The present study quantifies the possible genotype-environment interactions and determines the ideal number of test sites for rubber trees [Hevea brasiliensis (Willd ex Adr. de Juss.) Muell Arg] in the plateau region of São Paulo State. The study was based on the genetic correlation among progenies at three different sites and on estimates of genetic gains with indirect selection of rubber trees. Twenty-two half-sib progenies were planted at the Jaú, Pindorama and Votuporanga experimental stations using random blocks with five replications and 10 plants per plot. At three years of age, the plants were evaluated for their total number of latex ring vessels (NR), rubber production (RP), bark thickness (BT) and girth (SG). There was significant genetic variability in the characters RP, SG and BT, mainly among progenies from Pindorama and Votuporanga. The effects of genotype-site interactions were significant for RP and SG. The finding of significant interactions was not a complicating factor because of the large genetic correlation detected. These results indicate that the use of two sites is more profitable when the gains in efficiency of selection are greater than 10%. Thus, Pindorama and Votuporanga will satisfactorily attend the studied region.
