Meta-Analysis on grain yield effects of cereal-legume intercropping. Second Year’s Training Report. June-August 2009

Summary

In intercropping, more than one crop species are cultivated simultaneously on the same piece of land, to promote interaction between species. To determine the effects on yield performance of intercropping of legumes and cereals, a meta analysis has been set up based on data from publications on this kind of field trials as found from an extensive search on Web-of Science. The effect of intercropping has been assessed by comparing yield in pure stand and in the intercrop of each species. Two measures were considered : i) The LER (Land Equivalent Ratio), which is a simple way to compare the pure yield with the intercropping yield by a sum of ratios and ii) the overall effect decomposed into sum of a ”selection effect” and a ”complementary effect” (Loreau and Hector, Partitioning selection and complementary in biodiversity experiments, Nature, 2001). A large selection effect predicts a dominance of one of the species on the other, whereas a large complementary effect means cultures stimulate each other.
From the literature search, 8 papers were found to fulfil the requirements for a meta-analysis on yield effects of cereal/legume intercropping. For the estimation, a fixed-effects and a mixed-effects analysis of variance was used on 80 observations extracted from 8 papers. At first, an excess yield from intercropping was observed in all analysis, however, the effect varied depending on the cultivation practices, and the way in which the data were analysed in the papers.
Our analysis based on LER values revealed a significant yield effect of intercropping, but predicted also an important publication bias, implying that sometimes negative results have not been published. The simulataneous analysis of selection and complementary effects suggested this benefit was due to dominance of one of the two species, whereas the competition was
inefficient; this do not support the general principle of intercropping. However, one study
had a strong influence in the results and was responsible for a large negative complementary
effect. When omitting this study, we got an important positive complementary effect and
the selection effect remained positive but lower. The specific paper was the only one to involve oat+pea without chemical fertilizers and at a location with low annual rainfall.
In conclusion, meta-analysis was a valuable method for combining results from different studies. Unfortunately, many publications did not present enough details for been possible to include them in the analysis. Finally, the decomposed measure seems to give additional information about the mechanisms of the interactions observed.