The detection precision of soluble solids in apple fruit by near infrared reflectance (NIR) spectroscopy was affected by sample temperature. The NIR technique needs to be able to compensate for fruit temperature fluctuations. In the present study, it was observed that the sample temperature (2-42 ℃) affects the NIR spectrum in a nonlinear way. The temperature model was built with R2=0.985, RMSEC=1.88, and RMSEP=2.32. When no precautions are taken, the error in the SSC reading may be as large as 2.55%°Brix. Two techniques were found well suited to control the accuracy of the calibration models for soluble solids with respect to temperature fluctuations, such as temperature variable-eliminating calibration model and global robust calibration model to cover the temperature range. And an improved genetic algorithms (GAs) was used to implement an automated variables selection procedure for use in building multivariate calibration models based on partial least squares regression (PLS). The two compensation methods were found to perform well with RMSEP1=0.72/0.69 and RMSEP2=0.74/0.68, respectively. This work proved that the compensation techniques could emend the temperature effect for NIR spectra and improve the precision of models for apple SSC by NIR.