Abstract: Dye tracing experiments were carried out in apple orchards of 20 years old in Changwu tableland using Brilliant Blue FCF solution of 4 g·L^-1. Three water application methods (sprinkler, flood, stemflow simulation) were used to simulate regular 50 mm rainfall infiltration, intense 50 mm rainfall infiltration and stem flow infiltration (18 L of stemflow amount), respectively. The results showed that for sprinkler water application, the average dye-stained depth was 0.28 m and shallow. The average and median values of uniform infiltration fractions were larger than 0.65 and 0.7, respectively, indicating that the proportion of the uniform part was larger in the infiltration process, and piston flow was the main mechanism of rainfall infiltration in apple orchards on Loess Tableland. For flood water application, the average and maximum dye-stained depths were 0.53 m and 0.59 m, respectively. The depths were also shallow, indicating that it was difficult to recharge deep soil water for a short time under the conditions of intense rainfall and activation of macropores. For stemflow simulation water application, the maximum dye-stained depth was 0.46 m and also shallow, indicating that it was difficult for rainfall to recharge deep soil along the roots by stem flow.By comparing the dying of live and decayed roots, it was found that live roots (≥2 mm) had minor water conductivity, while the infiltration channels formed by degraded roots (≥2 mm) had better water conductivity. Most of the apple tree roots were horizontally distributed near the soil surface, and the lateral preferential flow was more likely to occur when these surface roots decomposed, thus preventing rainfall from infiltrating into the deep soil. In conclusion, piston flow was the main recharge mechanism of deep soil water recharge in apple orchards on Loess Tableland of China, and the channels such as soil macropores and degraded vertical roots might also be an approach to deep soil water recharge.