4. DISCUSSION

4.1 Effect of crack properties on preferential flow

During the infiltration event, the infiltration rate of Brilliant Blue FCF was over 1.5 times slower and the infiltration depth was 1.2–3.8 times lower than those of water, which may be attributed to the fact that the viscosity of water was different from that of Brilliant Blue FCF. A similar phenomenon was observed by Liu and She (2020a), who showed that a large organic molecule size and nonlinear adsorption behaviour within the soil matrix led to slow migration rates. This may be due to the retardation caused by the reaction between Brilliant Blue FCF and Ca2+ in carbonate laterite areas (Flury and Flühler 1995; Nobles et al., 2010). Therefore, using Brilliant Blue FCF as a staining tracer, the infiltration depth in the soil profile may not be the true water infiltration depth. The results of this experiment can be used as a basis for calibration in future field tests.
Soil cracks showed undulated reflections and individual features, whereas CK showed a more continuous interface, as derived from the envelope curves (Figure S3). Moreover, the morphological characteristics in the vertical direction of the I-shaped cracks showed a smaller envelope reflection amplitude than those of the other shapes. Cracks are always filled with sand grains or rock fragments, which had a significant role in providing high-amplitude reflections (Han et al., 2016; Dal et al., 2019) . Although the co-offset results only distinguished the envelope curves of different crack configurations, the soil cracks always had different widths and inclusions as well (Tsakiroglou et al., 2012; Yang et al., 2016), where the radar reflections provided by could not be distinguished.
The volume proportion of “manmade cracks” occupied approximately 0.16%, 0.35%, and 0.63%, corresponding to the 1-, 1.5-, and 2-cm cracks, respectively, as shown in the infiltration experiments. Although the volume of cracks was less than 1%, the existence of cracks considerably accelerated the water infiltration rate, while the infiltration rate, maximum dye-penetration depth, cumulative infiltration, and wetting advancing rate of CK were lower than those of other treatments containing cracks. Cracks provide channels or pores for fluid flow, even though only the I- and Λ-shaped cracks produce preferential flow. Our findings did not concur with those of Liu and She (2020a), who suggested that crack volume is not a good predictive parameter for infiltration characteristics. However, Yang (2020) concluded that the crack volume is an important indicator of the initial infiltration rate, and that they are positively correlated.
Preferential flow was observed only in the I- and Λ-shaped treatments, along the crack pore paths. These results are consistent with those of Liu and She (2020a) and Yang et al. (2016), who demonstrated that cracks influence preferential flow. Liu and She (2020a) showed that I-shaped cracks retained more water than V-shaped or non-cracks, in which preferential flow was the dominant flow type. Yang et al. (2016) revealed that the orientation of subsurface flow varied in accordance with the connectivity of the cracks or underlying structure, which resulted in lateral flows and deep percolation occupying approximately 32%–36% of the total rainfall.

4.2 Necessity of combining geophysical surveys and column experiments

GPR was successfully applied in karst topography to reveal structural information about the subsurface and achieve the transformation of invisible geological features into trended signals (Estrada-Medina et al., 2010; Fernandes et al., 2015), thus providing more detailed information about soil profiles (Busch et al., 2013). This geophysical technique retrieves information with the advantages of low-cost, efficiency, and less labour-intensive. We explored the distribution and characteristics of soil cracks on a large scale in the field, and verified the configuration by GPR. On the one hand, geophysical surveys provide a reference for the crack properties before the simulation experiments. The column experiments were conducted based on the results of GPR and excavations and visualise the flow infiltration process (Liu et al., 2020a). On the other hand, it provides a new insight for interpreting isolated cracks from radar signals. The combination of GPR with field observations and column experiments has shown great potential for investigate the effect of isolated soil cracks on preferential flow.