How widespread are yield declines in long-term rice experiments in Asia?

Abstract This paper analyzes yield trends in 30 long-term experiments (LTEs) conducted at 24 sites with intensive rice monoculture or rice-upland crop systems in tropical and subtropical regions of Asia. The objectives of the analysis were to (1) investigate how widespread yield declines are in experiments with constant levels of input and (2) study the cross-sectional relationship between the magnitude of yield trends and the initial yield level in the experiment. In rice–rice (–rice) systems, yields in the most favorable climatic season (dry season) have remained stable in many LTEs conducted outside of IRRI, mostly at yield levels of 4–7 Mg ha −1 with recommended rates of N, P and K applied. Statistically significant yield declines at p ≤0.05 were observed in only two of 21 data sets outside of IRRI, although all three LTEs at IRRI showed yield declines. One dry season experiment showed a statistically significant yield increase. In the less favorable climatic season (wet season), only two of 14 data sets outside of IRRI showed a statistically significant declining yield trend, while three of four at IRRI showed statistically significant yield declines. In rice-upland crop (typically wheat) LTEs, a significant yield decline for rice was found in only one of ten data sets. For wheat, there were no statistically significant yield declines, although there was one statistically significant yield increase. There was a negative correlation between the magnitude of yield decline and the level of initial yields, but it was not statistically significant. The data suggest that while yield declines exist in some LTEs, they are less common than previously thought, particularly at moderate yield levels. Where yield declines occurred, they were related to soil properties affected by prolonged soil wetness or soil nutrient depletion.