Intercrop overyielding weakened by high inputs: Global meta-analysis with experimental validation

The relative intensity of facilitative effect among coexisting plant species is frequently affected by soil water and nutrient availability, such as nitrogen (N), phosphorus (P), potassium (K), and organic fertilizer. However, to date, the general pattern of nutrient gradient effects on plant–plant interactions remains inconsistent or even opposite in the intercropping systems. To address this issue, we performed a global meta-analysis and three experiments to investigate the production efficiency of intercropping along fertilizer input gradient (low vs. high) by assessing the land equivalent ratio (LER). Trial 1 included five P gradient in the maize–grass pea intercropping system. Trial 2 consisted of two N gradient in the maize–wheat, maize–soybean and wheat–soybean intercropping systems. Trial 3 contained two water and three P gradients in the same intercropping systems as Trial 1 . The meta-analysis result indicated that the average LER for yield (LER Y ), biomass (LER B ), N uptake (LER N ), and P uptake (LER P ) were 1.49, 1.25, 1.35, and 1.66, respectively. It implied that intercropping would save 49 %, 25 %, 35 % and 66 % lands to achieve the same yield, biomass, N and P uptake amount as monoculture, respectively. Also, high fertilizer inputs significantly decreased the relative productivity and nutrient uptake. Average effect sizes of high inputs on LER Y , LER B , LER N , and LER P were up to − 0.0505 (−0.0550 to −0.0461), − 0.0267 (−0.0303 to −0.0230), − 0.0321 (−0.0395 to −0.0247) and − 0.0237 (−0.0370 to −0.0103), respectively. These negative values demonstrated that the LER markedly declined under the high-nutrient conditions. On the other hand, the results of Trial 1 confirmed that the LER Y , LER B , LER N , and LER P tended to decrease significantly with increasing P addition. In Trial 2 , the LER Y of intercropping systems tended to significantly decrease with increasing N application. In Trial 3 , the LER Y , LER B , LER N , and LER P also displayed a descending trend with increasing P addition and soil water availability. Collectively, the three trials confirmed the reliability and universality of the meta-analysis conclusions. This study provided a broad support for the hypothesis of the nutrient-dependent relative benefits in intercropping systems. The dynamic of net relative plant–plant interaction intensity in response to nutrient input gradients in the intercropping systems on the basis of conceptual model. • Average land equivalent ratio was greater than 1 in meta-analysis database. • High nutrient input decreased the land equivalent ratio relative to low input. • Field and pot-culture trials confirmed the reliability of meta-analysis results. • Relative productivity benefits were species specific and context dependent. • Results confirmed the universality of stress gradient hypothesis in intercropping.