Field capacity (θf) is an important parameter for assessing soil water-holding capacity and agricultural irrigation management, which is also a required input variable for many hydrologic models. The development of pedotransfer functions (PTF) for predicting θf is essential, because direct measurement of soil θf is both time consuming and complicated. Existing PTFs for predicting θf were developed mainly based on soil data from tropical and temperate regions, and their applicability to soils in subtropical regions is unknown. It is necessary to develop a new PTF which is applicable to subtropical soils in China. In this study, we measured the θf values and soil texture, bulk density (ρb), organic matter content (fom) information of soils from 175 runoff plots in typical subtropical regions of China. The accuracy of main existing PTFs were evaluated by comparing the prediction results with measured θf values of the investigated soils. Besides, we developed a new PTF suitable for subtropical soils using a multivariate stepwise linear regression analysis method. The results showed that soil θf was generally underestimated by the 11 main existing PTFs, of which only the PTF3 from Hunan, China produced relatively acceptable estimates with a root mean square error (εr) and a mean bias (ε) of 0.051 m3/m3 and -0.024 m3/m3, respectively. The new PTF based on soil data of 175 runoff plots was given as below: θf=0.502+1.30×10-3×fcl-0.115×ρb. The new PTF indicated that θf of subtropical soils was mainly affected by clay content (fcl) and ρb. By verifying with an independent dataset, the new PTF was shown to predict θf values accurately, with an R2 of 0.48, εr of 0.035 m3/m3, and ε of -0.024 m3/m3. The PTFs developed based on temperate and tropical soils performed poorly when applying to subtropical regions. The new PTF outperforms the existing models, and it only requires the two basic soil parameters (fcl and ρb), which are easy to measure.