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 similar to 0.2 Mio km(2) area. Samples from ten consecutive years (1998-2007) were obtained. Community biomass samples represented the carbon isotopic composition of standing vegetation on about 1000m(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 (similar to 5-10 km(2), "farm-scale"). Pair wise sampling of wool and vegetation revealed a C-13-enrichment of 2.7+/-0.7% (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 above-ground biomass (P-C4, %) was estimated with a two-member mixing model of C-13 discrimination by C3 and C4 vegetation ((13)Delta(3) and (13)Delta(4), respectively), in accounting for the effects of changing C-13 in atmospheric CO2 on sample isotope composition, and of altitude and aridity on (13)Delta(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 similar to 250 km. Importantly P-C4 was significantly higher above the 22 degrees C isotherm of the warmest month, which was obtained from annual high-resolution 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.