Growth anomaly of trees in some regions was detected under current episode of rapid warming. This raises a dilemma for temperature reconstructions by using tree-ring width which is believed to be the most important proxy on inter-annual temperature reconstruction during the past millenniums. Here we employed the tree-ring δ13C to reconstruct temperature variations for exploring their potential on capturing signals of rapid warming, and to test how its difference with the tree-ring width based reconstruction. In this study the mean May-July temperature (TM-J) was reconstructed over the past century by tree-ring δ13C of Chinese pine trees growing in the Nanwutai region. The explained variance of the reconstruction was 43.3% (42.1% after adjusting the degrees of freedom). Compared to a ring-width temperature reconstruction (May-July) from the same site, the tree-ring δ13C-based temperature reconstruction offered two distinct advantages: 1) it captured a wider range of temperature variability, i.e., at least May-July, even over a longer part of the year, January-September; and 2) the reconstruction preserved more low-frequency climate information than that of ring width did.The 20th century warming was well represented in the Nanwutai tree-ring δ13C temperature reconstruction, which implied that stable carbon isotope of tree rings potentially represents temperature variations during historical episodes of rapid warming. A spatial correlation analysis showed that our temperature reconstruction represented climate variations over the entire Loess Plateau in north-central China. Significant positive correlations (p<0.1) were found between the temperature reconstruction and ENSO, as well as SSTs in the Pacific and Indian Oceans. The reconstruction showed the periodicities of 22.78-, 4.16-, 3.45-3.97- and 2.04-2.83-year quasi-cycles at a 95% confidence level. Our results suggested that temperature variability in the Nanwutai region may be linked to Pacific and Indian Ocean SST variations and solar activity.