Comparison of selection indices to identify productive dry bean lines under diverse environmental conditions

This research was designed to quantify the effect of seasonal rainfall on production of dry bean (Phaseolus vulgaris L.) in the Rift Valley of East Africa, where insufficient soil moisture often Limits productivity, and compare methods to identify productive lines. Twenty dry bean lines were evaluated for seed yield in a randomized complete-block design under soil moisture stress and nonstress environments during 1988, 1989 and 1992. Soil moisture stress levels were imposed by altering sowing dates (early vs. late) after the onset of the rainy season. Mean seed yield increased 8 kg ha(-1) for each mm of seasonal rainfall that occurred during the growing season. Lines differed in response to soil moisture stress, and their response varied with environments. Six selection indices, including arithmetic mean (AM), geometric mean (GM), drought response index (DRI), susceptibility index (S), response to drought (RD), and percent reduction (PR), were calculated for each line to determine correlations between indices and yields in stress (Y-s) and nonstress (Y-ns) environments. AM and GM were the only indices positively correlated with both Y-s and Y-ns. AM had the largest correlation with Y-ns (r=0.90) whereas GM had the largest correlation with Y-s (r=0.93). DRI, PR and S were correlated with Y-s(r=0.86, -0.82 and -0.83, respectively) but not with Y-ns. Conversely, the index RD was correlated with Y-ns (r=0.73) but not Y-s. The indices PR and S evaluated change in Line performance across environments rather than line performance in either environment. GM and AM appear to be the most useful selection indices for identifying high yielding lines for both nonstress (early sown) and stress (late sown) conditions in the Rift Valley, while DRI has potential for use as a selection criteria only for stress environments. (C) 1998 Elsevier Science B.V. All rights reserved.