A sunflower testing network that included 25 commercial hybrids and 26 sites in Serbia was analyzed by the sites regression biplot based on grain yield data in 2006 and 2007. The scientific aims of our study were to determine the representativeness and discriminating ability of the test sites and to identify good test sites for selecting generally and specifically adapted sunflower hybrids. Correlations among test sites, revealed by biplot and with Spearman's nonparametric rank correlation coefficients for each pair of test sites, were compared. Discriminating but nonrepresentative test sites in our study were represented by Aleksa Santic (18.8 °C average temperature, 72.0 mm precipitation, and 269.2 h of sunshine) and Rimski Sancevi (18.4 °C average temperature, 79.2 mm precipitation, and 252.1 h of sunshine), based on a biplot analysis of the 2006 testing network. Sombor (20.1 °C average temperature, 52.5 mm precipitation, and 304.7 h of sunshine) was also a discriminating but nonrepresentative test site, based on a biplot analysis of the 2007 testing network. A test site that was both discriminating and representative was Kikinda (20.2 °C average temperature, 63.3 mm precipitation, and 313.7 h of sunshine), based on a 2007 biplot analysis. Sombor and Kikinda could be suitable test sites for selecting specifically and generally adapted hybrids of sunflower, respectively, for dry and hot areas and seasons, since 2007 had less precipitation and a higher mean temperature in comparison to 2006. The presence of close association between test sites Rimski Sancevi and Kikinda, based on the 2006, 2007, and combined data, and on biplot and Spearman's correlations, suggested that the same information about the genotypes could be obtained from either of these 2 test sites, and consequently testing costs could be reduced. The relationships among test sites revealed by biplot did not always coincide with Spearman's rank correlation coefficients for site pairs.

A sunflower testing network that included 25 commercial hybrids and 26 sites in Serbia was analyzed by the sites regression biplot based on grain yield data in 2006 and 2007. The scientific aims of our study were to determine the representativeness and discriminating ability of the test sites and to identify good test sites for selecting generally and specifically adapted sunflower hybrids. Correlations among test sites, revealed by biplot and with Spearman's nonparametric rank correlation coefficients for each pair of test sites, were compared. Discriminating but nonrepresentative test sites in our study were represented by Aleksa Santic (18.8 °C average temperature, 72.0 mm precipitation, and 269.2 h of sunshine) and Rimski Sancevi (18.4 °C average temperature, 79.2 mm precipitation, and 252.1 h of sunshine), based on a biplot analysis of the 2006 testing network. Sombor (20.1 °C average temperature, 52.5 mm precipitation, and 304.7 h of sunshine) was also a discriminating but nonrepresentative test site, based on a biplot analysis of the 2007 testing network. A test site that was both discriminating and representative was Kikinda (20.2 °C average temperature, 63.3 mm precipitation, and 313.7 h of sunshine), based on a 2007 biplot analysis. Sombor and Kikinda could be suitable test sites for selecting specifically and generally adapted hybrids of sunflower, respectively, for dry and hot areas and seasons, since 2007 had less precipitation and a higher mean temperature in comparison to 2006. The presence of close association between test sites Rimski Sancevi and Kikinda, based on the 2006, 2007, and combined data, and on biplot and Spearman's correlations, suggested that the same information about the genotypes could be obtained from either of these 2 test sites, and consequently testing costs could be reduced. The relationships among test sites revealed by biplot did not always coincide with Spearman's rank correlation coefficients for site pairs.