Determining the age of the cattle is important in the cattle industry especially for trading purposes. Dentition and ossification scores are currently used to estimate the age of cattle. Both these methods have limitations which can create significant differences in carcass value. In this study, a fibre-optic probe-based spatially resolved diffuse reflectance spectroscopy system was used to investigate the possibility of using visible-near-infrared spectra for predicting the chronological age of beef cattle. This investigation was carried out on hide samples taken from 80 cattle with accurate date of birth and of the same breed. Spectra of hide samples taken from the neck area were used to build partial least squares models to estimate the age of the cattle. Various empirical pre-processing methods and wavelength selection using genetic algorithm (GA) were used to investigate whether these approaches could enhance model performance. Model performance was evaluated using the repeated learning-training (RLT) method. A model with the lowest average root mean square error of prediction (ARMSEP) of 2.0 years was obtained when reflected intensity spectra collected from a source-detector distance of 1.0 mm was used after pre-processing with automatic Whittaker filter and applying wavelength selection using GA indicating the feasibility of this method for estimating the age of cattle. Two approaches for utilising measurements from all the source-detector distances collected by the spatially resolved measurement system were also considered. Both these approaches, co-adding and data fusion through augmentation, led to poorer model performance compared to using measurements from only one source-to-detector distance.