Publications des agents du Cirad


Comparison between randomized complete-block and alpha designs in sugarcane variety trials in the light of a spatial modelling of fertility trends

Hoarau J.Y., Thong-Chane A., Barau L., Dumont T.. 2019. In : Proceedings of the XXX International Society of Sugar Cane Technologists Congress. Tucuman : ISSCT, p. 121-127. ISSCT Congress. 30, 2019-09-02/2019-09-05, Tucuman (Argentine).

Among resolvable incomplete block (IB) designs, the alpha-design is a very convenient design for variety trials evaluating a large number of clones (n > 30). It is a flexible design in terms of number of clones, replicates and plots per IB. Alpha-design is considered a more recommendable design than the conventional Randomized Complete-Block (RCB) design, due to the sub-division of replicates into smaller IBs that are likely more efficient to control the risk of soil heterogeneity. The efficiency of RCB and alpha designs for the evaluation of cane yield (CY), sucrose content (SC) and economic index (EI) is compared in six case studies of large sugarcane replicated trials (64 = n = 120) with two complete replicates of eight to 12 IBs. Data were analyzed using mixed linear models. In the 18 comparative analyses (six trials x three traits) broad-sense heritability (H) and residual coefficients of variation (CVe) of traits were usually similar between the two design models. Moreover, the efficiency (E) of alpha-designs relative to RCB designs ranged between 100% and 107% only. In a third analysis using a geostatistical spatial model based on P-spline regressions, H values increased on average by 6% and 7% compared to RCB and alpha-designs, respectively, while CVe was reduced by 2.9% and 3.7%. The average efficiency of the spatial analysis compared to RCB and alpha designs exceeded 150%. Trend patterns of soil fertility revealed by the spatial analysis of CY showed that heterogeneity within some IBs in alpha-designs prevented them from being more efficient than RCBs. The use of spatial analyses is recommended in any large replicated trial aiming at the assessment of sugarcane yield traits. These analyses allow to evaluate the relevance of blocking pattern in trial designs.

Documents associés

Communication de congrès

Agents Cirad, auteurs de cette publication :