Meta-Analysis in Agricultural Research: Strengths, Limitations and Challenges.

Agricultural field experiments are often characterized by low replication and high spatial variability, making it difficult to draw general conclusions from individual studies.  Meta-analysis can overcome some of these problems, and form a valuable tool to increase statistical power and identify how treatment effects depend on experimental conditions. Here we present results from several recent meta-analyses on the effect of management practices on crop yield and greenhouse gas emissions. We also discuss some of the challenges in compiling datasets for our analyses, and suggest guidelines for reporting field data so that they can be incorporated in future efforts. Two recent analyses helped us to identify management practices that reduce greenhouse gas emissions and increase crop yield in rice agriculture. In two other meta-analyses, we found that no-till practices have only small effects on N₂O emissions, but decrease crop yield by 5.7% on average. In both analyses, treatment effects strongly depended on climate, experiment duration and interactions with other management practices.  Thus, for a meta-analysis to inform decision making, average overall treatment effects are arguably less important than the relation between treatment effects and experimental conditions. Missing details on experimental design prevented us from including many studies in our datasets, whereas missing standard deviation required us to use weighting approaches that differ from conventional approaches. To further our insight in the effects of management practices on cropping systems, meta-analyses benefit most from field studies that provide a detailed description of both their results and their experimental design.