In this research, hemicellulose contents of 78 wood meal samples of Acacia spp trees grown in Guangxi and another 33 wood meal samples of Acacia spp trees grown in Fujian were measured by wet chemistry. NIR spectra were also collected by a Bruker MPA spectrometer within 4 000-12 500 cm-1 of wavenumbers using a standard sample cup. Equations were developed using partial least squares (PLS) regression and cross validation for multivariate calibration in this study. High coefficients of determination (R2) and low root mean square errors of cross-validation (RMSECV) were obtained for hemicellulose content (R2=0.947, RMSECV=0.464) of Guangxi wood meal samples. Prediction produced high correlation coefficients between laboratory and predicted values, with R2 and RMSEP values being 0.925 and 0.455, respectively. A variable numbers of Fujian samples ranging from one to thirteen were used to enhance the Guangxi calibration so as to be widely used for routine assessment of wood chemistry. It was demonstrated that the addition of a single Fujian sample to the Guangxi calibration set was sufficient to greatly reduce predictive errors and that the inclusion of 3 Fujian samples in the Guangxi set was sufficient to give relatively stable predictive errors. The R2 is 0.904 and RMSEP is 0.759. The addition of different sets of 3 Fujian samples to the Guangxi calibration, however, caused predictive errors to vary between sets.