A 30 m long × 20 m wide capillary barrier cover (CBC) test site was constructed at the Jiangcungou landfill in Xi’an, China. The cover consisted of a compacted loess layer with a thickness of 0.9 m underlain by a gravel layer. After the cover surface was kept bare and exposed to natural climate conditions for nearly 5 months, one artificial rainfall event was implemented at the site. Vegetation was established at the test site after the first rainfall event. Four months later, a second artificial rainfall event was applied to the surface of the vegetated site. The pore-water pressures (PWPs) and volumetric water contents (VWCs) of the cover were monitored using jet-filled tensiometers and time-domain reflectometry moisture probes, respectively. Surface runoff and percolation were measured using field collection devices. The field measurements demonstrated a more rapid response of PWPs to the rainfall compared to the response of the VWCs. Percolation was observed when the PWPs near the interface reached the water-entry value of the gravel at local points. At that moment, the measured VWC near the interface was less than the VWC according to the water-entry value. The observation indicated that preferential flows took place in the compacted loess during the rainfall. As a result, the maximum water storage capacity was not reached at the onset of percolation. When percolation ceased, the average PWP near the interface decreased below the water-entry value, while the VWC near the interface was higher than that at the onset of percolation. Water storage at the completion of percolation was approximately 5% greater than that at the onset of percolation. Compared with the monolithic loess cover, the loess–gravel CBC increased the available water storage capacity by 41% at the completion of percolation. Vegetation had an insignificant influence on water storage capacity.