An Automatic Processing Framework for In Situ Determination of Ecohydrological Root Water Content by Ground-Penetrating Radar

Root water content (RWC) is a vital component in water flux in soilx2013;plantx2013;atmosphere continuum. Knowledge of RWC helps to better understand the root function and the soilx2013;root interaction and improves water cycle modeling. However, due to the lack of appropriate methods, field monitoring of RWC is seriously constrained. In this study, we used ground-penetrating radar (GPR), a common geophysical technique, to characterize RWC of coarse roots noninvasively. An automatic GPR data processing framework was proposed to (1) identify hyperbolic root reflections and locate roots in GPR images and (2) extract waveform parameters from the reflected wave of identified roots. These waveform parameters were then used to establish an empirical model and a semiempirical model to determine RWC. We validated the developed models using GPR root data at three antenna center frequencies (500 MHz, 900 MHz, and 2 GHz) that were produced from simulation experiments (with RWC ranging from 70x0025; to 150x0025;) and field experiments in sandy soils (with RWC ranging from 66x0025; to 144x0025;). Our results show that both the empirical and the semiempirical models achieved a good performance in estimating RWC with similar accuracy, i.e., the prediction error [root-mean-square error (RMSE)] was less than 8x0025; for the simulation data and 12x0025; for the field data. For both models, the accuracy of RWC estimation was the highest when applied to 2-GHz data. This study renders a new opportunity to determine RWC under field conditions that enhances the application of GPR for root study and the understanding and modeling of ecohydrology in the rhizosphere.

Automated summary

In this study, ground-penetrating radar (GPR) was used to noninvasively characterize root water content (RWC) of coarse roots. An automatic GPR data processing framework was proposed to identify root reflections and extract waveform parameters to determine RWC. The developed models achieved good performance in estimating RWC with high accuracy, especially when applied to 2-GHz data.