Yield improvement in a multistage breeding program for cassava

The role of cassava (Manihot esculenta Crantz) is rapidly changing from a traditional fresh human food commodity to an efficient crop for agro-industrial processing in many parts of Asia Varietal improve ment for higher yield and root dry matter content is bringing additional cash income to a great number of small farmers. We documented the yield component data of breeding populations through five evaluation stages with and without selection. We define a best selection scheme based on the genetic, interaction, and error parameters obtained. Important conclusions are (i) broad-sense heritability and phenotypic correlations obtained at a given selection stage may lead to erroneous selection schemes, (ii) regression across evaluation stages gives the most useful information, (iii) in early evaluation stages, eliminating inferior phenotypes is more beneficial than selecting superior phenotypes, (iv) selection for root dry matter content can be conducted without serious effects on other yield components, (v) harvest index has consistently high heritability at each evaluation stage, while biomass and yield have low heritability, (vi) genotype x evaluation stage interaction for yield is greatest between single-row and plot trials, while that of harvest index is much smaller [selection at single-row trials (SRT), usually the second stage of evaluation, is most crucial to the final success of selection for higher yield], and (vii) indirect selection for yield through harvest index is more effective than direct selection by yield itself especially in the early evaluation stages. Realized selection confirmed the conclusions. Actual cultivar dissemination testifies the usefulness of this methodology.