Diffuse reflectance spectroscopy (DRS) is increasingly attracting much interest in the soil science community because it has a number of advantages over conventional methods of soil analyses. The techniques are rapid, timely, relatively cheap and hence more efficient for obtaining data when a large number of samples and analysis are required. Moreover, a single spectrum may be used to assess various physical, chemical and biological soil properties. DRS is usually coupled with a multivariate calibration technique such as partial least squares regression (PLSR) for modelling and prediction of soil properties. In this paper we explore the use of bootstrap aggregation (bagging) (Breiman, 1996) with PLSR (Martens & Naes, 1989), or bagging-PLSR, to improve the robustness of the PLSR predictions. Using various data sets that include Visible-near infrared diffuse reflectance as well as hyperspectral gamma-ray spectrometric data, results show that bagging-PLSR does, on occasions improve on predictions made using PLSR alone. When it does not improve predictions, bagging-PLSR provides results that are equal to PLSR. However, in these instances, the added advantages of bagging-PLSR are: (i) the provision of a measure of the uncertainty of predictions and (ii) bagging-PLSR is less sensitive to over-fitting than PLSR alone.