The evidence of the shift of neighbour interactions from negative to positive had been accumulated along a gradient of environment stress for decades in plants, but both shift and positive coefficient of size-density-scaling have not been studied. Almost all of the debates on scaling relationship between size (mean individual mass, m) and density (d) were related with the values of the coefficient, -3/2, -4/3 or varies with them, which is also a key challenge for the theory of metabolic ecology. We predicted that the coefficient of size-density-scaling should shift following neighbour interaction from negative to positive along a environment gradient based on the quantitative relation between m-d-scaling and relative intensity of interaction (RII). The prediction was tested by a large scale experiments along a moisture gradient from south-east, Hangzhou, to north-west, Dunhuang, in China. The scaling exponent, RII and other related indexes were measured at 10 selected plots along the annual mean precipitation/ temperature gradient.

Results/Conclusions

     The data showed that (1) the coefficient of m-d-scaling shifted from negative to positive following RII with the decreasing of productivity along the moisture gradient, (2) the shift may be driven by the coupled increasing of root/shoot ratio induced by the moisture deficit along a moisture gradient, (3) The positive scaling and RII may be the main driving force to cause the patch distributions in arid vegetation. The results may be useful for resolving the debates of obvious vary of exponent value in evaluating theory of metabolic ecology and  understanding the response of vegetation to climate change. We argued that human use of vegetation, such as grazing, should be forbidden in the environments at which the m-d scaling exponent and RII are positive for preventing desertification.