The catchment scale application potential evaluation of GPR acquired using 50 MHz RTA and 500 MHz shielded antenna in detecting karst surface structure in the Houzhai catchment is carried out. The catchment is initially divided into three regions as low, moderate and high potential, based on field observations, the topography (Fig. 1) and vegetation cover data (see Appendix). Different potential degrees reflect the reliability of GPR data interpretation for soil distribution and epikarst depth by considering surface and subsurface influencing factors. Table 1 summarizes the findings of the present study as well as inferred from literature review.
Table 1
the potential degree evaluation of GPR applied to the Houzhai catchment
Potential degree | Corresponding geomorphic region and land use | Influencing factors | Effect of interpreting soil distribution for 50 MHz RTA and 500 MHz shielded antenna | Effect of interpreting epikarst depth for 50 MHz RTA and 500 MHz shielded antenna |
low | Forest land, High voltage line within 100 meters range | Tree trunk, tree root, high-voltage lines, moisture content, soil thickness | Bad effect for two kinds antennas, Great multiple solutions, High difficulty to interpret due to signals overlap of trunk, root and inclined interface reflection Poor methods to filter interference | Bad effect for two kinds antennas, Great multiple solutions, High difficulty to interpret with further considering the soil thickness and moisture content, High difficulty to confirm the epikarst depth, |
Moderate | karst slope, bottom of peak cluster depression, bottom of peak cluster valley | Moisture content, Soil layer thickness, Hills reflection | Interpret the soil structure in karst slope is moderate difficulty for 500 MHz but high difficulty for 50 MHz Moderate difficulty in the depression and valley for 50 MHz RTA due to the tilt signals overlap, 500 MHz antenna is not suitable in the depression and valley due to the soil thickness and moisture content | Shallow karst cave or pipelines under karst slope can be possibly detected by 50 MHz, more difficulty for 500 MHz; Interpretation difficulty of epikarst depth increases from top to foot of karst slope for both antennas; Reflection signals of epikarst inclined interface under valley and depression can overlap with the hill reflection signals for 50 MHz if soil layer is not too sick. |
high | Basin, plain, agricultural land use | Soil layer thickness, Moisture content | Soil layer depth and soil-rock interface can be quantitatively interpreted for 50 MHz RTA. The accuracy can be verified by auger. High difficulty for 500 MHz to penetrate soil bottom | High difficulty to interpret epikarst under the soil layer for two kinds antennas; Epikarst information may not be acquired when soil thickness is close to 4 meters for 50 MHz RTA, shallower for 500 MHz |
Low potential degree regions (e.g., forest lands and high voltage lines) are the areas where GPR data show multi-source interference as compared to signals of interests leading to difficulties in interpreting the soil and epikarst. Most of the forest lands are mainly distributed in the peak-cluster depression and peak-cluster valley of the catchment while some of forests are also grown at the flat sites. The influencing factors include tree trunks, roots system, high voltage lines, soil layer depth and moisture content. The diameter of tree trunk is usually more than 10 cm, reaching the resolution of the 50MHz and 500 MHz antennas. The effects of tree roots on GPR results as reflection of radar waves, have been documented in literature (e.g., Tardío et al. 2016; Alani and Lantini 2020). The reflection from trunks and roots are mixed with those from subsurface media, making the GPR results obscure and separation of real sources of either horizontal or tilt signals is difficult. Tall and dense growing trees imply a large amount of soil with a relatively large thickness. Whether radar wave can penetrate the maximum depth of soil or not has great uncertainty, especially for high frequency antenna. The electromagnetic field generated by the high voltage lines gives the GPR data great interference which is difficult to eliminate.
Moderate potential degree regions are the areas where partial information of the target is acquired while rest is lost because of the overlaping or attenuation of signals. The peak-cluster depression bottom, the bottom of peak-cluster valley and slope without forests can be classified into this catagory. The influencing factors include reflections from hills, soil thickness and moisture content.
In the first stage, the effects of karst slope on GPR results vary with the frequency antennas used as well as soil distribution on the slope. The general feature of karst slope is, the soil distribution becomes deeper and heterogeneous from upslope to downslope. The epikarst depth also increases gradually in downslope direction. The shallow and discontinuous surface soil is available for low frequency antenna to detect caves (Čeru et al. 2018; Hussain et al. 2020). Generally, shallow caves or pipelines are deeper than epikarst lower boundary and could be detected in favorable conditions as thinner and dry soil. The reflection from caves or pipelines can also account for GPR signals penetration till epikast bottom. But it is difficulty for 50 MHz data to interpret the upslope fissure soil distribution due to low resolution and little soil distribution. As for the 500 MHz (high frequency) antenna, the achievement of the interpretation of slope epikarst bottom requests lower moisture theoretically. The GPR coherence attribute is beneficial in interpreting the valid and invalid signals interface, which may correspond to the epikarst bottom (Gao et al. 2020). However, the level of difficulty in interpreting epikarst bottom depth increases from upslope to downslope due to the increase of soil thickness and moisture content.
Secondly, the use of low frequency antenna is recommended for the exploration of soil layer bottoms in peak-cluster depression and valley. In terms of unshielded 50 MHz RTA, the signals reflected by hills overlap with the signals reflected from the subsurface inclined interface due to the depression and valley close to hills, making difficult to accurately confirm the exact source of tilt signals. The tilt signals can be eliminated by F-K filtering, and the retained signals can reflect the depth of soil lateral layers and soil-rock interface in the depression and valley through extracting the average amplitude attribute and coherence attribute (Gao et al. under review). Unfortunately, the information of the inclined interface has been lost. The surface of depression and valley is usually entirely covered with the soil layer which is deeper than that of soil found on slope. Additionally, the bottom of karst peak-cluster depression and valley is a sink for the rainfall and runoff from the adjoining areas. Therefore, the these are the region of high soil moisture contents than that of slope soil. It is challenging to detect the epikarst depth under the depression and valley using low frequency antenna.
Summarizing the above GPR interpreting effects of high and moderate regions with high and low center frequency antennas, we sum up above results and draw the key information into one simple sketch as shown in Fig. 15.
High application potential is the region where GPR information is solely obtained from the subsurface structures without having any influence from the external factors as explained above. This region corresponds to the basin and plain of the catchment, excluding the sites covered by forest. Results of GPR data acquired from this region have high reliability in the soil and epikarst delineation. The soil distribution can be interpreted by analyzing the average amplitude attribute and coherence attribute and result accuracy can be verified by drilling. However, GPR acquiring the depth information of epikarst under relatively thick soil layer is most difficult. The limited thickness of soil layer can be referred to the example of the Zhongba depression. The thickness and water content of soil layer are usually higher than those of slope. The thickness of soil layer in the Zhongba depression is ~ 4 meters (Gao et al. 2020). The measured data was acquired in the premise of no rainfall event and hot weather lasting more than a week. The 50 MHz antenna just detected the maximum depth of the soil layer and did not receive the echo signals of the deeper epikarst. Thus, we simply consider the soil thickness close to 4 meters is not conducive for low frequency antenna to obtain the structure information of epikarst. As for the 500 MHz antenna with high resolution, its requirement for the soil thickness is much shallower.
The above information is summarized in one rough map about the application potential evaluation of GPR application in the Houzhai catchment based on the DEM, the vegetation data and locations of high voltage lines (Fig. 16).
Considering the seasonal variability of climatic conditions in the Houzhai catchment, the soil moisture content changes from low to high with dry to rainy seasons, receptively. The rainy period is not a good time for GPR survey because of high soil conductivity. We recommended the months of October and November as the best time for GPR acquisition in the catchment. On the one hand, this period is at the end of the rainy season and the soil in dry season has relatively low moisture content and conductivity (Yang et al. 2019). On the other hand, the majority place of the low uncertainty region is the farmland. October and November are the period between the summer crop being harvested and the winter crop not fully planted. It is also a good period to from the view of not destroying crops.