Regional importance of crop yield constraints: Linking simulation models and geostatistics to interpret spatial patterns uri icon

abstract

  • Over the next few decades, a central goal of agricultural research and policy will be to increase average regional crop yields in the face of diminished gains in genetic yield potential, likely climatic changes, decreased resource availability, and stricter environmental standards. Fundamental to the pursuit of effective investment strategies is an ability to quantify tradeoffs associated with potential policy and management changes. However, the data needed to predict regional yield responses to change, namely observations of yields and climatic, soil, and management conditions in farmers' fields, are often difficult to obtain. In this paper, we investigate the value of data on the spatial distribution of yields for understanding causes of landscape yield variability. Stochastic simulation models, which employ the CERES model to simulate crop yields across a landscape, are used to translate assumed spatial patterns of soil and management conditions into spatial pattern of yields. Monte Carlo simulation is then used to repeat this process for many different realizations of conditions, resulting in a modeled relationship between yield patterns and the relative importance of soil and management yield constraints, both of which can be computed in the controlled simulation environment. The derived relationship then allows one to infer from observed yield patterns the true proportion of yield variability explained by soil and management.
  • This procedure was tested for wheat in the Yaqui Valley, an intensive agricultural region in Sonora, Mexico, where yield patterns have been previously estimated with remote sensing. Comparison of simulated and observed yield patterns indicated that roughly 80% of spatial yield variance in 2001-2002 was attributable to management variations. The ability of simulation models to aid interpretation of landscape patterns is potentially invaluable for understanding yield constraints in many agricultural regions, where direct observations of soil and management variables are infeasible.

publication date

  • 2006
  • 2006
  • 2006