Integrating rice and wheat productivity trends using the SAS mixed-procedure and meta-analysis

Abstract

Long-term trends of crop yields have been used as a means to evaluate the sustainability of intensive agriculture. Previous studies have measured yield trends from long-term rice–rice and rice–wheat experiments in different sites from the slopes of individual site regressions of yield over time. The statistical significance of each site regression was determined but not that of the aggregate trend, which could give an indication of the magnitude and significance of global yield change. The random regression coefficient analysis (RRCA) and meta-analysis were used in this study to analyze the aggregate yield trend from several long-term experiments (LTE) across the Indo-Gangetic Plains (IGP) and outside the IGP. Both methods show that there has been a significant ( p < 0.05) declining trend in rice yield in rice–wheat LTEs in South Asia including China with the recommended rates of nutrients, but that there has been no significant change in wheat and system (rice + wheat) yields. There was no significant year × region (IGP versus non-IGP) interaction in rice and wheat yields. However, RRCA showed that the average yield trend was significantly negative (−41.0 kg ha −1 yr −1) only in the IGP. In the rice–rice LTEs, there was a significant year × site (IRRI versus non-IRRI sites) interaction during the dry season but not the wet season. Rice yields declined throughout Asia in the wet season. The average system (dry + wet season rice) yield trends were significantly negative in both IRRI and non-IRRI sites (−170.1 and −52.8 kg ha −1 yr −1, respectively) but the magnitude of yield decline was significantly greater in the IRRI sites than in the non-IRRI sites. Rice in the rice–wheat LTEs showed a significantly positive yield trend with the addition of farmyard manure (FYM) but the initial yield was generally lower with FYM than without FYM. After 15 years, yield increase due to FYM was not evident in most of the LTE.