Résumé
Agriculture is already facing in several regions of the globe a warmer, more prone to drought and more variable climate. At the same time, the demand for food and non-food products increases while resources are limited. So, crops for the next decades should be more (productive, efficient) with less (water, nutrients, pesticides), a challenge that prompts the design of better varieties. Conventional or marker assisted breeding, introduction of ancestor alleles, as well as genetic engineering technologies are now able to provide us at an increasing rate with new genotypes that can be genotyped at an ever decreasing cost. All this orientate the next frontier towards our capacity to assess the benefit of this new germplasm, ie. in our capacity to perform efficient and relevant phenotyping. Because yield, the most widely assessed trait in crop breeding is also one of the most dependent upon the environment where the crop is growing, traditional schemes include the evaluation of germplasm in ‘target populations of environments’ ie. in networks of field trials representative of climate types.However, this empirical strategy faces an obvious limitation with (i) climates that are not yet present on earth and (ii) an ever increasing number of genotypes to test. An alternative is thus to use modeling to help select the right alleles for the targeted climate, either present or future. The models used are based on robust biophysical laws in which parameters are genotypic traits. National and international initiatives now promote phenotyping and modelingefforts to tackle this complex and crucial challenge.