eRcane currently operates in Reunion several decentralized concurrent sub-selection progr'ammes located in the major areas under cane differing in their agro-climatic conditions. Each year about 100,000 new seedlings are dispatched among the different sub-programs which receive mainly the same families. This early decentralized scheme of selection corresponds to a selection strategy for local adaptation with the objective of increased genetic gains in each environment. Seedlings enter into the same 14-years selection scheme across all sub-programs: two non-replicated stages followed by three replicated yield trials (RYT) stages dedicated to the _evaluation of cane yield (CY), estimable recoverable sugar (ERS) and an economic index (EI). The selection scheme is strictly the same across all sub-programmes (trial designs, rate of selection between stages, selection criteria and procedures). A __ database of 10 years trial was used in order to compare the genetic variability of agronomic traits of interest (CY, ERS and EI) in the second R YT between four regional sub-programmes. Two selection sites are located in the wet windward coast (La Mare: LM and Saint-Benoit: SB) of the island and two others in the dry leeward coast (Vue-Belle: VB and Etang-Sale: ES). The objective of this retrospective study was to compare the effect of selection pressure exerted on genotypes by the different agro-climatic environments of the selection sites. Mean genetic coefficient of variation (GCV%) across the ten series and range [min-max] of variety performances for CY distinguished clearly SB and VB sites from LM and ES sites: (i) mean GCV% for CY was · higher at SB (15.5%) and VB (15.8%) compared to LM (13.7%) and ES (11.3%); (ii). congruently, the [ min-max] range of CY performance of candidates was wider and better centered on the standard cultivar at SB ([45%-152%]) and VB ([35%-173%]) compared to LM ([42%-123%]) and ES ([50%-116%]). These results reflect higher chances of identifying