Objective  Ground-penetrating radar (GPR) of 1 500 MHz was used to detect the underground coarse roots and rock fragments of the typical broadleaved forest of Laotudingzi National Nature Reserve in eastern Liaoning Province of northeastern China to provide theoretical reference for forest establishment and its function of soil and water conservation.

  Method  Based on GPR detection technology in the test sample site, controlled experiments were conducted to analyze the correlations between GPR reflected wave parameters and root diameter and rock fragment diameter, and an effective estimation model was established. The coarse roots and rock fragments in the study stand were identified and their distribution patterns were determined.

  Result  (1) In the controlled experimental condition, the coarse roots and rock fragments can be effectively identified. The coarse roots with root diameter > 1 cm can form obvious hyperbolic waveform in the radar profile, while the rock fragments were the “black-white-black” waveform. (2) The correlation between total time interval and root diameter was the most significant (P < 0.01), and the fitting model Y = 0.2 862x + 1.18 (R² = 0.7098) was obtained. The horizontal and vertical diameters of rock fragments had no significant correlation with the parameters of each radar reflection wave. (3) The profile test showed that the GPR with 1 500 MHz frequency antenna had a recognition rate of 36.7% for coarse roots and 17.9% for rock fragments. (4) The results of profile test showed that the GPR effect on the identification and distribution of coarse roots was affected by the diameter and buried depth of roots, but the effect of GPR on the identification of rock fragments was not significantly correlated with its buried depth. The coarse roots with root diameter < 3 cm were mainly distributed in the 0−30 cm soil layer, and the coarse roots with root diameter > 3 cm were mainly distributed in the 20−40 cm soil layer. On the basis of effective recognition, the recognition accuracy of each root diameter in the soil layer with the maximum distribution density could reach more than 50%. The GPR identification accuracy of gravel quantity reached the maximum of 26.9% in the 20−30 cm soil layer, and the minimum recognition accuracy was only 8% in the 30−40 cm soil layer.

  Conclusion  The root system can be effectively identified by GPR under specific conditions, and the total time interval parameter is better for coarse root diameter estimation and less effective for rock fragments and rock fragment identification than for coarse roots.