Sensitivity analysis for a complex crop model applied to Durum wheat in the Mediterranean.

2010 - European Journal of Agronomy, 32, 127-136
Richter, G.M., Acutis, M., Trevisiol, P., Latiri, K., Confalonieri, R.


Complex crop models with large parameter sets and high temporal resolution need a parsimonious screeningmethodto identify the most important parameters for the genotype by environment interaction in a specific context. Morris’ method of sensitivity analysis, changing parameters “One-at-a-Time” (OAT) was applied to yield formation of Durum wheat (DW) in two different Mediterranean sites (Italy, Tunisia) and rainfall regimes (average, −25%). 46 crop and site-specific parameters were screened and ranked for their sensitivity, and our objective was also to determine the minimum requirements for the parameter sampling strategy. The number of levels and trajectories for sampling the multi-dimensional parameter space were varied between 4 and 8 and 4 and 10, respectively.
Key results are that (1) Parameters for development and early light interception ranked higher than those for photosynthesis and assimilate partitioning, of which the latter also showed some redundancy. (2) Parameters referring to transpiration and water stress response are ranked in 11th position or greater; however, their rank differed greatly between sites in response to natural and imposed water shortage. (3) Ranking for sensitivity is stable within each site, but differs among sites; the choices of level and trajectory number have a similar impact on the ranking stability. Finally, first order sensitivity rank (strength, m*) is similar to 2nd order sensitivity (spread, s), which indicates that parameters with a high effect on output show also high interaction with other parameters or non-linearities. Overall, Morris’ method proves an effective and reliable tool to identify key phenotypic crop and management parameters and to select traits for optimizing yield in different environments.

Keywords: Crop model, genotype, environment, Morris sensitivity, Multi-dimensional scaling
DOI: 10.1016/j.eja.2009.09.002