spatial variation of yield is influenced by season and is often temporally unstable. Parts of a field may yield well relative to the remainder of the field in one season and poorly in another, suggesting that regions in the field vary in their response to season type. We evaluate the capacity of two analytical techniques, the regression on the field mean and the regression on growing-season rainfall, to capture the variation in responsiveness of yield variation across a field. We applied these indices to a commercial 134-ha field that had been sown to wheat, Triticum aestivum cv. Calingiri, in 1996, 1999, 2001, 2003, and 2005. The slope from the regression on field mean was variable with a mean of 1 and standard deviation of 0.67. The technique successfully identified regions that were responsive and unresponsive to variations in the cropping environment. In contrast, the average slope derived from the regression on growing-season rainfall was just -0.003 +/- 0.003 t/ha. mm of growing-season rainfall. This approach failed to capture the spatial-temporal dynamic of yield variation, and implies that the overall cropping environment was poorly characterised by growing-season rainfall. Crop yields, derived from 10 soils in this field, were simulated from 1900 to 2006. The analytical techniques were applied to these simulated yields and revealed that the spatial-temporal dynamic observed in the field is partially explained by the interactions between soil type and climate. In addition, the spatial-temporal dynamic is best captured when mean field yields vary temporally by more than 1.2 t/ha if the assessments are made with 5 years of data. We further discuss the application and interpretation of these indices and the role they play in identifying soils that are responsive to season type.