Model Calibration & Evaluation

Before using a model for research or decision support applications, the user should understand how well a given model will perform for the planned work. This requires some form of model evaluation. The most preferred approach is to compare simulated values of crop traits or soil properties of interest with measured values that correspond to the range of environments and management practices of interest for the planned model application. A highly complementary approach is to vary specific inputs, including model parameters, to assess the modeled response. Such sensitivity analyses are a powerful means for ensuring that the model behaves as expected.

Comparisons with measured responses

“cross-validation”

Problem of independence of observations (White et al., 2007).

Sensitivity analysis

 

 

References

White, J.W., Hoogenboom, G., Hunt, L.A., 2005. A structured procedure for assessing how crop models respond to temperature. Agron. J. 97, 426-439.

White, J.W., Boote, K.J., Hoogenboom, G., Jones, P.G., 2007. Regression-based evaluation of ecophysiological models. Agron. J. 99, 419-427.