We used the weighted wavelet method to perform spectral analysis of observed long-term precipitation, streamflow, actual evapotranspiration, and soil water storage at a sub-humid mountain catchment near Tucson, Arizona, USA. Fractal scaling in precipitation and the daily change in soil water storage occurred up to a period of 14 days and corresponded to the typical duration of relatively wet and dry intervals. In contrast, fractal scaling could be observed up to a period of 0.5 years in streamflow and actual evapotranspiration. By considering long-term observations of hydrologic fluxes and storages, we show that, in contrast to previous findings, the phase relationships between water balance components changed with component period and were not perfectly in or out of phase at all periods. Self-averaging behavior was apparent, but the temporal scales over which this behavior was applicable differed among the various water balance components. Conservative tracer analysis showed that this catchment acted as a fractal filter by transforming white noise in the precipitation input signal to a 1/f flicker in the streamflow output signal by means of both spatial and temporal subsurface advection and dispersion processes and soil wetting properties. This study provides an improved understanding of hydrological filtering behavior in mountain critical zones that are critical sources of water and ecosystem services throughout the world.