Ground-penetrating radar allows the acquisition of many images for investigation of the pavement interior and shallow geological structures. Accordingly, an efficient methodology of detecting objects, such as pipes, reinforcing steel bars, and internal voids, in ground-penetrating radar images is an emerging technology. In this paper, we propose using a deep convolutional neural network to detect characteristic hyperbolic signatures from embedded objects. As a first step, we developed a migration-based method to collect many training data and created 53510 categorized images. We then examined the accuracy of the deep convolutional neural network in detecting the signatures. The accuracy of the classification was 0.945 (94.5%)–0.979 (97.9%) when using several thousands of training images and was much better than the accuracy of the conventional neural network approach. Our results demonstrate the effectiveness of the deep convolutional neural network in detecting characteristic events in ground-penetrating radar images.

Neural network has been successfully used for field-scale temperature estimation based on resistivity data. Although the methodology is able to estimate temperature distribution based on geophysical data with a limited number of temperature data, it has also been shown that the estimation accuracy can decrease as distance increases from temperature measurements. In this letter, we developed a resistivity-based neural kriging approach to improve the accuracy of estimated temperatures. The neural kriging method in this letter incorporated a variogram of temperature data, which constrained the spatial distribution of temperature. We examined the accuracy of standard neural network approaches, as well as the neural kriging approach, using a cross-validation test. The examination showed that when neural kriging was used, the estimation error and variogram error were reduced by 6%-87% and 191%-285%, respectively. The improvement was also significant at areas distant from well logs. The pattern of estimated temperatures matched the underlying geology and was in accordance with previous studies. The proposed methodology can incorporate other types of geophysical data for estimating other physical parameter distributions.

We estimated recent surface displacements around Bangkok by means of persistent scatterer interferometry with ALOS/PALSAR images acquired from November 2007 to December 2010. Land subsidence due to excessive groundwater pumping has been reported in this region. However, we detected ground surface uplift around the mega-city, along with seasonal surface displacement, with high spatial resolution. We then discriminated long-term natural rebound and seasonal displacement by fitting exponential and sinusoidal functions to displacement time-series, and mapped their spatial distributions. This mapping allowed us to infer that the second and third shallowest aquifers are laterally continuous, whereas the shallowest aquifer has lateral discontinuities. The temporal decay rate of the long-term rebound might reflect spatial changes of the Chao Phraya River watershed or the magnitude of the preceding groundwater extraction. We demonstrated that our method of decomposing the displacement time series into different spatial and temporal patterns is useful for understanding aquifer connectivity and the elastic response pattern in an aquifer system.

Deep learning has gained attention as a potentially powerful technique for modeling natural-state geothermal systems; however, its physical validity and prediction inaccuracy at extrapolation ranges are limiting. This study proposes the use of transfer learning in physics-informed neural networks to leverage prior expert knowledge at the target site and satisfy conservation laws for predicting natural-state quantities such as temperature, pressure, and permeability. A neural network pre-trained with multiple numerical datasets of natural-state geothermal systems was generated using numerical reservoir simulations based on uncertainties of the permeabilities, sizes, and locations of geological units. Observed well logs were then used for tuning by transfer learning of the network. Two synthetic datasets were examined using the proposed framework. Our results demonstrate that the use of transfer learning significantly improves the prediction accuracy in extrapolation regions with no observed wells.