Abstract. This work explored the spatial variation of C3/C4 distribution in the Inner Mongolia, P. R. China, steppe by geostatistical analysis of carbon isotope data of vegetation and sheep wool. Standing community biomass (n=118) and sheep wool (n=146) were sampled in a âź0.2 Mio km 2 area. Samples from ten consecutive years (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) were obtained. Community biomass samples represented the carbon isotopic composition of standing vegetation on about 1000 m 2 ("community-scale"), whereas the spatio-temporal scale of wool reflected the isotope composition of the entire area grazed by the herd during a 1-yr period (âź5-10 km 2 , "farm-scale"). Pair wise sampling of wool and vegetation revealed a 13 C-enrichment of 2.7Âą0.7â° (95% confidence interval) in wool relative to vegetation, but this shift exhibited no apparent relationships with environmental parameters or stocking rate. The proportion of C4 plants in aboveground biomass (P C4 , %) was estimated with a two-member mixing model of 13 C discrimination by C3 and C4 vegetation ( 13 3 and 13 4 , respectively), in accounting for the effects of changing 13 C in atmospheric CO 2 on sample isotope composition, and of altitude and aridity on 13 3 . P C4 averaged 19%, but the variation was enormous: full-scale (0% to 100%) at community-scale, and 0% to 85% at farm-scale. The farm-scale variation of P C4 exhibited a clear regional pattern over a range of âź250 km. Importantly P C4 was significantly higher above the 22 ⢠C isotherm of the warmest month, which was obtained from annual high-resolution Correspondence to: K. Auerswald (auerswald@wzw.tum.de) maps and averaged over the different sampling years. This is consistent with predictions from C3/C4 crossover temperature of quantum yield or light use efficiency in C3 and C4 plants. Still, temperature gradients accounted for only 10% of the farm-scale variation of P C4 , indicating that additional factors control P C4 on this scale.