Detect the size of tree roots by using ground-penetrating radar(GPR) 作者：万芳源

      author：万芳源

      Ground-penetrating radar (GPR) is a  mean of detecting buried objects with electromagnetic waves. It has been applied to detect coarse woody roots, estimate biomass, root diameter, and spatial distribution of roots.
      The aim of this blog is to give a basic understanding of GPR in its application to detecting and determining the size of roots.Above all,the antenna propagates a wave into the ground and reflected signal is detected as an induced alternating voltage and this trace of alternating voltage with time is recorded. The antenna is then moved slightly and another pulse emitted . For GPR systems used in this way, the antenna is designed so that the pulse propagates in the plane of the direction of travel along a transect with no directional information available, only the time taken for the pulse to return from all directions simultaneously. The radargrams can be viewed side by side and soil layers are seen as horizontal features across the traces while buried objects such as stones or roots show up as characteristic hyperbolas . The apex of the hyperbola occurs when the antenna is directly above the buried object. The depth to the object can be determined from the time it takes the pulse to travel to the object and back and the speed of the pulse through the soil, which depends on a number of properties of the soil, including clay content, water content, and salinity. This approach has successfully been used to locate simple buried objects such as pipelines and landmines and in principle could be a way of determining the location of roots and mapping root systems; however, root systems tend to be complex with branching, interlocking, crossing over, and growing in all three dimensions. The resulting GPR radargrams tend to be complex and difficult to de-convolve into a reliable map of a root system.
      Despite some current methodological limitations, progress has been made toward utilizing GPR to locate and determine the size of roots using simplified test beds. These controlled environment test beds allow users to explore the potential validation in site-specific conditions and gain better understanding of some of the parameters involved. The frequency of the emitted pulse has an inverse correlated influence on the size of the smallest object detectable: higher frequencies can resolve smaller objects, while low frequencies can only resolve larger objects. scientists demonstrated the use of a 1.5 GHz GPR to detect and visualize a small twig 2.5 mm diameter buried 25 cm deep in a sand box in the lab. One drawback is that high-frequency signals are attenuated rapidly and so can only be used to detect shallow objects; the maximum depth penetration of a 1 GHz antenna is about 1 m or less in typical soils .The transit time for the reflected signal provides information on the depth of the buried object, but there is additional information in the return pulse such as the signal strength and the shape of the waveform that may contain information about the size of the object. A study under controlled conditions where roots of known size were buried at known locations within a 4 m × 4 m × 2 m box of homogenous sand demonstrated a good correlation between parameters extracted from the waveform of the returned signal and the diameter of the root . Roots of the same size were buried at various depths and of different sizes (1–8 cm diameter) at one depth (50 cm), and three GPR antennas were tested (500, 800, 1,000 MHz). An unsupervised maximum-convexity algorithm  was used to de-convolve the information contained in the hyperbolas, effectively focusing the signal back to the source of the reflector in order to determine the location of the roots. The GPR trace through the center of each root was then analyzed and the zero crossing times, the time taken for the signal to switch polarity as the waveform progressed, were extracted . These timings correlated well with the size of the roots and a multivariate model was used to estimate root diameter from GPR data. Half of the dataset was used to calibrate the model, which was then tested on the other half of the dataset . Surprisingly, the 500 MHz antenna gave the most accurate estimates of root diameter and under near optimal conditions it was possible to determine the diameter of the roots ranging in size from 1 to 8 cm with a root mean squared error of 0.56 cm. The calibration derived in this example is site-specific; it is not transferable to another site or applicable if the soil moisture content changes. Calibration may be required at the time of measurements until these effects are better understood. These early studies have shown some promising results but they have tended to be under optimal conditions with single roots placed in uniform soil. The goal of using GPR to map complex root systems in difficult real world conditions such as stony, multilayered soils will require advances in antenna technology and data processing algorithms such as those that are now commonplace in medical imaging.