Prediction of ungauged catchments refers, in a general sense, to predictions of streamflows, and associated quantities such as water quality indices, in catchments devoid of any streamflow (and water quality) measurements. Therefore, in a strict sense, the use of models that require calibration for the estimation of their parameters on the basis of streamflow (or other) measurements is precluded. ... [Show full abstract] The prediction problem therefore requires much better understanding of the climate, soil, vegetation and topography interactions contributing to streamflow variability than we have at the present moment. Likewise it increases the requirement to collect existing and new datasets in different climatic and physiographical settings, and to analyse observed space-time variabilities of runoff processes (within and between catchments), including extremes, and to interpret these in terms of the underlying climate-soil-vegetation-topography interactions. The new understandings obtained as a result of such investigations can be the basis of new process-based models of both water quantity and quality, which upon validation in gauged catchments subject to natural and/or human-induced changes to climate, soil and vegetation, can then begin to be applied in ungauged catchments for predictions. This paper will present the theoretical and experimental advances that will need to be made to tackle the problem of "prediction of ungauged catchments", and will survey the progress that has been made and the efforts that are currently under way to tackle this problem at a global scale.