Search

Geospatial artificial intelligence (GeoAI) has emerged as a subfield of GIScience that uses artificial intelligence approaches and machine learning techniques for geographic knowledge discovery. The non-regularity of data structures has recently led to different variants of graph neural networks in the field of computer science, with graph convolutional neural networks being one of the most prominent that operate on non-euclidean structured data where the numbers of nodes connections vary and the nodes are unordered. These networks use graph convolution – commonly known as filters or kernels – in place of general matrix multiplication in at least one of their layers. This paper suggests spatial regression graph convolutional neural networks (SRGCNNs) as a deep learning paradigm that is capable of handling a wide range of geographical tasks where multivariate spatial data needs modeling and prediction. The feasibility of SRGCNNs lies in the feature propagation mechanisms, the spatial locality nature, and a semi-supervised training strategy. In the experiments, this paper demonstrates the operation of SRGCNNs with social media check-in data in Beijing and house price data in San Diego. The results indicate that a well-trained SRGCNN model is capable of learning from samples and performing reasonable predictions for unobserved locations. The paper also presents the effectiveness of incorporating the idea of geographically weighted regression for handling heterogeneity between locations in the model approach. Compared to conventional spatial regression approaches, SRGCNN-based models tend to generate much more accurate and stable results, especially when the sampling ratio is low. This study offers to bridge the methodological gap between graph deep learning and spatial regression analytics. The proposed idea serves as an example to illustrate how spatial analytics can be combined with state-of-the-art deep learning models, and to enlighten future research at the front of GeoAI. [ABSTRACT FROM AUTHOR]

Microvibration induced by natural disturbance and human activities has an adverse effect on the operation of the large-scale and ultraprecise facilities in the world. Under such circumstances, a passive vibration control method is generally deployed for such vibration-sensitive facilities, taking the High Energy Photo Source (HEPS) in Beijing as an example, a 3 m-thick mass concrete layer forming a ring foundation was cast at the facility, where a 1 m-thick reinforced concrete slab (RC slab) lies. Since microvibration control plays a crucial role in the operation of such large-scale scientific and ultraprecise facilities and few studies have been reported for large-scale concrete layer as antimicrovibration devices, this paper presents four field tests in Beijing, China, to evaluate the vibration control capacity of a mass concrete layer. Based on a large number of field tests, the effect of applying the concrete layer is discussed, and a reference is provided for the construction of similar facilities. The vibration signals, generated by shock excitation and ambient excitation, are measured through a highly sensitive and high-accuracy vibration acquisition system. It is concluded that the existence of the 1 m-thick RC slab has little influence on the microvibration signal frequency distribution in the vertical direction and that the signals from the concrete layer and subsoil differ by approximately 10 Hz in the vertical direction while differing by approximately 5 Hz in the horizontal direction. The microvibration control ability of the concrete layer is favorable in a higher frequency band over 20 ~ 30 Hz and more than 50% attenuation can be gained through the concrete layer; however, the microvibration control ability is not significant below 20 ~ 30 Hz. The vibration levels across different heights of the concrete layer section are the same. To prevent adverse vibration disturbance below 20 ~ 30 Hz, it is suggested that the traffic and road surface conditions should be taken into consideration when choosing the construction location. In addition, a long-term monitoring shows that 75% vibration energy at the site is firmly related to the construction activities which are approximately 1.4 km from the site. [ABSTRACT FROM AUTHOR]

To study urban air quality, this paper proposes a novel categorical time series model, which is based on a linear combination of bounded Poisson distribution and discrete distribution to describe the dynamic and systemic features of air quality, respectively. Daily air quality level data of three major cities in China, including Beijing, Shanghai and Guangzhou, are analyzed. It is concluded that the air quality in Beijing is the worst among the three cities but is gradually improving, and its dynamics is also the most pronounced. Theoretically, the design of our model increases the flexibility of the probabilistic structure while ensuring a dynamic feedback mechanism without high computational stress. We estimate the parameters through an adaptive Bayesian Markov chain Monte Carlo sampling scheme and show the satisfactory finite sample performance of the model through simulation studies. [ABSTRACT FROM AUTHOR]

The catastrophic Sanyanyu and Luojiayu debris flows, which were induced by heavy rain, struck Zhouqu County in Gannan Prefecture, Gansu Province, at approximately midnight, 7 August 2010 (Beijing time, UTC + 8), causing 1765 fatalities and huge economic loss. The ZHQ seismic station is located approximately 170 m west of the outlet of the Sanyanyu gully, and its power system was destroyed by the Sanyanyu debris flow when its leading edge reached the vicinity of the seismic station. In this paper, seismic signals recorded approximately 10 min before its termination are collected and analyzed to study the Sanyanyu debris flow. A double-exponential model is first proposed to quantitatively characterize seismic energy distributions in the frequency domain, which reveals that the peak frequency of seismic signals is around 5 Hz. Influenced by the Doppler effect, the peak frequency of the N–S component is the highest, and the U–D component is the lowest. Time–frequency analysis is applied to the seismic signals. From the spectrogram, it is easily observed that the formation time of the Sanyanyu debris flow is around 23:33:10. The entire debris flow is divided into three phases with distinct frequency characteristics, using 23:36:20 and 23:37:35 as crucial times. The frequency energy distributions in the first two phases are relatively stable, and are constrained in 0–8.8 Hz and 0–17.8 Hz, respectively. For the third phase, the upper boundary of frequency energy increases in a nearly linear manner, reaching approximately 35 Hz at the end. We calculate synthetic seismograms of the Sanyanyu debris flow. Generally, synthetic seismograms have morphological features and characteristics in key stages similar to those of the actual seismic records, and their maximum values are of the same magnitude. Our results suggest that the seismic source of a debris flow can be represented by a single-force model, and reveal that real-time monitoring and rapid identification of potential debris flows using broadband seismic network records is possible, which can provide approximately 15 min of pre-event warning for local residents and hopefully save many lives. [ABSTRACT FROM AUTHOR]