Among environmental degradation of wood, photodegradation originated mainly by ultraviolet light is the fastest and the strongest chemical reaction process. This work on the base of near-infrared spectroscopy (NIRs) technology studied color changes of Larch wood surface irradiated by UV light with 340 nm wavelength. At different times (180, 540, 900, 1 080 h), the wood surface colorimetric index was measured and NIRs information was collected after ultraviolet light. Qualitatively chemical group changes of the wood surface were discussed according to collecting NIRs 2nd derivative spectra and their difference spectra. Quantitatively, the prediction of color the UV-irradiated wood surface were constructed using partial least squares regression method combined with leave-one-out cross-validation process. The results can be drawn as follows: (1) The color changes for UV light irradiated larix wood surface showed that lightness (ΔL^*) decreases, whereas a^* and b^* showed increasing then decreased slowly, indicating chromatic groups formed and then decreased with prolonged irradiation time. ΔE^* increase with increasing irradiation time. (2) The amorphous, semi-crystalline and crystalline wood surface occurred at the wavenumbers of 6 996, 6 773 and 6 287 cm^-1, respectively increased with prolonged UV-irradiation time. The wavenumber at 5 986 cm^-1 assigned to lignin, however, decreased with increasing UV light irradiation time to some degree. Furthermore, the difference spectra between 1 080 h UV light irradiation and control found that the positive values representing cellulose and hemicellulose showed quantitative increment, the negative value representing lignin showed a quantitative decrement in relative content, respectively. (3) The color prediction models established by NIRs information in combination with measuring the L^*, a^*, b^* showed that the determined coefficient (R²) and Ratio of performance to deviation (RPD) were 0.949 and 4.42, 0.928 and 3.73, 0.831 and 2.43 for L^*, a^* and b^*, respectively. These results confirmed that the proposed models, especially L^* and a^* models, were perfectly suitable for the in-process inspections of the UV-irradiation wood surface color and chemical content changes.