Using Automatic Differentiation to study the sensitivity of a crop model

Claire Lauvernet
(Irstea, 3 bis quai Chauveau - CP 220, 69336 Lyon, France)
Laurent Hascoët
(INRIA, BP93, 06902 Sophia-Antipolis, France)
Francois-Xavier Le Dimet
(Université de Grenoble, France)
Frédéric Barret
(INRA, Avignon, France)


Proceedings of the AD2012 Conference, Fort Collins (CO), USA, july 2012 (11 pages)

Abstract: Automatic Differentiation (AD) is often applied to codes that solve partial differential equations, e.g. in geophysical sciences or Computational Fluid Dynamics. In agronomy, the differentiation of crop models has never been performed since these models are more empirical than derived from equations. This study shows the feasability of constructing the adjoint model of a crop model referent in the agronomic community (STICS) with the TAPENADE tool, and the use of this accurate adjoint to perform some sensitivity analysis. This paper reports on the experience from AD users of the environmental domain, in which AD usage is not very widespread.

Keywords: adjoint mode, agronomic crop model, sensitivity analysis

Full text (pdf)

@inproceedings{LauvernEtAl12,
  author = {Lauvernet, C. and Hasco\"et, L. and Le Dimet, F.-X. and Barret, F.},
  title = {{U}sing {A}utomatic {D}ifferentiation to study the sensitivity of a crop model},
  booktitle = {{R}ecent {A}dvances in {A}lgorithmic {D}ifferentiation},
  series = "Lecture Notes in Computational Science and Engineering",
  note = "Selected papers from AD2012 Fort Collins, july 2012",
  publisher = "Springer",
  pages = "59-70",
  year = 2012
}